Perspectives on Causation 9781472561022, 9781849460866

Citation preview

Foreword The law has to deal with the relationship which must exist between the actions or omissions of one person and harm happening to another in order that the first person should have responsibility for that harm. This problem arises in many fields of law: for example, in criminal law, in the law of tort or negligence and in relation to damages for breach of contract. The basic approach is to hold that the relationship is established if the first person’s conduct has caused the damage, because such a link suggests that the first person’s conduct brought about the harm to the other. This formulation does not deal with the situation where the actions of two or more persons have brought about the harm, so it is modified to create the necessary link where the conduct of the first person contributed to bringing about the harm. In order to establish liability this link must be proved to the standard required in the relevant field of law. But what if the first person employs the other to do work which carries danger unless certain precautions are taken? He does not take these precautions and the employee suffers harm of the type caused by the danger; but such harm might occur from other sources than the danger the employer has failed to protect against and the relevant science cannot establish from which source the harm suffered by the employee emerged. It is difficult to say that the employer’s conduct is proved to have caused the harm suffered by the employee. On the other hand, the employer’s conduct may have caused the harm. Should the relationship required for liability be altered to include this type of case? Would it be necessary to establish in such a case that the relevant science could not identify the source of the harm? If not, should the test be that the impugned actions may have caused the harm suffered? If the person suffering the harm was about to embark on a highly remunerative career which the harm prevented him from doing, should the person whose conduct is impugned be liable for the consequent loss? If so, how definite would the injured person’s prospects have to be for this to be so? There are also difficult questions about the methods which can be used to establish causation. In short, the use of the concept of causation in the law gives rise to philosophical legal and scientific problems. This collection of essays discusses many of these in a very thoughtful way and I believe it is a very constructive contribution to the development of the law in many fields. Rt Hon Lord Mackay of Clashfern House of Lords May 2011

Preface The central issue at the heart of tort law over the last decade has been the problem of establishing causation. This area of law has seen itself subject to rapid development, with several landmark decisions in the UK over the last few years, as well as significant developments in the United States, Canada, Commonwealth countries and France. This has taken place against a backdrop of an increasingly important role being played by complex scientific evidence in the resolution of judicial controversies concerning causation. During this period, and partly in response to these challenges, there has been a rise in the development of the legal theory of causation. In the light of these developments, the School of Law of the University of Aberdeen convened an international conference at the University entitled ‘Perspectives on Causation’, which sought to determine how British law should seek to influence and be influenced by developments in other countries. The problems of causation were approached from the three perspectives of law, scientific evidence and legal theory. It was felt that publication of a book on the developing law of causation from these three perspectives would be a topical and valuable contribution to legal scholarship. With the encouragement of Richard Hart and the support of Hart Publishing, I have been able to bring together this collection of essays, written by a distinguished group of scholars in the causation field. The conference was itself was generously sponsored by the Clark Foundation for Legal Education (primary sponsor), the Society of Legal Scholars, the London Chambers of Four New Square, the Medicines and Health Care Products Regulatory Agency (MHRA), the Faculty of Advocates, Simpson and Marwick Solicitors and Stronachs. The production of a volume of essays of this magnitude has inevitably taken a considerable period of time to put together. I am most grateful for the assistance I have received from the contributors and the production team at Hart Publishing; in particular, I wish to thank Richard Hart and Rachel Turner for their patience and support. The collection endeavours to state the law as at 1 January 2011, but it has been possible to add some additional references to more recent material, including that of the Supreme Court’s decision in Sienkiewicz v Grief (UK) Ltd. Richard Goldberg King’s College Old Aberdeen April 2011

Notes on the Contributors LORD HOFFMANN has been Honorary Professor of Intellectual Property Law, Queen Mary, University of London since June 2009, following his retirement as a Lord of Appeal in Ordinary. He attended the University of Cape Town and then Queen’s College, Oxford, as a Rhodes Scholar and won the Vinerian Scholarship. He was an advocate of the Supreme Court of South Africa 1958–60, called to English Bar by Gray’s Inn in 1964 and appointed Queen’s Counsel in 1977. He was appointed a judge of the High Court (Chancery Division) 1985–1992, elevated to the Court of Appeal 1992–1995 and appointed a Lord of Appeal in Ordinary 1995–2009. From 1980 to 1985 he was a (part-time) member of the Courts of Appeal of Jersey and Guernsey. Lord Hoffmann remains an active arbitrator and mediator at Brick Court Chambers and continues to be a non-permanent judge of the Court of Final Appeal of Hong Kong. In his judicial career he has given judgments which have had a profound effect in shaping modern English law, ranging from the reading of arbitration clauses and patent specifications to the scope of the Convention rights under the Human Rights Act. Of his many influential speeches, some have concerned the problems in establishing causation in the context of indeterminate defendants and of medical negligence. His publications include: ‘The Universality of Human Rights’ (2009) 125 LQR 416; ‘Patent Construction’ (2006) 35 CIPA J 727; ‘Tax Avoidance’ (2005) BTR 197; ‘Causation’ (2005) 121 LQR 592; ‘Separation of Powers’ (2002) 7 JR 137; and ‘Bentham and Human Rights’ (2001) 54 CLP 61. JOSEPH SANDERS JD, PhD (Northwestern) is AA White Professor of Law at the University of Houston. His research and writing focus on both socio-legal and traditional legal topics, including jury decision-making, the attribution of responsibility across cultures, torts, mass torts and scientific evidence. With respect to the current book, his relevant writings are: Bendectin On Trial: A Study of Mass Tort Litigation (The University of Michigan Press, 1998) and Modern Scientific Evidence (with Edward Cheng, David Faigman, David Kaye and Michael Saks) (West/Thomson Publishing Co, 2007–2008) as well as a number of articles on scientific evidence and expert witnessing, including most recently, ‘The Insubstantiality of the Substantial Factor Test For Causation’ (with Michael Green and William Powers Jr), (2008) 73 Missouri Law Review 399, and ‘Science, Law and the Expert Witness,’ (2009) 72 Law and Contemporary Problems 63. He has served on the advisory panel for the Law and Social Science Program at the National Science Foundation. He is a member of the American Law Institute and the Law and Society Association. Professor Sanders edited the Law and Society Review from 2001 to 2004. He teaches courses in torts, products liability, scientific evidence, the jury, and law and social science. MARTIN HOGG LLB, LLM, PhD (Edin) is a Senior Lecturer at the Edinburgh Law School. His main areas of research lie in all aspects of the law of obligations, with a current particular emphasis on obligations theory, causation and the concept of damage. He has published on a number of aspects of causation and the law and in relation to the idea of damage,

xii  Notes on the Contributors including that of asbestos related injuries. In September 2008, Dr Hogg gave oral evidence before the Justice Committee of the Scottish Parliament in relation to the proposed Scottish legislation relating to liability for asbestos-related pleural plaques. He is the author of Promises and Contract Law: Comparative Perspectives (Cambridge University Press, 2011). JONATHAN MORGAN MA (Oxon), PhD (Cantab), read jurisprudence at Balliol College, and wrote a doctoral thesis on contract law theory at Corpus Christi College, Cambridge. He was formerly Fellow and Director of Studies in Law, Christ’s College, Cambridge (2004– 9). Since 2009 he has been Fellow and Tutor in Law at St Catherine’s College, and CUF lecturer in the Faculty of Law in the University of Oxford. He has published articles on various aspects of tort, contract and human rights law. Most pertinent for present purposes are his notes of Fairchild v Glenhaven in (2003) 66 MLR 277, Gregg v Scott in [2005] LMCLQ 281 and the Pleural Plaques Litigation in [2006] CLJ 269. He has also reported the historical response of English law to accidents caused by new technologies (railway fires, boiler explosions and asbestos), for the Cambridge European Legal Development Project (‘Technological change and the development of liability for fault in England and Wales’ in Miquel Martin-Casals (ed), The Development of Liability in Relation to Technological Change (CUP 2010)). He is one of the editors of Hepple and Matthews: Cases and Materials on Tort, 6th edn, (Oxford University Press, 2008). RUSSELL BROWN BA(Brit Col), LLB (U Vic), LLM, SJD (Toronto) joined the Faculty of Law at the University of Alberta in 2004 and is currently an Associate Professor. A recipient of the Faculty’s Tevie H. Miller Teaching Excellence Award, Professor Brown currently teaches Tort Law and Civil Procedure. His research interests lie generally within tort and public authority common law liability for takings. He has been called to the Bars of British Columbia (1995) and Alberta (2008). Professor Brown is a Fellow of the Institute for United States Policy Studies at the University of Alberta, a member of the governing board of the Canadian Forum for Civil Justice and of the advisory board to the Canadian Constitution Forum. He holds a master’s and doctoral degree in law from the University of Toronto. DUNCAN FAIRGRIEVE LLB (Lond), DPhil, MJur (Oxon), Maitrise, (Paris I, Sorbonne) is a Senior Fellow in Comparative Law at the British Institute of International and Comparative Law, London. He is also responsible for the Product Liability Forum at the Institute. As an academic, he has published widely, and his work has been cited extensively by the Courts, including the Court of Appeal, House of Lords and the French Conseil d’Etat. He is a Visiting Professor at the Institute of Law, Jersey. Dr Fairgrieve’s research interests focus upon comparative law, spanning both private and public law. In the sphere of public law, he has a particular interest in the liability of public authorities in tort law, which is reflected in his monograph, State Liability in Tort (Oxford University Press, 2003). In the sphere of comparative private law, Dr Fairgrieve works on a variety of topics, in particular tort law and product liability. He is the founding Director of the Tort Law Centre at the Institute and is responsible for projects on product liability, clinical negligence and public authority liability. In the sphere of product liability, Dr Fairgrieve has edited a leading analysis of the comparative law of product liability: Product Liability in Comparative Perspective (Cambridge University Press, 2005). He is a door tenant at One Crown Office Row.

Notes on the Contributors  xiii FLORENCE G’SELL-MACREZ is a Lecturer in Law at University of Paris I, a member of the l’Observatoire du Principe de Précaution and serves on the editorial board of the European Review of Private Law. PHILIP DAWID is Professor of Statistics at the University of Cambridge. He is Chartered Statistician and Fellow of the Royal Statistical Society, which has awarded him the Guy Medal in Bronze and in Silver; Fellow of the Institute of Mathematical Statistics; Member of the International Statistical Institute; and an organiser of the Valencia International Meetings on Bayesian Statistics. He was President of the International Society for Bayesian Analysis in 2000. He has served as Editor of the Journal of the Royal Statistical Society (Series B) and of Biometrika, and is currently an Editor of Bayesian Analysis. Much of Philip Dawid’s research is focused on the logical foundations of probability and statistics, with emphasis on the Bayesian approach. His co-authored book Probabilistic Networks and Expert Systems (Springer, 1999) won the first DeGroot Prize for a Published Book in Statistics. His decision-theoretic framework for statistical causality has proved valuable, both conceptually and practically. He has a long-standing interest in the rational analysis of legal evidence, and led an international research project applying Bayesian networks to complex cases of forensic identification from DNA profiles. He recently directed a multidisciplinary research programme ‘Evidence, Inference and Enquiry’ at University College London. RICHARD GOLDBERG LLB (Strath), LLM, PhD (Lond) is Reader in Law at the University of Aberdeen. Prior to taking up his Readership in January 2006, Dr Goldberg previously taught at the London School of Economics, King’s College London, Queen Mary and Westfield College and, most recently, the University of Birmingham. He has published widely on tort law, product liability and medical liability, including: his monograph Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Hart Publishing, 1999); Pharmaceutical Medicine, Biotechnology and European Law (co-editor, with Julian Lonbay) (Cambridge University Press, 2001); Product Liability, 2nd edn (co-authored, with Professor CJ Miller) (Oxford University Press, 2004); ‘Licensing of Medicines’ in The New Oxford Companion to Law (Oxford University Press, 2008), ‘Causation and Defences’ in A Grubb, J Laing, J McHale (eds) Principles of Medical Law (Oxford University Press, 2010); and ‘Breach of Confidence’ in J M Thomson (Gen Ed), Delict (Scottish Universities Law Institute), (W Green/Sweet & Maxwell) (2008). In addition to editing Perspectives on Causation, he is currently working on a monograph entitled Medicinal Product Liability and Regulation for Hart Publishing. He is consultant editor of the Medicinal Products and Drugs entries in Halsbury’s Laws of England and coordinator of the Delict Title for the Stair Memorial Encyclopaedia. Dr Goldberg serves on the editorial board of the Medical Law Review. In 2011, Dr Goldberg is a Visiting Scholar at the Faculty of Law, McGill University, Montreal. PETER FELDSCHREIBER is dually qualified as a barrister (2000 call) and physician (Kings College Hospital 1974) and Fellow of Faculty of Pharmaceutical Medicine, Royal College of Physicians. He specialises in medical and healthcare law including medical products liability, pharmaceutical and medical devices regulatory law, and clinical negligence. He is a Tenant in the Chambers of Roger Stewart QC, Four New Square, Lincolns’ Inn. Peter holds appointments as Senior Medical Assessor and Special Litigation Coordinator to the

xiv  Notes on the Contributors Medicines and Healthcare Products Regulatory Agency (MHRA). He is General Editor of The Law and Regulation of Medicines (Oxford University Press, 2008) and co-author of the chapter on the regulation of healthcare products in Butterworths’ Healthcare Law and Practice. LEIGH-ANN MULCAHY QC MA (Cantab), LLM (Osgoode), Dip ECLaw (Lond) specialises in professional negligence (particularly claims involving lawyers, accountants and construction professionals), product liability law (especially pharmaceutical cases, multiparty litigation, cases involving complex scientific and technical evidence and commercial product liability claims), insurance law, general commercial litigation and energy/utility law. Until her appointment to silk, Leigh-Ann held the appointment of Junior Counsel to the Crown (A Panel) and acted for a wide range of Government departments and bodies on issues including public law, EC law and human rights. Leigh-Ann is Consultant Editor and contributor to The Regulation and Law of Medicines (Oxford University Press, 2008), Editor of Jackson & Powell on Professional Liability (Thomson, 2007) and General Editor of Human Rights and Civil Practice. She is a CEDR accredited mediator. SIMON DAY PhD (Napier) is a Statistical Expert of Roche Products Limited, advising on statistical, clinical trials and regulatory matters throughout the company. Simon is a former president of the International Society for Clinical Biostatistics; he is an associate editor of Statistics in Medicine, on the editorial board of Pharmaceutical Statistics and is the Read Papers Consulting Editor for the Journal of the Royal Statistical Society (JRSS). He has also served on the editorial boards of Controlled Clinical Trials and JRSS Series A. He has published widely in statistical and medical journals, is author of one book, Dictionary for Clinical Trials, 2nd edn (Wiley 2007) and is joint editor of the Textbook of Clinical Trials, 2nd edn, (Wiley 2006). He is an Honorary Research Fellow at Napier University in Edinburgh; a member of the External Advisory Panel for the Department of Statistics at the University of Oxford; an Associate on the Faculty of the Johns Hopkins University in Baltimore; and a regular lecturer and member of the Educational Advisory Board at the Vienna School of Clinical Research in Austria. He also serves on the Hammersmith and Chelsea Hospitals Research Ethics Committee in London. RICHARD WRIGHT is Distinguished Professor of Law at the Illinois Institute of Technology’s Chicago-Kent College of Law. He has been a visiting professor or fellow at the University of Texas, the University of Canterbury in New Zealand, the University of Melbourne, the Universidad Torcuato di Tella in Argentina, and Brasenose College, Oxford and the Law Faculty of the University of Oxford. He is an elected member of the American Law Institute and has been an active participant in its revision of the Restatement of the Law of Torts. He has served as chair of the Section on Torts and Compensation Systems of the Association of American Law Schools and is a member of the Advisory Boards of the Torts, Product Liability and Insurance Law Journal of the Social Science Research Network, the Journal of Tort Law, and the Center for Justice and Democracy. His published work, including extensive writings on causation in the law, appears in several international collections of leading scholarship on tort law and legal philosophy. ARIEL PORAT JSD, LLB (Tel Aviv) is the former Dean of Tel Aviv University Faculty of Law. He is Alain Poher Professor of Law at Tel Aviv University, Global Visiting Professor of

Notes on the Contributors  xv Law at NYU Law School, and Fischel-Neil Distinguished Visiting Professor of Law at the University of Chicago (on an annual basis). He was also a Visiting Professor at the University of California at Berkeley, Columbia University, and the University of Virginia. He is a member of the American Law Institute, a board member of the American Law and Economics Association and a former president of the Israeli Law and Economics Association. From 1997–2002, Ariel Porat was the Director of the Cegla Center for Interdisciplinary Research of the Law. He is the founder of the journal Theoretical Inquiries in Law and was its editor in chief in the years 1999–2003, author of Contributory Fault in the Law of Contracts (Hebrew University Press, 1997), co-author (with Alex Stein) of Tort Liability under Uncertainty (Oxford University Press, 2001), and author of many articles in torts and contracts, most recently, ‘Total Liability for Excessive Harm’ 36 Journal of Legal Studies (2007) (with R Cooter), ‘Offsetting Risks’ 106 Michigan Law Review (2007), ‘A Comparative Fault Defense in Contract Law’ 107 Michigan Law Review (2009) and ‘Private Production of Public Goods: Liability for Unrequested Benefits’ 108 Michigan Law Review (2009). ALEX STEIN is a Professor of Law at Cardozo Law School and George W Crawford Visiting Professor of Law at Yale Law School. He specialises in Evidence and, more broadly, in the mechanisms for handling factual uncertainties in adjudication. His other specialties are Criminal Process and Torts. He is the author of Foundations of Evidence Law (Oxford University Press, 2005), Tort Liability under Uncertainty (Oxford University Press, 2001) (with Ariel Porat), and many other academic publications. Before joining Cardozo Law School in 2004, Professor Stein served for more than a decade at the Hebrew University of Jerusalem Faculty of Law. Professor Stein is on the editorial board of the International Journal of Evidence and Proof and was one of the founding editors of Theoretical Inquiries in Law. IAN FRECKELTON SC is a Senior Counsel at the Victorian Bar in Melbourne, Australia, undertaking a mixture of cases in the personal injury, professional discipline and criminal areas. He is also a Professor in the Law Faculty, the Department of Forensic Medicine and the Department of Psychological Medicine at Monash University. He has been a member of nine statutory tribunals, ranging from regulation of medical practitioners and psychologists to Victoria’s Mental Health Review Board. Ian is the author and editor of over 400 peer reviewed articles and chapters, as well as some three dozen books, including Expert Evidence, Law, Practice, Procedure and Advocacy (Thomson, 2009, with Hugh Selby), Death Investigation and the Coroner’s Inquest (Oxford University Press, 2006, with David Ranson) and Causation in Law and Medicine (Ashgate, 2002, with Danuta Mendelson). Ian is the Editor of the Journal of Law and Medicine and the Editor-in-Chief of Psychiatry, Psychology and Law. His teaching is principally on evidence and litigation and on interdisciplinary areas involving law and medicine; mental health law; and law and forensic science. CARL F CRANOR PhD (UCLA), MSL (Yale) is Professor of Philosophy at the University of California, Riverside. His generic research interests are in legal and moral philosophy. In recent years he has focused on philosophic issues concerning risks, science and the law, writing on the regulation of carcinogens and developmental toxicants, the use of scientific evidence in legal decisions, the idea of acceptable risks, protection of susceptible populations, and how society might approach the regulation of new technologies and toxicants. He is the author of Legally Poisoned: How the Law Puts Us at Risk from Toxicants (Harvard,

xvi  Notes on the Contributors 2011), Toxic Torts: Science, Law and the Possibility of Justice (Cambridge University Press, 2006) and Regulating Toxic Substances: A Philosophy of Science and the Law (Oxford University Press, 1993), as well as co-authoring a report for the Office of Technology Assessment, Identifying and Regulating Carcinogens (1987), and a study by an Institute of Medicine Committee, Valuing Health: Cost Effectiveness Analysis for Regulation (2006). His research has been supported by grants from the National Science Foundation and the University of California Toxic Substances Research and Teaching Program. He has served on science advisory panels (California’s Proposition 65 Panel, its Electric and Magnetic Fields Panel, and is currently on its Nanotechnology Panel) as well as on Institute of Medicine and National Academy of Sciences Committees. CHRIS MILLER is Emeritus Professor of Environmental Law at the University of Salford. His interest in causation in the law of personal injury was first prompted by an interest in toxic tort (and the Sellafield leukaemia litigation in particular). It has subsequently widened to embrace medical negligence, asbestos, tobacco, defective drug and other cases which rely primarily on epidemiological evidence. His writing reflects his background in physics and his earlier experience in probabilistic risk analysis in the nuclear power industry. His recent articles have attempted to demonstrate his belief that there remains a deeply rooted incoherence in the English common law of liability for negligent increase in the risk of injury. HORACIO SPECTOR is Professor of Law and Provost of Universidad Torcuato Di Tella in Buenos Aires, Argentina. He obtained a law degree and a doctor’s degree in legal philosophy from the University of Buenos Aires. He has been awarded prestigious research fellowships including those of the John Simon Guggenheim Memorial Foundation and the Alexander von Humboldt Foundation. He has been visiting scholar at the universities of Oxford, Heidelberg, and Mannheim, and visiting professor at Louisiana State University, the University of Toronto, the University of San Diego, and the University of Alicante in Spain. He is a member of the editorial board of Law and Philosophy, and has refereed for various academic journals (Philosophy and Economics, Australasian Journal of Philosophy, etc) and academic publishers (Kluwer, Springer, etc). He has published various articles in the areas of moral, legal and political philosophy, comparative law, and law and economics in journals like Mind, Law and Philosophy, Chicago-Kent Law Journal, Florida State University Law Review, and many others. He is the author of two books: Autonomy and Rights, The Moral Foundations of Liberalism (Oxford University Press, 1992, 2008) and Analytische und postanalytische Ethik (Freiburg-München, Alber Verlag, 1993). He has recently edited two influential symposia at the intersection of jurisprudence and law and economics: ‘Promises, Commitments, and the Foundations of Contract law’, (2006) 81 Chicago-Kent Law Review, and ‘Law and Economics and Legal Scholarship’, (2004) 79 Chicago-Kent Law Review. At present he is writing on the concepts of freedom, theories of rights, and causation in morality and the law. RODERICK BAGSHAW is Tutor and Fellow in Law at Magdalen College, University of Oxford, and has been teaching law in Oxford for the last 16 years. He is the co-author (with Nicholas McBride) of a textbook on Tort Law 3rd edn (Longman Law Series, 2008) and writes mainly in the areas of tort law and evidence law.

Notes on the Contributors  xvii JOHN PATERSON LLB, (Abdn), LLM (Edin), PhD (EUI) is Reader in Law at the University of Aberdeen. After studying at the EUI, Florence, Italy, he was a Research Assistant at the Centre de Philosophie du Droit, Université Catholique de Louvain, Belgium, where he worked principally on the Governance Project with the Forward Studies Unit of the European Commission. His research has covered systems theory, the regulation of risk, governance in the EU, corporate governance and energy law. He has been involved in a number of international projects, both in research and teaching, and has acted as a consultant for the OECD’s Nuclear Energy Agency. He is series editor (together with Professor Julian Webb) of the Law, Science and Society series published by Routledge-Cavendish.

Introduction This collection of essays seeks to examine the developing law of causation, and how British law should seek to influence and be influenced by developments in other countries. The essays are clustered in three groups – the law, scientific evidence and legal theory. In black letter law scholarship, major arguments have emerged about how legal doctrine will develop in cases involving indeterminate defendants and evidential gaps in causation. Various chapters examine the ways in which legal doctrine should develop over the next few years, in particular in England, Scotland, Canada and the US, including the problem of causation in asbestos cases. The chapter by Lord Hoffmann provides us with an account of the way in which law as utilised by judges employs causal concepts. He explains that where rules of law include a requirement of a form of causal relationship, the nature of this relationship is a matter of the interpretation of the rule: this is a question of legal interpretation. Such a question is answered like another question of legal interpretation, namely by reference to the language of the rule (if a statute) and its policy and purpose. He observes that this account of the way in which the law employs causal concepts goes some way to explaining why judges find it difficult to understand the academic approach to the question of whether the causal requirements are satisfied, which involves a two-stage process in which one determines whether a particular cause amounts to a cause in fact, and then, if so, whether it counts as a cause in fact for the purpose of the particular rule. The issue is compounded by the fact that there is no agreement as to what amounts to being a cause in fact, opinions ranging from Jane Stapleton’s requirement of having some historical connection with the relevant outcome to the more exacting requirements of the NESS test. But in any event no judge in fact adopts the two-stage test. Lord Hoffmann notes the operation of the concept of ‘cause in fact’ as acting as a ‘kind of filter’ which one must get through to qualify for a final round of being selected as legal causation. It is here that he posits a dilemma for academic lawyers in this field who accept the need at the cause in fact stage to deem something to have to be ‘really a cause’, according to criteria outside the law. He legitimately asks on what basis are legal academics ‘entitled to say that judges should take into account a philosophically privileged form of causation which satisfies criteria not required by law’.1 Lord Hoffmann then proceeds to look at recent cases which adopt his account of the way law as used by judges employs causal concepts which use a range of different causal relationships to give effect to the policies of different rules. In each case, different causal requirements of a rule of law are being interpreted by the courts. In doing so, he notes that substantive law may attach legal consequences to the fact that something may have   L Hoffmann, ch 1, 5.

1

xx Introduction happened, where that is all that is required by the rule, and illustrates this by the wellknown case of Fairchild v Glenhaven Funeral Services Ltd.2 He suggests that Fairchild may cause difficulties for the adherents of the cause in fact doctrine since the facts which the House of Lords held were sufficient to satisfy the causal requirements of the relevant rule did not pass the filter. In his view, a better approach would be to avoid speaking of ‘proof of causation’ but to speak of the ‘causal requirements’ of a legal rule. This would make it clear that such causal requirements ‘are creatures of the law and nothing more’.3 Joseph Sanders examines the scope, limits on, and potential application of what he terms the ‘risk rule’, ie the determination of causation in terms of risk rather than injury, in asbestos cancer cases as developed in the United States under Rutherford v Owens-Illinois Inc4 and Fairchild v Glenhaven Funeral Services,5 in the UK. He focuses on the causation (causal mechanism) and proof (which defendant causes which part of the claimant’s injury) issues that transpire when the claimant’s injury occurs after exposure to asbestos products of multiple defendants. In particular, he examines how the potential causal mechanisms by which an asbestos exposure produces an injury and proof of causation issues interact. In a comprehensive comparison of Rutherford and Fairchild in the context of their respective national jurisprudence, he concludes that the practical arguments for limiting the scope of the risk rule to single causal agents have merit, but inevitably create a rather arbitrary line. While neither Rutherford nor Fairchild offers a principled reason why the risk rule should be restricted to asbestos cases, in his view we would be better served by doing so. He supports the existing regime in the US (substantial factor) or at least is opposed to expansion of the risk rule beyond asbestos cases. Sanders deduces that the risk rule buys us very little at the cost of considerable uncertainty concerning the nature of causal questions in an array of toxic tort cases and potentially beyond. Martin Hogg provides an analysis of the Scottish doctrinal perspective on causation. He stresses that there is nothing unique about causal doctrine in Scots law and that Scots law shares the same problems with other jurisdictions in determining its understanding about causation. He expresses disappointment that Scottish judicial thinking has largely been unaffected by the recent international academic debate in this field. Nevertheless, he suggests ways in which greater academic influence could be brought to bear on judicial thinking in Scots law. He accepts Stapleton’s view that there should be a context-specific understanding of causation and that there should be reform of both the causal terminology employed by the courts and the context of the causal enquiry. In his view, the normative issues relevant to deciding what causes in the real world ought to attract liability (those matters traditionally dealt with under so-called ‘legal causation’) should not be considered as touching on the term ‘causation’ simpliciter. Hogg suggests that it is crucial for the academic community to try to agree terms. Given the inability of the sine qua non test to deal with common causal difficulties, he suggests that the NESS test should be adopted by the courts. He argues that while academic presentation of the NESS test may ‘perhaps have been somewhat complex and seemingly esoteric in the past’,6 it should now be presented to the courts in a simple practicable and more relevant manner.   [2002] UKHL 22, [2003] 1 AC 32.   L Hoffmann, ch 1, 9. 4   Rutherford v Owens-Illinois Inc 941 P2d 1203 (Cal 1997). 5   Fairchild v Glenhaven Funeral Services (n 2). 6   M Hogg, ch 3, 56. 2 3

Introduction  xxi Finally, he posits whether all different types of indeterminate causation might be mapped to provide a coherent taxonomy explaining why certain solutions are appropriate in certain cases, and other solutions are appropriate elsewhere. While not attempting the whole of this exercise, Hogg tries to map one particular type of inherent causal uncertainty, namely where the causal uncertainty relates to the inherent unpredictability of human behaviour. Discussion is restricted to cases of counterfactual human behaviour, where what is at issue is how a person would have behaved in the past in the absence of the defender’s negligence. This is a counterfactual analysis of human decision-making. He illustrates the extent to which courts have shown willingness to fashion solutions to some of the problems of counterfactual analysis through, eg, careful definition of the duty undertaken on the facts of the case, through a general rule that in misrepresentation cases a careful representation should be posited rather than some other counterfactual scenario and through application of lost chance analysis in certain cases. He notes that there remains an unresolved question as to whether, in considering different possible worlds, the counterfactual analysis should have an objective or subjective element to it. Hogg commends the adoption by the judiciary of a consistently subjective approach in all cases where determining counterfactual human decision-making is in issue. Jonathan Morgan examines both the common law and recent legislation on causation in English and Scots law. He explains why it is extremely difficult to stake out robust ‘exceptional’ categories as a matter of judicially-developed common law, and why extending the law for exceptional cases should thus be left to legislation. He uses this framework to criticise the decisions in Fairchild and Barker v Corus:7 the role of ‘policy’ reasoning in Fairchild is also considered. He then considers the radical proposals by ‘corrective justice’ theorists that judges, at least, should not consider the effects and consequences of their decisions when developing the law. He seeks to demonstrate that corrective justice, shorn of any engagement with such ‘policy’ concerns, fails to provide satisfactory answers to the issues raised by cases such as Fairchild. While he accepts the role of ad hoc legislation in this field (in particular, the Compensation Act 2006, and the Damages (Asbestos-related Conditions) (Scotland) Act 2009), he notes the strict limits of such ad hoc legislative interventions. A key point made by Morgan in respect of the legislation is that ‘binding precedent simply does not operate in the science of legislation,’8 and that legislation of this sort (pleural plaques) should not be taken to lay down any general principles at all. Nevertheless, he concedes that given the eventual form of the Scots legislation, there is merit in Hogg’s criticism that this legislation is in clear breach of established Scottish principles concerning the nature of actionable injury. Morgan’s chapter concludes with the view that both judges and legislatures have generally failed to engage with the questions of principle raised by the asbestos compensation claims. Direct engagement by the courts with such questions is necessary to promote coherent legal development, alongside a more rigorous analysis of policy questions by both judges and legislators. In his chapter, Russell Brown supports the use of inference-drawing as a means of finding (or not finding) cause in fact. This has particular relevance to Canadian torts jurisprudence, in that since the Supreme Court of Canada’s 1990 decision in Snell v Farrell,9 findings of cause in fact in cases characterised by ‘gaps’ in the plaintiff ’s evidence have been said to 7   Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572. He also gives some consideration to take account of the Supreme Court’s decision in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523. 8   J Morgan, ch 4, 86. 9   [1990] 2 SCR 311.

xxii Introduction rest upon whether a causal link between the defendant’s negligent conduct and the plaintiff ’s harm can be inferred. He stresses, however, that inference is not a licence for emotivism in fact-finding, since inference drawing necessarily involves both deliberation and explicit justification of a determination of whether cause in fact is demonstrated on the evidence. In so doing, Brown supports the use of statistics to furnish a legitimate basis for assessing putative explanations of evidence. This attempt at reconciling the seemingly polarised position between law and science is significant. Indeed, it may help to justify an approach which, it has been suggested ‘sanctions a benign “interference” ’10 in medical malpractice cases to compensate the parties’ inequality in access to evidence. Duncan Fairgrieve and Florence G’sell-Macrez examine the modern French approach to the issue of causation. They demonstrate that the French approach reveals interesting issues about the role and function of the French judiciary, the functioning of the French civil justice system, as well as the policy issues underlying the French substantive law. In particular, they examine the implications of the French jurisdictional divide between public and private law, which has meant that the administrative and civil courts have had autonomous approaches to the notion of causation. Fairgrieve and G’sell-Macrez then focus on pro­ cedural aspects of the causation issue, especially the law of evidence and the use of presumptive reasoning by French judges. Finally, they illustrate the way that French courts manipulate the concept of causation for policy reasons. In the area of scientific evidence its role in the assessment of causation in civil litigation has never been greater. The extent to which such evidence can be admitted and used in causation disputes is controversial. This section of the book is therefore devoted to exploring the role of statistical evidence in resolving causation problems, including recent trends in litigation in the UK, US, Australia and in France and the question of liability for future harm. Philip Dawid introduces lawyers to the statistical ways of thinking about causality and tries to suggest how statistical evidence could be utilised in resolving legal problems of causation. He distinguishes between two different types of causal questions. The first type of causal questions – Effects of Causes (EoC) – have the form ‘Does [will] A cause B’ and such questions are known as problems of type or general causation. Science is more typically concerned with these queries. The second type of questions – Causes of Effects (CoE) – have the form ‘Did A cause B’ and such questions are known as problems of token or individual causation. In general, law is more concerned with this type of query than with general EoC type queries. By virtue of this mismatch, Dawid notes that we must be cautious in the way we try to bring scientific evidence and reasoning to bear on questions of legal causation. In both types of analysis, he explains the ‘need to specify an alternative “foil” ’.11 Thus in CoE analysis, the ‘alternative “foil” ’ is a necessary counterfactual alternative, similar to the ‘but for’ test. In EoC analysis, the appropriate foil is a future hypothetical. In respect of effects of causes (EoCs), he explains that before we can sensibly interpret the data, it is essential to know and take account of the nature, protocol and properties of the study leading to the generation and/or collection of the data. Interpretation is easiest when the data arise from ‘a properly designed and conducted experimental study’.12 However, while experimentation is the scientific ideal, in many cases such information is 10   J Fleming, ‘Probabilistic Causation in Tort Law: A Postscript’ (1991) 70 Canadian Bar Review 136, 139; see further, R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing, 1999), 16–17. 11   AP Dawid, ch 7, 134. 12   ibid 135.

Introduction  xxiii simply not available. It therefore becomes a ‘pragmatic necessity’13 to have recourse to pure observation. However, the problems of taking observational data from observational studies at face value are immense (particularly because of the existence of ‘confounding’ variables that disturb the relationship between a putative causal factor and its measured effects). This is illustrated in two examples – one concerning Berkeley graduate admissions,14 and the other concerning hormone replacement therapy and coronary artery disease.15 The real legal issue concerns causes of effects (CoEs). Dawid stresses that any attempt to bring statistical and scientific evidence to bear on legal issues must be aware of the problems of data interpretation. Even when these are resolved, the implications of such ‘effects of causes’ evidence for the ‘causes of effects’ task of assigning causality in an individual case remains problematic. Dawid explains that applying this scientific evidence to inform issues of causality in cases at law is very subtle. The issue remains as to how one is to estimate the probability of causation on the basis of available statistical evidence. When the evidence is based on observational rather than experimental studies, there remains plenty of scope for drawing a wide range of conflicting conclusions from the same data. Even with the best experimental evidence, the implications of such evidence for a particular case at hand are far from clear-cut, and they normally only support bounds on the probability of causation at best. However, he stresses that it may be possible to sharpen these bounds by taking into consideration more detailed scientific understanding of the cause-effect process. Proof of causation in product liability litigation is an inherently difficult problem, which regularly requires time-consuming analysis of complex scientific evidence. Richard Goldberg examines recent trends in such litigation in the UK, US and in France and asks to what extent have the courts in these countries been pragmatic and fair in their inter­ pretation and utilisation of epidemiological evidence. He explores the difference between evidence of causation for purposes of science and for the law, and the difficulties in reconciling the standards of proof in law and science, including the theory that causation can be proved on the balance of probabilities by reference to the doubling of risk of injury. However, he notes that the majority of the Supreme Court in Sienkiewicz v Greif  16 appears to be sceptical of introducing a threshold for the use of epidemiological evidence and remains of the view that epidemiological evidence can be useful but must be viewed with caution; without further non-statistical evidence there is reluctance for courts to proceed to find the existence of a causal relationship. Goldberg considers that a danger otherwise is that the Legal Services Commission, in assessing whether to fund multi-party product liability litigation, may regard this doubling of risk theory as the sole basis on which to allow or prevent cases from going forward to trial, even in cases where epidemiological evidence is lacking. This could potentially prejudice access to justice in future cases. The distinction between association and causation and the difficulty in proving general and specific causation between a product and damage using epidemiological evidence is reviewed in the context of the controversial Scottish case of McTear v Imperial Tobacco   ibid 137.   Peter J Bickel, Eugene A Hammel and J William O’Connell, ‘Sex Bias in Graduate Admissions: Data from Berkeley’ (1975) Science 187, 398–404. 15   Ross L Prentice, Robert Langer, Marcia L Stefanick, Barbara V Howard, Mary Pettinger, Garnet Anderson, David Barad, J David Curb, Jane Kotchen, Lewis Kuller, Marian Limacher and Jean Wactawski-Wende, ‘Combined Postmenopausal Hormone Therapy and Cardiovascular Disease: Toward Resolving the Discrepancy between Observational Studies and the Women’s Health Initiative Clinical Trial’ (2005) American Journal of Epidemiology 162, 404–14. 16   Sienkiewicz v Greif (n 7). 13 14

xxiv Introduction Limited,17 which takes a cautious approach to the use of epidemiological evidence. Goldberg subjects to criticism the way that the epidemiological evidence was received by the trial judge. The problem of utilising statistics deriving from trends in general populations to prove causation in an individual case is highlighted in McTear, and a possible solution in the form of utilisation of the Bayes’ Theorem is discussed. The role of scientific evidence in determining causation in product liability suits is wellestablished in the US. Here the emphasis is often on the relevance and reliability of evidence since the landmark Supreme Court decision of Daubert v Merrill Dow Pharmaceuticals Inc.18 The Daubert framework has been utilised in exercising the gatekeeping role to the admissibility of scientific evidence – in some cases in order to reduce the use of so-called ‘junk science’ by juries. However, a developing issue has been the assessment of the significance of scientific evidence admitted in the determination of proof of causation in complex product liability cases, which often involve prescription drugs and vaccines. In particular, of recent interest is the assessment of the value of scientific evidence in six test cases of the Omnibus Autism Proceeding (OAP) under the National Childhood Vaccine Injury Act of 1986 (NVIA), decided in February 2009 and March 2010. These cases essentially explored two causation theories, viz that MMR vaccines and thimerosal containing vaccines could combine to cause autism, and that thimerosal vaccines alone can cause autism. The Special Masters in these proceedings, having considered all the available scientific evidence, concluded that there was no merit in either of the two general causation theories. Goldberg provides an analysis of the implications of this complex and lengthy litigation. The final section of Goldberg’s chapter explores the much more liberal approach to causation established in France by the Cour de Cassation for medicinal product liability cases in the context of injury allegedly caused by the Hepatitis B vaccine through the use of presumptions of causation. In an extremely controversial judgment, the Cour de Cassation granted a patient’s claim against the manufacturer of the Hepatitis B vaccine, by finding that causation had been proven even in the absence of general causation, but where such a causal link could not be excluded.19 The Cour de Cassation now appears to be retreating somewhat from that position.20 The inconsistency of these decisions has been unhelpful in generating uncertainty for both claimant and defendant. However, Goldberg submits that, without scientific evidence of general causation, there should be no question of overcoming the burden of proof of causation in such cases. He considers that the Cour de Cassation would be wise to study the factors required to overcome that burden as established in the National Vaccine Injury Compensation Program and utilised by the OAP test cases, which, he believes, generate more clarity and consistency in approach. The chapter by Peter Feldschreiber, Leigh-Ann Mulcahy and Martin Day reviews the principles of epidemiology and biostatistics used by the regulator of medicinal products to evaluate causation when assessing their safety and efficacy. It then attempts to assess the relevance and practicality of these principles to the way courts determine causation in alleged cases of drug-induced injury. They explain that while the attribution of causality by regulators is generally an exercise in prospective forecasting of the magnitude of the safety issues and risks inherent in allowing the widespread use of medicine in the population at   2005 2 SC 1.   509 US 579 (1993). 19   Civ 1, no 08–11.073, 9 July 2009. 20   Civ 1, no 09–16.556, 25 November 2010. 17 18

Introduction  xxv large, the attribution of causation in clinical negligence and product liability actions requires a retrospective assessment of clinical and biological data by the court in determining whether an association is sufficiently strong to satisfy the legal tests for causation. The chapter notes the differences in the standard of proof required to establish causation in each context and divergent approaches to risk as a test for establishing causality. It questions the distinction countenanced by the Court of Appeal in Fairchild 21 between divisible and indivisible injuries as a legal distinction which is not known to or easily reconcilable with science. In a consolidation of the thoughts of several of his seminal articles in this field,22 Richard Wright compares the conviction, belief in truth standard of proof of civil jurisdictions with the preponderance of the evidence or a balance of probability standard of proof of common law systems in civil cases. In general, the common law as well as that of the civil law countries continues to regard the applicable standard of persuasion in both civil and criminal actions as ‘requiring the formation of a belief in the truth of the facts at issue in the particular case’;23 mere statistical probability, no matter how high, generally is insufficient. ‘[C]oncrete “particularistic evidence” specific to the particular case is necessary to support such a belief ’.24 Wright concludes that ex ante causal probabilities are non-individualised, class-based probabilities which are insufficient and generally unhelpful for the formation of belief, which instead requires particularistic evidence, since ‘only it is capable of converting possibly applicable causal generalisations (with their associated statistical frequencies) into actually instantiated causal laws.’25 Supporting the major thrust of his arguments is the Supreme Court decision in Sienkiewicz v Greif (UK) Ltd,26 which he considers has gone a long way towards clarifying proof of causation in UK law by, for the first time, squarely recognising the error in interpreting ‘balance of probability’ literally as a mere 50+ per cent statistical probability, rather than as a belief standard. While recognising the validity of Bayes’ Theorem per se, he rejects the the use of abstract initial base rates, whether naked statistics or ex ante causal probabilities, as initial base rates in Bayes’ Theorem to adjust the judgment regarding causation derived from the reliability of the particularistic evidence.27 He applies these conclusions to several types of particularistic causal situations, viz indistinguishable alternative tortious causes; the doubling of the risk doctrine in toxic exposures; lost chances in professional malpractice cases; and the res ipsa loquitur doctrine. The clear recognition and proper resolution of the issues in these cases is often hampered by an assumption that the standard of persuasion in civil actions is a mere statistical probability standard. The chapter by Ariel Porat and Alex Stein addresses the possibility of imposing liability in tort for a stand-alone wrongfully created risk of future harm. They examine the American and British court decisions pertaining to this, in particular Norfolk & Western Railway   [2001] EWCA Civ 1881, [2002] 1 WLR 1052, [21], [27].  See RW Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (1988) 73 Iowa Law Review 1001, 1044–67; RW Wright, ‘Liability for Possible Wrongs: Causation, Statistical Probability, and the Burden of Proof ’ (2008) 41 Loyola Los Angeles Law Review 1295; RW Wright, ‘Proving Facts: Belief versus Probability’ in H Koziol and BC Steininger (eds), European Tort Law 2008 (Vienna, Springer-Verlag, 2009) 79. 23   R Wright, ch 10, 196. 24   ibid 196–97. 25   ibid 197. 26   [2011] UKSC 10, [2011] 2 WLR 523. 27   cf R Goldberg, ch 8, 162–63. 21 22

xxvi Introduction Company v Ayres28 and Johnston v NEI International Combustion Ltd,29 and consider whether a probability-based compensation for the victim’s expected – albeit not yet materialised – harm is just and efficient. They demonstrate how the virtues of a legal regime that allows a tort victim to recover compensation for expected harm overshadow its vices. They conclude that a person’s risk of sustaining harm in the future should be actionable whenever the risk is substantial, but that small risks of future illness should not be actionable. The prospective victim should have the option to choose between immediate recovery of compensation for his expected harm and a postponed entitlement to recover full compensation in the event of illness. The victim should be able to recover from the wrongdoer compensation for the wrongfully imposed risk. The amount of this compensation would equal the harm associated with the illness multiplied by the victim’s probability of becoming ill due to the wrongdoing. They observe that allowing victims to make this choice might create a collective action problem. Since expedited compensation for a victim’s expected harm erodes the wrongdoer’s ability to compensate future claimants, victims would opt for an early recovery for expected harm even when their substantive remedial preferences are different (victim migration), which would increase the risk of the wrongdoer’s insolvency. They demonstrate, however, that this problem can be resolved. Ian Freckelton surveys several examples from Australian case law to illustrate the issues posed by the proof of causation on the basis of expert evidence. An attempt is made to explore some of the dynamics and issues which have the potential to lead to appealable error and miscarriages of justice in the context of determinations of causation. He concludes that the litany of difficulties encountered by Australian courts in cases relating to expert evidence as to causation highlights several issues. In a number of cases30 the expert evidence was characterised by questionable and methodologically dubious expert evidence. In others,31 the existence of multiple potential aetiologies created both theoretical and practical problems for fact-finding, calling into question what the criteria should be for reception of evidence about causation which was non-specific and amounted only to ‘possibility evidence’. He notes that the Australian Courts traditionally have struggled with ‘novel’ scientific and medical evidence – especially how to deal with iconoclastic or heterodox theories.32 In such decisions, where there are juries, there is a risk both of hindsight error and also sympathy for a victim leading to an assumption of causation in the face of potentially contrary expert evidence. When the law is unclear on crucial issues of causation (eg on admissibility of possibility evidence) or loss of chance reasoning,33 Freckelton cautions that the risks are magnified that trial courts will be distracted by less than definitive expert evidence which seeks to compensate plaintiffs whose circumstances are adverse, resulting in outcomes which are disadvantageous for defendants. The importance of tort law responding to developing science is explored in the chapter of Carl Cranor. He reviews several causal models proposed or found by scientific research to pose the major question of whether law can appropriately accommodate scientific discoveries.   Norfolk & Western Railway Company v Ayres 538 US 135 (2003).   Johnston v NEI International Combustion Ltd [2007] UKHL 39, [2008] 1 AC 281. 30   See eg, Commissioner for Government Transport v Adamcik (1961) 106 CLR 292; R v Parenzee [2007] SASC 143, [2007] SASC 316; R v Karger [2001] SASC 64. 31   See eg Seltsam v McGuiness (2000) 49 NSWLR 262, [2000] NSWCA 29; Amaca Pty Ltd v Ellis [2010] HCA 5. 32   See eg, Commissioner for Government Transport v Adamcik (1961) 106 CLR 292; R v Parenzee [2007] SASC 143, [2007] SASC 316. 33   Tabet v Gett [2010] HCA 76. 28 29

Introduction  xxvii In the section of essays clustered on legal theory, the so called NESS (necessary element in a sufficient set) test of causation, originating from Hart and Honoré in their seminal Causation in the Law34 and developed by John Mackie (INUS) (‘insufficient but necessary part of an unnecessary but sufficient condition’)35 and Richard Wright (NESS),36 is discussed and defended. The concept of cause which ought to be used by tort lawyers where one person caused another to act in some way is investigated. Finally, the importance of tort law responding to developing science, while respecting the limitations imposed by precaution and indeterminate causation is examined by case study, in the context of legal theory. Richard Wright addresses the objections that have been raised to the claim that the NESS account fully captures the concept of natural causation and properly handles all types of situations, in particular: the creation of spurious necessity; the problem of causal directionality; epiphenomena (collateral effects of a common cause); the inability to handle properly some types of preemptive causation situations; the inability to handle a condition that was not, or could not be proven to be, either strongly necessary or independently strongly sufficient; the treating of human decisions and actions, which often are based on multiple reasons, as causes; the problem of indeterminism and ‘probabilistic causation;’37 the problem of omissions and negative causation; the problem of overdetermined negative causation; and, finally, the criticism that the NESS account greatly increases the proliferation of causes. He distinguishes between his NESS account, Herbert Hart’s and Tony Honoré’s ‘causally relevant factor’ account and John Mackie’s INUS account, which differ in significant ways that make the latter two accounts vulnerable to objections to which the NESS account is immune. He offers an account of causal laws that he considers rebuts the claim that the NESS account is viciously circular. While the NESS account’s definition of causal sufficiency relies upon the concept of a causal law, the concept of a causal law is defined without any use of causal language. He considers that there is therefore no conceptual circularity here, vicious or otherwise. While conceding that his initial elaborations of the NESS account were too demanding, he explains that in singular instances of causation, a NESS condition need only be a part of the instantiation of one of the abstract conditions in the completely instantiated antecedent of a causal law; it need not be necessary for such instantiation.38 He concludes that consideration of the various objections to the NESS account has reinforced his belief that it is not merely a very useful tool for identifying singular instances of causation, but that it does so precisely because it captures the essence of causation and gives it a comprehensive specification and meaning. Chris Miller’s chapter articulates that both of the two main criticisms of Richard Wright’s NESS test – that it is unable to deal adequately with multiple omissions and that his treatment of necessity and sufficiency lacks philosophical rigour – are neither fatal. The main 34  HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Oxford University Press, 1985) 13–22, 111–13; 122–28, 206–07, 235–53. 35   JL Mackie, ‘Causes and Conditions’ in E Sosa and M Tooley (eds), Causation (Oxford, Oxford University Press, 1993) 33, 34–37; JL Mackie, The Cement of the Universe: A Study of Causation (Oxford, Clarendon Press, 1974), ch 9. 36   RW Wright, ‘Causation in Tort Law’ (1985) 73 California Law Review 1735, 1788–91. 37   See, further, RW Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (n 22), on which this section is based. 38   RW Wright, ‘Acts and Omissions as Positive and Negative Causes’ in JW Neyers et al (eds), Emerging Issues in Tort Law (Oxford, Hart Publishing, 2007) 297–98; RW Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (n 22) 1001, 1045.

xxviii Introduction reason why NESS can survive these criticisms is that its aims are modest in that it offers a coherent approach to attributing cause in situations where the traditional ‘but for’ test of causation leads to paradox. He also argues that Wright’s NESS approach assists in an understanding of the existence of multiple, independent causal pathways to a single instantiation of physical harm, an all too common example of which is the problem inherent in the multiple asbestos employers’ litigation. He argues that while the House of Lords in Fairchild39 and Barker40 did not explicitly employ the NESS test, their Lordships approach can be interpreted as making the increased risk of injury which has transpired as a recoverable head of damages, and then treating each defendant’s negligence as a necessary element in a set which is independently sufficient for that head. In effect, this is an application of the NESS test which could have offered a principled justification of the application of joint and several liability. Horacio Spector presents a new theory of causation in moral and legal contexts which he calls the minimal-maximal 3-set (MMTS) analysis. This analysis involves three sets of propositions: the set of antecedent conditions, the set of causal generalisations, and the set of empirical consequences. He takes antecedent conditions in MMTS causal relations to be ‘causal factors in a generic sense’. ‘Causes stricto sensu’ are events that instantiate some positive MMTS conditions according to a rule of correspondence that he calls denominalisation. Spector shows that his MMTS analysis responds to the objections raised against the nomonological, regularity or covering law theory of causation, the best known version of which is the so-called INUS/NESS account, originating from Hart and Honoré in their seminal Causation in the Law and developed by John Mackie (INUS) and Richard Wright (NESS). The objections to this theory are that it cannot exclude spurious conditions as causes and that it cannot resolve cases of preemption (‘early’ and ‘late’) as well as cases of ‘alternative causation’, omission and double preemption. He also explains how this MMTS analysis must be supplemented by a counterfactual dependence theory that explains how agents can relate to negative or positive MMTS causal factors. In so doing, he uses the term ‘agent-conditioning’ to denote the ways in which agents can condition changes. The MMTS account bears on ‘commissive’ agent-conditioning (making things happen), whereas counterfactual dependence is the proper account for ‘omissive’ agent-conditioning (letting things happen). Roderick Bagshaw’s chapter investigates the concept of cause which ought to be used by tort lawyers where one person caused another to act in some way. He uses Stapleton’s writings as an essential foundation for a discussion of how tort lawyers ought to understand cause and how they ought to use causal language. In particular, he refers to her argument that lawyers should choose to use ‘causal language’ only when reporting an enquiry into whether a specific factor was ‘involved’ in the existence of a particular phenomenon, where involvement extends to ‘necessity’, ‘duplicate necessity’ and ‘contribution’.41 He identifies two demands made by Stapleton in support of her proposal that in law ‘cause’ should be understood as ‘involvement’, viz that the legal concept of cause should be one that can be described wholly in terms of ‘objective facts’ and that it should serve the purposes of the law generally. While not rejecting these demands, he adds a third, viz that the legal concept   Fairchild v Glenhaven Funeral Services Ltd (n 2).   Barker v Corus (UK) plc (n 7). 41   See J Stapleton, ‘Choosing what we mean by “Causation” in the Law’ (2008) 73 Missouri Law Review 433 and J Stapleton, ‘Causation in the Law’ in H Beebee, C Hitchcock and P Menzies (ed), The Oxford Handbook of Causation (Oxford, Oxford University Press, 2009). 39 40

Introduction  xxix of ‘cause’ should not be inconsistent with what we know about natural processes that can plausibly be described as involving ‘one thing bringing about another’.42 Bagshaw then addresses his first major question as to whether there is something special or different about ‘causing’ someone to act as opposed to ‘causing’ a machine to work. He examines Hart and Honoré’s view that we are not reporting the result of ‘the central type of causation of physical events’.43 Instead, we are reporting the existence of a different type of causal connection and possible objections to it.44 He suggests that Hart and Honoré’s noncentral test conforms to the pattern of allowing a litigant to succeed where he or she proves the possibility of causation and an additional condition makes it fair to resolve any doubts in a particular direction. Bagshaw then proceeds to address his next major question, viz how Stapleton’s specific recommendation as to how lawyers should use ‘causal language’ copes with mixtures of ‘causes’. In order to do so, he considers what counts as ‘causal contribution’ within a physical system. In particular, he reviews Stapleton’s recommendation: that lawyers ought to choose to say that a factor ‘causes’ an injury by way of ‘contribution’ where it is ‘a necessary element for the sufficiency of a sufficient set’ of factors once any excess has been ‘disaggregated’ and left out of account.45

He examines the ‘problem of “disaggregation” ’ and whether there should be limits on the type of disaggregation possible. Bagshaw argues that an account of ‘causal contribution’ which permits ‘disaggregation’ without limits is likely to disappoint, and he draws attention to some of the difficulties for this failing by defining limits on when disaggregation is permissible. He submits that the obstacles in the way of a satisfactory ‘disaggregation’ account provide support for a rival account to Stapleton, which defines ‘causal contribution’ as ‘playing an active role in an operative mechanism’. This is what he terms ‘the natural process account of causal contribution’.46 Bagshaw demonstrates how this rival account of causing by contribution can help to explain what is special about causing someone to act through the general suggestion that a satisfactory account of cause ought to be consistent with what we know about natural mechanisms. The final section of Bagshaw’s chapter considers whether a ‘causal explanation’ might underpin the ‘intervention’ doctrine. He argues that none of the most plausible non-causal explanations for limiting liability (non-foreseeability, coincidence, fairness and intention) seem sufficient to explain the intervention doctrine, and that the concept of ‘causal potency’47 might be useful in developing a ‘causal’ explanation of the doctrine. He submits that the attempts to resolve the questions posited in this chapter suggest that the causal analysis which the law ought to be willing to utilise ‘should be far richer than “causal minimalists” generally acknowledge’.48 In the final chapter, John Paterson examines how law confronts harm in the context of scientific uncertainty from two perspectives – the perspective of precautionary decisionmaking and the perspective of indeterminate causation. The discussion of precautionary decision-making considers the way in which the courts have addressed the presence of scientific uncertainty in such situations, using the High Court and Court of Appeal judicial   R Bagshaw, ch 17, 362–66.   See HLA Hart and T Honoré, Causation in the Law (n 34) 51–52.  ibid. 45   J Stapleton, ‘Choosing what we mean by “Causation” in the Law’ (n 41) 433, 476. 46   R Bagshaw, ch 17, 373–74. 47   ibid 379–80. 48   ibid 381. 42 43 44

xxx Introduction review decisions in Downs v Secretary of State for Environment, Food and Rural Affairs49 as a case study. Utilising David Resnik’s characterisation of decision situations for the operation of the precautionary principle,50 Paterson concludes that the Court of Appeal in Downs proceeded on the basis of an incomplete understanding of the precautionary principle. By contrast, the first instance judgment represents ‘a quite sophisticated understanding’ of Resnik’s tests, which he prefers. He then examines the leading decision of indeterminate causation, that of Fairchild v Glenhaven Funeral Services Ltd.51 Both cases, he suggests, are best characterised as decisions under ignorance, and that the superficial differences between precautionary decisions and cases of indeterminate causation only serve to mask similarities at a deeper level, in particular with regard to the nature of scientific uncertainty in each case. Paterson then considers why the courts in each case seem to display different degrees of willingness to engage with that uncertainty and to reach substantive decisions. He suggests that the approach of the courts in the context of indeterminate causation can provide support for a more robust judicial approach in the context of the review of precautionary decision-making, without which the precautionary principle risks being rendered meaningless. He also submits that understanding law’s approach to harm in the context of scientific uncertainty in this way could assist in any future extension of the courts’ willingness to engage substantively in situations of indeterminate causation beyond those currently covered by the McGhee/Fairchild exception. The issues which are discussed in these chapters are a representative selection of the current difficulties faced by the developing law of causation. It is hoped that they will encourage readers into appreciating the wider significance of these problems, which bestride law, scientific evidence and legal theory.

49   [2008] EWHC 2666 (Admin); Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664. 50   D Resnik, ‘Is the precautionary principle unscientific?’ (2003) 34 Studies in history and philosophy of biological and biomedical sciences 329–44, 332. 51   Fairchild v Glenhaven Funeral Services Ltd (n 2).

1 Causation RT HON LORD HOFFMANN

The relationship between judges and academic writers in the United Kingdom has much improved over the past 20 years or so. Although few judges are regular readers of academic publications, that is because judges seldom do any legal research of their own. Nor do they have, on an individual basis, the assistance of law clerks.1 For the most part, however, judges rely upon counsel to do the research and refer them to the relevant materials. When I was at the Bar, many judges appeared allergic to law review articles. More recently, however, the House of Lords has encouraged counsel to produce any relevant academic work. Citation of academic writers has greatly increased. It has, I think, improved the quality of appellate judgments. Why does causation appear to be an exception to the current practice of taking note of academic work? Why have no judges heard of the NESS (‘necessary element of a sufficient set’) test? Why do they obstinately refuse to apply a ‘two-stage test’ and distinguish between their findings of ‘cause in fact’ and their subsequent decision about ‘legal causation’? None of these concepts appears to have attracted any judicial interest. One side or the other must be missing something. Let me start with how it looks to a judge. Causation is of interest to a judge only because many legal rules, attributing liability (civil or criminal) for wrongful acts, or entitling people to payments (for example, under insurance policies) contain causal requirements. They say that you are guilty of murder only if (among other requirements) you have caused someone’s death. To be liable in damages for negligence, you must have caused loss to someone, and so on. The application of all such legal rules involves applying the rule to facts. One examines the facts: A fired a gun, a bullet identified as having been fired by that gun penetrated B’s body, the medical evidence shows that B died from the injury caused by the bullet. The judge (or jury) concludes that on these facts, the causal requirement in the crime of murder has been satisfied A has caused B’s death within the meaning of that concept in the common law of murder. Exactly the same process applies to other requirements (‘elements’) of the crime. Murder, it used to be said, required ‘malice aforethought’. That has come to be interpreted to mean an intention to kill or cause grievous bodily harm. So, again one examines the evidence to apply this rule to the facts. B had been having an affair with A’s wife and 1   For the last five or six years, four research assistants have been engaged on a yearly basis by the House of Lords. The use made of them by the Law Lords varied. There has also been legal assistance available in the Court of Appeal.

4  Rt Hon Lord Hoffm ann A had threatened to kill him. He had bought a gun and bullets and gone to B’s house. He pointed the gun and fired. There was nothing to show that A would not have been aware that the gun was a lethal weapon. Applying the rule to the facts, the jury decides that the killing was intentional within the meaning of that concept in the law of murder. Notice, now, that in each case I have said that the facts satisfied the causal requirements of the rule of law because they came within the meaning of a particular concept (‘causing death’, ‘intending to kill’) used by the rule. In using concepts such as causing death or intending to kill, lawyers ordinarily intend that such concepts should have at least the core meaning which would be attached to them in ordinary speech. And in the examples I have given, the facts fall squarely within the core meanings of the relevant concepts. By the standards of ordinary speech, A caused B’s death and intended to do so. But of course there are many cases which are not so straightforward. A burns down his factory for the insurance money, knowing that the caretaker’s elderly mother will probably be unable to escape. Does he intend to kill her? A shoots at B and wounds him, but B would have recovered if he had not been treated negligently at the hospital. Has A caused his death? These cases involve interpretation. If the rule is contained in a statute, the usual techniques of interpretation must be applied. You read the statutory language in the context of its apparent purpose, the background of related parts of the legal system, the practical consequences of one interpretation rather than another. In the case of a common law rule, you apply similar techniques to the principles stated in earlier authorities. In both cases, the object of the exercise is to ascertain what should count as causing death, or intending to kill, for the purposes of the rule in question. You are declaring what the rule means, what its proper interpretation should be. This process is exactly the same whether you are deciding what should count as ‘causing’, ‘intending’ or as falling within any other concept employed by the rule. This account of the way in which the law employs causal (or indeed any other) concepts should explain why judges find it so difficult to understand why academics claim that the question of whether the causal requirements of some legal rule have been satisfied involves a ‘two-stage process’ in which you first decide whether the putative cause amounted to a ‘cause in fact’ and then, if it passes that test, whether it counted as a cause in law for the purposes of the particular rule. There is no agreement on what amounts to being a cause in fact. Professor Stapleton says that it means having some historical connection with the relevant outcome. Others are more exacting and prefer the NESS test: that it means being a necessary element in a set of conditions which are jointly sufficient to produce the outcome. But no judge in fact adopts such a two-stage test. Of course the application of the legal rule is always a two-stage process in the sense that you find the facts and then decide whether they answer to the requirements of the rule, or (which comes to the same thing) you decide as a matter of interpretation what are the requirements of the rule, and then decide whether the facts satisfy those requirements. That is the natural process of decisionmaking when applying any legal concept. But that two-stage process is not what the advocates of ‘cause in fact’ or the NESS test have in mind. Their theory is that when you have ascertained the facts, you do not go straight to the question of whether they satisfy the requirements of the particular legal rule, having resolved any questions of interpretation which may be necessary to answer this question. For them, there is a two-stage process after you have decided the facts, so that the decision-making process really has three stages. First, you ascertain the facts. Then you decide whether they count as ‘cause in fact’ by being part of the history of the outcome or by satisfying the NESS test or whatever, and thirdly, you

Causation  5 decide whether they count as ‘legal causation’ for the purposes of the particular rule. That means that the concept of ‘cause in fact’ acts as a kind of filter which you have to get through in order to qualify for the final round of being selected as legal causation. If, heaven forbid, the rule should be construed, upon its true interpretation, as treating something which did not qualify as ‘cause in fact’ as nevertheless satisfying its causal requirements, then it should be regarded as deeming something to be a cause when it was not really a cause. It is this concept of something having to be really a cause according to criteria lying outside the law which puzzles lawyers. On what basis are academic writers entitled to say that judges should take into account a philosophically privileged form of causation which satisfies criteria not required by the law? Of course, anyone is entitled to say that in treating X in some particular context as having caused Y, the courts are stretching the ordinary meaning of ‘cause’. But this is engaging in a legitimate argument over interpretation and not introducing the concept of ‘real’ causation as a preliminary test which has to be satisfied before the question of interpretation arises. I would not deny that the discussions about ‘cause in fact’ can sometimes throw useful light on the ordinary meaning of the causal language used in legal rules. The great contribution made by Hart and Honoré to this part of the law was to identify the principles by which, as a matter of ordinary language, one would say that X had caused Y: the importance attached to voluntary human acts, deviations from the ordinary course of events and extraordinary occurrences. These observations are valuable as aids to the interpretation of causal language. A recent example of their being used as such is R v Kennedy (No 2).2 The case turned upon the interpretation of the words ‘caused to be administered’ in section 23 of the Offences against the Person Act 1861. The accused had prepared a syringe of heroin for the deceased and given it to him. The deceased injected himself and subsequently died. Did that count as causing the heroin to be administered to the deceased, for the purposes of the Act? The House of Lords said no. They cited a passage from Hart and Honoré:3 The free, deliberate and informed intervention of a second person, who intends to exploit the situation created by the first, but is not acting in concert with him, is normally held to relieve the first actor of criminal responsibility.

The conclusion therefore was that the facts did not count as causing the heroin to be administered. The deceased had made a free and informed decision to exploit the situation created by the accused, ie being put in possession of the syringe, by injecting himself. The decision was based upon what would ordinarily be regarded as satisfying a requirement of ‘causing’ for the purpose of attributing responsibility. As Hart and Honoré explain, this ordinary meaning reflects commonly held moral positions in our society. As the House of Lords said in the Kennedy case,4 ‘the criminal law generally assumes the existence of free will’. Note, however, that the citation from Hart and Honoré says that that the deliberate intervention of another person is ‘normally’ held to relieve the first actor of responsibility. But the law does not always apply this principle. Causal requirements may be given a different interpretation when one is construing a rule which imposes a duty to guard against injury caused by the deliberate acts of others, either to third parties or, very exceptionally, to themselves. A duty to protect premises against thieves and burglars is a straightforward   R v Kennedy (No 2) [2007] UKHL 38, [2008] 1 AC 269.   HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Oxford University Press, 1985) 326.   [2008] 1 AC 269, 275.

2 3 4

6  Rt Hon Lord Hoffm ann example of such a case. It would negate such a duty if one denied liability on the ground that the loss had been caused by the thief and that his deliberate act relieved the negligent keeper of the premises of his responsibility.5 The common law assumption of free will makes it very unusual to find a duty to protect people against injury deliberately inflicted on themselves, but one does find it in cases in which the person in question lacks capacity (as in the case of children) or is under psychological pressure from external circumstances (as in the case of prisoners). Where such a duty does exist, it would be equally absurd to say that failure to take precautions did not cause the injury because the claimant inflicted it upon himself.6 A similar construction may be given to a regulatory statute, intended to impose strict liability to encourage high standards of prevention of harm by others rather than (as is normally the case in the criminal law) to reflect what would ordinarily be considered individual moral responsibility. That was the case in Environment Agency (formerly National Rivers Authority) v Empress Car Co (Abertillery) Ltd,7 where the question was whether the car company had caused oil to enter a river, contrary to section 85(1) of the Water Resources Act 1991. The company had a diesel tank in its yard. One night a vandal opened the tap and let the oil flow into the river. The House said that maintaining the tank of oil which gave the vandal the opportunity to let it out counted for the purposes of the Act as causing it to enter the river. Despite the fact that the escape was caused by the voluntary act of the vandal, the fact that it would not have happened unless the oil had been there was enough to enable the company to be treated as having caused it within the meaning of the Act. As the intervention of a third party or some unusual natural event are in practice the only ways in which oil is likely to escape from someone’s tank into a river without any negligence on his part, any other interpretation of the causal requirements of the statute would have destroyed the character of the offence as one of strict liability.8 Notice, however, that the interpretation given to section 85(1) did not dispense altogether with causality. That would have been a strong thing to do with a provision which said that the accused must have ‘caused’ the oil to enter the river. Keeping the oil was not sufficient to count as causing it to enter the river if it was the result of some very unusual event like an earthquake or terrorist attack. In some cases, however, the rule is given an interpretation which does not require any element of causation. Instead, it is interpreted as having the character: ‘If A does X, he becomes B’s insurer in respect of risk Y’. It is unnecessary for X to have caused Y. Such an interpretation was given to the rule of liability for conversion by the House of Lords in In Kuwait Airways Corporation v Iraqi Airways Co (Nos 4 and 5).9 The Iraqi Airways received from the Iraqi government some aeroplanes which belonged to Kuwait Airways but had been brought to Iraq by the Iraqi armed forces after the invasion of Kuwait in 1990. Taking possession of the aircraft was a conversion by Iraqi Airways. Afterwards they were destroyed in the American bombing of Mosul. When Kuwait Airways claimed their value as damages, Iraqi Airways said the conversion had not caused the loss of the aircraft. They would have been destroyed anyway. But the House of Lords said that it was not necessary that the   Stansbie v Troman [1948] 2 KB 48.   Reeves v Metropolitan Police Commissioner [2000] 1 AC 360.   Environment Agency (formerly National Rivers Authority) v Empress Car Co (Abertillery) Ltd [1999] 2 AC 22. 8   Compare the rule in Rylands v Fletcher (1868) LR 3 HL 330, which was almost immediately neutered by decisions that it did not apply when the escape was caused by a third party or Act of God: see Transco plc v Stockport Metropolitan Borough Council [2004] 2 AC 1, 17. 9   Kuwait Airways Corporation v Iraqi Airways Co (Nos 4 and 5) [2002] 2 AC 883. 5 6 7

Causation  7 conversion should have caused the loss. If you convert someone’s property, you have to pay for it or give it back. You become insurer of the chattel. It does not matter what would otherwise have happened to it. Conversion is an ancient tort of strict liability. It was however somewhat startling when the House of Lords in Chester v Afshar10 gave a similar construction to the modern law of negligence, where it has always been assumed that a causal relationship between the negligent act and the damage needs to be shown. A doctor did not warn a patient of a possible complication which might be caused by an operation she was about to undergo. The chances of the complication occurring were very small and she would have had the operation anyway, but, without any negligence on the part of the doctor, it did in fact occur. The doctor’s failure to warn neither affected the patient’s choice to have the operation nor increased the risk of the complication occurring. It was simply a coincidence that the risk happened to eventuate in a case in which the patient had not been warned about it.11 Nevertheless, the House of Lords held the doctor liable for the damage caused by the complication. The rule laid down by the House was that if a doctor fails to warn of a given risk, he becomes the patient’s insurer against the occurrence of that risk. The precise scope of this form of insurance liability remains to be discovered. Cases like Empress Cars cause distress to fundamentalists who believe that the deliberate acts of third parties should always relieve the original actor from liability under any rule of criminal law, irrespective of the purpose or policy of the rule.12 But they cause no difficulty for proponents of the theory that there must be ‘cause in fact’ because they easily pass the filter imposed by that concept. The presence of the oil in the tank is part of the history which explains how it got into the river. In fact, one suspects that Professor Stapleton chose ‘part of the history’ as her test for ‘cause in fact’ to give the filter as wide a mesh as possible and to avoid having to declare that cases in which the law regarded the causal requirements of a rule as satisfied were nevertheless not really cases of causation because they did not pass the NESS or some other test. In fact, the oil in the tank would pass the NESS test as well, since its presence is a necessary part of the set of conditions which were sufficient to produce the spillage. Problems about ‘cause in fact’ do however arise when the causal requirements of a rule of law are interpreted in a way which takes into account the law’s somewhat artificial system of epistemology. The law operates a binary system in which the only values are 0 and 1. If the evidence that something happened satisfies the burden of proof (in criminal law,   Chester v Afshar [2004] UKHL 41, [2005] 1 AC 134.   Judges and writers have made remarkable efforts to argue that the failure to warn did cause the damage. One argument, originating with Gaudron J in the similar case of Chappel v Hart (1998) 195 CLR 332, is that if the warning had led the patient to postpone the operation, or have it performed by another doctor, the chances of the complication occurring on the later occasion would have been very small. Therefore the failure to warn meant that the patient actually suffered the injury while on the other, hypothetical, occasion, she would very likely have escaped it. That is like arguing that if the taxi to take you to Oxford arrives half an hour late and you are injured when it is struck by a meteorite on the M40, its late arrival caused your injury because if it had arrived on time, the chances of being struck by a meteorite during the hypothetical alternative journey would have been very small. In fact it is simply a coincidence that the car struck by the meteorite happened to have set out late. The lateness of the car did not cause the meteorite to strike it. Similarly, it was a coincidence that the complication happened to a patient who had not been warned. 12  ‘[P]rofoundly unsatisfactory . . . bad principle, bad law and bad reasoning’ (A Simester and R Sullivan, Criminal Law Theory and Doctrine, 3rd edn (Oxford, Hart Publishing, 2007) 92, 94; ‘almost perverse’ (MJ Allen, Criminal Law, 10th edn (Oxford, Oxford University Press, 2009) 44; ‘aberrant decision’ (A Ashworth, Principles of Criminal Law, 6th edn (Oxford, Oxford University Press, 2009) 107; ‘appears to confuse culpability with causation’ (D Ormerod, Smith and Hogan’s Criminal Law, 12th edn (Oxford, Oxford University Press, 2008) 81. 10 11

8  Rt Hon Lord Hoffm ann beyond reasonable doubt, in civil law, more probable than not) then it is assigned a value of 1 and treated as definitely having happened. If the evidence does not discharge the burden of proof, the event is assigned a value of 0 and treated as definitely not having happened. There is no forensic space for the conclusion that something which has to be proved may have happened. This binary system does not of course mean that the substantive law may not attach legal consequences to the fact that something may have happened. In that case, what has to be proved, to the necessary standard of proof, is that the event may have happened. That is all that the rule requires. The binary system still operates, but the facts which have to be proved are different. This distinction was applied to the causal requirements of the law of negligence in the well-known case of Fairchild v Glenhaven Funeral Services Ltd.13 The deceased had developed mesothelioma caused by the effect of an asbestos fibre on a pleural cell. He had been negligently exposed to asbestos by more than one employer but was unable to prove whose asbestos had caused his disease. The Court of Appeal applied the standard causal requirements of the law of negligence, namely that the claimant must prove that, but for the negligent act, the damage would not have occurred. The claimant was unable to discharge this burden in respect of any employer, even though it was clear that one of them must have been responsible. His action was therefore dismissed. The House of Lords decided that this result was unfair and that in cases of mesothelioma and analogous situations, the substantive causal requirements of an action for negligence should be different. It should be sufficient to prove that the negligent exposure increased the risk of developing mesothelioma and may have caused it in the sense that the fibre which was responsible for the development of the disease was inhaled during that particular exposure. There are two ways in which one could characterise this change in the substantive law of negligence. One is to say that the causal requirements for an action for damages for mesothelioma have been changed: all that is necessary is to prove that the risk has been increased and that the specific exposure may have provided the actual agent. The other is to say that the House created, exceptionally, a cause of action for the increased risk of mesothelioma rather than for the disease itself. In the former case, satisfying the new causal requirement would entitle the claimant to sue for the whole injury caused by contracting the disease. In the latter case, he would be able to sue only for the loss caused by the risk of his contracting the disease having been increased. That would be a proportion of the injury caused by the disease, depending on the extent to which the risk had also been created by other causes. In Barker v Corus14 the House of Lords (Lord Rodger of Earlsferry dissenting) adopted the second explanation of what had happened in Fairchild. Parliament almost immediately reversed this decision by a statute giving effect to the first explanation,15 which had been advocated by Lord Rodger in his dissenting speech. The result is that a person who negligently exposes someone to asbestos so as to increase the risk that he will develop mesothelioma and may have provided the agent which caused the disease will be liable to compensate the victim for the injury caused by the disease and will be jointly and severally liable with anyone else who may be liable on a similar basis.   Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32.   Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572.   S 6 Compensation Act 2006.

13 14 15

Causation  9 Fairchild was avowedly an exceptional case, although the limits of the exception remain to be tested. But the subsequent decision in Gregg v Scott16 shows that whether one adopts the first explanation or the second (increase of the risk of an adverse outcome or loss of the chance of avoiding it as the cause of action), it does not apply outside the limits which Fairchild laid down. In other cases, including other cases of medical negligence, it is still necessary to prove that the injury would not otherwise have happened. Fairchild does however cause difficulty for the adherents of the cause in fact doctrine. The facts which the House of Lords held sufficient to satisfy the causal requirements of the relevant rule do not pass the filter. Applying the standard rule of the burden of proof, the defendant is not part of the history because it has not been proved more likely than not that he had anything to do with the mesothelioma. Therefore the epistemological value returned by the rule is 0. He is treated as having had nothing to do with the disease. By the same token, exposure to asbestos by any one employer does not pass the NESS test. So the subscribers to this method of analysis are obliged to say that the House of Lords has been playing fast and loose with the true meaning of causation. The defendant did not ‘really’ cause the mesothelioma but is treated for the purposes of liability as if he had. Caught between their definition of ‘cause in fact’ and the unforgiving epistemology of the law, supporters of such a ‘first stage’ requirement have to invent fictions to patch the holes in their theory. In my view, all this wringing of hands is quite unnecessary. It might be easier if, instead of speaking of proof of ‘causation’, which makes it look as if we are dealing with one monolithic concept which can be defined as ‘part of the history’ or by NESS, we spoke of the ‘causal requirements’ of a legal rule. That would make it clear that causal requirements are creatures of the law and nothing more. The causal requirements of one rule may be different from those of another. The causal requirements of section 23 of the Offences against the Person Act 1861 (Kennedy) are different from those of section 85(1) of the Water Resources Act 1991 (Empress Car) even though they both use the verb ‘cause’. It is true that, as Hart and Honoré pointed out, the courts, following common usage, have normally given a standard meaning to a requirement that X must have ‘caused’ Y, such as the ‘but-for’ test, selection of deliberate human acts and abnormal events as causes rather than necessary conditions, and so on. Departures from the standard case have to be explained. But different causal requirements are nevertheless causal requirements. To say, as in Fairchild, that the exposure to asbestos by the defendant must have increased the risk of mesothelioma is to state a causal requirement. To say, the absence of omniscience, that X increased the chances of Y happening and may have been a necessary condition of Y happening, is a statement about causal relations. Whether the rule about liability for mesothelioma should be interpreted as containing such a departure from the standard case is a question of legal policy. But the concept of cause in fact seems to me to add nothing of value to the discussion of this question.

  Gregg v Scott [2005] UKHL 2, [2005] 2 AC 176.

16

2 Risky Business: Causation in Asbestos Cancer Cases (and Beyond?) JOSEPH SANDERS*

I. Introduction

In this chapter, I discuss a set of opinions in the United States and Great Britain that address the difficult causal issue that arises in a subset of asbestos cases. The fundamental decision in the two leading cases, Rutherford v Owens-Illinois Inc1 and Fairchild v Glenhaven Funeral Services,2 was to determine causation in terms of risk rather than injury. In this chapter, I will call this the ‘risk-rule’.3 The three central questions in this chapter are: (i) how do the two courts define the scope of their risk rule, (ii) how do they attempt to justify limits on the scope of the rule, and (iii) is the risk rule something we should encourage in similar situations? If so, the risk rule could become a normal test for resolving the causal question in a variety of settings. In the following section, I provide some background on the asbestos problem. Section III reviews the general nature of the causal question in asbestos cases. Section IV discusses the application of the risk-rule in Rutherford and a few other American cases. Section V focuses on the adoption of the risk-rule in Fairchild and a few other United Kingdom cases. In section VI, I compare the scope of Rutherford and Fairchild. In section VII, I discuss the likelihood that the risk rule will be applied to other fact situations. I do so from a pragmatic perspective – what are the courts likely to do – and from a normative perspective – are there principled reasons to limit the application of the rule to some particular set of cases? In the conclusion, I express my own view concerning whether we should encourage the spread of the risk rule.

*  I wish to thank Richard Goldberg and Jane Stapleton for very helpful comments on earlier drafts of this article. 1   Rutherford v Owens-Illinois Inc 941 P2d 1203 (Cal 1997). 2   Fairchild v Glenhaven Funeral Services [2002] UKHL 22, [2003] 1 AC 32. 3   Calls to apportion liability on the basis of risk predate these cases. See GW Boston, ‘Toxic Apportionment: A Causation and Risk Contribution Model’ (1995) 25 Environmental Law 549.

12  Joseph Sanders

II. The Asbestos Mess

Asbestos is the common name given to minerals that occur naturally in rock formations.4 More precisely, ‘[a]sbestos is a broad commercial term for a group of naturally occurring hydrated silicates that crystalize in a fibrous habitat’.5 The minerals fall into two varieties, serpentine and amphibole fibres. Serpentine fibres, such as chrysotile, are pliable curly fibrils resembling scrolled tubes. Amphibole fibres are longer rod or needle-shaped spikes.6 Individual fibers are invisible to the naked eye. Asbestos produces almost all of its adverse health effects through inhalation of air-borne fibers into the lungs. Natural processes release some asbestos fibres into the air and water.7 However, the explosive growth in asbestos-related diseases followed the widespread commercial use of the mineral in the early 1900s owing to its very useful properties such as high tensile strength, heat and chemical resistance, and flexibility.8 Asbestos-containing materials found their way into buildings, ships and many other applications. The chrysotile mineral is the most abundant type of asbestos and has accounted for over 90 per cent of the world’s asbestos production.9 There is much disagreement about who had knowledge of the dangers posed by airborne asbestos and when this knowledge was acquired.10 There is evidence that the asbestos industry knew the mineral was harmful years before it was forced to concede this fact.11 However, only with the publication of the seminal research by Selikoff and his colleagues in the 1960s was it universally accepted that asbestos posed serious health risks.12 In the United States, early litigation by injured plaintiffs met with little success. The beginning of successful products liability litigation in the United States can be traced to the 1973 Fifth Circuit opinion in Borel v Fibreboard Paper Products Corp.13 Asbestos litigation became the first and is still the largest of the mass toxic torts. In the years following Borel, litigation involving this substance has involved hundreds of thousands of personal injury claims.14 It is, ‘the mass tort that dwarfs all others’.15 4   They are: actinolite, amosite, anthophyllite, chrysotile, crocidolite and tremolite. LS Siegel, ‘As the Asbestos Crumbles: A Look at New Evidentiary Issues in Asbestos Related Property Damage Litigation’ (1992) 20 Hofstra Law Review 1139, 1149. 5   Mossman and Gee, ‘Asbestos-Related Diseases’ (1989) 320 New England Journal of Medicine 1721, 1723. 6   Siegel (n 4) 1149. 7   ibid 1148. 8   Matter of Celotex Corp 196 B.R. 973, 980 (Bankr MD Fla 1996). 9   Siegel (n 4) 1149. 10   See P Brodeur, Outrageous Misconduct: The Asbestos Industry on Trial (New York, Pantheon Books, 1985); P Bartrip, Beyond the Factory Gates: Asbestos and Health in Twentieth Century America (London, Continuum, 2006); P Bartrip, The Way from Dusty Death: Turner and Newall and the Regulation of Occupational Health in the British Asbestos Industry, 1890–1970 (London, Athlone, 2001); G Tweedale, Magic Mineral to Killer Dust: Turner & Newall and the Asbestos Hazard (Oxford, Oxford University Press, 2000); J McCulloch and G Tweedale, Defending the Indefensible: The Global Asbestos Industry and Its Fight for Survival (Oxford, Oxford University Press, 2008). 11  See In re Joint Eastern and Southern Dist Asbestos Litigation, 129 BR 710 (ED NY 1991), judgment vacated on other grounds, 982 F2d 721, opinion modified on rehearing, 993 F2d 7 (2nd Cir 1993).‘When these studies found that asbestos was a carcinogen this information was suppressed. The ensuing cover-up, effected through industry associations and research compacts, resulted in thousands of deaths’. 12   See eg, Irving J Selikoff et al, ‘The Occurrence of Asbestosis Among Insulation Workers in the United States’ (1965) 132 Annals of the New York Academy of Science 139. 13   Borel v Fibreboard Paper Products Corp, 493 F2d 1076 (5th Cir 1973). Borel, like the vast majority of asbestos cases, was brought on a products liability failure to warn theory. 14   Boston (n 3) 291. 15   D Hensler and M Peterson, ‘Understanding Mass Personal Injury Litigation: A Socio-Legal Analysis’ (1993) 59 Brooklyn Law Review 961, 1004.

Causation in Asbestos Cancer Cases (and Beyond?)  13 The end of the asbestos cases is not in sight. A Rand Corporation summary in 2001 indicated that there had been over half a million claimants and the number of claims filed annually had risen sharply in recent years for both malignant and non-malignant injuries. It reported United States insurers had paid approximately $21.6 billion, and major defendants had spent more than $1 billion apiece on litigation costs.16 A follow-up report in 2005, discussing data through 2002, indicated a continued explosion in claims, mostly due to case filings involving non-malignant injuries.17 Paul Carrington notes that by 2002 three-­ quarters of a million claims had been filed and more than $50 billion had been paid to claimants.18 Today, that number approaches $60 billion. Between 75 and 100 defendant corporations have been forced into bankruptcy, many in the last decade.19 However, beginning around 2003, the number of new filings in the United States declined.20 The decline is primarily the result of a smaller number of non-cancer cases being filed. As a result, there is considerable disagreement about where we are in the claims process in the United States, and estimates of the number of as yet unfiled claims vary widely.21 Court claims and expenditures have been much lower in the UK as, of course, have been mean recoveries.22 The composition of the defendants has changed as well. As traditional defendants fall into bankruptcy they have been replaced by non-traditional defendants who increasingly carry the burden of liability. Carrington notes that more than 8,000 businesses have been named as defendants.23 Many of these defendants are far less culpable than earlier defendants. In the United States, various attempts to find a global solution to the asbestos problem have been relatively unsuccessful. Efforts at global settlements within the judicial system appear to be very unlikely.24 Despite repeated judicial and academic calls for legislative action, the Congress has by-and-large sat on its hands and there are no present indications it will do otherwise in the near future.25 As a result, courts in both the United States and the United Kingdom continue to be confronted with very difficult causal questions. 16   D Hensler, S Carroll, M White, and J Gross, Asbestos Litigation in the U.S.: A New Look at an Old Issue (Santa Monica, CA, RAND, Institute for Civil Justice, 2001). The Rand report generated two special law review issues published in 2003 in the South Texas Law Review and Pepperdine Law Review. 17   SJ Carroll, D Hensler, J Gross, EM Sloss, M Shonlau, A Abrahamse and JS Ashwood, Asbestos Litigation (Santa Monica, CA, RAND, Institute for Civil Justice, 2005) 71. 18   P Carrington, ‘Asbestos Lessons: The Consequences of Asbestos Litigation’ (2007) 26 Review of Litigation 583, 593. 19   J Barnes, ‘Rethinking the Landscape of Tort Reform: Legislative Inertia and Court-Based Tort Reform in the Case of Asbestos’ (2007) 28 Justice System Journal 157, 169. 20   P Hanlon and A Smetak, ‘Asbestos Changes’ (2007) 62 New York University Annual Survey of American Law 525, 593–95; M Behrens, ‘What’s New In Asbestos Litigation?’ (2009) 28 Review of Litigation 501, 510. 21   JL Stengel, ‘The Asbestos End-Game’ (2006) 62 New York University Annual Survey of American Law 223; MA Behrens, ‘The Asbestos Litigation Crisis: The Tide Appears to be Turning’ (2006) 12 Connecticut Insurance Law Journal 477; D Hensler, ‘Asbestos Litigation in the United States: Triumph and Failure of the Civil Justice System’ (2006) 12 Connecticut Insurance Law Journal 255, 279. 22   S Jasanoff and D Perese, ‘Welfare State or Welfare Court: Asbestos Litigation in Comparative Perspective’ (2004) 12 Journal of Law and Policy 619. In general, levels of tort litigation are higher in the United States than in Britain. See RA Posner, ‘Explaining The Variance in the Number of Tort Suits Across U.S. States and Between the United States and England’ (1997) 26 Journal of Legal Studies 477. There is some evidence that the incidence of asbestos-related diseases in the UK trails that in the United States by a decade or two. J Peto et al, ‘The European Mesothelioma Epidemic’ (1999) 79 British Journal of Cancer 666. 23   Carrington (n 18) 593. 24   Carrington (n 18). 25   In recent years, the proposed legislation has carried the title the Fairness in Asbestos Injury Resolution Act (FAIR). One version of the Act would create a trust fund of approximately $140 billion to compensate workers exposed to asbestos. Exposed workers would be paid under a no-fault system based on the severity of their injury. Workers with mesothelioma would receive $1.1 million. The bill would bar all lawsuits alleging injury from asbestos exposure unless the fund runs out of money. Businesses and insurers would pay into the fund for an estimated

14  Joseph Sanders

III. The Causal Question in Asbestos Cases

In two respects, the causal issue in asbestos cases is relatively easy. First, plaintiffs almost never confront questions of general causation, ie, is the substance capable of causing the disease in question. Second, because two of the three most serious ailments caused by the mineral – mesothelioma and asbestosis – are ‘signature diseases’, their very presence implicates asbestos as the cause. Only lung cancer is an ailment with multiple rival causes.26 Nevertheless, asbestos cases do pose causal problems as well as problems of proof of causation. First, problems are generated by the latency period between exposure and injury. Both lung cancer and mesothelioma have quite long latency periods. For mesothelioma the average latency period is on the order of 30 to 40 years.27 However, sometimes the diseases develop after even longer passage of time.28 This creates causal problems in two ways. First, some plaintiffs may be ‘sick-too-soon’ for a given defendant’s product to have caused their injury. Second, and more important, the plaintiff may have a difficult time discovering to whose asbestos he was exposed, a problem that becomes more acute with the demise of all of the primary asbestos manufacturers. Second, because of the long latency period, most plaintiffs have been exposed to the asbestos-containing products of many defendants. Courts must decide how to resolve causal questions among the defendants. This involves the intersection of both causation and proof of causation. Let’s discuss causation first. How do asbestos-related injuries come about? Jane Stapleton notes three ways to understand the causal mechanism by which an exposure produces an injury. First, the disease may occur via a single-hit mechanism where a single ‘insult’ such as the inhalation of a single fibre of a mineral results in the total injury suffered. Infectious diseases are typically caused in this way . . . Secondly, the mechanism may be cumulative. Here each exposure, including the first, results in some actual injury. Each exposure is by itself a but-for factor to some actual injury even though it is not a but-for factor to the entire condition . . . Asbestosis is caused by such a mechanism . . . Thirdly, a disease may operate by a threshold mechanism where there is no injury at all until the accumulated dose exceeds some threshold. Here a pre- and less-than-threshold dose is not by itself a but-for factor to any actual injury. Where the threshold is passed, any pre-­threshold dose is a cause of the entire injury which is triggered. Noise-induced deafness seems to be caused by such a mechanism.29 30 years. For a review of the proposed legislation, see P Hanlon, ‘An Elegy for the FAIR Act’ (2006) 12 Connecticut Insurance Law Journal 517. At the state level, a number of jurisdictions have adopted the strategy of creating inactive dockets and similar case management plans to deal with the large number of claimants who are not currently sick. Medical Criteria statutes designed to achieve similar results have passed in at least six states. See J Sanders, ‘Medical Criteria Acts: State Statutory Attempts to Control the Asbestos Litigation’ (2008) 37 Southwestern Law Review 671; P Zimmerly, ‘The Answer is Blowing in Procedure: States Turn to Medical Criteria and Inactive Dockets to Better Facilitate Asbestos Litigation’ (2008) 59 Alabama Law Review 771. 26   Within the asbestos arena, the biggest threat is to plaintiffs with lung cancer who have been lifetime smokers. Some sources suggest that smoking is implicated in 90 per cent of all lung cancers in the United States. V Rogglie, ‘Asbestos’, in DL Faigman, MJ Saks, J Sanders and EK Cheng, Modern Scientific Evidence § 26:14 (St Paul, Thompson/West 2008–09). 27   Faigman et al Modern Scientific Evidence (2008-09) § 26:24. 28   P Mustacchi, ‘Lung Cancer Latency and Asbestos Liability’ (1996) 17 Journal of Legal Medicine 277. However, latency periods are dose sensitive: the greater the exposure the shorter the latency period. 29   J Stapleton, ‘Two Causal Fictions at the Heart of U.S. Asbestos Doctrine’ (2006) 122 LQR 189, 191.

Causation in Asbestos Cancer Cases (and Beyond?)  15 Next, there is the issue of proof of causation. When the plaintiff sues multiple defendants, we must determine whether it is possible to prove which defendant caused which part of the plaintiff ’s injury. If so, the injury is divisible and most courts will ask the fact-finder to allocate damages to each responsible defendant according to the harm that defendant caused. What if this is not possible? That is, it is impossible to untangle which of two or more defendants caused the harm. Most courts will then resort to the idea of an indivisible injury and each defendant will be held responsible for the entire harm. A traditional example would be an automobile accident where the plaintiff is struck simultaneously by two other automobiles and suffers a head injury. If it is possible to sort out which defendant caused the injury, then traditionally that defendant is responsible for the harm. If, however, this is not possible, then both defendants are responsible for the injury.30 This does not mean, of course, that either of them will end up paying the entire amount. That depends on several other factors, such as whether responsibility is apportioned among the parties, whether the jurisdiction retains joint and several liability for joint tortfeasors, and the nature of the contribution rights the defendants may have against each other.31 One additional proof issue should be mentioned at this point. Must the plaintiff prove that the defendant’s product was the ‘but for’ cause of his injury or will some other type of proof suffice? In the United States, many state courts pose the causal question not as ‘but for’ causation but whether the defendant’s tortious behaviour is a ‘substantial factor’ in producing the plaintiff ’s harm.32 All of these issues interact in the asbestos cases. In this chapter, I focus my attention on the causation and proof issues that arise when the plaintiff’s injury occurs after exposure to the asbestos products of multiple defendants. Specifically, how do the possible causal mechan­ isms discussed by Stapleton interact with the proof of causation issues discussed above?

IV. The Risk Rule in the United States: Rutherford and Related Cases

This section focuses on the California Supreme Court opinion in Rutherford v OwensIllinois, Inc.33 However, in order to place the opinion in its proper context I must briefly discuss three earlier opinions. I also discuss the single American state case that purports to adopt the Rutherford approach.

A.  Borel Borel v Fibreboard Paper Products Corp34 began the surge of asbestos litigation in the United States. Clarence Borel suffered from asbestosis. The jury awarded him $79,000 in damages, which was reduced by prior settlements to $58,000 against six defendants. The Fifth Circuit agreed that each was jointly and severally liable for Mr Borel’s injury. One preliminary fact about Borel is worth mentioning. Following the approach of the Second Restatement of Torts, it judged causation using the substantial factor test:   Boston (n 3) 557.   See Restatement (Third) of Torts: Apportionment of Liability §§ A18, C18, D18, E18, and 23. 32   Stapleton (n 29) 191. 33   Rutherford v Owens-Illinois Inc 941 P2d 1203, 1215–20 (Cal 1997). 34   Borel v Fibreboard Paper Prods Corp 493 F2d 1076 (5th Cir 1973), cert denied, 419 US 869 (1974). 30 31

16  Joseph Sanders We next consider whether there was substantial evidence to support the jury’s finding that each defendant was the cause in fact of injury to Borel. The traditional rule is that a defendant’s conduct is the cause of the event if it was a substantial factor in bringing it about . . . Second Restatement of Torts, §§ 431, 433 . . . Whether the defendant’s conduct was a substantial factor is a question for the jury, unless the court determines that reasonable men could not differ.35

One should not make too much of the use of the idea of ‘substantial factor’ as it is applied in Borel. The substantial factor test of causation had been adopted by the American Law Institute as a solution to the ‘two fire’ cases and also to the problem of trivial causes. However, many courts commonly understood it to be synonymous with the traditional but for test in ordinary cases. It is not clear that Judge Wisdom intended to use it in any other sense in Borel. That is, if asked, he may well have said that the asbestos of each defendant was a ‘but for’ cause of the plaintiff ’s asbestosis. However, as I discuss below, the substantial factor test does play a significant role in some later US cases. Borel was a diversity case, decided using Texas law.36 At the time Texas still retained the contributory negligence rule. The plaintiff was not barred by this rule because it did not apply to his successful strict products liability claim. However, the rule had another effect on the case. Before US jurisdictions moved to a comparative fault, the rule for the apportionment of damages among equally responsible defendants was a pro rata rule. The only way there could have been apportionment in any other fashion would be for the courts to conclude that the plaintiff ’s injury was divisible. Normally during this period, divisibility was thought of as an all or nothing affair. For example, in the automobile accident example above, for the injury to be thought of as divisible the fact-finder would have to conclude that one of the defendants caused the entire injury. The Restatement (Second) of Torts did recognise the possibility of dividing an injury based on one’s relative contribution to an entire harm.37 Examples included crop destruction due to the trespass of cattle from multiple defendants, apportionment being based on the relative number of cattle owned by each defendant; and pollution of a stream due to oil discharges from several factories, apportionment being based on the amount of oil coming from each factory. In Borel, however, neither the court nor the parties seem to have attempted to draw a parallel with these situations. Here is what the Fifth Circuit said about causation: In general [the disease] does not manifest itself until ten to twenty-five or more years after initial exposure. This latent period is explained by the fact that asbestos fibers, once inhaled, remain in place in the lung, causing a tissue reaction that is slowly progressive and apparently irreversible. Even if no additional asbestos fibers are inhaled, tissue changes may continue undetected for decades. By the time the disease is diagnosable, a considerable period of time has elapsed since the date of the injurious exposure. Furthermore, the effect of the disease may be cumulative since each exposure to asbestos dust can result in additional tissue changes. A worker’s present condition is the biological product of many years of exposure to asbestos dust, with both past and recent exposures contributing to the overall effect. All of these factors combine to make it impossible, as a practical matter, to determine which exposure or exposures to asbestos dust caused the disease.38

 ibid.   Under the United States Constitution, Art 3, s 2, when a suit arises between citizens of different states either party may remove it to the federal courts. In the seminal case of Erie RR v Tompkins, 304 US 64, 58 SCt 817 (1938), the Supreme Court adopted the principle that a federal court exercising diversity jurisdiction over a case that does not involve a federal question must apply the substantive law of the state where the claim arose. 37   Restatement (Second) of Torts § 433A. 38   Borel v Fibreboard Paper Prods Corp 493 F2d 1076, 1083 (5th Cir 1973), cert denied, 419 US 869 (1974). 35 36

Causation in Asbestos Cancer Cases (and Beyond?)  17 The key point is that although the Borel opinion spoke of asbestosis as a cumulative disease, it treated it as an indivisible injury. In hindsight, one could argue that the Borel court could have been more open to apportioning damages based on relative exposure to each defendant’s product. But this is hindsight. Recall that the case was governed by Texas law and the court did follow Texas precedent concerning indivisible injuries. The Texas rule, similar to that in many states, was that when an injury is indivisible defendants are jointly and severally liable for the total damages.39 In Borel only two defendants attempted anything along these lines. They argued that their asbestos could not have caused the plaintiff ’s injury because his exposure to their product was too recent. In response, the Fifth Circuit said, ‘Even the most recent exposures could have added to or accelerated Borel’s overall condition’.40 Perhaps the courts would have been more receptive if the defendants had attempted not to completely absolve themselves but to argue they were responsible for a small percentage of the total injury. We’ll never know. But, truth be told, it is not clear that either the trial court or the Fifth Circuit was really open to such an effort. Dividing causation along such lines without a clear unit of apportionment, ie, the defendant’s percentage of the total number of cows that ate the plaintiff ’s crop, would have been novel during this period. Many courts picked up on the indivisible injury rationale. Indeed, they made the plaintiff ’s case even easier by not requiring expert testimony that more likely than not the defendant’s product caused any part of the plaintiff ’s ailment. Co-worker testimony was often sufficient proof of exposure to a particular defendant’s product. A defendant found to be responsible faced the possibility of being liable for the entire injury although whether a given defendant would actually confront liability for the entire injury turned on each state’s apportionment and contribution rules.41 In this relaxed proof environment, the only key issue was whether a plaintiff could indicate significant exposure to the defendants’ product.42 Eventually, most courts resolved this proof issue by adopting the ‘frequency, regularity and proximity’ test generally attributed to Lohrmann v Pittsburgh Corning Corp.43

B. Lohrmann Mr Lohrmann also suffered from asbestosis. The trial court granted three defendants directed verdicts because there was insufficient evidence that the pipefitter plaintiff came into contact with their products. The appellate court affirmed. It rejected the argument that if the plaintiff can prove the defendant’s asbestos-containing product was anywhere at the workplace while the plaintiff was there, a jury question existed as to whether the product was a cause of the plaintiff ’s disease. Instead, it concluded that ‘to support a reasonable inference of substantial causation from circumstantial evidence, there must be evidence of exposure to a specific product on a regular basis over some extended period of time in proximity to where the plaintiff actually worked’.44 Some form of this test has been adopted   Landers v East Texas Salt Water Disposal Co 248 SW2d 731 (Tex 1952).   Borel v Fibreboard Paper Prods Corp 493 F2d 1076, 1094 (5th Cir 1973), cert denied, 419 US 869 (1974).   Stapleton (n 29). 42  J Stapleton, ‘The Two Explosive Proof-of-Causation Doctrines Central to Asbestos Claims’ (2009) 74 Brooklyn Law Review 1011. 43   Lohrmann v Pittsburgh Corning Corp 782 F2d 1156 (4th Cir 1986). 44   ibid 1162. 39 40 41

18  Joseph Sanders by most courts. It has been used to prove exposure to a particular defendant’s product and to prove exposure to asbestos in general. In nearly all these cases, the courts view the test as an instantiation of the Second Restatement of Torts § 431 substantial factor test of causation.45

C.  Migues The Borel approach was not restricted to asbestosis cases. It was also applied to mesothelioma (and lung cancer) cases. Mesothelioma is not a cumulative disease. There is a doseresponse relationship between asbestos exposure and the likelihood one will contract the disease; but this is commonly understood to be true because the more asbestos fibers in one’s lungs, the greater the probability some fibre or set of fibres will cause the genetic damage that generates the disease. There are documented cases of people contracting the disease with very low levels of exposure, and the severity of the disease once contracted is not correlated with one’s level of exposure. The Borel approach was applied to these cases as well. In Migues v Fibreboard Corp,46 the plaintiff ’s husband suffered from mesothelioma. After his death, his wife sued 14 defendants. Prior to trial, she settled with all but one, Nicolet. On appeal, Nicolet argued that the plaintiff did not meet her burden of proving that the products of this particular defendant, Nicolet, were a cause of her husband’s death. To this, the court responded: Three of [Migues’] former co-workers testified that the decedent had worked with defendant’s asbestos products while employed as an insulator at Texaco, and that the job of cutting insulation pieces to fit pipes produced dust which the insulators, including the decedent, inhaled . . . In sum, plaintiff introduced uncontroverted evidence that Mr. Migues died of mesothelioma, that asbestos fibers were present in his lungs at the time of death, that asbestos inhalation is the only known cause of mesothelioma, that Nicolet produced insulation products containing asbestos, and that Mr. Migues worked with Nicolet asbestos products during the course of his employment as an insulator. We find that there was more than sufficient circumstantial evidence to support the jury’s conclusion that Nicolet’s products were a producing cause of Mr. Migues’ death. Borel v Fibreboard, 493 F.2d at 1094.47

The Migues court did not discuss the relative exposure to Nicolet’s products as compared to the products of the settling 13 defendants. Rather, it proceeded as if Mr Migues’s injury was indivisible. At the time Migues was decided, Texas still had not adopted a comparative responsibility regime for products liability cases.48 Thus, the defendant was responsible for the entire injury. Similar results have been reached in much more recent cases.49 In many ways, mesothelioma cases much more nearly conform to the indivisible injury model than do asbestosis cases. If we view the choice between viewing something as a divisible or an indivisible injury as a question of proof, the mesothelioma plaintiff truly 45   J Sanders, WC Powers Jr and MD Green, ‘The Insubstantiality of the “Substantial Factor” Test for Causation’ (2008) 73 Missouri Law Review 399. 46   Migues v Fibreboard Corp 662 F2d 1182 (5th Cir 1981). 47   ibid 1185. 48   A comparative causation scheme was formulated by the Texas Supreme Court in Duncan v Cessna Aircraft Co 665 SW2d 414 (Tex 1984). After Duncan, juries in similar cases were asked to assign a percentage of liability to each defendant and each settling party but the injury was apparently still treated as indivisible. See King v Armstrong World Industries Inc 906 F2d 1022 (5th Cir 1990). 49  See John Crane Inc v Jones 604 SE2d 822 (Ga 2004).

Causation in Asbestos Cancer Cases (and Beyond?)  19 confronts an indivisible injury situation. That is, no one knows or can know which defendant caused the cancer just as no one can know which defendant caused the head injury in our automobile accident hypothetical. This state of affairs is most fully recognised in Rutherford.

D. Rutherford In order to understand Rutherford v Owens-Illinois Inc50 one must appreciate the posture of the case when it came to the California Supreme Court. The lower courts in Rutherford and other California asbestos cases winding their way through the system were confronted with defence arguments that mesothelioma conforms to the single-hit model of causation. Almost certainly, other defendants had made similar arguments, but without much success in persuading courts that this created a causal problem for the plaintiff.51 In California, the home of alternative liability, they found a more favourable audience. If we believe that mesothelioma is caused by a particular, specific injury from a single asbestos fibre, then cases such as Migues are choosing to ignore this fact or are sub rosa adopting something like a Summers v Tice 52 approach to the problem. Recall, that in this well-known case, two hunters shot at the plaintiff, but the pellet of only one of the defendants struck the plaintiff in the eye. The court reversed the burden of ultimate persuasion in the case, requiring each defendant to prove by a preponderance of the evidence that his pellet was not the one that injured the plaintiff. This approach has come to be called ‘alternative liability’. If the asbestos-mesothelioma relationship really is just like the Summers situation, ie, one and only one defendant actually causes the injury, then we could recognise this fact and apply the Summers solution. One California appellate court agreed and adopted a variation on Summers in a mesothelioma case.53 In Rutherford itself, the trial court adopted this approach, but had been overruled by the court of appeals.54 Thus, unlike almost all other appellate courts, Rutherford was compelled to decide whether alternative liability should be extended to the asbestos-cancer cases. Because of this procedural posture, Rutherford v Owens-Illinois Inc55 is in many ways one of the most coherent asbestos opinions from the 1990s. Mr Rutherford died of lung cancer that the jury found was caused by his inhalation of asbestos fibers over a near 40-year career working at the Naval Shipyard at Mare Island.56 He sued numerous asbestos manufacturers. All except Owens-Illinois settled prior to trial.57 Over the defendant’s objection, the trial court had instructed the jury that once the plaintiff establishes that the defendant manufactured or sold defective asbestos-containing products to which he was exposed, the burden on causation shifts to the defendants to prove its

  Rutherford v Owens-Illinois Inc 941 P2d 1203 (Cal 1997).   An exception is Menne v Celotex Corp 861 F2d 1453 (10th Cir 1988). In a case involving Nebraska law, the 10th Circuit concluded the plaintiff could not meet his burden on causation and approved a burden-shifting instruction in a mesothelioma case. However, it reversed the trial court on other grounds. 52   Summers v Tice 199 P2d 1 (Cal 1948). See MD Green, ‘Second Thoughts About Apportionment in Asbestos Cases’ (2008) 37 Southwestern Law Review 531, 539. 53   Coughlin v Owens-Illinois Inc 27 Cal.Rptr.2d 214 (Cal App 1993). 54   Rutherford v Owens-Illinois Inc 941 P2d 1203, 1206 (Cal 1997). 55   Rutherford v Owens-Illinois Inc 941 P2d 1203 (Cal 1997). 56   ibid 1208. 57   ibid 1208. 50 51

20  Joseph Sanders product was not the cause of the plaintiff ’s injury.58 The Supreme Court agreed with the court of appeals that this instruction was in error, because it found that ‘no insuperable barriers prevent an asbestos-related cancer plaintiff from demonstrating that exposure to the defendant’s asbestos products was, in reasonable medical probability, a substantial factor in causing or contributing to his risk of developing cancer.’59 The plaintiff was free to establish that his disease was the result of many separate exposures, each constituting a ‘substantial factor’ that contributed to his risk of injury.60 The key phrase in all of this is ‘contributed to his risk’ of injury. Because the opinion does a good job of setting out the issue, a lengthy quote is in order. At the most fundamental level, there is scientific uncertainty regarding the biological mechanisms by which inhalation of certain microscopic fibers of asbestos leads to lung cancer and mesothelioma. Although in some cases medical experts have testified that asbestos-related cancer is the final result of the fibrosis (scarring) process (see Armstrong World Industries, Inc. v. Aetna Casualty & Surety Co. (1996) 45 Cal.App.4th 1, 37–39, 52 Cal.Rptr.2d 690), a general reference on the subject describes the link between fibrosis and carcinogenesis as ‘a debated issue for which further extensive analysis is needed’. (1 Encyclopedia of Human Biology (1991) Asbestos, p. 423.) An answer to this biological question would be legally relevant, because if each episode of scarring contributes cumulatively to the formation of a tumor or the conditions allowing such formation, each significant exposure by the plaintiff to asbestos fibers would be deemed a cause of the plaintiff ’s cancer; if, on the other hand, only one fiber or group of fibers actually causes the formation of a tumor, the others would not be legal causes of the plaintiff ’s injuries. If, moreover, the question were answered in favor of the latter (single cause) theory, another question-apparently unanswerable-would arise: which particular fiber or fibers actually caused the cancer to begin forming. Because of the irreducible uncertainty of the answer, asbestos-related cancer would, under the single-fiber theory of carcinogenesis, be an example of alternative causation, i.e., a result produced by a single but an indeterminable member of a group of possible causes. The disease would thus be analogous to the facts of the hunting accident in Summers, supra, 33 Cal.2d 80, 199 P.2d 1. . . . [T]he burden of proof as to the mechanism of carcinogenesis [is not] disputed here; defendant concedes that plaintiff does not bear such a burden to ‘connect the manufacturer and the fibers’. Asbestos plaintiffs, Owens-Illinois acknowledges, ‘are not required to identify the manufacturer of specific fibers’ that caused the cancer. We agree: Plaintiffs cannot be expected to prove the scientifically unknown details of carcinogenesis, or trace the unknowable path of a given asbestos fiber. But the impossibility of such proof does not dictate use of a burden shift. Instead, we can bridge this gap in the humanly knowable by holding that plaintiffs may prove causation in asbestosrelated cancer cases by demonstrating that the plaintiff ’s exposure to defendant’s asbestos-­ containing product in reasonable medical probability was a substantial factor in contributing to the aggregate dose of asbestos the plaintiff or decedent inhaled or ingested, and hence to the risk of developing asbestos-related cancer, without the need to demonstrate that fibers from the defendant’s particular product were the ones, or among the ones, that actually produced the malignant growth.61

  ibid 1206.   ibid 1206.   ibid 1207. The substantial factor test is embodied in California Civil Jury Instruction 3.76, which reads: ‘The law defines cause in its own particular way. A cause of [injury] [damage] [loss] [or] [harm] is something that is a substantial factor in bringing about an [injury] [damage] [loss] [or] [harm]’. CALIFORNIA JURY INSTRUCTIONS – CIVIL 3.76 (2007). 61   Rutherford v Owens-Illinois Inc 941 P2d 1203, 1218–19 (Cal 1997) (emphasis added). 58 59 60

Causation in Asbestos Cancer Cases (and Beyond?)  21 At the liability phase of the trial, the jury was instructed to assign percentages of fault adding up to 100 per cent among the plaintiff, Owens-Illinois, other manufacturers of asbestos to which the decedent was exposed, and each employer that contributed to the exposure.62 The jury apportioned 1.2 per cent of the fault to Owens-Illinois, 2.5 per cent to Rutherford, and the rest to all the remaining entities. The percentage of responsibility assigned to Rutherford presumably was based on the fact that he had smoked approximately a pack of cigarettes a day over a period of 30 or more years until he quit smoking in 1977.63 The California Supreme Court seized upon this allocation as proof the jury was not mislead by the trial court’s erroneous burden-shifting instruction: To the contrary, the jury’s verdict suggests that, regardless of the burden shift, it accepted much of the defense’s factual theory, concluding that exposure to [the defendant’s product] contributed a relatively small amount to decedent’s cancer risk, but rejected defendant’s argument that such a small contribution should be considered insubstantial.64

Rutherford’s discussion of causation and proof of causation in mesothelioma cases is much clearer than other contemporary opinions. The opinion steers a delicate course between the Scylla of a Summers v Tice solution and the Charybdis of disregarding the causal mechanisms by which asbestos causes lung cancer and mesothelioma, but the posture of the case causes it to focus on Scylla. It goes to considerable pains to explain why the more radical Summers solution is not called for, but it avoids a principled discussion of the question that is central to the British cases I will turn to next: What is the reach of the risk rule? In which cases are courts to substitute ‘risk’ for ‘cause’? In lieu of a discussion of the rule’s scope the court simply states that this is a rule for asbestos cases. In conclusion, our general holding is as follows. In the context of a cause of action for asbestosrelated latent injuries, the plaintiff must first establish some threshold exposure to the defendant’s defective asbestos-containing products, and must further establish in reasonable medical probability that a particular exposure or series of exposures was a ‘legal cause of his injury, i.e., a substantial factor in bringing about the injury. In an asbestos-related cancer case, the plaintiff need not prove that fibers from the defendant’s product were the ones, or among the ones, that actually began the process of malignant cellular growth. Instead, the plaintiff may meet the burden of proving that exposure to defendant’s product was a substantial factor causing the illness by showing that in reasonable medical probability it was a substantial factor contributing to the plaintiff ’s or decedent’s risk of developing cancer.65

What should we make of this statement? Perhaps the court simply meant to save for some future case the question of Rutherford’s scope. If, on the other hand, the court really meant to restrict the risk rule to asbestos cases, it does not explain why. A principled explanation would not be easy, especially given the fact that Rutherford itself is not purely an asbestos rule. Recall that the plaintiff suffered from lung cancer, not mesothelioma and, as the jury found, his smoking was undoubtedly a risk factor leading to his disease. In Rutherford, smoking only played a role with respect to the plaintiff’s contributory negligence, but one may imagine the same case in which the plaintiff joined tobacco manufacturers as defendants. In that case,   ibid 1209.   ibid 1208. Because California retains joint and several liability for economic, but not for non-economic damages, the Rutherford plaintiffs were entitled to $177,047 in economic damages and $2,160 in non-economic damages from the defendant. ibid 1209–10; see also CAL CIV CODE § 1431.2 (West 2007). 64   Rutherford v Owens-Illinois Inc 941 P2d 1203, 1225 (Cal 1997). 65   ibid 1223. 62 63

22  Joseph Sanders would the Rutherford rule apply? Even more radically, could the risk rule ever apply in situations where the mechanism of causation is understood? As Jane Stapleton notes, does the risk rule relate only to the issue of proof of factual cause or could it stand for ‘an even more radical move, that risk creation is itself actionable?’66 For an answer to these and similar questions, one must look elsewhere than the California Supreme Court. Indeed, one must look elsewhere than the United States. To my knowledge, there are no California cases that have applied the Rutherford rule in a non-asbestos setting. Outside California, only a handful of opinions have even cited Rutherford and none have explicitly adopted its risk rule.67 The closest any court has come to adopting this approach is in the Texas Supreme Court opinion in Borg-Warner v Flores.68 In Borg-Warner, the Court acknowledged that it would not require a plaintiff to prove the scientifically unknowable details of which fibre caused what part of his injury. As to what a plaintiff must prove, the court quoted Rutherford for the proposition that: [W]e can bridge this gap in the humanly knowable by holding that plaintiffs may prove causation in asbestos-related cancer cases by demonstrating that the plaintiff ’s exposure to defendant’s asbestos-containing product in reasonable medical probability was a substantial factor in contributing to the aggregate dose of asbestos the plaintiff or decedent inhaled or ingested, and hence to the risk of developing asbestos-related cancer, without the need to demonstrate that fibers from the defendant’s particular product were the ones, or among the ones, that actually produced the malignant growth.69

However, in Borg-Warner the plaintiff suffered from asbestosis, not mesothelioma or lung cancer and thus the risk rule formulated by Rutherford was not needed to resolve the factual question in the case.70 In fact, the Borg-Warner court was focused far less on the risk rule than it was on Rutherford’s support for the substantial factor test of causation.71 Ultimately, the Texas Supreme Court decided that the plaintiff failed to prove that BorgWarner’s asbestos was a substantial factor contributing to plaintiff ’s disease. A substantial factor analysis was also employed to the same effect by the Pennsylvania Supreme Court in Gregg v V-J Auto Parts Inc.72   Stapleton (n 42) 1029.   In Kennedy v Southern California Edison Co 268 F3d 763 (9th Cir 2001), a case involving chronic myelogen­ous leukemia (CML) allegedly caused by exposure to radiation from the defendant’s nuclear power plant, the Court of Appeals held that the trial court’s failure to give the jury a Rutherford instruction was harmless error because uncontroverted expert testimony indicated that there was only a one in 100,000 chance the plaintiff ’s illness was caused by exposure at the plant. Therefore, it concluded that ‘it is not necessary for us to decide whether there is an obligation to give such an instruction in cases involving exposure to substances other than asbestos’. ibid 770. 68   Borg-Warner v Flores 232 SW3d 765 (Tex 2007). For a discussion of the opinion, see J Offerman, ‘ “The Dose Makes the Poison”: Specific Causation in Texas Asbestos Cases After Borg-Warner’ (2009) 41 Texas Tech Law Review 709. 69   Borg-Warner, 232 SW3d, 772–73 (quoting Rutherford, 941 P2d, 1219) (alteration in original). 70   See MD Green, ‘A Future for Asbestos Apportionment?’ (2006) 12 Connecticut Insurance Law Journal 315. 71   Rutherford made the following comment concerning the substantial factor test. Finally, at a level of abstraction somewhere between the historical question of exposure and the unknown biology of carcinogenesis, the question arises whether the risk of cancer created by a plaintiff ’s exposure to a particular asbestos-containing product was significant enough to be considered a legal cause of the disease. Taking into account the length, frequency, proximity and intensity of exposure, the peculiar properties of the individual product, any other potential causes to which the disease could be attributed (e.g., other asbestos products, cigarette smoking), and perhaps other factors affecting the assessment of comparative risk, should inhalation of fibers from the particular product be deemed a ‘substantial factor; in causing the cancer? Rutherford v Owens-Illinois Inc 941 P2d 1203, 1218 (Cal 1997). 72   Gregg v V-J Auto Parts Inc 943 A2d 216 (Pa 2007). See also Vanaman v DAP Inc 966 A2d 603 (Pa Super, 2009). But see John Crane Inc v Jones 604 SE2d 822 (Ga, 2004). 66 67

Causation in Asbestos Cancer Cases (and Beyond?)  23 Both Borg-Warner and Gregg are an indication that the substantial factor test continues to play a role in US asbestos cases. However, it is a far different role from the test originally played. In the early years of asbestos litigation in the United States, if the substantial factor test meant anything other than a synonym for the ‘but for’ test, its role was to liberalise causal proofs. The Lohrmann test, and the even more liberal ‘fiber-drift’ test,73 when coupled with the substantial factor jury instruction, made plaintiffs’ proof relatively easy. However, in Borg-Warner and Gregg the substantial factor test is invoked to mark a more demanding exposure requirement and thus to restrict recovery against some defendants. Used in this way, it is a type of proximate cause analysis designed to bar recovery as a matter of law.74

V. The Risk Rule in the United Kingdom: Fairchild and Related Cases

Asbestos litigation in Great Britain does not have the long and tortured history that exists in the United States, and in any case I am not the person to review all of the case law leading up to Fairchild. It is necessary, however, to mention two cases that preceded Fairchild: McGhee v National Coal Board75 and Wilsher v Essex Area Health Authority.76

A.  McGhee In McGhee the pursuer was employed at a brickworks and claimed that his dermatitis was caused by the fact that his employer failed to provide washing facilities that would have permitted him to clean up immediately after work. Before the House of Lords, the defence was on causal grounds. Failure to provide an opportunity to bathe immediately after work rather than after a rather short commute home did not cause the ailment. The pursuer’s expert was unwilling to say that if the appellant had been able to wash off the dust immediately after work, he would not have contracted the disease. He would only opine that the failure to provide showers materially increased the chance, or risk, that dermatitis might occur. The Court of Session held that the plaintiff had to prove his additional exposure caused by bicycling home unwashed caused the disease, ie, it was more probable than not that this additional exposure to injury was the cause of it. This, it said, he was unable to do. In the House of Lords, Lord Reid disagreed with this approach. From an American perspective, McGhee is often lumped with the lost-chance cases. However, there is a key connection to the asbestos exposure cases as well. In the following quote, Lord Reid draws a causal distinction that is parallel to that drawn in Rutherford; a distinction between a single-hit model and what Professor Stapleton calls a threshold  See Eagle-Picher v Balbos 578 A2d 228 (Md App, 1992).   For discussions of the substantial factor test in US tort law see J Stapleton (n 29); Sanders et al (n 45). This shift to a more pro-defendant point of view may simply reflect the different philosophical makeup of the courts in the 2000s. However, it may also reflect the fact that the defendants in more recent cases are not nearly as culpable as the defendants from the time of Borel. 75   McGhee v National Coal Board [1972] 3 All ER 1008. For a discussion of McGhee and other cases leading up to Fairchild see C Miller, ‘Judicial Approaches to Contested Causation: Fairchild v Glenhaven Funeral Services’ (2002) 1 Law, Probability & Risk 119, and R Goldberg, ‘Causation and Defences’ in A Grubb, J McHale and J Lang (eds), Principles of Medical Law 3rd edn (Oxford, Oxford University Press, 2010) paras 6.13–6.22. 76   Wilsher v Essex Area Health Authority [1988] AC 1074 HL. 73 74

24  Joseph Sanders model. Moreover, as in Rutherford, he avoids making a choice between the models, but he does recognise that the uncertain aetiology of the injury influences how we should describe the causal question presented in the case: It may be that an accumulation of minor abrasions of the horny layer of the skin is a necessary precondition for the onset of the disease. Or it may be that the disease starts at one particular abrasion and then spreads, so that multiplication of abrasions merely increase the number of places where the disease can start and in that way increases the risk of its occurrence. I am inclined to think that the evidence points to the former view. But in a field where so little appears to be known with certainty I could not say that that is proved. If it were then this case would be indistinguishable from Wardlaw’s case. But I think that in cases like this we must take a broader view of causation . . . I can not accept the distinction drawn by the Lord Ordinary between materially increasing the risk that the disease will occur and making a material contribution to its occurrence. There may be some logical ground for such a distinction where our knowledge of all the material factors is complete. But it has often been said that the legal concept of causation is not based on logic or philosophy. It is based on the practical way in which the ordinary man’s mind works in the every-day affairs of life. From a broad and practical viewpoint I can see no substantial difference between saying that what the defender did materially increased the risk of injury to the pursuer and saying that what the defender did made a material contribution to his injury.77

The import of McGhee is its indication that the Lords were prepared to substitute proof of risk for proof of causation when it was uncertain whether the pursuer’s dermatitis was caused by the workplace exposure itself or by the exposure due to the delayed opportunity to bathe.78

B. Wilsher Martin Graham Wilsher, an infant born three months prematurely, suffered from retro­ lental fibroplasia (RLF), and as a result was almost totally blind. After his birth, the plaintiff was given supplemental oxygen. In addition, through the defendant’s negligence, an umbilical catheter was incorrectly placed in a vein instead of an artery. The parents claimed that this resulted in their child receiving excessive amounts of oxygen and that this was the cause of his RLF. The case presented two causal questions, (1) the causes of RLF and (2) whether any negligent excess of oxygen was likely to have been the cause in the plaintiff ’s case. The Court of Appeal concluded that excess oxygen can cause or contribute to RLF but so can a number of other factors. The question presented, therefore, was whether the case is governed by McGhee. Mustill LJ concluded that it was. There is, however, one problem which must be tackled. I had at one time believed that the present case is on all fours with the McGhee case [1973] 1 W.L.R. 1 and that any apparent difference between the two simply stemmed from the way in which the problem happened to be expressed. I am now persuaded that this is not so, and that the two situations really are different. In the McGhee case there was only one risk operating, namely that contact of a sweaty skin with brick dust would lead to dermatitis. The fact that such contact did cause the injury was not in dispute . . . The uncertainty was whether the fault had tipped the scale. In the present case there is a greater uncertainty.   McGhee v National Coal Board [1972] 3 All ER 1008, 1011.   This is Lord Hoffmann’s understanding of the case. Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 75

77 78

HL.

Causation in Asbestos Cancer Cases (and Beyond?)  25 Instead of a single risk factor known to have caused the injury there is a list of factors, which cannot be fully enumerated in the current state of medical science, any one of which might have caused the injury. What the defendants did was not to enhance the risk that the known factor would lead to injury, but to add to the list of factors which might do so . . . The question is whether this makes a crucial difference. The root of the problem lies in the fact that, for reasons of policy, their Lordships’ House mitigated the rigour of the rule that the plaintiff must prove that the breach caused the loss, in the interests of achieving a result which was considered to be just. Given that this was a decision based on policy, rather than a chain of direct reasoning, the difficulty is to know whether a similar approach can properly be adopted in the different circumstances of the present case. After much hesitation I have come to the conclusion that it can. Reading all the speeches together, the principle applied by the House seems to me to amount to this. If it is an established fact that conduct of a particular kind creates a risk that injury will be caused to another or increases an existing risk that injury will ensue; and if the two parties stand in such a relationship that the one party owes a duty not to conduct himself in that way; and if the first party does conduct himself in that way; and if the other party does suffer injury of the kind to which the risk related; then the first party is taken to have caused the injury by his breach of duty, even though the existence and extent of the contribution made by the breach cannot be ascertained. If this is the right analysis, it seems to me that the shape taken by the enhancement of the risk ought not to be of crucial significance. In the McGhee case [1973] 1 W.L.R. 1, the conduct of the employers made it more likely that the pursuer would contract dermatitis, and he did contract dermatitis. Here, the conduct of those for whom the defendants are liable made it more likely that Martin would contract RLF, and he did contract RLF. If considerations of justice demanded that the pursuer succeed in the one case, I can see no reason why the plaintiff should not succeed in the other.79

Sir Nicholas Browne-Wilkinson, VC dissented from this view. He said: The position, to my mind, is wholly different from that in the McGhee case [1973] 1 W.L.R. 1 where there was only one candidate (brick dust) which could have caused the dermatitis, and the failure to take a precaution against brick dust causing dermatitis was followed by dermatitis caused by brick dust. In such a case, I can see the common sense, if not the logic, of holding that, in the absence of any other evidence, the failure to take the precaution caused or contributed to the dermatitis. To the extent that certain members of the House of Lords decided the question on inferences from evidence or presumptions, I do not consider that the present case falls within their reasoning. A failure to take preventative measures against one out of five possible causes is no evidence as to which of those five caused the injury.80

The case was appealed to the House of Lords which sidestepped the difficult causal question by ordering a new trial to resolve the question of whether the oxygen negligently administered to Martin was capable of having caused or materially contributed to his RLF.81 However, in passing Lord Bridge noted that he was ‘quite unable to find any fault’ with the position of the Vice-Chancellor.82 McGhee and Wilsher set the stage for Fairchild.

  Wilsher v Essex Area Health Authority [1987] QB 730, 771–72.   ibid 779. 81   Wilsher v Essex Area Health Authority [1988] AC 1074 HL. 82   ibid 1090. 79 80

26  Joseph Sanders

C. Fairchild In Fairchild v Glenhaven Funeral Services,83 the plaintiffs and plaintiffs’ decedents84 suffered from mesothelioma. As in Rutherford, the products of multiple entities – here multiple employers – may have caused the harm. The Lords adopted the Rutherford description of the nature of the disease. There is a dose-response relationship between exposure and injury but the disease is equally severe no matter what the level of exposure. In addition, no-one understands the mechanism by which asbestos exposure produces the disease. One thing is clear. As Lord Bingham notes, The medical evidence does not permit a finding that mesothelioma is a divisible or cumulative injury and it is not open to a court to infer that it is such an injury or to find that the tortious element of the exposure contributed to the injury. Liability for mesothelioma, unlike other asbestos-related illnesses, cannot therefore be susceptible to apportionment.85

At least not apportionment based on causation. In answer to the question of whether under these circumstances the plaintiff may prevail, Fairchild like Rutherford says yes and for the same reason. One may recover upon the proof of risk, not causation.86 Once again, the key issue is to which circumstances does this rule apply? On this point, Fairchild is only a little more forthcoming than Rutherford. At the outset of his speech, Lord Bingham sets forth six criteria necessary for the application of the risk rule. If (1) C was employed at different times and for differing periods by both A and B, and (2) A and B were both subject to a duty to take reasonable care or to take all practicable measures to prevent C inhaling asbestos dust because of the known risk that asbestos dust (if inhaled) might cause a mesothelioma, and (3) both A and B were in breach of that duty in relation to C during the periods of C’s employment by each of them with the result that during both periods C inhaled excessive quantities of asbestos dust, and (4) C is found to be suffering from a mesothelioma, and (5) any cause of C’s mesothelioma other than the inhalation of asbestos dust at work can be effectively discounted, but (6) C cannot (because of the current limits of human science) prove, on the balance of probabilities, that his mesothelioma was the result of his inhaling asbestos dust during his employment by A or during his employment by B or during his employment by A and B taken together, is C entitled to recover damages against either A or B or against both A and B?87

Together, these six criteria produce the Rutherford answer. The risk rule is a rule for asbestos cases only. And what of Wilsher? Lord Bingham notes that on its facts Wilsher fails to meet his fifth criterion and that this failure is adequate to distinguish the case.88 As to whether the rule may apply to other, unspecified cases in the future, Lord Bingham had this to say:   Fairchild v Glenhaven Funeral Services [2003] 1 AC 32 HL.   Fairchild involved appeals from three cases. For simplicity, I will simply refer to the Fairchild plaintiff.   Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 38 HL. See Holtby v Brigham & Cowan (Hull) Ltd [2000] ICR 1086 CA (Civ Div) for a case apportioning responsibility in an asbestosis case. 86   Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 68 HL. 87   ibid 40. 88   ‘It is plain, in my respectful opinion, that the House was right to allow the defendants’ appeal in Wilsher, for the reasons which the Vice-Chancellor had given and which the House approved. It is one thing to treat an increase of risk as equivalent to the making of a material contribution where a single noxious agent is involved, but quite another where any one of a number of noxious agents may equally probably have caused the damage’. Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 57 HL [22]. 83 84 85

Causation in Asbestos Cancer Cases (and Beyond?)  27 I would in conclusion emphasise that my opinion is directed to cases in which each of the conditions specified in (1)–(6) of paragraph 2 above is satisfied and to no other case. It would be unrealistic to suppose that the principle here affirmed will not over time be the subject of incremental and analogical development. Cases seeking to develop the principle must be decided when and as they arise. For the present, I think it unwise to decide more than is necessary to resolve these three appeals which, for all the foregoing reasons, I concluded should be allowed.89

Before I conclude the discussion of Fairchild I wish to comment on the speeches by Lord Hoffmann and Lord Rodger. Lord Hoffmann agrees that Wilsher was correctly decided but he does so on much more pragmatic grounds.90 Alone among the Law Lords who wrote opinions in Fairchild, he is unsympathetic to the idea of distinguishing McGhee and Wilsher based on the multiple-causes single-cause distinction. That distinction would leave the present case on the right side of the line because the agent of injury was the same – asbestos dust. But I do not think it is a principled distinction. What if Mr. Matthews had been exposed to two different agents – asbestos dust and some other dust – both of which created a material risk of the same cancer and it was equally impossible to say which had caused the fatal cell mutation? I cannot see why this should make a difference.91

He concludes with a reference to Rutherford. The Californian Supreme Court stated the principle specifically in relation to asbestos-related cancer cases. No doubt it could also apply in other cases which were thought to have sufficient common features, but that was left for decision on a case-by-case basis. Likewise I would suggest that the rule now laid down by the House should be limited to cases which have the five features I have described.92

  Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 68 HL.   ‘It is true that actions for clinical negligence notoriously give rise to difficult questions of causation. But it cannot possibly be said that the duty to take reasonable care in treating patients would be virtually drained of content unless the creation of a material risk of injury were accepted as sufficient to satisfy the causal requirements for liability. And the political and economic arguments involved in the massive increase in the liability of the National Health Service which would have been a consequence of the broad rule favoured by the Court of Appeal in Wilsher’s case are far more complicated than the reasons given by Lord Wilberforce for imposing liability upon an employer who has failed to take simple precautions’. Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 76–77 HL [69]. 91   Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 77 HL. 92   These five features are, ‘First, we are dealing with a duty specifically intended to protect employees against being unnecessarily exposed to the risk of (among other things) a particular disease. Secondly, the duty is one intended to create a civil right to compensation for injury relevantly connected with its breach. Thirdly, it is established that the greater the exposure to asbestos, the greater the risk of contracting that disease. Fourthly, except in the case in which there has been only one significant exposure to asbestos, medical science cannot prove whose asbestos is more likely than not to have produced the cell mutation which caused the disease. Fifthly, the employee has contracted the disease against which he should have been protected’. Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 74 HL. Jane Stapleton lists six common factors that distinguish McGhee and Fairchild. 89 90

(1) the risk of a person contracting the claimant’s condition was ‘dose-related’ in the sense that the greater the exposure to the dangerous agent (in McGhee this was brick dust) the greater the risk that the disease will be triggered; (2) but, unlike a disease with a cumulative mechanism such as the pneumoconiosis in Wardlaw’s case where the severity of the disease is dose-related, the diseases in McGhee and Fairchild were not ‘dose-related’ but were ‘indivisible’ in the sense that once triggered the disease would not vary in severity according to the victim’s exposure history; (3) both cases involved a disease linked to only a single type of agent, brick dust in McGhee and asbestos in Fairchild; (4) it was accepted that all relevant exposures to risk occurred in the victims’ workplace;

28  Joseph Sanders Lord Rodger, like Lord Bingham, set forth a set of criteria for the application of the Fairchild rule. Importantly, his criteria do not in any way suggest that the risk rule is restricted to asbestos cases and draws on the facts in McGhee and Wilsher to demonstrate the application of his criteria. Here is the relevant part of his speech: First, the principle is designed to resolve the difficulty that arises where it is inherently impossible for the claimant to prove exactly how his injury was caused. It applies, therefore, where the claimant has proved all that he possibly can, but the causal link could only ever be established by scientific investigation and the current state of the relevant science leaves it uncertain exactly how the injury was caused and, so, who caused it. McGhee and the present cases are examples. Secondly, part of the underlying rationale of the principle is that the defendant’s wrongdoing has materially increased the risk that the claimant will suffer injury. It is therefore essential not just that the defendant’s conduct created a material risk of injury to a class of persons but that it actually created a material risk of injury to the claimant himself. Thirdly, it follows that the defendant’s conduct must have been capable of causing the claimant’s injury. Fourthly, the claimant must prove that his injury was caused by the eventuation of the kind of risk created by the defendant’s wrongdoing. In McGhee, for instance, the risk created by the defenders’ failure was that the pursuer would develop dermatitis due to brick dust on his skin and he proved that he had developed dermatitis due to brick dust on his skin. By contrast, the principle does not apply where the claimant has merely proved that his injury could have been caused by a number of different events, only one of which is the eventuation of the risk created by the defendant’s wrongful act or omission. Wilsher is an example. Fifthly, this will usually mean that the claimant must prove that his injury was caused, if not by exactly the same agency as was involved in the defendant’s wrongdoing, at least by an agency that operated in substantially the same way. A possible example would be where a workman suffered injury from exposure to dusts coming from two sources, the dusts being particles of different substances each of which, however, could have caused his injury in the same way. Without having heard detailed argument on the point, I incline to the view that the principle was properly applied by the Court of Appeal in Fitzgerald v Lane [1987] QB 781. Sixthly, the principle applies where the other possible source of the claimant’s injury is a similar wrongful act or omission of another person, but it can also apply where, as in McGhee, the other possible source of the injury is a similar, but lawful, act or omission of the same defendant. I reserve my opinion as to whether the principle applies where the other possible source of injury is a similar but lawful act or omission of someone else or a natural occurrence.93

D.  Barker A review of the British cases would be incomplete without a discussion of Barker v Corus.94 Barker is also a mesothelioma case but with the added fact that Mr Barker was found to be (5) in both cases the defendant’s tort was responsible for only part of the total exposure of the victim, although in McGhee the other sources of exposure were created by the sole defendant-employer’s business, while in Fairchild sources of exposure in addition to that of a defendant were created by other parties; and (6) the contribution of the defendant’s tortious conduct to the risk of the disease could not be medically assessed, not even crudely on the basis, for instance, of length of that tortious exposure relative to other exposures to the risk. J Stapleton, ‘Lords a’leaping Evidentiary Gaps’ (2002) 10 Torts Law Journal 276, 285–86. 93   Fairchild v Glenhaven Funeral Services, [2003] 1 AC 32, 118–19. In Barker v Corus UK Ltd [2006] UKHL 20, [2006] 2 AC 572, [97], Lord Rodger agreed that the Fairchild principle applies when the other possible source of the injury is a similar but lawful act. 94   Barker v Corus (UK) plc [2006] 2 AC 572. As in Fairchild, Barker involves the appeal of three plaintiffs. For a detailed discussion of the case and its aftermath, see C Miller, ‘Liability for Negligently Increased Risk: The

Causation in Asbestos Cancer Cases (and Beyond?)  29 contributorily negligent when he failed to take adequate precautions to avoid asbestos dust while working as an independent plasterer. In Barker, Lord Hoffmann noted that in Fairchild, (1) the causal agent (asbestos dust) was the same in every case, (2) the claimants had all been exposed in the course of employment and all the exposures which might have caused the disease involved breaches of duty by employers or occupiers and (3) although it was likely that only one breach of duty had been causative, science could not establish which one it was. He then turned to two issues left undecided in Fairchild. The first question is whether all three factors must be present for the plaintiff to invoke the risk rule? Second, are all defendants jointly and severally liable for all of the plaintiff ’s damages or is each defendant ‘liable only for his aliquot contribution to the total risk of the claimant contracting the disease-a risk which is known to have materialised?’95 As to the first question, Lord Hoffmann concluded that at least factor (2) is not necessary. All the sources need not be breaches of duty by the employer or occupier. Therefore, Barker’s comparative fault would not, by itself, bar recovery under the risk rule.96 On the question of distinguishing Wilsher, however, Lord Hoffmann expresses a change of heart from his position in Fairchild. In a delightful piece of self-reinterpretation, he declares that the question he raised about different kinds of dust was not about the principle that the causal agent should be the same but merely a quibble about what counts as the same agent.97 Fairchild applies as long as the agents involved ‘operate in substantially the same way’. Lord Hoffmann concludes this part of his opinion with the following passage: If the distinction between Fairchild and Wilsher does not lie in the fact that in the latter case a number of very different causative agents were in play, I think it would be hard to tell from my Fairchild opinion what I thought the distinction was. In my opinion it is an essential condition for the operation of the exception that the impossibility of proving that the defendant caused the damage arises out of the existence of another potential causative agent which operated in the same way. It may have been different in some causally irrelevant respect . . . but the mechanism by which it caused the damage, whatever it was, must have been the same. So, for example, I do not think that the exception applies when the claimant suffers lung cancer which may have been caused by exposure to asbestos or some other carcinogenic matter but may also have been caused by smoking and it cannot be proved which is more likely to have been the causative agent.98

Finally, Lord Hoffmann addressed the question of joint and several liability, or, as the UK courts would say, liability in solidum: In my opinion, the attribution of liability according to the relative degree of contribution to the chance of the disease being contracted would smooth the roughness of the justice which a rule of joint and several liability creates. The defendant was a wrongdoer, it is true, and should not be allowed to escape liability altogether, but he should not be liable for more than the damage which he caused and, since this is a case in which science can deal only in probabilities, the law should accept that position and attribute liability according to probabilities. The justification for the joint Repercussions of Barker v Corus UK (PLC)’ (2009) 8 Law, Probability & Risk 39 and R Goldberg, ‘Causation and Defences’ in A Grubb, J McHale and J Lang (eds), Principles of Medical Law, 3rd edn (Oxford, Oxford University Press, 2010), paras 6.25–6.28. I have purposefully skipped over Gregg v Scott [2005] 2 AC 176, in part owing to space limitations. 95   Barker v Corus (UK) plc [2006] 2 AC 572, 580 HL. 96   ibid 584. 97   ibid 587. 98   ibid 587.

30  Joseph Sanders and several liability rule is that if you caused harm, there is no reason why your liability should be reduced because someone else also caused the same harm. But when liability is exceptionally imposed because you may have caused harm, the same considerations do not apply and fairness suggests that if more than one person may have been responsible, liability should be divided according to the probability that one or other caused the harm.99

In support of this position, Lord Hoffmann cites two California DES cases, Sindell and Brown as examples where the courts in the United States have taken this path.100 He does not refer to the fact that in Rutherford the defendant was jointly and severally liable for the plaintiffs economic (but not his non-economic) damages. In the event, the majority decision was reversed by Parliament. In the Compensation Act 2006, they reimposed joint and several liability in mesothelioma cases.101

VI. Comparing Rutherford and Fairchild–Barker

This review of American and British cases should make it clear that comparing Rutherford and Fairchild is a complex matter, for they vary on several dimensions. First, there is the question of joint and several liability. In my opinion, Lord Hoffmann makes a persuasive case for imposing only several liability under the risk rule.102 The Rutherford opinion did not even attempt to wrestle with this problem. Rather, it simply followed California’s general apportionment statute, which imposes joint and several liability for economic harms but only several liability for non-economic harms, ie, pain and suffering.103 It chose not to follow its own DES precedents which impose only several liability on defendants held responsible based on their share of the DES market.104 Of course, after parliamentary intervention, UK mesothelioma defendants face full joint and several liability. In the United States, the outcome is contingent on the law in each state. If other US states were to follow California’s lead and simply apply their general apportionment statute, in many if not most US jurisdictions Owens-Illinois would be only severally liable and would owe relatively little compensation to the plaintiff.105   ibid 592.   I discuss the DES cases at greater length below. 101   Compensation Act 2006 s 3: Mesothelioma: damages. This leaves undecided the question of whether the Barker rule on joint and several liability would apply to a case involving a disease other than mesothelioma. For a discussion of the legislation see Jonathan Morgan, ch 4, 000. 102   On this point at least one commentator also believes the House of Lords had it right in Barker when it called for apportionment based on the relative risk contribution of each defendant. See Green (n 52). In this article, Professor Green adopts the position that the risk rule is the correct rule in asbestos-cancer cases. This is contrary to the position he took earlier when he was the co-reporter for the Restatement (Third) of Torts: Apportionment of Liability. At that time, he was in favour of apportionment in the indivisible injury cases based on comparative responsibility. Professor Green adopts his new position in large part on legal efficiency grounds. 103   See CAL CIV CODE § 1431.2 (West 2007). Because the jury apportioned 1.2 per cent of the fault to OwensIllinois, it was required to pay $177,047 in economic damages but only $2160 in non-economic damages. Rutherford v Owens-Illinois Inc 941 P2d 1203, 1209–10 (Cal 1997). 104   Brown v Superior Court 751 P2d 470 (Cal 1988). 105   eg, the most widespread statutory apportionment scheme in the United States imposes only several liability if the defendant’s percentage of liability is below some threshold, often 50 per cent. See Restatement (Third) of Torts: Apportionment of Liability § 17 tbls to rptrs note (2000). Under such a statute, Owens-Illinois would only be responsible 1.2 per cent of both economic and non-economic damages, which would come to a total of approximately $4300. 99

100

Causation in Asbestos Cancer Cases (and Beyond?)  31 The more intriguing criterion upon which we may compare Rutherford and Fairchild is the ultimate reach of the risk rule. Here, I believe we need to distinguish two questions. Does the risk rule extend to more complex fact patterns and second does it extend to nonasbestos cases. On the first question – whether the rule may be extended to more complex fact patterns – Rutherford is clearly more liberal. With Lord Hoffmann’s reversal of direction in Barker the House of Lords has drawn a line in the sand. Fairchild does not apply to multi-agent fact situations.106 Rutherford, however, involved this very type of situation. Recall that Rutherford involved lung cancer, not mesothelioma. Lung cancer is a disease with many known causes, asbestos being only one. The substance most strongly associated with lung cancer is tobacco smoke. Mr Rutherford was a heavy smoker and among smokers asbestos is not even the most likely cause of the disease and the impact of smoking and asbestos exposure on the probability of developing lung cancer is complex. Thus Rutherford presents the very kind of complex fact pattern the House of Lords excludes from the risk rule. On the second question – whether the risk rule may apply to non-asbestos setting – the current state of affairs is reversed. Rutherford limits its holding to the ‘context of a cause of action for asbestos-related latent injuries’. Thus far, the Rutherford holding has not been applied to non-asbestos cases in California or elsewhere. In the UK, this limitation is not likely present. It is the case that Lord Bingham’s speech lists the inhaling asbestos dust as one of the criteria for the application of the rule. However, Lord Bingham recognised that McGhee was an application of the same exception to the proof of ‘but for’ causation, and clearly this is not an asbestos case. Lord Hoffmann also acknowledged that McGhee was an avant la lettre application of Fairchild. Finally, as I noted above, Lord Rodger’s criteria for the application of the risk rule has no reference to an asbestos requirement.107 106   Even here there is some vagueness. Lord Hoffmann agreed that the Fairchild rule might apply where the agent ‘may have been relevant in some causally irrelevant respect’. and he cited Lord Rodger’s example in Fairchild of a workman suffering injury from exposure to particles of dust coming from two different substances, each of which could have cause the injury in the same way. Fairchild v Glenhaven Funeral Services [2003] 1 AC 32, 119. See also Novartis Grimsby Ltd v Cookson [2007]EWCA Civ 1261 [72] where the court adopts the position that if an illness is caused by the same substance, in this case amines, deriving from different sources, in this case cigarette smoke and occupational exposure, then the Fairchild rule is applicable. 107   This is the reading of Fairchild adopted by the Court of Appeal in Sanderson v Hull [2008] EWCA Civ 1211, [2009] PIQR P7. The Court focused on Lord Rodger’s Fairchild criteria and determined that Fairchild did not apply because Lord Rodger’s first criterion – the inherent impossibility of proving exactly how an injury was caused – was not met. Sanderson v Hull [2009] PIQR P7 [63]. Sanderson v Hull adopts a particularly conservative definition of ‘impossibility’. The plaintiff worked as a turkey plucker and as a result contracted campylobacter enteritis. Ultimately, the Recorder held that she could not prove by a preponderance of the evidence that she contracted the illness due to a breach of duty on the part of the defendant or because she became infected in a way not associated with any breach of duty. He initially found in favour of the defendant but, after further submissions by the parties, he concluded that this was a case that fell within the Fairchild principle. Smith LJ concluded the facts met all of Lord Rodger’s criteria save the first. Why? Because choosing between a responsible and an innocent cause was not ‘inherently impossible’; it was merely difficult. It seems to me that the recorder’s difficulties in reaching a conclusion on causation were created not by any impossibility of proof but by his failure to make crucial findings of fact and to exercise his judgment as to the probabilities as to what happened. I accept that it can be difficult for a judge to decide whether on the balance of probabilities a claimant would have heeded a warning. Often there is no direct evidence on which the judge can rely and he has to base his decision on his impression of the claimant as a personality. I accept also that it can be difficult for a judge to make a value judgment about the causative potency of various factors in play. But if the findings of fact that I have outlined above had been made, the recorder would have been able to make a decision on the usual ‘but for’ basis. Sanderson v Hull [2008] EWCA Civ 1211, [2009] PIQR P7, [61]. The distinction is not altogether clear to me. Both the Recorder and the Court of Appeals seem to accept the idea that this illness is akin to mesothelioma in the sense that most likely a single exposure ultimately lead to the

32  Joseph Sanders

VII. Limitations on the Risk Rule

But what of the future? One may address this question from two perspectives, the real­ politik of these cases and the theoretical justification for limiting the reach of the risk rule. First, what is the likelihood that the risk rule will be applied to other situations? If one looks only at the Rutherford opinion one must conclude that chances the rule will be applied to many other situations in the United States are small. More than a decade after the decision in Rutherford, it has not been adopted by other US jurisdictions and it has not been applied to other situations in California. Borg-Warner, the one non-California court that has referred favourably to the Rutherford risk rule, did so in the context of finding for the defendant using the substantial factor test of causation. If we step beyond the Rutherford opinion itself and look at the situation from a somewhat broader perspective, the likelihood the rule may be applied to other situations is somewhat greater. As noted by Lord Hoffmann, the rule in Fairchild and in Rutherford bears a close resemblance to the rule in Sindell v Abbott Laboratories.108 The plaintiff in Sindell was one of many daughters of mothers who took diethylstilbestrol (DES), a synthetic oestrogen to reduce the likelihood of a miscarriage. The drug was prescribed between 1938 and 1971. DES was not patented and thus anyone could manufacture it. Approximately 300 companies sold the drug during some or all of this period. The drug was usually prescribed by its generic name. Many years after the mothers took the drug their daughters developed various ailments, the most serious being vaginal cancers. In many cases there was no way the plaintiff (or anyone else) could discover which company or companies manufactured the DES taken by her mother.109 In Sindell, as in Rutherford, the California Supreme Court refused to apply the Summers alternative liability theory. However, it did hold each defendant liable based on its proportionate share of the DES market.110 Market share liability is not identical to the Rutherford rule, but this should not hide their similarity. In Sindell, as in Rutherford, defendant’s liability is based on a probability that the defendant caused the plaintiff ’s harm and the probability is calculated by compardisease. Apparently, Smith LJ believed that given what we know about the different possible ways the plaintiff might have been exposed, a judge could decide whether the plaintiff had met her burden. But the Recorder’s point in applying Fairchild was not that he could not make up his mind. Indeed, he already did so when he initially found that the plaintiff had not met her burden. His point was that one could not know whether the infection was the result of the defendant’s negligence. And it seems to me his point is that resolving this issue with any direct scientific evidence is inherently impossible. How is this any different from a mesothelioma case where we can guess as to the source of the plaintiff ’s disease based on the quantum of exposure to asbestos from each defendant but it is scientifically impossible to know which fiber cause the plaintiff ’s illness? I confess that I am not well enough read in English cases to fully assess the merits of Smith LJ’s distinction. Had the opinion been written by a court in the United States, I would say that the judge is seizing any available rationale to avoid expanding the reach of the Fairchild rule. 108   Sindell v Abbott Labs 607 P2d 924 (Cal 1980). See Miller (n 74) 134, CJ Miller and RS Goldberg, Product Liability, 2nd edn (Oxford, Oxford University Press, 2004) paras 17.48–17.56; and R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing, 1999) 55–101 (including discussion of DES and causation in UK context). 109   Sindell v Abbott Labs 607 P2d 924 (Cal 1980). 110   The court sidestepped the difficult question of which market. The local city market? California? The United States? Apparently, the parties in the case stipulated to using a national market. RW Wright, ‘Liability for Possible Wrongs: Causation, Statistical Probability, and the Burden of Proof ’ 41 Loyola of Los Angeles Law Review 1295, 1328 (2008); CJ McGuire, ‘Market-Share Liability After Hymowitz and Conley: Exploring the Limits of Judicial Power’ (1991) 24 University of Michigan Journal of Law Reform 759. Hymowitz v Eli Lilly & Co 539 NE2d 1069 (NY 1989) dictated a similar solution.

Causation in Asbestos Cancer Cases (and Beyond?)  33 ing any given defendant to other defendants. In Sindell, as in Rutherford, liability is based on the creating of a risk, not on causing a harm. In Sindell, as in Rutherford, apparently the only way a defendant can avoid liability is to prove the plaintiff could not have been exposed to its product or perhaps that the exposure was so small that it was not a substantial factor. If the market share cases and Rutherford are peas from the same pod, this suggests that the risk rule might have better prospects in the United States than one would anticipate by looking only at Rutherford. However, if Sindell suggests a route by which the Rutherford rule could be expanded, it also teaches a cautionary tale about how difficult it is to persuade courts to broaden the rule to encompass more complicated fact patterns. The Sindell market share theory has been well cabined by most American courts. Even within the DES context, as many courts have rejected the concept as have adopted it.111 Outside the DES context, market share liability has been adopted very sparingly.112 Courts often justify their refusal to extend the DES rule to other situations by noting that unlike DES most other products are not fungible. Every DES pill is like every other pill, but this is not true of most substances, including asbestos. Courts that reject the expansion of market share liability to other contexts also cite the practical problems of actually estimating the level of exposure to each defendant’s product.113 Note that these are the very same types of concerns that have caused the House of Lords to restrict Fairchild to single-agent fact patterns. So, even with the DES alternative in mind, substantial expansion of the scope of Rutherford’s risk rule seems unlikely in the United States. In California, it is likely to be limited to a small set of cases, perhaps only asbestos cases. In the rest of the United States, Rutherford appears more and more to be an outlier. And what are the prospects that the reach of the risk rule will be expanded in the UK? I am wise enough to know that I am not the best person of whom to ask that question. In Fairchild and Barker the court has adopted the position that this is a single-causal-agent rule but the rule might well be expanded to other single-cause settings. Unlike the situation in the United States, however, these two cases are too recent to base any prediction on a review of what the courts have actually done. The best answer to the question of whether the reach of the risk rule will be expanded in the UK is that we will have to wait and see. Looking on from afar I suspect that the Lords will be very reluctant to permit their risk rule to travel very far from its asbestos home. At least part of the original motivation for adopting such a rule may well have been to sidestep the normal liability in solidum rules that

111   MA Geistfeld, ‘The Doctrinal Unity of Alternative Liability and Market-Share Liability’ (2006) 155 University of Pennsylvania Law Review 447, 451. 112   In re Methyl Tertiary Butyl Ether (MTBE) Prods Liab Litig 175 F Supp 2d 593, 621 (SDNY 2001). Donald G Gifford, ‘The Challenge to the Individual Causation Requirement in Mass Products Torts’ 62 Washington & Lee Law Review 873, 903 (2005). Thomas ex rel Gramling v Mallett, 285 Wis2d 236, 701 NW 523 (2005) is a rare case that expanded market share to a non-DES case. In Thomas, the court concluded that the substance in question, white lead carbonate pigment, is a fungible product. Interestingly, however, in Gibson v American Cyanamid Co 719 FSupp2d 1031 (ED Wis 2010) and Gibson v American Cyanamid Co, 750 FSupp2d 998 (ED Wis 2010) a federal judge ruled that applying Wisconsin’s risk contribution rule to a group of defendants whose predecessors in interest manufactured white lead carbonate pigment was unconstitutional because it constituted retroactive liability. This issue may be moot because in January 2011 Wisconsin passed a tort reform statute that eliminates the cause of action established in the Gramling case. It does so by restricting market share liability to cases where the claimant proves that he or she has suffered an injury ‘that can be caused only by a manufactured product chemically and physically identical to the specific product that allegedly caused the claimant’s injury’ and the product was ‘distributed or sold without labeling or any distinctive characteristic that identified the manufacturer, distributor, seller, or promoter’. See 2011 Wis Act 2, 2011 SB1. 113   Green (n 52) 546.

34  Joseph Sanders apply with respect to indivisible torts in the United Kingdom.114 The intervention of Parliament, reimposing joint and several liability rules in mesothelioma cases, may cause the Lords to ask whether the game is worth the candle. If this interpretation is correct, we would anticipate that the risk rule will remain limited in the UK as well, perhaps through the device of a narrow construction of the ‘impossibility criterion’ as occurred in Sanderson. This leads me to the normative question. Is this a principled result? If a claimant asks these courts to expand the risk rule to other situation could they provide good practical or theoretical reasons to decline the opportunity? Again, it is important to separate the question into two parts: (1) are there good reasons for refusing to apply the risk rule to multiple-causal-agent cases, and (2) are there good reasons for restricting the rule to asbestos cases? The two-part nature of the question is particularly important here, because if we were to decide that on principle the risk rule should be applied to all similar cases, we would like to know how many such situations exist. How big a hole in the common law causal proof dike does this represent? The answer to that question turns to a large extent on whether one agrees with the House of Lords that the rule should apply only to single-agent cases. If one agrees with this limitation, the exception will find a more limited application. If, on the other hand, one is willing to begin down the road marked out by Rutherford then the exception may cut a much wider swath. As to the first question – are there reasons to restrict the rule to single-causal-agent cases – the best place to start is Lord Hoffmann’s changing position. Recall that in Fairchild he agreed that both the Wilsher and the McGhee cases were correctly decided but he was not persuaded that excluding recovery to cases involving single-causal-agents was principled. However, in Barker he specifically rejects the application of the rule to facts such as those in Rutherford. He concludes that the existence of a single causal agent is ‘an essential condition for the operation of [the Fairchild rule]’.115 He justifies his new position by noting, If the distinction between Fairchild and Wilsher does not lie in the fact that in the latter case a number of very different causative agents were in play, I think it would be hard to tell from my Fairchild opinion what I thought the distinction was.116

But even if Lord Hoffmann is correct that one cannot deduce from his Fairchild opinion what he thought the distinction between McGhee and Wilsher was, this does not mean that he was incorrect when he concluded that the distinction offered up by the Vice-Chancellor was unprincipled. If, indeed, it is principled, Lord Hoffmann does not choose to set forth this principle in Barker. Almost certainly, however, he would agree with the other speeches in Fairchild and with the following statement by Lord Scott of Foscote in Barker: One reason why I would not [extend Fairchild to multi-agent cases] is that the identification of the proportion of risk of the eventual outcome attributable to each particular agent would, to my mind, be well nigh impossible and highly artificial.117

This, of course, is the type of practical, administration-of-justice rationale employed in the United States to oppose the expansion of Sindell to other, more complex fact patterns. Perhaps this is the best that the courts have to offer and perhaps it is good enough for it is

  See Green (n 52) 538 for this interpretation of Fairchild and Barker.   Barker v Corus (UK) plc [2006] 2 AC 572, 587 HL. 116  ibid. 117   ibid 599–600. 114 115

Causation in Asbestos Cancer Cases (and Beyond?)  35 an argument with considerable merit.118 Rutherford is a case in point. Recall that in this case the jury assigned 1.2 per cent of the liability to the defendant and 2.5 per cent to the plaintiff, presumably because of his use of tobacco. Most experts would probably agree that this is a substantial underestimation of the role tobacco use played in Mr Rutherford’s disease, but how one should apportion risk between these two causes is a complex question with multiple reasonable answers. Epidemiological studies reveal a strong relationship between lung cancer and both asbestos exposure and smoking. On average, the relative risk between significant asbestos exposure and the disease is approximately 5.0, while the relative risk between smoking and lung cancer is closer to 10.0.119 However, the relationship is dose related. For example, in one study the relative risk of lung cancer for those smoking one to 10 cigarettes daily was 5.5 compared to non-smokers, while for those smoking more than 31 cigarettes a day the relative risk was 22.0.120 As I mention above, the same dose-response relationship exists with respect to asbestos exposure. Thus, measuring relative risk contributions involves knowledge of dosage for both agents, knowledge that may well be hard to come by. Even more confounding is the fact that asbestos and tobacco act synergistically.121 Although the mean relative risk of developing lung cancer from tobacco use is approximately 10 and the mean relative risk of developing lung cancer from asbestos exposure is approximately five, the relative risk for someone who is exposed to asbestos and uses tobacco as compared to someone who has neither risk factor is on the order of 50.122 Under these circumstances how should one apportion the risk? Should we say that because the independent relative risks are ten and five that two-thirds of the risk came from tobacco use? On the other hand, is Mr Rutherford wrong in saying that he had increased his own risk by a factor of 10 because of his smoking, but owing to the defendants’ act in exposing him to asbestos his risk increased fiftyfold and, therefore, the defendants are collectively responsible for 80 per cent of the total risk? Would the defendants be wrong taking the opposite view and saying that their tortious behaviour increased Mr Rutherford’s risk fivefold but that his smoking raised the risk to 50 times greater than the background risk and, therefore, they should be held responsible for only 10 per cent of the total risk? In the face of this conundrum, can we do much more than what the California courts did, which was to unquestioningly accept whatever the fact-finder concludes?123 118   See MD Green, ‘The Future of Proportional Liability: The Lessons of Toxic Substances Causation’ in MS Madden (ed), Exploring Tort Law (New York, Cambridge University Press, 2005) 353. 119   Faigman et al (n 26) § 26:12. The relative risk is a ratio that divides the prevalence of injury among people exposed to a risk factor by the prevalence of injury among those who were not exposed. For example, imagine that in a certain area there were 5,000 workers and 1000 of them were exposed to asbestos. Among the 1000 exposed workers, 60 contract lung cancer and 940 do not. Among the 4000 unexposed workers we find that 160 contract lung cancer and 3840 do not. The prevalence among the exposed workers is 60/(60+940) = .06. The prevalence rate among non-exposed workers is 160/(160+3840) = .04. The relative risk is .06/.04 = 1.5. 120   Faigman et al (n 26) § 25:29. 121   See D Hamer, ‘Mind the ‘Evidential Gap’: Causation and Proof in Amaca Pty Ltd v Ellis’ (2009) 31 Sydney Law Review 465. 122   Faigman et al, (n 26) § 26:29. 123   Professor Green attempts to sidestep this issue by abandoning the risk rule for all but the asbestos defendants. Speaking of the Rutherford situation, ie the jury finds the plaintiff was contributorily negligent for smoking, he states that his inclination is to use comparative responsibility to apportion risk between the asbestos defendants and the plaintiff. Presumably, he would be prepared to adopt a similar position if the other responsible parties were defendants who exposed the plaintiff to a different risk, eg tobacco companies. In this situation, he does not comment on whether he would be prepared to use the risk rule to allocate responsibility among the tobacco defendants before applying comparative responsibility principles to allocate responsibility between the two groups of defendants. Green (n 52) 544–45.

36  Joseph Sanders The Rutherford example suggests there is substantial merit to restricting recovery under the risk rule to problems that are somewhat tractable but we must recognise that even the asbestos-only cases apportionment may be a difficult task. Lord Hoffmann’s opinion in Barker recognises that basing allocation solely on the basis of the plaintiff ’s length of employment with each defendant may not be adequate and that ‘allowance may have to be made for the intensity of exposure and the type of asbestos’.124 As difficult as these complications may be when the employer is the defendant, they are likely to be substantially greater when, as would occur in the United States, the defendant is one of a number of suppliers of asbestos to multiple work sites. Moreover, we would have to decide what to do with cases where the substance itself is a chemical cocktail. Benzene exposures may be similar to asbestos exposures, but what about a case involving exposure to diesel exhaust?125 Inevitably, therefore, the practical justifications for the single-agent rule create an arbit­ rary line. Moreover, it is a moving line. Presumably if knowledge progresses in some area an empirical question that appeared unanswerable in the past may now be tractable. In that case, the practical arguments against applying the risk rule evaporate.126 This, I take it, is Lord Hoffmann’s point in Barker when he says that we should treat a ‘two dust’ problem the same way we treat a ‘one dust’ problem. In sum, the practical arguments for limiting the reach of the risk rule have merit but inevitably they create a rather arbitrary line. Now let me turn to the second question, are there good reasons for restricting the risk rule to asbestos cases? Neither case offers a principled reason why we should do so. As recent United Kingdom cases indicate, it is hard to justify limiting the rule to asbestos exposures and not other similar exposures; that is all cases in which the exact causal mechanism is poorly understood and where, once triggered, the severity of the disease does not vary according to the victim’s exposure history.127 Nevertheless, in my opinion, we would be better served by restricting the rule to its asbestos home. In the final section I explain why I think this is the better course of action.

VIII. Conclusion: What Should the Courts Do?

If the presence of persuasive substantive arguments were the only barrier to the expansion of the risk rule, Fairchild and Rutherford might well presage a shift away from proof of causation in some non-trivial set of tort cases. It is more likely, however, that the risk rule will have a small influence. In my view, this is not an unhappy outcome. Allow me to explain my position. I begin near where this chapter started by reviewing questions of causation and proof. I believe it is important to understand that the single-hit (or multiple-hit) model of the Whatever the merits of this approach, it means that Professor Green agrees with Lord Hoffmann and that riskrule liability ‘should be limited to a single toxic substance whose risk profile is established (asbestos is surely that) and in which multiple defendants contributed to the risk but plaintiff is unable to prove which one(s) is the actual cause of her harm’. ibid 546. 124   Barker v Corus (UK) plc [2006] 2 AC 572, 594 HL. 125  See Summers v Missouri Pacific RR System, 897 F Supp 533 (ED Okla, 1995), aff ’d, 132 (10th Cir, 1997). 126   Indeed, if knowledge advances far enough, the justification for the application of the risk rule may itself evaporate for we will know the exact mechanism that produced the plaintiff ’s disease. See GE Marchant, ‘Genetic Data in Toxic Tort Litigation’ (2006) 14 Journal of Law and Policy 7. 127   The burden should be on the claimant to show that the disease operates in this fashion.

Causation in Asbestos Cancer Cases (and Beyond?)  37 relationship between asbestos exposure and cancer is a question of causation. It is separate from whatever rules we create concerning the proof of causation. Separate but related. Consider once again the classic example of the application of the indivisible injury rule, the three-car automobile accident resulting in multiple injuries to the plaintiff. Some injuries may be attributed to one of the other drivers, ie it is clear that the plaintiff ’s broken arm was caused by her collision with one other driver. In this divisible injury condition, there is no occasion to apply anything but the standard ‘but for’ test of causation. With respect to other injuries, this division may not be possible. The injuries are indivisible. Traditionally, courts have not felt compelled to resolve such cases in terms of risk con­ tribution. Why not? I think part of the answer is that the indivisible injury rule is applied in several different causal contexts. Imagine three additional injuries in the car wreck with two defendants. The plaintiff suffers an internal injury, a broken leg, and subsequent serious emotional depression. All are deemed to be indivisible. That is no one can prove which defendant caused which injury. The internal injury may well be like asbestosis: cumulative. Each defendant caused some of it and together they made it worse. The emotional distress claim is particularly difficult to parse. It may well be that the crash with either of the defendants would have been sufficient to produce the result, or it may be that it was the combination of the two collisions, one immediately following the other that caused the problem and either, by itself might not have been sufficient.128 Finally, we have the broken leg. We may well be reasonably certain that it was broken by one of the two collisions, but not both.129 At bottom, Rutherford and Fairchild stand for the proposition that when we encounter situations where the injury is not divisible we should take an additional step and determine as best we can whether the underlying facts reflect the emotional distress scenario. If so, we should employ the risk rule. If not, or, according to Fairchild, if the risk rule would require a complex calculation, we should follow some other course. And what course is that? Importantly, it is not necessarily to leave the plaintiff without a remedy because she cannot prove that any defendant ‘caused’ her injury. Courts most likely will continue to apply indivisible injury rules to such facts. In the United States, that means we are likely to remain with the Borel position. All the defendants, or at least all the defendants who are a ‘sub­stantial cause’ of the plaintiff ’s injury, will be treated as if each of them caused the injury. That is, most courts will continue to treat these cases and other similar cancer cases as occasions of an indivisible injury and not enquire further into the specific causal mechan­ism involved.130   If this is the case, it is an example of Professor Stapleton’s ‘threshold’ example. See text accompanying n 29.   When I say we are reasonably certain, I mean to imply that we are as certain about this as we are that the plaintiff ’s mesothelioma was caused by the asbestos fibre of a single defendant. 130   Courts have taken different positions on how to determine whether an injury is divisible. Restatement (Third) of Torts: Apportionment of Liability ‘26, comment h reflects one widely adopted approach: Whether damages are divisible is a question of fact. A party alleging that damages are divisible has the burden to prove that they are divisible. Whether there is sufficient evidence to prove divisibility is determined by each jurisdiction’s applicable rules. The magnitude of each divisible part is also a question of fact. The burden to prove the magnitude of each part is on the party who seeks division. One may ask if this rule is the same as the rule in Rutherford and Fairchild? If we view those cases as merely permitting plaintiffs to substitute proof of risk for proof of causation, perhaps not. However, if we focus on what the defendant would have to demonstrate to avoid liability, we observe a certain similarity. In order to avoid liability under Rutherford and Fairchild the defendant must prove either that he did not contribute in any way to the risk, ie he did not cause the plaintiff ’s injury or, in the case of Rutherford that his risk creation did not rise to the level of a substantial factor. 128 129

38  Joseph Sanders How will this affect the resolution of cases? In the United States, the answer will vary depending on individual state law. However, most states now give their cases to juries with an instruction that requires the jury to assign a percentage of liability to each defendant and to the plaintiff if she bears some of the legal responsibility for her injury. They are told to base this allocation on the relative responsibility of each party. Therefore, the actual difference in case outcome between the risk rule and the traditional approach to indivisible injuries may turn on the degree to which an allocation based on ‘relative responsibility’ differs from an allocation based on contribution to the total risk. This, in turn, may depend on how ‘responsibility’ is defined. The Restatement (Third) of Torts: Apportionment of Liability suggests the fact-finder should consider the following factors when apportioning responsibility: (1) the nature of the person’s risk-creating conduct, including any awareness or indifference with respect to the risks created by the conduct and any intent with respect to the harm created by the conduct; and (2) the strength of the causal connection between the person’s risk-creating conduct and the harm.131 Some states place greater, even exclusive emphasis on the nature of the risk creating conduct. In my state, Texas, the proportionate responsibility statute calls on the fact-finder to ‘determine the percentage of responsibility, stated in whole numbers . . . with respect to each person’s causing or contributing to cause in any way the harm for which recovery of damages is sought’,132 With instructions based on provisions such as these, the actual allocation in cases may be similar whether they are tried using the risk rule or simply as an undifferentiated individual injury case. I do not mean to argue that the allocations will be identical. However, they are likely to overlap to a substantial degree. And we should keep in mind that as the precision with which we can allocate liability based on risk erodes, courts employing the risk rule will themselves make different allocations based on identical facts. As variation employing the risk rule increases, the overlap between outcomes using the risk rule and outcomes using comparative responsibility is likely to increase. If actual allocations may be similar under either a risk rule or a comparative responsibility rule, the remaining fundamental question is whether those found responsible may be subjected to joint and several liability. As I noted earlier, I believe Lord Hoffmann makes a compelling argument in Barker that when one can only be said to have contributed to a risk, liability in solidum is unfair. If whenever a court adopted a risk rule it simultaneously abolished any joint and several liability rule, the choice of approaches would matter. But this is not what has occurred. We have seen what happened to such an effort in Great Britain. In the United States, we should expect that few courts would even attempt to alter the joint and several liability rules for a special group of risk rule cases. The rules concerning joint and several liability have been codified in most states. As the Rutherford case

131  Restatement (Third) of Torts: Apportionment of Liability § 8. Likewise, The Uniform Model Product Liability Act Section 111(B)(3) provides: ‘In determining the percentages of responsibility, the trier of fact shall consider, on a comparative basis, both the nature of the conduct of each person or entity responsible and the extent of the proximate causal relation between the conduct and the damages claimed’. The Uniform Comparative Fault Act s 2(b) states: ‘In determining the percentages of fault, the trier of fact shall consider both the nature of the conduct of each party at fault and the extent of the causal relation between the conduct and the damages claimed’. 132   Vernon’s Texas Civil Practice and Remedy Code § 33.003. Determination of Percentage of Responsibility § 33.003.

Causation in Asbestos Cancer Cases (and Beyond?)  39 indicates, even if a court adopts the risk rule it is likely to conclude that it cannot alter the legislatively mandated apportionment rule.133 All of this causes me to arrive at a conclusion different from where I have been in the past and where I thought I would end up when I began this essay. In earlier articles and my classes I have been a casual supporter of Rutherford.134 No longer. I do not pretend to speak for what should occur in the United Kingdom. In the United States, however, I find myself marginally in favour of the existing regime and opposed to the risk rule, or at least opposed to any expansion of the risk rule beyond asbestos cases. In terms of the actual allocation of liability in single-hit or quasi single-hit cases it buys us very little at the cost of considerable uncertainty concerning the nature of the causal question in an array of toxic tort cases and perhaps beyond.135 This, it now seems to me, is risky business. As the British cases amply demonstrate, the reach of the risk rule is quite uncertain. Finding principled reasons for limiting its scope is an unenviable task. And to what end? If the risk rule is only to be applied to indivisible injury cases, why create this subspecies of indivisibility? To my mind, the best reason for doing so is to make clear that the liability among defendants is to be determined by their relative contribution to risk. However, in the United States it seems very unlikely that courts will alter their apportionment and contribution rules for risk rule cases. Finally, there is the small, but not zero risk that the risk rule may escape its indivisible injury confines. Undoubtedly, both Rutherford and Fairchild confine the rule to these situations.136 However, as Lord Hoffmann notes in Barker: Consistency of approach would suggest that if the basis of liability is the wrongful creation of a risk or chance of causing the disease, the damage which the defendant should be regarded as having caused is the creation of such a risk or chance. If that is the right way to characterize the damage, then it does not matter that the disease as such would be indivisible damage. Chances are infinitely divisible and different people can be separately responsible to a greater or lesser degree for the chances of an event happening, in the way that a person who buys a whole book of tickets in a raffle has a separate and larger chance of winning the prize than a person who has bought a single ticket.137

Calls for altering the traditional ‘but for’ test of causation in the toxic tort era wisely have been resisted by the courts.138 Insofar as the risk rule presents a new rubric for easing the ‘but for’ test in all toxic tort cases, it is a risky concept. 133   Indeed, the Rutherford case highlights the potential for inconsistency in the application of joint and several liability rules. In the DES cases such as Sindell, the California Supreme Court mandated a several liability rule. Arguably, the asbestos lung cancer cases present the same causal conundrum. In both situations, it may well be that one and only one defendant caused the plaintiff ’s injury and our problem is that we do not know which defendant that is. Nevertheless, in Rutherford the court did not consider altering California’s general rule concerning joint and several liability. 134   Sanders et al (n 45); DA Fischer, MD Green, WC Powers and J Sanders, Cases and Materials on Products Liability, 4th edn (St Paul, Thompson West, 2004). 135   I realise that my position may call into question the use a risk-contribution approach even in cases where the single-hit model is absolutely certain, ie Summers v Tice. That is an issue that goes well beyond the scope of this chapter. 136   Rutherford explicitly limits itself to asbestos cases. In Fairchild, criteria 6 in Lord Bingham’s speech and feature 5 in Lord Hoffmann’s speech clearly indicate that this is a rule for one sub-species of indivisible injury cases. See also factors 1 and 2 in Professor Stapleton’s statement of Fairchild’s reach. 137   Barker v Corus (UK) plc [2006] 2 AC 572, 589-90 HL. 138   See e.g., D Rosenberg, ‘Individual Justice and Collectivizing Risk-based Claims in Mass Exposure Cases’ (1996) 71 New York University Law Review 211; MA Berger, ‘Eliminating General Causation: Notes Towards a New Theory of Justice and Toxic Torts’ (1997) 97 Columbia Law Review 2117.

40  Joseph Sanders Why then was I originally attracted to Rutherford and Fairchild? I believe it was because of the intellectual honesty expressed in these opinions. They attempt to come to grips with the difficult causal questions presented by situations where single-hit or quasi-single-hit explanation for a toxic disease seems likely but where the mechanism of injury is unknown. They highlight the fact that the approach the courts took in Borel and its progeny is a fiction. Ceteris paribus, the truth is better than fictions. But legal fictions generally arise for a reason. One reason is that they allow us to intentionally ignore complications that will lead us into making arbitrary distinctions. To me, all of the distinctions designed to rein in the reach of the risk rule feel arbitrary. Another reason for fictions is that they permit us to maintain a level of conceptual simplicity. The risk rule introduces a substantial degree of conceptual complexity and uncertainty into how courts should deal with the causal question in mass torts. Both of these reasons seem to justify an approach such as that in Borel where are indivisible injury cases are handled in the same way. When applied to meso­ thelioma, Borel’s causal approach may be based on a fiction. On balance, on this occasion I prefer fiction to reality.

3 Developing Causal Doctrine MARTIN HOGG

I.  Introduction

Though the focus of this chapter will be on causal doctrine in Scots law, it is important to begin by stating emphatically that there is nothing peculiar about the Scottish legal system which requires it to use a special test of causation different from that employed in other legal systems, whether common law, civilian or mixed. Causation is one of those aspects of private law where comparative analysis is readily applicable, and where ‘legal transplants’ are to be welcomed, especially given that the small size of the Scottish jurisdiction means that litigation suitable for developing causal doctrine occurs relatively infrequently. That does not mean, however, that there is nothing interesting or distinctive to say about the Scottish doctrinal position on causation: on the contrary, as the following discussion attempts to show, there is. It is also important to add that the discussion which follows is confined to the law of delict, and largely to what may be called causation in fact (or causation simpliciter, if one prefers, though questions of terminology are given a fuller discussion below).

II. The Current Scottish Doctrinal Position on Causation

A general assessment of causation in delict cases indicates that Scottish judicial thinking has been largely unaffected by recent academic writing in the field. Scottish courts still conceive of causation in the terms expounded by Hart and Honoré,1 that is as a unified doctrine comprising the two component elements of factual and legal causation. Factual causation, it is said, tests whether something caused or contributed to an injury, while legal causation tests whether something which qualifies as a factual cause is deemed significant enough to attract liability at law. In terms of factual causation, courts recognise the basic test as that of sine qua non (or ‘but for’), to be applied by asking whether, in the absence of the defender’s behaviour, the damage complained of would still have occurred. If it would not, then the defender’s behaviour is considered a factual cause of the damage. Such a test identifies causes which were   HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Oxford University Press, 1985).

1

42  Martin Hogg necessary for an injury’s occurrence. In addition to this basic test of necessity, there is said to be a second, alternative test, the so-called material contribution test, which asks whether, without being a necessary cause of an injury, a factor nonetheless contributed to the injury. In fact, this alleged second test is really just as a gloss on the sine qua non test, because it simply clarifies that, in order for sine qua non causation to be demonstrated, it is not necessary that the pursuer, P, demonstrate that the defender, D, caused the totality of the harm suffered by P: rather it is sufficient for P to demonstrate that D ‘materially contributed’ to the harm by having been a ‘but for’ cause of a not insignificant portion of the totality of the harm. That courts do in practice recognise material contribution as a gloss on the sine qua non test may be seen from any one of a number of uncomplicated reported cases of divisible injury (those where different elements of the overall harm sustained, or different degrees of its severity, may be separated out) – cases which therefore do not raise difficult questions of indeterminacy or duplicate causation (as to which, see further below) or indivisible injuries. Courts often express the causal enquiry to be undertaken in such simple cases as an investigation of whether the defender’s behaviour ‘caused or materially contributed’ to the outcome, indicating that the identification of either a sole cause or a contributing cause of an injury will be sufficient to establish causation in fact. Such, for instance, was the stipulated enquiry in the recent case of Kerr v Stiell Facilities Ltd,2 an industrial injury claim in which the pursuer suffered a minor injury at work while moving heavy machinery with some colleagues. The judge in the case, Lord Hodge, expressed the causal enquiry to be ‘whether it has been established that the accident caused or materially contributed to the pursuer’s long term pain and debility’,3 holding on the facts that the defender’s negligence had ‘triggered’ the chronic pain syndrome from which the pursuer suffered and was thus properly considered as a factual cause of that injury. It used also to be the case, until Barker v Corus (UK) plc,4 that Scots law employed the idea of ‘material contribution’ in a further way, such usage being that established in the Scots appeal McGhee v National Coal Board,5 a usage which held that to have materially increased the risk of an injury occurring was equivalent to having materially contributed to the injury. This conflation was a wholly fictional one, given that a consideration of any case where a risk is created but no injury occurs is sufficient to demonstrate that mere risk creation is not the same thing as causing an injury. The illogicality entrenched by the McGhee decision bedevilled the law for some considerable time, the resulting mess not being resolved until the Barker decision redefined materially increasing the risk of an injury as meaning causing the loss of a chance of avoiding the injury.6 This redefinition restored a degree of orthodoxy to the law, though it has meant that Scottish courts have now reverted to a position where they essentially have only the sine qua non test, fleshed out with the material contribution gloss, to rely upon when testing causation in fact. The problems for a legal system’s relying solely on such a test are notorious, sine qua non being unable adequately to deal with cases of overdetermined outcomes (eg two simultaneously operating   [2009] CSOH 67.   Para [82]. 4   [2006] UKHL 20, [2006] 2 AC 572. 5   1973 SC (HL) 37, 1973 SLT 14. 6   Subsequently, as a result of the enactment of s 3 of the Compensation Act 2006, ‘material increase in risk’ was restored as a test of causation for mesothelioma (not merely the loss of a chance of avoiding mesothelioma), and thus continues to operate as a test of causation in relation to that restricted class of injury. See, for recent application of the test in a case of mesothelioma, the decision of the Supreme Court in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 11, [2011] 2 WLR 523. 2 3

Developing Causal Doctrine  43 causes either of which alone would have caused the injury) or indeterminate outcomes (those where it cannot be shown which of a number of possible causes was in fact the operative one). The frequently published criticisms of academics about the lack of a test sufficient to deal with such cases have as yet failed to spur the judiciary to action. It is a fruitless endeavour seeking to identify a single reported Scots judgment of recent years where the court has referred to academic writing on the subject of causation. If, for instance, one types in ‘Stapleton’ to the Scottish Courts database of reported cases,7 three references come up to Professor Stapleton’s writing, but none of them are to her published writing on causation; if one types in ‘NESS’, one will find many references to Loch Ness, but none to the NESS test of causation (as to which, see further below). In terms of judicial appreciation of academic writing on so-called ‘legal causation’, Scotland is even further behind. To recap, traditionally courts have talked of something being a legal cause of an injury if, in addition to its having been shown to be a ‘but for’ cause of that injury (whether by virtue of being a necessary or contributing cause), it is considered to have been a significant or important cause of the injury. Such significance or importance has been tested by asking whether it was a proximate cause or a causa causans (a ‘causing cause’), an enquiry sometimes answered in the negative by a finding that its ‘causal potency’ was overwhelmed by a later cause considered to be more significant in the chain of events (a so-called novus actus interveniens). Such language suggests that the issue being tested is properly part of the causal enquiry, an approach favoured in the writings of Hart and Honoré. However, recent academic writing has tended to argue that adopting such causal language when attempting to assess the significance of causes masks the fact that what is going on is a normative exercise, one in which courts are applying often unspoken policies in determining whether or not they believe that causes ought to attract liability. It has been argued by a number of the leading commentators in the field that it would be more honest to use explicitly normative language for this exercise, recognising that it has nothing to do with the type of exercise being undertaken when assessing so-called factual causation, and to replace the language of ‘legal causation’ with that of the ‘scope of liability for consequences’, dropping in the process the old language of causa causans and novus actus interveniens.8 Such academic attempts to abandon the obscurantist language of legal causation in favour of more honest and accurate language have, however, thus far failed to impact upon Scottish judicial thinking. Courts routinely continue to employ the traditional terms in their decisions. This unwillingness to employ the new academic thinking seems somewhat puzzling given recent excellent and wide-ranging engagement by courts with academic writing in other fields. What is different with causation? The impression is given that Scottish courts are perhaps a little intimidated by causation, perceiving it as a complex issue to be glossed over whenever possible. Perhaps more problematic for the academic com­ mun­ity, however, is the sense that part of the judicial reticence to refer to and employ new academic doctrinal thinking stems from the fact that, as with the subject of error in contract law, there are still too many competing theories and classifications of causation, often expressed in highly complicated ways, to give the judiciary sufficient confidence to  www.scotcourts.gov.uk/opinionsApp   See eg, J Stapleton, ‘Cause-in-fact and the scope of liability for consequences’ (2003) 119 LQR 388; J Stapleton, ‘Legal cause: Cause-in-Fact and the Scope of Liability for Consequences’ (2001) 54 Vanderbilt Law Review 941; J Stapleton, ‘Unpacking Causation’ in P Cane and J Gardner (eds), Relating to Responsibility (Oxford, Hart Publishing, 2001); R Wright, ‘Once More into the Bramble Bush: Duty, Causal Contribution, and the Extent of Legal Responsibility’ (2001) 54 Vanderbilt Law Review 1071; M Hogg, ‘The Role of Causation in Delict’ 2005 Juridical Review 89. 7 8

44  Martin Hogg abandon established modes of thinking in favour of something new. If that is part of the problem, then it behoves the academic community both to agree new terminology and to explain that terminology in a way which shows it to be practicable and uncomplicated, a matter which is discussed further below.

III. The Future Doctrinal Development of Causation in Scots Law

The above comments indicate that development of Scottish causal doctrine is desirable. What form should such development take? The following suggestions are offered as part only of such desirable development; not every possible development is touched upon. The suggestions are largely linked by three causal issues requiring attention if doctrine is to develop in a helpful way: (1) the appropriate test for causation; (2) the appropriate termin­ ology to be employed; and (3) the understanding of what properly falls to be considered part of causation and what ought to be put in to a different legal box (a matter which also clearly affects issue (2)).

A.  Accepting that Causation has a Legally Specific Content Preceding the legal development of causal doctrine should be an acceptance of the idea that, for the purposes of legal enquiry, causation can and ought to mean precisely what lawyers decide it should mean. This point – arguing for a context-specific understanding of causation – has been developed in much greater detail by others, especially Professor Stapleton,9 so it is unnecessary to repeat it in detail here. It is worth making here again even briefly however, because causal theorists can too often agonise as to whether their attempts at exposition of doctrine measure up to some supernatural concept of causation. A good start to sensible doctrinal development would be an acceptance that we should not be hampered by a belief that there is some holy grail of causation, a platonic and pan-disciplinary understanding of that concept which we must struggle to discover and distil. For lawyers, just as ‘duty’ and ‘damage’ mean what we decide they should mean in a legal context, so it should be with the concept of causation. What counts as a sufficient causal connection for the purposes of the law is quite properly a decision to be taken with legal considerations in mind. The fact that particle physicists or surgeons may have a different understanding of causation, employed quite properly for their own purposes, should not worry lawyers, even though they will wish to be aware of such different understandings, not least when expert witnesses from other disciplines are giving evidence before a court.

B.  Reforming Causal Terminology and Agreeing the Content of the Causal Enquiry Any sensible development of doctrine must include reform of the causal terminology employed by courts and the legal profession. Reference has already been made to the grow9   See, most recently, J Stapleton, ‘Choosing what we mean by “Causation” in the Law’ (2008) 73 Missouri Law Review 434.

Developing Causal Doctrine  45 ing academic consensus that the traditional language of factual and legal causation should be abandoned, with those matters traditionally lumped under the latter heading being recognised as normative issues having no relation to causation properly understood (but rather referred to as ‘matters affecting the scope of liability for consequences’).10 Additionally, consideration should be given to whether it remains desirable to utilise the term ‘causation in fact’ (which in academic circles has come to replace the previously preferred ‘factual causation’) to refer to causation as it operates in the real world. While some jurisdictions have already adopted the language of ‘causation in fact’ (the mixed legal jurisdiction of Louisiana being one example)11 an alternative approach would be to talk of ‘causation’ simpliciter. Such an approach has recently found favour with, amongst others, Professor Stapleton and Lord Hoffmann, though, in the case of the latter, rather than using ‘causation’ to refer merely to the factual exercise embodied in the sine qua non or NESS tests,12 his Lordship would bring within ‘causation’ simpliciter many of the normative and policy issues that, it has been argued here, ought not to be considered as touching on causation properly understood. Settling on an agreed causal terminology is crucial to developing doctrine: the words we use give form to the underlying theory being conveyed. While there is probably no fundamental objection to dropping ‘causation in fact’ in favour simply of ‘causation’, such a development would be a sensible one only, it is suggested, if it were to be clearly understood that the term ‘causation’ simpliciter did not include those matters traditionally dealt with under so-called ‘legal causation’, ie the normative issues relevant to deciding which causes in the real world ought to attract liability. The counterargument, employed by Lord Hoffmann, is that courts are often compelled to consider everything relevant to the determination of a case under a single, composite head of causation, and that one cannot therefore exclude normative issues from the causation box. An example where it is argued that this is so would be the case where a statute imposed liability on persons causing dangerous drugs to be administered to another. In considering whether A, who passed a syringe containing heroin, to another, B, who injected it and then died of a drug overdose was liable under the relevant statutory provision, how else (it is said) is a court to introduce consideration of the normative issue that it was the free decision of B which appears to have been determinative of B’s death, other than giving an appropriate (and restricted) meaning to the word ‘cause’ in the statute? Just such an issue arose for determination by the House of Lords in R v Kennedy,13 where the statute in question made it an offence to ‘administer to or cause to be administered to or taken by any other person any poison or other destructive or noxious thing, so as thereby to endanger the life of such person’.14 The conundrum this creates for the content of the concept of causation stems from the drafting of the provision in question. Parliament has chosen a wording which forces both issues of causation in fact and matters affecting the scope of liability for consequences into the interpretation of the single statutory term ‘cause’. Such drafting makes the issue of causation seem as if it is an interpretative exercise into which one might properly bring consideration not just of real world causal questions   See the literature referred to at n 8 above.   Scots and Louisiana causal thinking have recently been compared in M Hogg, ‘Causation as an element of delict/tort in Scots and Louisiana law’ in V Palmer and E Reid (eds), Mixed Jurisdictions Compared: Private Law in Louisiana and Scotland (Edinburgh, Edinburgh University Press, 2009). 12   As to the NESS test, see further below at II.C. 13   [2007] UKHL 38, [2008] 1 AC 269. The case is discussed by Lord Hoffmann in ch 1of this volume. 14   Offences against the Person Act 1861, s 23. 10 11

46  Martin Hogg but also matters which, in a common law case, might happily be treated separately as normative ones affecting liability for consequences (in this case, the consideration that the defendant’s behaviour had been superseded by the voluntary and freely undertaken conduct of the victim). This impetus to consider these two separate enquiries together is, however, a mere seeming, created by the specific statutory word in question here. The statutory word at issue in another case might just as easily be the word ‘kill’, the word’ boil’, or the word ‘release’. In none of those cases would the word at issue also denote a stage in the delictual equation (duty + breach + causation + harm – defence = liability), so in none of those cases would we think we had hit upon an argument against separating normative matters out of causation. What this suggests is that the fact that the term requiring interpretation in Kennedy was ‘cause’ should not lead us to believe that causation as a delictual concept (rather than as, in this case, a statutory term requiring to be interpreted) should incorporate both the counterfactual exercise inherent in tests of causation as well as normative questions. The answer to the ‘Kennedy objection’ to separating out causation proper from factors affecting the scope of liability for consequences is thus to consider what we would think if the relevant statutory wording in the case had not been ‘administer to or cause to be administered to’ but rather, say, ‘administer to or facilitate the administration to’: replacing the word ‘cause’ with another term, as the parliamentary draftsman might have done to achieve the same end, would rob the ‘Kennedy objection’ to keeping causation free from normative considerations of its apparent force. Unfortunate statutory wording is not a valid argument against restricting the concept of causation to real world, factual questions. One further point about causal terminology is worth making. In Professor Stapleton’s most recent writing on causation, she has argued that the term ‘involvement’ merits use as an umbrella term for explaining the idea of causation as it operates in a legal context. Involvement, as she conceives it, encompasses those factors which are either (1) necessary for; (2) ‘duplicately necessary’ for (meaning operating as one of a number of causes of an overdetermined outcome); or (3) ‘contribute’ to an outcome (contribution essentially covering any NESS condition other than one which is a simple sine qua non cause or a duplicately necessary cause). ‘Involvement’ so defined has the merit of being an accessible term, which might be less productive of causal anxiety in the minds of judges. On the other hand, its use may not make the causal debate that much simpler. Why not? Well, ‘involvement’ is arguably a somewhat imprecise term. People talk of being ‘involved’ in things, in a legal context, in quite a wide variety of different ways. For instance, someone might describe himself as being ‘involved’ in the process of government by virtue of his exercise of his legal right to vote in an election, but we would not therefore think that that meant he had caused any of the individual policies and actions implemented by the government he helped to elect. So, there may be a concern that using the word ‘involvement’ could give rise to too wide a notion of causal connection, and that it might conceivably start to lead us back to relying on ‘common sense’ in deciding if a factor is ‘involved’ in an outcome. Moreover, ‘involvement’ is also arguably no more than a synonym for causation, in that to ask ‘Was A involved in outcome X?’ appears to be simply a restatement in different terms of the question ‘Did A cause X?’. The former interrogative merely employs a less causal sounding term. So it might be questioned whether we really need to employ this additional idea of involvement to explain causation. These minor criticisms of Stapleton’s latest view are however essentially semantic: in substance we both agree, as we do with the leading American causal theorist, Richard Wright, that the NESS test provides a good basis for developing the law.

Developing Causal Doctrine  47 If there is largely an academic consensus about what direction, in substance, legal development should take, the terms employed in the causal debate by different academics do still, however, require to be reconciled. We have gone from talking of factual causation, to causation in fact, and now perhaps to involvement, in the space of only a few years. It is crucial for the academic community to try to agree terms, or else it is going to be difficult to persuade the courts to abandon their traditional, flawed classificatory language, given that judges will be unsure as to which proposed new terminology to adopt. The settlement of this terminological issue is, moreover, as the foregoing comments on the Hoffmann view show, not simply a question of linguistic preference, but also a matter of convincing judges about what should go in to the causal box and what should go in to other boxes. On that matter, we are still some way from a resolution.

C.  A New Test of Causation Further development of causal doctrine should also be based upon an acceptance that the present reliance by the courts on the ‘but for’ test, with its material contribution gloss, cannot continue. Given the inability of sine qua non to deal with common causal difficulties, such as duplicate causation, a new and more comprehensive test needs to be adopted by the courts. Such a more comprehensive test has been developed within the academic community, winning wide acceptance as the best alternative to the sine qua non test. The test in question will be familiar to academics (but few practising lawyers) as the NESS test, known in such an acronymic form because of its identification of all factors preceding an outcome which were necessary elements of a set of conditions sufficient for the outcome to occur (the italicised letters combining to form the acronym). This test, while it contains an element of necessity within it, locates that necessity element within a larger consideration of a set of conditions sufficient for an outcome (thus rightly recognising that there can be more than one set of conditions capable of producing, and thus being sufficient for, a specific outcome). Despite courts and practising lawyers tending to view NESS as very complicated to apply, it is not. Its application can be explained relatively simply as follows: Assemble a set of all the conditions which preceded an outcome (eg conditions A, B, C, D, E). If you want to see whether one condition – let’s say it is A – was a cause of the outcome, remove any conditions from the set which were not required for the outcome to occur, eg if the outcome would still have occurred without D and E, they are removed. At this point a set of conditions (A, B, C) will be left which was minimally sufficient for the outcome to occur, ie all the conditions in the set were required on this occasion to produce the outcome. Removing the condition in question, A, will therefore produce a set of conditions which is no longer sufficient for the outcome to occur. A is thus held to have been a cause of the outcome.

This improves on the sine qua non test because it allows us to identify duplicate causes as each having been a cause of an outcome. So, if two vehicles V1 and V2 each struck a pedestrian crossing the road, killing him, and the medical evidence is that each blow on its own would have been fatal, while application of the sine qua non test results in the bizarre conclusion that neither vehicle caused the accident (because the impact of the other vehicle would have caused the death in any event), the NESS test correctly identifies each vehicle as a cause. The three conditions present in the set which preceded the death of the pedestrian are: (1) the impact of V1, (2) the impact of V2, and (3) the presence of the pedestrian, P, on

48  Martin Hogg the road. If one wants to see if V1 is a cause of P’s death, one removes V2 from the list of conditions and one still has a set minimally sufficient for the outcome; however removing V1 means one no longer has such a set, so V1 was a cause of the death. The same result is reached if one separately tests V2 as a possible cause using the NESS test: V2 is also shown to be a cause of the death. NESS also works to allow proper identification of what can be styled ‘pre-emptive’ causes, such causes being those which, through their supervening operation, prevent another condition which would have resulted in an outcome from achieving that effect. Thus, negligent exposure of an employee to a fatal dose of radiation, which exposure would have resulted in the employee’s death several months later, is not identified as a cause of death if the employee is killed by a lightning strike one week later: the lightning strike is identified using NESS as the (pre-emptive) cause of the death. This is so because the set of conditions minimally sufficient for the death of the employee from the radiation exposure are: (1) the radiation exposure, and (2) the employee affected by it. If the employee is no longer in existence (having been killed by lightning), then this set of minimally sufficient conditions cannot exist. This of course is not to say that the employer’s conduct cannot be a cause of the pain and suffering caused between the point of exposure to the radiation and death (indeed such conduct can be shown to be such a cause), but so putting matters identifies a different harmful outcome in relation to which causation is being tested: rather than the death of the employee, we are testing for causes of his pain and suffering up to the point of death. This is not the place to explain further the application of the NESS test, as that has been done elsewhere in much greater detail.15 However, even this brief explanation indicates, it is hoped, that the NESS test provides a largely comprehensive algorithm for testing causation. It adequately deals, as briefly shown above, with both duplicate and pre-emptive causation, and it can also correctly identify omissions to act as causes in appropriate cases (ie in those cases where a failure to act is properly identified as a possible cause because there was a duty to act in the circumstances). It cannot, however, deal with indeterminate causation (about which, see further below), but then no test of causation is capable of doing that. Despite its improvement upon the simple and inadequate sine qua non test, the NESS test has yet to be taken up by the Scottish or English courts. This does appear largely to be so because courts perceive it to be complex and impractical (possibly because academics are apt to choose fanciful examples when demonstrating its application, like that of the infamous poisoned desert traveller). Yet the NESS test can be presented in a concise and not overly complex way, as it is hoped it has been here, thereby demonstrating its practicability and utility to the courts.

D.  Mapping Cases of Causal Uncertainty No test for causation developed to date has been able to provide a satisfactory answer in cases of indeterminate causation, that is in those cases where removing a condition from the set of conditions preceding an outcome fails to produce a clear answer to the question whether one is left with a minimally sufficient set for the outcome still to occur. Such cases of indeterminate causation are not infrequent, common examples being personal injury 15  See ch 14 of this volume. See also, for instance, R Wright, ‘Once More into the Bramble Bush’ (n 8); J Stapleton, ‘Choosing what we mean by “Causation” in the Law’, (n 9); M Hogg, ‘The Role of Causation in Delict’ (n 8).

Developing Causal Doctrine  49 cases where the cumulative effect of exposure to chemical or other agents cannot be broken down into individual causal effects (often because medical science has not yet determined the aetiology of the medical condition in question) or where too much time has passed since the injury to determine the causal impact of the various possible conditions on the patient’s present state. Indeterminacy also has the potential for arising in cases involving the uncertainty of human decision-making, such as those where what is at issue is what a pursuer would have chosen to do in the absence of the defender’s negligence. In cases such as these, courts may be faced with whether to, and if so how to, allow a pursuer to overcome an evidentiary gap created by such inherent causal uncertainty in order to allow his claim to succeed, a course of action which requires the deployment of non-causal solutions to achieve the desired result. At the present time, this area is rapidly developing, and the law requires to develop a clearer understanding of which circumstances will justify the overcoming of an evidentiary gap by employment of a non-causal solution, and what the specific solution ought to be in the type of case in question. An interesting question is whether one might comprehensively map all of the different types of indeterminate case and provide a coherent taxonomy explaining why certain solutions are thought appropriate in certain types of case, but others require different solutions. No such grand enterprise can be attempted here, but it is intended directly below to select one particular type of inherent causal uncertainty – that of the inherent unpredictability of human behaviour – to see whether a consistent and coherent approach can at least be developed in all cases of that type. The prospect that a rational and comprehensive map of the whole area of causally uncertain cases might be produced is, for those working within a jurisdiction like Scotland – which conceives of private law as capable of being comprehensively mapped – a desirable, and one would hope achievable, one, though one which will have to await further research and exposition.

E.  A Case Study of Causal Uncertainty: the Assessment of Counterfactual Human Decision-making As a final suggested development of causal doctrine, consideration is given below to one specific area of causal uncertainty, that being cases where the causal uncertainty relates to the inherent unpredictability of human behaviour. Discussion is restricted to cases where what is at issue is how a person would have behaved in the past in the absence of the defender’s negligence. Judicial development in this area is ongoing, though the various solutions adopted by courts in different types of case can give the appearance of a somewhat piecemeal approach. i.  The Nature of the Indeterminacy Problem as it Relates to Human Behaviour How does the unpredictability of past human behaviour create causal problems? The difficulty is that, when one has to consider how a pursuer would have behaved in the past in the absence of the defender’s negligence, there may be uncertainty as to which imaginary, alternative set of facts to postulate, or, to put it in causal terms, as to which counterfactual world to postulate when testing causation using the NESS algorithm. This of course assumes that one does use counterfactual analysis with NESS, which is taken here as a given.

50  Martin Hogg However, at least one leading NESS scholar, Professor Wright, disputes this.16 He argues that, in applying NESS, one merely removes the condition being tested, while leaving all the others unchanged.17 That is a workable approach with simple causal sets where, having removed one condition, it does not seem fanciful to leave the others as they are. A good example of such a simple causal set, proposed by Chris Miller,18 is the electrical circuit, where, if one has a set minimally sufficient to produce light comprising (1) a switch, (2) a battery, and (3) a bulb, then removing, say, the battery, from the set, one can simply leave the other members of the set (the switch and the bulb) as they are in order to test whether light is still produced (which, we know, it is not). One can just about see how counterfactualism need not be a necessary part of such an exercise, though it is surely the case that most of us, when considering this example, are still conjuring up an image of a world in which the battery is absent in order to imagine what the outcome would be. Indeed, conjuring up such counterfactual worlds is surely the way that most judges go about thinking what would have happened in the absence of a defender’s wrongdoing. However, in any event, if one goes beyond simple examples like the electrical circuit, and moves on to cases involving human decision-making, it seems impossible not to posit counterfactual scenarios: in the absence, for instance, of D’s negligent misrepresentation, it does not make sense to leave unchanged the remaining condition (P’s detrimental actings in reliance on the misrepresentation), as that condition only exists in that form as a result of the now absent misrepresentation. Without the misrepresentation, the other condition would not have existed in the form in which it did, so that simply eliminating the misrepresentation but leaving the other condition unchanged makes no sense. These examples bring home the point that, in anything other than the simplest types of case (and even perhaps for those), one has to use counterfactual analysis in order to test causation using NESS. But that brings us back to the difficulty that there will be cases where there is more than one such counterfactual world which might reasonably be posited, thus posing the dilemma of which such world to posit. Thus there arises the problem of multiple conceivable parallel universes beloved of science fiction writers. The root of the problem is that all counterfactual behaviour of human beings is indeterminate – we cannot know how someone would have behaved if we change the other surrounding factors – or at least what may be called ‘quasi-indeterminate’19, that is to say it is not able to be determined given the current state of human knowledge. As we are not omniscient beings (and not likely to become so in the immediate future), we do not possess that intimate knowledge of the complexity of individual human decision-making which would allow us to determine how a particular victim would have behaved in the absence of a particular wrongdoer’s negligent conduct. Does this mean therefore that all cases involving questions about how people would have behaved are treated by courts as indeterminate and thus unable to be solved by applying the NESS algorithm, given that the NESS test presupposes that we are able to input a given counterfactual world against which to apply 16   See ch 14 of this volume. See also R Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (1988) 73 Iowa Law Review 1001. 17   So Wright, explaining his position, states: ‘We hypothetically eliminate only the condition being tested . . . from the sufficient set of actual antecedent conditions. Then, without adding or subtracting any other conditions, we determine – by matching the remaining conditions in the set against the applicable causal generalization – whether the set still would be sufficient for the occurrence of the result’. (‘Causation, Responsibility, Risk, Probability, Naked Statistics and Proof ’ (n 16) 1041). 18   See C Miller, ‘NESS for beginners’ (ch 15 of this volume) p 324. 19   On quasi-indeterminacy, see H Reece, ‘Losses of Chance in the Law’ (1996) MLR 188.

Developing Causal Doctrine  51 the test? Certainly not. In practice, courts regularly treat many cases as if it can be determined how someone would have acted in the past, albeit that this deemed determinability is in strict terms a fiction. Courts in fact often take it for granted that we can prove, on the balance of probabilities, how someone would have behaved. Take, as an example, the Scottish case of Keith v Davidson Chalmers.20 The pursuer claimed that the defenders, a firm of solicitors, had been negligent in failing to advise him that, were he to undertake a transaction he was contemplating, this would result in his being in breach of a fiduciary duty owed to a company of which he was a director and employee, and his thereby incurring liability towards that company. He sued the defenders for the amount for which he had had to settle such a claim by the company against him. On appeal, the Inner House of the Court of Session upheld the decision at first instance that the pursuer had failed to show that he had suffered any loss as a result of the defender’s breach of duty, stating that in a case such as this the pursuer had to prove what he would have done had he been given the correct advice at the material time. In the present case he would undoubtedly have aborted the entire operation. This is a clear and unequivocal finding by the Lord Ordinary which was not satisfactorily challenged at the appeal. In these circumstances the pursuer has wholly failed to prove that he has suffered any real loss arising out of the transaction in respect of which the advice should have been tendered.21

The causal process in Keith v Davidson Chalmers was treated as if it were deterministic: the court felt able to conclude what the pursuer would have done in the absence of the negligent conduct of the defender.22 What, however, if the counterfactual behaviour of a particular party in the absence of the defender’s negligent behaviour seems not to be determinable? Orthodox principles suggest that, in such a case, an ordinary damages claim by a pursuer for real world harm should not be allowed to succeed (although the case might conceivably be reframed as one of loss of a chance, an alternative solution touched on briefly below). Although there may be a temptation to skate over apparent problems of indeterminacy when drafting pleadings, such a temptation should be avoided. A bald averment in pleadings that claimed losses flow from the defender’s negligent behaviour, without narration of how, may result in an otherwise sound case being dismissed for lack of specification of the claim. Such was the outcome in Henderson v The Royal Bank of Scotland,23 where the pursuer averred simply that substantial damages flowed from the defender’s misrepresentation, without saying how. The case was dismissed for lack of specification. However, even in cases where, at first glance, the potential problem of the indeterminacy of a party’s behaviour seems to arise, courts may be willing to employ one of a number of techniques to overcome the apparent indeterminacy. Such techniques are considered below in cases involving either the indeterminacy of the pursuer’s or defender’s behaviour.

  2004 SC 287, 2003 SCLR 941.   Opinion of the Inner House, delivered by Lord Wheatley, at [29]. The pursuer had eventually been unable to proceed with the contemplated transaction. Instead, he was able to sell his interest in the company which he had set up to undertake the transaction for a much higher sum than the settlement he had had to pay in respect of his breach of fiduciary duty, thus resulting in his making a substantial net gain. 22   A similar case is that of Leeds & Holbeck Building Society v Alex Morison & Co (No 2) 2001 SCLR 41, where the Lady Ordinary (Paton) concluded that, had the defender properly warned the pursuer of risks relating to a third party, the pursuer would not have made a loan to such third party. 23   [2008] CSOH 146, 2008 SCLR 823. 20 21

52  Martin Hogg ii.  Potentially Indeterminate Behaviour of the Pursuer So far as cases involving the potentially indeterminate behaviour of the pursuer’s conduct is concerned, the problem can be exemplified by further consideration of misrepresentation claims, of which there have been a number of recent reported Scottish examples. In such cases, a court first needs to give proper consideration to the nature of the duty which the defender ought to be held to have come under: how this duty is framed will affect the losses which may conceivably be claimed as having flowed from the breach of such duty. Was the duty to provide information only, or was it to advise on a particular course of action? That important distinction was made by Lord Hoffmann in the SAAMCO case,24 a distinction commented on further by the House of Lords in Aneco Reinsurance v Johnson & Higgins,25 and one which has been followed and applied in subsequent Scottish cases such as Hamilton v Allied Domecq plc,26 Preferred Mortgages v Shanks,27 and, most recently, in a case mentioned earlier, Henderson v Royal Bank of Scotland.28 How a court frames the nature of the duty undertaken by the defender may exclude some potential counterfactual problems, as a particular type of loss to which a causal link seems to be indeterminate may in any event be excluded from the claim by virtue of its falling outside the scope of consequences covered by the duty in question. Correct stipulation of the duty undertaken may not, however, solve all potential counterfactual problems in a misrepresentation case (or indeed in other types of case). Even given the permissible types of loss claimable by virtue of the nature of the duty in question, it may still be unclear what the pursuer would have done had the defender not made the misrepresentation on which the pursuer relied, because a number of alternative reasonable courses of action might then have been open to him and he cannot demonstrate which he would have adopted. In such cases, failure to demonstrate which course of action would have been taken by the pursuer will usually result in dismissal of the action.29 iii.  Potentially Indeterminate Behaviour of the Defender In fact, misrepresentation claims may also give rise to questions of the defender’s apparently indeterminate counterfactual behaviour: should a court counterfactually posit that the defender made no representation at all, or should a careful representation be posited? What if a number of careful representations might have been made by the defender, which one should be posited? For instance, if the careless information to invest in bad concern X were not given, thus avoiding P losing £50,000, should it be posited that D would instead   South Australia Asset Management Corporation v York Montague Ltd [1997] AC 191.   [2001] UKHL 51, [2001] 2 All ER (Comm) 929. 26  The SAAMCO decision was cited in the judgments of both Outer and Inner Houses (see 2001 SC 89 (OH) and 2006 SC 221 (IH)), the House of Lords eventually holding that the defender had neither assumed any duty towards the pursuer nor made any misrepresentation to him. 27   [2008] CSOH 23. 28   [2008] CSOH 146, 2008 SCLR 823. 29   Exceptionally a failure by P to state what he would have done will not be fatal to a claim, however. In Chester v Afshar [2004] UKHL 41, [2005] 1 AC 134, a failure by P to say what she would have done absent D’s breach of duty did not preclude a successful claim by P against a doctor who had failed to warn her of the inherent risks of surgery when these risks materialised in injury. There is an ongoing debate among academics as to why the claimant in Chester was permitted to overcome uncertainty about her counterfactual behaviour when claimants in other types of case have not been so permitted, a debate largely the result of academic disagreement as to the ratio of the Chester decision (compare, for instance, M Hogg, ‘Duties of Care, Causation and the implications of Chester v Afshar’ 2005 Edinburgh Law Review 156 with J Stapleton, ‘Occam’s Razor reveals an Orthodox Basis for Chester v Afshar’ (2006) 122 LQR 426). 24 25

Developing Causal Doctrine  53 have advised investment in gilts, or in property speculation, or in good, albeit differently performing, concerns Y or Z  ? A degree of assistance in such cases of apparent counter­ factual indeterminacy has been provided by the judicial determination that one should counterfactually posit that the defender made a careful representation, rather than that no representation was made or treating the representation which was made as if it were accurate.30 As a general approach, this commends itself logically: as the making of the representation was undertaken voluntarily, and it is the issue of whether or not the representation was given carefully, it seems correct as a general approach to posit a counterfactual alter­ native where information is given carefully, rather than some other alternative. There may be exceptions, however. In a recent case, Halifax Life Ltd v DLA Piper Ltd,31 an owner of property sued for damages in misrepresentation on the basis that the defender, a firm of solicitors, made a successful offer to purchase commercial property for £8.8 million on behalf of a syndicate of buyers which, it transpires, did not in fact exist. On discovering this, the owners resold the subjects at a substantially reduced price. In such a case, how can one sensibly posit a careful counterfactual representation by the defender? It would have to be something like: ‘We hereby offer to purchase subjects on behalf of a non-existent client’. An offer framed thus is clearly nonsensical, and no such offer would ever be issued. Surely, in such a case, where the idea of a carefully issued offer seems nonsensical, it seems correct counterfactually to posit that, in the absence of negligence on the defender’s part, no offer would have been made at all; had that been so, the pursuers would have tried to sell the property to another buyer. They may be able to indicate a second potential buyer to whom the property would likely have been sold; if not, the circumstances might nonetheless be appropriate for a claim for loss of a chance, given that courts have been sympathetic to pursuers who cannot prove what a third party might have done in the absence of the defender’s negligence.32 Indeed, somewhat similar circumstances arose in an English misrepresentation case, First Interstate Bank of California v Cohen Arnold,33 in which a lost opportunity to sell a property for a higher sum was successfully pled. iv.  The Default Rule in Cases of Indeterminacy As mentioned at the beginning of this section, the problem about the potentially indeterminate human behaviour arising in misrepresentation cases is that the NESS algorithm presupposes that we are able to input a given counterfactual world, but doesn’t tell us how to choose between competing, reasonable counterfactual worlds. Although the NESS test can’t help with this problem, the above discussion shows that, on some occasions, potential difficulties might be resolved through the nature of the duty in question or through the assumption that a careful representation was made (rather than some other assumption). However, where such solutions cannot provide an answer, then, approaching the problem as a matter of principle, the rule must be that a pursuer who claims that, absent D’s mis­ representation, outcome X would have prevailed should be required to demonstrate, on the   See Lord Hoffmann in SAAMCO, [1997] AC 191, 216D-E.   [2009] CSOH 74. Following this reported judgment of Lord Hodge, the case was put out By Order to allow discussion of further procedure, but the claim was eventually settled extra-judicially. 32   See, for instance, the English cases Allied Maples v Simmons & Simmons [1995] 1 WLR 1602 and Spring v Guardian Assurance plc [1995] 2 AC 296. 33   [1996] 5 Bank LR 150, 140 Sol Jo LB 12, [1996] PNLR 17. Another type of misrepresentation case where arguably no representation ought counterfactually to be posited would be where it could be shown that the extra time and effort required to give information carefully would have put D off giving any information at all. 30 31

54  Martin Hogg balance of probabilities, that it is indeed X which would have prevailed, rather than some other outcome.34 If P cannot do so, because it is equally or more likely that Y or Z would have prevailed, then P’s claim based upon outcome X must fail. The same will hold if it cannot be said that any particular counterfactual outcome would have prevailed. If, however, P can demonstrate, on the balance of probabilities, that X was more likely than not to have occurred, but this cannot be said of Y or Z, then P’s case alleging the losses arising from outcome X should succeed. In applying this rule, and thus in deciding which of these outcomes is proven in any given case of misrepresentation, much will depend on assessing the credibility and reliability of the evidence presented to the court as to who would have done what. As to this point, it seems that courts view the matrix of facts which they are entitled to consider (for instance, what the previous business decisions and dealings of parties suggest about their likely counterfactual behaviour in the case before the court) to be a wide one. It also seems that, while courts expect P to be able to demonstrate what course of action he would have chosen absent the misrepresentation, they are more willing to infer which course of action D would have chosen if D pleads uncertainty as to what he would have done: a pursuer is not to be prejudiced, if at all possible, by pled uncertainties about the defender’s counterfactual behaviour, given that this is a matter over which he has no control. v.  Road Traffic Accidents The problems of potentially indeterminate human behaviour are not restricted to misrepresentation cases, though such cases do focus those problems most clearly. One further common type of problematic case relating to the counterfactual behaviour of a defender, a type previously discussed by both Professor Stapleton and me, is that of the pedestrian injured in a road traffic accident. Suppose a pedestrian to have been struck by a driver in a hurry who was travelling at 50 mph in a 40 mph zone, and the pedestrian to have suffered certain injuries in consequence. In asking whether, had D not been negligent in what he did, would P still have suffered those injuries, is one counterfactually to posit that safe speed which D would most likely have driven at given his usual driving habits (perhaps he usually drives 10 mph below the limit), or should it be the speed which is the nearest careful speed to that at which D was actually driving? Or what if, if one takes D’s negligence out of the equation, D would not have been driving at all, because the evidence indicates that he always refuses to drive if he is made to stick within the speed limit – should one then posit a counterfactual scenario where D was not driving at all? Professor Stapleton has recently said that in such an example: [t]he Law determines what would have been the highest speed a reasonable person would have been going in the circumstances, say 45 mph . . . the defendant’s behaviour is altered just enough to bring it into conformity with his duty as mandated by law, namely 45 mph.35

The same approach was advocated by Becht and Miller in their 1961 work Factual Causation.36 However, another alternative would be to posit the most likely alternative behaviour of this 34   For a recent example of a case where the pursuers were unable to discharge this burden of proof see Blower & others v Edwards & Scottish Widows plc [2010] CSOH 34, where the Lord Ordinary (Hodge) held that the pursuers had failed to demonstrate that their counterfactual behaviour in the absence of the alleged negligent advice of the defenders would have been any different to their factual behaviour. 35   J Stapleton, ‘Choosing what we mean by “Causation” in the Law’ (n 9) 451. 36  A Becht and W Miller, The Test of Factual Causation in Negligence and Strict Liability Cases (St Louis, Washington University Publications Committee, 1961) 34.

Developing Causal Doctrine  55 defender had he not been speeding. As suggested above, that might be shown to be 40 mph because, other than on this aberrant occasion, he invariably drove at 10 mph below the speed limit. Such an approach might be said to be more consistent with what the courts do when dealing with misrepresentation cases, where they posit not what reasonable parties would have done, but what this defender and this pursuer would have done had the misrepresentation not been made. That is certainly what the Inner House of the Court of Session professed to be doing in the passage quoted earlier from Keith v Davidson Chalmers. vi.  Conclusion on Cases of Indeterminacy What may one conclude from the above discussion of the potential problem of the indeterminacy of counterfactual human behaviour? First, it is important to recognise that courts have shown willingness to fashion solutions to some potential problems through, for instance, careful definition of the duty undertaken on the facts of the case, through a general (though not, it has been suggested, invariable) rule that, in misrepresentation cases, a careful representation should be posited rather than some other counterfactual scenario, and through application of lost chance analysis in certain cases. More work needs to be done in analysing and explaining why certain solutions (lost chance claims, for instance) will be available in some cases but not in others. Second, however, it remains clear that those framing pleadings are sometimes not sufficiently aware of the pitfalls which may arise. This lack of awareness may simply be the result of a failure to appreciate the import­ ance of the counterfactual exercise which must be conducted when following through claimed losses to a demonstrable counterfactual outcome. If that is so, then greater willingness to attend to this exercise would benefit pleaders. Not all problems may be resoluble, but some might be with sufficient care and attention. Third, as the road traffic accident example shows, there remains an unresolved question as to whether, in considering different possible counterfactual worlds, one should be positing one in which the party or parties are supposed to have behaved as reasonable individuals – in which case, one would be, to some extent, ‘objectifying’ the issue – or whether one is trying to determine what would have been the most likely alternative behaviour of the actual party or parties in question, in which case the counterfactual exercise has a more subjective element to it. An objective approach seems more suited to the issue of whether there has been a breach of a duty of care, where the standards of the reasonable person are in issue, but less suited to the application of counterfactual analysis, where one ought to be attempting to ascertain what the actual parties would have done (the exercise, after all, purports to be a factual one). A satisfactory resolution of this issue requires the clear adoption by the judiciary of a consistently subjective approach in all cases where determining counterfactual human decision-making is in question.

IV. Conclusions

There is nothing unique about causal doctrine in Scotland. The problems Scotland faces in developing its understanding about causation in the field of delict are shared by other jurisdictions throughout the world. It has been a little disappointing then that recent international academic debate in this field appears to have penetrated so little in to current

56  Martin Hogg judicial thinking. However, ways in which greater academic influence might be brought to bear on judicial thinking have been suggested in this chapter. Crucially, it has been argued that the superior NESS test ought to be presented to the courts in a simple, practicable and more relevant manner. Academic presentation of the test has perhaps been somewhat complex and seemingly esoteric in the past. This may explain the absence of reference to the test in reported decisions of the courts. In Scots law at least, there has also been a tendency in recent times to wait for the lead in developing causal doctrine to be shown by the English courts. In a jurisdiction like Scotland, which gave birth to both the Wardlaw v Bonnington Castings37 and McGhee v National Coal Board38 decisions, it would be heartening to think that Scottish courts might once again be bold in their causal thinking and willing to blaze a trail. That of course presupposes a willingness first to revise thoroughly their causal ideas and terms. While cases where causation is the evident focus of the parties’ dispute do not arise often, they are not so scarce that Scottish courts could not, if they so choose, be at the forefront of developing causal analysis in both Scotland and England.

  1956 SC (HL) 26, [1956] AC 613.   1973 SC (HL) 37, 1973 SLT 14.

37 38

4 Causation, Politics and Law: The English – and Scottish – Asbestos Saga JONATHAN MORGAN*

I. Introduction

This is an essay about legal reasoning. How should courts, and legislators, approach controversial issues in tort law? What should be the proper respective roles of these two institutions? What part should ‘policy’ play in adjudication and legislation? No area of tort law has been more controversial in England over the past decade than causation. The seminal asbestos case of Fairchild v Glenhaven seems to have been the catalyst for many other controversial questions about causation (and actionable damage, and proof) to come before the House of Lords.1 Therefore, we will examine Fairchild, the follow-up decisions in Barker v Corus and the pleural plaques litigation (Rothwell),2 and the legislative response to these cases at Westminster and Edinburgh. The episodes have all concerned situations in which a claimant cannot prove, in the orthodox way, that he has suffered actionable damage (in the orthodox sense) from the defendant’s breach of duty. Such questions have, of course, been much debated in the literature as well as in the courts,3 but space does not permit another detailed substantive consideration of the responses that *  I am especially grateful to Mr Sandy Steel of King’s College London. It has been possible to take some account of the Supreme Court’s decision in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523, handed down just before this book went to press. The Supreme Court of the UK heard argument in the challenge to the Damages (Asbestos-related Conditions) (Scotland) Act 2009 as the chapter was in proof (June 2011): Axa General Insurance Ltd v Lord Advocate. 1   Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32 [Fairchild]. 2   Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572 [Barker]; In re Pleural Plaques Litigation: Rothwell v Chemical and Insulating Company Ltd; Johnston v NEI International Combustion Ltd; Grieves v F T Everard & Sons [2007] UKHL 39, [2008] 1 AC 281 [Rothwell]. 3   See eg, JH King Jr, ‘Causation, valuation, and chance in personal injury torts involving preexisting conditions and future consequences’ (1981) 90 Yale Law Journal 1353; D Rosenberg, ‘The causal connection in mass exposure cases: A “public law” vision of the tort system’ (1984) 97 Harvard Law Review 851; TM Dworkin, ‘Fear of disease and delayed manifestation injuries: A solution or a Pandora’s box?’ (1984) 53 Fordham Law Review 527; RW Wright, ‘Causation, responsibility, risk, probability, naked statistics, and proof: Pruning the bramble bush by clarifying the concepts’ (1988) 73 Iowa Law Review 1001; S Levmore, ‘Probabilistic recoveries, restitution and recurring wrongs’ (1990) 19 J Legal Studies 691; DA Fischer, ‘Proportional liability: Statistical evidence and the probability paradox’ (1993) 46 Vanderbilt Law Review 1201; MA Berger, ‘Eliminating general causation: Notes towards a new theory of justice and toxic torts’ (1997) 97 Columbia Law Review 2117; AR Klein, ‘Causation and Uncertainty: Making Connections in a Time of Change’ (2008) 49 Jurimetrics 5; E Voyiakis, ‘The great illusion: Tort law and exposure to danger of physical harm’ (2009) 72 MLR 909.

58  Jonathan Morgan the law might take (including relaxing burdens of proof, and designating the creation of risk – rather than the infliction of harm – as actionable damage).4 Our primary concern here will be the kinds of arguments that might properly be used, by courts and legislatures, in approaching the problem. If such a meta-analysis (an argument about arguments) brings upon us the charge of ‘postmodernism’ then so be it.5 These are important questions with resonance for the entire law of torts, and indeed legal development in common law systems generally. The argument may be sketched as follows. First, we explain why it is extremely difficult to stake out robust ‘exceptional’ categories as a matter of judicially-developed common law, and why extending the law for exceptional cases should therefore be left to legislation. This framework will be used to criticise the decisions in Fairchild and Barker. The role of ‘policy’ reasoning in Fairchild will also be considered. We then consider the radical proposals by ‘corrective justice’ theorists that judges, at least, should not consider the effects and consequences of their decisions when developing the law. Against this, we will argue that corrective justice, shorn of any engagement with such ‘policy’ concerns, fails to provide satisfactory answers to the issues raised by cases such as Fairchild. From the common law and its allied theory of corrective justice, we then turn to recent legislation. The reversal of Barker by the Compensation Act 2006, and of Rothwell by the Damages (Asbestos-related Conditions) (Scotland) Act 2009, shows that tort liability cannot be considered only at the abstract level of principle. The politics of compensation often intrude. But the strict limits of such ad hoc legislative interventions must be emphasised. Ultimately, in this saga, it seems that both judges and legislatures have generally failed to engage with the questions of principle raised by the asbestos compensation claims. Direct engagement by the courts with such questions is necessary to promote coherent legal development, alongside a more rigorous analysis of policy questions by both judges and legislators.

II. Common Law and Statute

The English lawyer has traditionally viewed common law and statute law as separate and immiscible legal categories: the ‘oil and water’ school of thought. It has been coupled with the characteristic common law approach to statutory interpretation, in which legislation is taken to enunciate precise, isolated legal rules which apply strictly to the situations intended by Parliament, but have no wider effect. In other words, statutory rules, unlike common law ‘rules’,6 may not be generalised, or form the basis for reasoning by analogy. Increasingly, these fundamental assumptions have come under attack.7 It has been argued that statutory law is no more a ‘wilderness of single instances’ than the judge-made common law; that it, too, must be woven into a ‘seamless web’ of legal principle; and so   cf Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523 [Sienkiewicz] [140] (Lord Rodger).   Postmodernism could ‘take as its motto that delightful quip: anything you can do, I can do meta’ according to Philip Pettit, The Common Mind: An Essay on Psychology, Society, and Politics (Oxford, Oxford University Press, 1996) xii. 6  Ronald Dworkin persuasively criticised HLA Hart’s treatment of the ‘rule-ness’ of the common law in R Dworkin, Taking Rights Seriously (London, Duckworth, 1978) ch 2. 7   See eg, J Beatson, ‘Has the common law a future?’ [1997] CLJ 291; R Zimmermann, ‘Statuta sunt stricte interpretanda? Statutes and the common law: A continental perspective’ [1997] CLJ 315; J Beatson, ‘The role of statutes in the development of common law doctrine’ (2001) 117 LQR 247. 4 5

The English – and Scottish – Asbestos Saga  59 statutory rules must not be taken as the singular products of the sovereign legislator’s will but seen as proceeding, like the common law, from a deeper, coherent set of system-wide principles.8 However, we firmly maintain that the traditional view correctly recognises distinctive features of the common law and of legislation, and is greatly to be preferred. As will be illustrated in this essay, the unfashionable view of statutory rules as ‘isolated irruptions’ offers great advantages when we seek to create a special category to deal with a ‘hard case’ felt to be particularly meritorious. A notorious failing of judge-made law is that ‘hard cases make bad law’. Despite the House of Lords reminding itself of this maxim in Fairchild, it will be argued below that their Lordships fell foul of it nonetheless. If the Law Lords’ determination of the merits of Fairchild was correct, ie if this truly was a case in which an exception to general tort/causation principles needed to be made in the interests of justice, such an exception ought to have been made by legislation. It is very difficult to carve out a truly exceptional category by judicial decision, since the common law method supposes that the reasons given for one decision must (for consistency’s sake) be applied with equal force in analogous cases. If ‘bright’ dividing lines are to be created and upheld, it must be by legislation, interpreted and applied in the traditional ‘strict’ fashion (rather than liberally and ‘teleologically’). As will be seen, the English and Scottish statutory interventions in the Fairchild saga have created just such exceptional and precisely-defined categories of asbestos claimants. This legislation, unlike Fairchild, has not set any far-reaching precedents. The categories should prove robust in future litigation, although there may of course be political pressure to extend the generous treatment afforded to certain categories of asbestos victims, through further legislation.

III. Causation and Judicial Reasoning: Fairchild v Glenhaven

The common problem raised by the recent asbestos cases has been the inability to prove either what has happened to the victim (or in one case, what will happen to the victim) on the balance of probabilities. The problems correspond to archetypical hard cases in causation that have been much discussed. In Fairchild: who injured me? This is like the ‘two hunters’ cases in which the claimant is shot by a stray pellet and two people have negligently discharged shotguns, but it is impossible to say which ‘hunter’ fired the relevant shot.9 In Rothwell: what will happen to me? This is like the pollution cases in which emissions increase the risk of a certain cancer developing in a given population: who (if anyone) can claim, and do they have to develop the cancer before they have a claim or is the fact of exposure enough? In Fairchild the House of Lords faced a dilemma. The judges felt that it would be a great injustice to dismiss the claims, but equally they did not wish to recognise, as a general principle, that proving that negligence has increased the claimant’s risk of sustaining a harmful outcome is the same as proving that the negligence actually caused harm. So, reversing the Court of Appeal which had (with impeccable orthodoxy) dismissed the claims for want of proof of causation,10 the House of Lords consciously took an exceptional approach. On the   cf R Dworkin, Law’s Empire (London, Fontana, 1986).   cf Cook v Lewis [1951] SCR 830 (Supreme Court of Canada).   [2001] EWCA Civ 1881, [2002] 1 WLR 1052.

8 9 10

60  Jonathan Morgan allegedly special facts of Fairchild, it was sufficient to prove that the defendants had increased the risk of the claimant contracting mesothelioma (by exposing him to asbestos) for the claimant to recover damages in respect of that mesothelioma. While relying on (and indeed, rehabilitating) another, earlier ‘exceptional’ decision,11 the court in Fairchild simultaneously emphasised the correctness of the other leading House of Lords decisions in which there was no liability for simply increasing the risk of the claimant’s injury.12 We suggest that for the court to decide Fairchild in the way that it did while purporting to uphold the general principle ultimately proves impossible. Their Lordships have not been able to provide sufficiently convincing reasons for the ‘specialness’ of Fairchild.13 Therefore, the predictable consequence is that continuing attempts have been made to extend the Fairchild ‘principle’ to other ‘exceptional’ cases. An arbitrary line is always difficult to hold in the common law. In certain pathological cases, the courts have declared that the law is so unprincipled that it cannot be extended by analogy, as will be seen below. This surely risks undermining respect for the common law process altogether: the legitimacy of legal development by unelected judges derives in part from the constraints of deciding, and explaining decisions, in accordance with the extant principles of the common law. A common problem appears in all of the causal uncertainty cases, up to and including Fairchild. The scientific evidence was not sufficient to determine whether the defendant’s admittedly negligent conduct had caused or contributed to the claimant’s injury. Uncertainty about the aetiology of the conditions meant that the injury might have been caused by non-tortious causes in McGhee and Wilsher. In Fairchild itself, however, negligent asbestos exposure was (assumed to be) the only possible cause of the claimant’s injury,14 but since science was unable to rule out the ‘single fibre theory’ (in which just one fibre might have triggered the malignant cell development), it was impossible to decide whose negligent exposure (ie which defendant) had caused the mesothelioma. All members of the House of Lords shared a strong instinct that justice required that the claimant should recover on these facts.15 The real difficulty with the case followed immediately afterwards. This is captured in the opening paragraph of Lord Nicholls of Birkenhead’s speech: My Lords, I have no hesitation in agreeing with all your Lordships that these appeals should be allowed. Any other outcome would be deeply offensive to instinctive notions of what justice requires and fairness demands. The real difficulty lies is elucidating in sufficiently specific terms the principle being applied in reaching this conclusion. To be acceptable the law must be coherent. It must be principled. The basis on which one case, or one type of case, is distinguished from another should be transparent and capable of identification. When a decision departs from principles normally applied, the basis for doing so must be rational and justifiable if the decision is to avoid the reproach that hard cases make bad law.16

It is suggested that Lord Nicholls did not live up to his own strictures. In the end, he concluded:   McGhee v National Coal Board [1973] 1 WLR 1 (HL).   See especially Hotson v East Berkshire Area Health Authority [1987] AC 750 (HL) and Wilsher v Essex Area Health Authority [1988] AC 1074 (HL). 13   cf Sienkiewicz (n 4) [186] per Lord Brown. 14   cf facts of Sienkiewicz (environmental exposure to asbestos exceeding tortious exposure). 15   cf Fairchild (n 1) [9], [32], [33] (Lord Bingham); [63] (Lord Hoffmann); [108], [114] (Lord Hutton); [155] (Lord Rodger). 16   Fairchild (n 1) [36]. 11 12

The English – and Scottish – Asbestos Saga  61 There must be good reason for departing from the normal threshold ‘but for’ test. The reason must be sufficiently weighty to justify depriving the defendant of the protection this test normally and rightly affords him, and it must be plain and obvious that this is so. Policy questions will loom large when a court has to decide whether the difficulties of proof confronting the plaintiff justify taking this exceptional course. It is impossible to be more specific.17

But of course, ‘policy’ is not a meaningful category in itself.18 The question is which policies are used, and whether they have been convincingly weighed and applied to the facts of the case. In Fairchild itself, the ‘policy’ for allowing recovery seemed to be no more and no less than the alleged ‘injustice’ that would befall the claimants if they failed to recover through their inability to say which of a group of admittedly negligent defendants had caused their mesothelioma.19 The difficulties with this reasoning are obvious enough. It is too broad because in every case where the rules on proof of causation mean that the claim must fail, there has been similar ‘injustice’ (assuming that, as in Fairchild, the defendants admit lack of care or are held to have been negligent).20 So this would equally be true of Wilsher and Hotson, cases that the House of Lords in Fairchild nevertheless approved. It is surely noteworthy that in a recent leading case on the duty of care, Lord Rodger of Earlsferry felt moved to comment that ‘I do not actually find it helpful to bear in mind – what is in any event obvious – that the public policy consideration which has first claim on the loyalty of the law is that wrongs should be remedied’.21 As his Lordship observed, ‘the world is full of harm for which the law furnishes no remedy’.22 The so-called ‘policy’ concern at the heart of Fairchild would lead to limitless liability if taken literally. If it is not meant seriously, then it is no more than a makeweight, and does little to identify or justify the making of an exception – the test that the House of Lords had correctly set itself. How then did the House of Lords limit the fertile, indeed explosive, potential for ‘justice’ to supplant the need for proof of causation altogether? As noted, the court approved its earlier decision in Wilsher as correctly stating the normal approach: showing that the defendant’s negligence increased the risk of the claimant’s injury is not sufficient to show that the defendant caused or contributed to the injury. The preferred basis for reconciliation of McGhee, Wilsher and Fairchild was the ‘single agency’ theory: there was but one causal agent at play in McGhee (brick dust) and Fairchild (asbestos fibres), but several possible causes of the damage to the claimant’s eyesight in Wilsher. While it is true that this point does reconcile the cases, its normative relevance is rightly said to be ‘obscure’.23 As Dr Beever comments, ‘the rule seems to be nothing more than an arbitrary restriction on the application of Fairchild and, moreover, appears to be intended as such’.24   Fairchild (n 1) [43].   Particularly if defined negatively, to signify all non-doctrinal considerations to which the court may properly pay regard. 19   See eg, Fairchild (n 1) [33] per Lord Bingham of Cornhill, under the subheading ‘policy’. 20   There is also more than a hint of question-begging in the repeated assertion that denial of liability would be ‘unjust’. Counsel for the defendants had conversely argued: ‘Compensation under schemes which do not depend on proof of causation is a matter for Parliament, but it is unjust to visit liability on a defendant in the absence of proof of causation’ (emphasis added) [2003] 1 AC 32, 39. 21   JD v East Berkshire Community Health NHS Trust [2005] UKHL 23, [2005] 2 AC 373 [100]. cf X (Minors) v Bedfordshire CC [1995] 2 AC 633 (CA) 663 (Bingham MR). 22   JD v East Berkshire Community Health NHS Trust [2005] UKHL 23, [2005] 2 AC 373 [100]. 23   R Stevens, Torts and Rights (Oxford, Oxford University Press, 2007) [Stevens] 51. 24   A Beever, Rediscovering the Law of Negligence (Oxford, Hart Publishing, 2007) [Beever] 475, citing, for the latter point, Barker (n 2) [24]. See also Beever 484: ‘normatively bizarre’; Chris Miller, ‘Causation in personal 17 18

62  Jonathan Morgan It is surprising that Lord Hoffmann, having deplored the single agency theory as unprincipled and irrelevant in Fairchild,25 in the later case of Barker v Corus declared that this minority view had been ‘wrong’ although without really explaining why.26 Lord Scott in Barker did, at least, identify a pragmatic reason for thus limiting Fairchild: extending the ‘principle’ to multi-agent causation would make ‘the identification of the proportion of risk of the eventual outcome attributable to each particular agent . . . well nigh impossible and highly artificial’.27 It is suggested that such an ‘ease of application’ argument does little to reassure those who view the single agent rule as ‘arbitrary’, ‘obscure’ or ‘bizarre’. In fact, Lord Hoffmann (despite his protestation in Barker that ‘it would be hard to tell from my Fairchild opinion what I thought the distinction was’28) had suggested a different distinction, squarely on policy grounds. His Lordship alluded to the ‘complicated’ ‘political and economic arguments involved in the massive increase in the liability of the National Health Service which would have been a consequence of the broad rule favoured by the Court of Appeal in Wilsher’s case’.29 Although, as we have written elsewhere, this remark is rather tantalising and raises many further questions,30 it is at least a recognisable policy argument. How cogent it is, and whether policy should even be justiciable, is an important question to which we will return. In Sanderson v Hull, the Court of Appeal has emphasised another restriction on the application of Fairchild.31 To follow the exceptional course of equating increased risk with causal contribution, it must be ‘inherently impossible for the claimant to prove exactly how his injury was caused. It must be impossible because of the current state of scientific knowledge’.32 On the facts of Sanderson, there had been difficulty in showing exactly what had caused the claimant turkey-plucker’s food poisoning, and in particular whether her illness would have been averted had the defendant employer performed his duty to provide gloves and to warn her not to eat or smoke prior to washing her hands. But this probative difficulty was neither ‘inherent’ nor ‘scientific’: simply, the trial judge had failed ‘to make crucial findings of fact and to exercise his judgment as to the probabilities as to what happened’.33 Therefore, Sanderson was not a case in which Fairchild could be used to equiparate increasing the risk of food poisoning with causing food poisoning. However, it is not clear why ‘inherent’ or ‘scientific’ obstacles to discharging the burden of proof should be treated differently.34 In the classic ‘two hunters’ case there is no ‘scientific’ uncertainty – the causal mechanism is perfectly understood (cf Fairchild, McGhee – and Wilsher). But the gaps in the evidence mean, just as ineluctably, that causation cannot

injury: legal or epidemiological common sense?’ (2006) 26 Legal Studies 544, 560–63. Lord Brown also criticises the ‘supposedly critical’ distinction in Sienkiewicz [187], although his Lordship’s dismissal of undue ‘academic focus’ upon it seems unfair when the judges themselves elevated it to its dubious prominence! 25   Fairchild (n 1) [72]. 26   Barker (n 2) [23]. 27   Barker (n 2) [64]. See now Sienkiewicz (n 4) [104] per Lord Phillips. 28   Barker (n 2) [24]. 29   Fairchild (n 1) [69]. 30   J Morgan, ‘Lost causes in the House of Lords’ (2003) 66 MLR 277. 31   Sanderson v Hull [2008] EWCA Civ 1211, [2009] PIQR P7. 32   ibid [53] per Smith LJ. cf Gregg v Scott [2005] 2 AC 176, 196, [79] per Lord Hoffmann (‘inherent uncertainty’). 33   Sanderson v Hull (n 31) [61]. 34   Sandy Steel raises the inconsistency between Sanderson v Hull and Fitzgerald v Lane [1987] QB 781 in his case-note of Sienkiewicz [2010] QB 370 (CA), which also contains valuable analysis of the crucial concept of ‘impossibility’: (2010) 73 MLR 646.

The English – and Scottish – Asbestos Saga  63 be proved in the normal way. Fairchild, it seems, would not then ‘solve’ this prototypical hard case.35 Again, Dr Beever points out that the restriction seems ‘arbitrary’.36 It is difficult to escape the conclusion that despite Lord Nicholls’s stated determination ‘to avoid the reproach that hard cases make bad law’, Fairchild was the hardest of cases and has made the worst of law. The Law Lords showed themselves to have generous instincts, for which they might well be applauded.37 But they clearly realised that strict limits must be placed on the exceptional approach, or the entire law of causation might unravel.38 The ‘single agent’ and ‘scientific uncertainty’ limitations have been stressed both in Fairchild itself and in later cases concerning its application, but ultimately these fail to convince.39 Without very good reasons for such lines to be drawn, and to be drawn where they have been drawn, the courts will find themselves under continual pressure to expand the exception. Such pressure will be hard to resist.40 The jurisprudential basis of the common law is in principles applicable to analogous cases. It is extremely difficult to limit common law rules by drawing arbitrary lines around them.41 A notorious area in which the courts have tried to do this is liability for nervous shock or psychiatric injury.42 The key case is Alcock, in which the Law Lords were candid in their project to draw arbitrary lines around the categories of ‘secondary victims’ to keep liability under control.43 As Lord Oliver of Aylmerton admitted, for example, it would be ‘logically analogical’ to extend claims from sudden shocking events to a more gradual realisation that a loved one had been injured, but: the law in this area is not wholly logical and whilst having every sympathy with the plaintiffs, whose suffering is not in doubt and is not to be underrated, I cannot for my part see any pressing reason of policy for taking this further step along a road which must ultimately lead to virtually limitless liability.44 35   cf AR Klein, ‘Causation and Uncertainty: Making Connections in a Time of Change’ (2008) 49 Jurimetrics 5, who suggests that the distinction should be the other way round. Scientific discovery means that cases which currently cannot be solved using the burden of proof should, in time, become resolvable, and therefore we should encourage science to supply the answer rather than changing the legal rules. For Klein, contra Fairchild, the exceptional situation in which compensation should be made available without proof of fault would be precisely that of ‘inherent difficulty proving specific causation regardless of recent scientific developments’ (emphasis added). 36   Beever (n 24) 484. Dr Beever further asks (ibid): if Fairchild (as (re-)interpreted in Barker) recognises an exceptional right not to be exposed to the risk of injury, why should scientific discoveries (ie those leading to better understanding of the aetiology of diseases) rob us of that right? 37   Contrast the storm of criticism which met the later decisions curtailing asbestos claimants’ (thought-to-be) vested rights to compensation in Barker and Rothwell, below. Would either of these decisions have been necessary, however, had Fairchild been decided in the orthodox way (so as to dismiss the claims)? 38   Lord Phillips MR clearly feared that this process was underway, in Gregg v Scott [2005] UKHL 2, [2005] 2 AC 176, [172]. Having cited the decision ‘in the interests of justice’ in Fairchild, and Lord Nicholls’s dissenting opinion in Gregg itself, the Master of the Rolls continued: ‘My Lords it seems to me that there is a danger, if special tests of causation are developed piecemeal to deal with perceived injustices in particular factual situations, that the coherence of our common law will be destroyed’. 39  In Barker, Lord Scott said at [57] that Fairchild had been a ‘pragmatic judicial response to what would otherwise have been an unjust and unsatisfactory denial of a remedy . . . [and] cannot, therefore, be taken to have established an overarching principle in the law of tort’. 40   Barker (n 2) [5] (Lord Hoffmann). 41   cf Lord Walker’s thwarted desire to recognise an ‘exceptional’ liability for banks who negligently breach Mareva injunctions: HM Commissioners of Customs and Excise v Barclays Bank plc [2006] UKHL 28, [2007] 1 AC 181 [75]. Lord Walker at [77] accepted that such liability would inevitably extend to ‘any other person affected by notice of a freezing order’, and therefore reluctantly concluded that the defendant bank was not liable. 42   On the latter term, were the miscarriages suffered by the plaintiffs in the leading cases Dulieu v White & Sons [1901] 2 KB 669 (KBDC) and Bourhill v Young [1943] AC 92 (HL) ‘psychiatric injuries’? 43   Alcock v Chief Constable of South Yorkshire Police [1992] 1 AC 310 (HL) [Alcock]. 44   ibid 417.

64  Jonathan Morgan His Lordship added that ‘I cannot, for my part, regard the present state of the law as either entirely satisfactory or as logically defensible’.45 But there had to be a limit somewhere, ‘based upon policy rather than upon logic’. Hence the notorious, and much criticised, ‘control devices’ for nervous shock claims.46 Crucially, Lord Oliver in Alcock also said that: ‘Policy considerations such as this could, I cannot help feeling, be much better accommodated if the rights of persons injured in this way were to be enshrined in and limited by legislation’.47 A very similar argument had previously been used by Lord Scarman to suggest that judges should stick to the ‘principled’ Donoghue v Stevenson approach in nervous shock cases, ignoring policy questions such as indeterminate liability,48 but the House of Lords in Alcock unanimously took the decision to ‘draw the line somewhere’ rather than leave the problem for the legislature to solve. This might seem brave, or foolhardy, or just realistic according to taste: the courts are well aware of Judge Friendly’s law reform dilemma between ‘judges who can’t and legislators who won’t’.49 But the result was a common law ‘patchwork quilt of distinctions’ which proved so painfully ‘difficult to justify’ that when the second round of Hillsborough litigation reached the House of Lords, their Lordships refused to heed arguments for any further principled development of the law.50 It was declared to be frozen into its unsatisfactory shape until such time as Parliament might undertake radical reform, since ‘there are no refined analytical tools which will enable the courts to draw lines by way of compromise solution in a way which is coherent and morally defensible’.51 Lord Hoffmann, of course, put it with exhilarating clarity: the search for principle was called off in Alcock v. Chief Constable of South Yorkshire Police [1992] 1 A.C. 310. No one can pretend that the existing law, which your Lordships have to accept, is founded upon principle.52

Could the same not be said of the ‘exceptional category’ created by Fairchild? If Fairchild is to be kept under control, that can only be by treating the decision as so lacking in principle that it should not be followed outside the rather arbitrary limits laid down (‘single causal agent’ and ‘scientific uncertainty’). This, however, is likely to bring the coherence of the law into disrepute, as did Alcock and Frost. The Law Lords in Fairchild should have recognised that they were setting an inevitably fertile precedent for future situations in which proving causation was difficult or impossible,53 and therefore faced up to the general question of when the burden of proof might be relaxed, or whether increasing risk might of itself form the gist of an action in negligence. As we will shortly see, the latter approach was belatedly adopted by the House of Lords in Barker. But no attempt was made to ease the strict (essentially arbitrary) limits on Fairchild itself, even under its reinterpreted guise. Fairchild, then, was and remains a disfiguring episode upon the face of the common law: an   ibid 418.   cf J Stapleton, ‘In restraint of tort’ in PBH Birks (ed), The Frontiers of Liability, Volume II, (Oxford, Oxford University Press, 1994). 47   Alcock (n 43) 419. 48   McLoughlin v O’Brian [1983] 1 AC 410, 430 (HL); Donoghue v Stevenson [1932] AC 562 (HL). 49   (1963) 63 Columbia Law Review 787 50   Frost (sub nom White) v Chief Constable of South Yorkshire Police [1999] 2 AC 455 [Frost] (per Lord Steyn at 500.) 51  ibid. 52   Frost (n 50) 511. 53   The ‘gravitational force’ of common law precedents, in the famous metaphor of R Dworkin, Taking Rights Seriously (London, Duckworth, 1978) 111. 45 46

The English – and Scottish – Asbestos Saga  65 essentially legislative attempt to carve out an exceptional category, doomed to failure by the common law’s ineluctable method of reasoning by analogy. The House of Lords should have followed the orthodox logic of the Court of Appeal and dismissed the claims,54 leaving it to Parliament to create an exceptional category. Alternatively, their Lordships could have recognised a new form of actionable damage (increase in risk), acknowledging that this would have an impact far beyond the facts of Fairchild. The attempt to find a via media (a judicial, but unprincipled and limited, ‘exceptional’ category) has failed. These arguments are supported by certain observations in the latest Fairchild case to reach the highest court, Sienkiewicz v Greif UK Ltd.55 The deceased had occasionally been exposed to asbestos by the defendant employer, but had also respired atmospheric asbestos in her home town, Ellesmere Port. The trial judge held that in order to show that her fatal mesothelioma had been caused by the defendant’s exposure rather than by the environmental exposure (for which the defendant was not responsible), the claimant administratrix had to show that the defendant’s exposure exceeded that from the environment – ie that the defendant had more than doubled the background risk of mesothelioma. (On the evidence, the claimant had failed to do this and so at trial, the claim failed.) But in the Supreme Court this decision was reversed. Fairchild had laid down a special rule to deal with the impossibility of proof in mesothelioma cases; and proof of the origin of the fatal fibre(s) was no easier in ‘single exposure’ cases – medical science not having advanced in the decade since Fairchild was decided. Accordingly, there was no room for a doubling-ofrisk approach. The ‘policy reasons’ behind the Fairchild rule were equally applicable in single exposure cases.56 All that had to be shown was material exposure to asbestos, viz a material increase in risk of mesothelioma. ‘Material’ connotes more than de minimis: it could not mean ‘doubling of the background risk’. The judgments contain interesting and important discussions of the nature and utility of epidemiological evidence, and some revealing observations about Fairchild. Baroness Hale said that ‘Fairchild kicked open the hornets’ nest’ and she found it ‘hard to believe that their Lordships there foresaw the logical consequence of abandoning the “but for” test’ – ie what Lord Phillips described as the ‘draconian consequences’ of Fairchild as applied in Sienkiewicz.57 But even if the judges wanted to go back on Fairchild it was now too late: ‘Parliament would soon reverse us’.58 Lord Brown was even more critical. Although entertaining no doubt that Fairchild must lead to the success of the claim, he said: I think it only right to indicate just how unsatisfactory I for my part regard this position to be and how quixotic the path by which it has been arrived at.59

The House of Lords in Fairchild cannot have foreseen such wide liability;60 nevertheless, it was in Fairchild that ‘the die was inexorably cast’.61 There was no logical stopping-point between Fairchild and the ‘extreme (“draconian”) position now arrived at’ in Sienkiewicz,   [2001] EWCA Civ 1881, [2002] 1 WLR 1052.   [2011] UKSC 10 [Sienkiewicz]. 56   ibid [203] (Lord Kerr), [212] (Lord Dyson). 57   ibid [167] (Baroness Hale), [58] (Lord Phillips). See also [168]: ‘the inevitable result of Barker, made even more severe through the intervention of Parliament’ (Baroness Hale). 58   ibid [167] (Baroness Hale). Lord Mance at [189] expressly agreed. See also [58] per Lord Phillips: ‘Parliament has willed it so’ – cf discussion of Compensation Act 2006 in section VI, below. 59   Sienkiewicz (n 4) [174]. 60   ibid [178]–[179]. 61   ibid [185]. 54 55

66  Jonathan Morgan ‘no rational basis for confining the special rule within narrow bounds, whatever may have been contemplated by the House in Fairchild’.62 Lord Brown was at pains to emphasise that mesothelioma claims are in a ‘special’ ‘category all their own’.63 The lesson to be drawn from the claimant’s success in Sienkiewicz was that ‘courts should in future be wary indeed before adding yet further anomalies’ to the law on causation.64 These points seem well made, but there is tension between Lord Brown’s attempt to confine ‘mesothelioma cases’ as sui generis, and his recognition that the rationale of Fairchild must be allowed to expand into its natural conceptual space. Faced in future with a case that fits the stated of rationale of Fairchild (single causal agent, inherent impossibility),65 will the courts really be able to dismiss the claim on the basis that it is ‘not about mesothelioma’? It might be the only thing that the judges can do to prevent ‘turn[ing] our law upside down and dramatically increas[ing] the scope for what hitherto have been rejected as purely speculative compensation claims’.66 ‘Thus far and no further’ as Lord Steyn put it, rather despairingly, contemplating the courts’ inability to place defensible limits on nervous shock claims.67

IV.  Fairchild Reinterpreted: Barker v Corus

A question left open in Fairchild concerned the nature of the defendant employers’ liability. It was assumed by the House of Lords that the defendants would all be (jointly and severally) liable for the whole damage suffered by the claimant – liability in solidum for the mesothelioma. However, the defendants reserved the right to argue on a future occasion that their liability should be for a proportionate share only.68 This question duly reached the House of Lords again some four years later, in Barker v Corus.69 On the main question, the House held that liability under Fairchild was proportionate to the degree of risk created by each defendant. This major practical setback for asbestos claimants was retrospectively reversed by Parliament, with remarkable celerity. The argument of the defendants’ counsel in Barker seems to have been a highly pragmatic one. Jeremy Stuart-Smith QC noted that in the United States ‘proportionate liability has been adopted in the interests of an equitable trade off against the modified rule of causation’.70 He commended this to their Lordships as ‘the most fair, just and reasonable   ibid [184].   ibid [174]. cf Lord Phillips at [1]: ‘special rule’ for ‘[u]nusual features’ of mesothelioma; Lord Mance at [188]. 64   ibid [187]. 65   Lord Brown observes at [186] that the courts are often faced with ‘comparable rocks of uncertainty in a wide variety of other situations’. (For the ‘rock of uncertainty’ cf Lord Bingham of Cornhill, Fairchild [7].) 66   ibid [186] (Lord Brown). 67   Frost (n 50) 500. 68  WE Peel, in ‘Lost Chances and Proportionate Recovery’ [2006] Lloyd’s Maritime and Commercial Law Quarterly 289, suggests that this division of the question was a deliberate tactical move by the defendants’ liability insurers: by presenting the House of Lords in Fairchild with the stark choice of holding defendants (who may not have caused any damage) liable in full for the claimants’ injuries, they hoped (unsuccessfully as it proved) to deter the court from finding any liability at all. Lord Brown makes the same point about Barker (and also Bonnington Castings Ltd v Wardlaw [1956] AC 613) in Sienkiewicz (n 4) [180]. 69   Barker (n 2). 70   See eg, Hymowitz v Eli Lilly & Co (1989) 539 NE 2d 1069, 1078 per Wachtler CJ, quoted by Lord Hoffmann, Barker (n 2) [45]: ‘We understand that, as a practical matter, [proportionate liability] will prevent some plaintiffs from recovering 100% of their damages. However, we eschewed exculpation to prevent the fortuitous avoidance of 62 63

The English – and Scottish – Asbestos Saga  67 balance’.71 But Lord Hoffmann’s reasoning (with which Lord Scott of Foscote and Lord Walker of Gestingthorpe agreed) was notably different.72 Lord Hoffmann based his decision for proportionate liability on the principle that liability in Fairchild was for increasing the risk of disease rather than actually causing disease (viz the mesothelioma).73 Lord Rodger of Earlsferry, in dissent, complained that this was an argument that counsel for the defendants ‘was at very considerable pains to say that he was not advancing’.74 Lord Scott similarly explained liability in Fairchild by ‘the fact’ that all the defendants had exposed the claimant to the risk of contracting mesothelioma rather than ‘the fiction that each Fairchild defendant had actually caused the eventual outcome’.75 Lord Hoffmann maintained that his interpretation of Fairchild was consistent with the reasoning in that case, but stated: In the end, however, the important question is whether such a characterisation would be fair. The Fairchild exception was created because the alternative of leaving the claimant with no remedy was thought to be unfair. But does fairness require that he should recover in full from any defendant liable under the exception?76

Lord Hoffmann answered the question (of whether characterisation of the gist of the damage as the increased risk of disease was fair) affirmatively. Proportionate liability would ‘smooth the roughness of the justice which a rule of joint and several liability creates’:77 The justification for the joint and several liability rule is that if you caused harm, there is no reason why your liability should be reduced because someone else also caused the same harm. But when liability is exceptionally imposed because you may have caused harm, the same considerations do not apply and fairness suggests that if more than one person may have been responsible, liability should be divided according to the probability that one or other caused the harm.78

The conclusion that liability was proportionate to each defendant’s share of the risk neatly solved the other problem raised by the facts of Barker – whether Fairchild applied in a case where part of the exposure was not tortious, because the claimant had (during a period of self-employment) exposed himself to asbestos fibres. For at least three of the Law Lords,79 the majority’s decision that liability was not joint and several permitted the application (or extension) of Fairchild to such a case.80 Lord Rodger dissented vigorously over joint and several liability. The problem with Lord Hoffmann’s reasoning was not only that it reinterpreted Fairchild, and indeed arguably liability, and thus, equitably, we decline to unleash the same forces to increase a defendant’s liability beyond its fair share of responsibility’. 71   Barker (n 2) 576. 72  In Sienkiewicz (n 4) [181] Lord Brown said of Barker: ‘one finds the House having to face up to some of the problems it had left open with Fairchild and, as it seems to me, beginning to have second thoughts both as to the juristic basis for this special rule of causation . . . and as to where the abandonment of the “but for” principle was taking the law’. 73   Barker (n 2) [31], [35]. 74   ibid [68]. 75   ibid [61]. 76   ibid [40]. 77   ibid [43]. 78  ibid. 79   Lord Rodger [101], Lord Walker [117] and Baroness Hale of Richmond [128]. 80  This extension of Fairchild invalidates one (pre-Barker) suggested justification for the decision: that the claim would otherwise fall down the gap between tortfeasors (cf Baker v Willoughby [1970] AC 467 (HL)): J Morgan, ‘Lost causes in the House of Lords’ (2003) 66 MLR 277.

68  Jonathan Morgan rewrote it.81 More fundamentally, Lord Rodger pointed out that the reinterpretation served to heighten the differences within and without the ‘enclave’ of Fairchild liability.82 Since the House of Lords had only recently denied liability for increased risk of dying of cancer owing to negligent diagnosis,83 to explain Fairchild in this way made it an even more radical and unusual exception to the normal rules on proof of causation (or rather actionable damage). Thus, the criticisms of the rather mechanistic conditions laid down for Fairchild liability made in the last section above become more pertinent still: do such dubious factors justify such a radical revision of liability for exposure to risk?84 By giving an apparently ‘principled’ explanation of Fairchild, but keeping in place the essentially arbitrary limits on Fairchild’s application, Barker makes holding the line over Fairchild harder still. Finally, Lord Rodger did not accept Lord Hoffmann’s (and the defendants’) argument that ‘fairness’ required proportionate liability. The law does not normally seek to ‘smooth’ the ‘rough justice’ which joint and several liability can visit upon defendants, ‘preferring instead, as a matter of policy, to place the risk of the insolvency of a wrongdoer or his insurer on the other wrongdoers and their insurers’.85 Any departure from this approach, Lord Rodger concluded, should come from Parliament rather than the courts.86 Lord Hoffmann in Barker suggested that the majority’s decision in Gregg v Scott had not been that ‘there was some conceptual objection’ to loss of chance/increase in risk claims,87 but merely that to allow recovery in Gregg ‘would in effect have extended the Fairchild exception to all cases of medical negligence, if not beyond, and would have been inconsistent with Wilsher’.88 In Gregg v Scott itself, Lord Hoffmann had argued (consistently with our argument here concerning Fairchild) that approving loss of a chance in Gregg v Scott would have to be general rather than fact-specific, ‘wholesale adoption of possible rather than probable causation as the criterion of liability’.89 This was not a step Lord Hoffmann was prepared to take, since it would have ‘enormous consequences for insurance companies and the National Health Service’; thus, such a radical change in the law would ‘amount to a legislative act’ and ‘should be left to Parliament’.90 It is interesting to note that in counsel’s reported argument, while the defendant had indeed submitted that extension of medical negligence to lost-chance cases would be a far-reaching (and expensive) step that only Parliament should take, the claimant had argued conversely that the common law should develop without fearing for the NHS – since Parliament could always intervene to counter-

  See especially Barker (n 2) [71], [83].   Barker (n 2) [84]–[85] 83   Gregg v Scott [2005] UKHL 2, [2005] 2 AC 176. 84   According to Lord Brown in Sienkiewicz (n 84) [181], in Barker ‘Lord Hoffmann explained that Fairchild had recognised a new tort, that of negligently increasing the risk of personal injury’ (emphasis in original). cf criticism by K Amirthalingam (2010) 126 LQR 162, 166 of ‘the ongoing attempt artificially to rationalise Fairchild in order to turn what was in fact a policy-based exception into a general principle of causation’. 85   Barker (n 2) [90]. 86   ibid [91]. 87   NB that Lord Phillips MR in Gregg v Scott [2005] UKHL 2, [2005] 2 AC 176, [190] seemed to leave open the question of liability where (unlike Mr Gregg’s claim on Lord Phillips’s interpretation of the evidence) ‘medical treatment has resulted in an adverse outcome and negligence has increased the chance of that outcome’. Lord Phillips MR’s refusal to deal with the issue of law raised by the agreed statement of fact is rightly criticised by Stevens (n 23) at 49. 88   Barker (n 2) [39]. 89   Gregg v Scott [2005] UKHL 2, [2005] 2 AC 176, [39]. 90   Lord Phillips MR pointed out at [174] that the Law Commissions would be better placed than the courts to consider the implications of such a change. 81 82

The English – and Scottish – Asbestos Saga  69 act the ‘impact of a proliferation of actions on the National Health Service’.91 Lord Hoffmann’s policy argument was therefore not as clear-cut as he seemed to think.92 The difficulty of assessing the ‘deference to Parliament’ argument in Gregg v Scott, given that it could plausibly be used to strengthen or to weaken the case for judicial intervention, leads to a wider point. It has been argued that it is universally inappropriate (for the courts, at least) to use policy arguments to determine the boundaries of liability in tort cases. Instead, liability is said to turn upon the rights of claimants, as protected by the principles of corrective justice. In the next section, we critically assess these claims.

V. The Seduction of Purity

The kinds of criticisms of Fairchild and Barker in this essay are twofold. First, that Fairchild recognises a highly unstable, and potentially legally disruptive, ‘exceptional’ liability to meet the merits of a perceived hard case; Barker, while purporting to rationalise Fairchild, does nothing to justify its arbitrary boundaries, and if anything (therefore) makes them even harder to defend, heightening the anomaly with cases where liability has been denied such as Gregg v Scott (or, arguably, Rothwell). Such ‘exceptions’, if they are to be made at all, should be made by legislation not by judicial decision. Secondly, we disapprove of the rather cursory examination of policy arguments in Fairchild and Barker. Lord Hoffmann, for example, originally suggested that claims brought against employers (Fairchild) should be treated differently from those against the NHS (Wilsher, Hotson), before recanting in Barker and pledging allegiance to the ‘single agent’ theory. We view this as disappointing. Lord Hoffmann, to be sure, identified a possible deleterious effect of liability in say Wilsher (disruption of NHS budgets) which did not apply to Fairchild. This ‘policy’ argument would therefore justify the distinction between those two cases which is otherwise (with respect) difficult to understand. But Lord Hoffmann did not explore it in any detail; certainly not with the careful critical attention that such a controversial reason (pertaining to the identity of defendants, and distributive concerns93) deserved. As seen, Lord Hoffmann subsequently turned away from this line of policy reasoning altogether. Some commentators would applaud Lord Hoffmann’s conversion. ‘Corrective justice’ theorists argue that courts should never take policy considerations into account when developing the law of torts. We examine such arguments in this section, and conclude that it is highly questionable whether corrective justice can of itself provide answers to the vexing questions raised by Fairchild or the other problematic causation cases. The formal/structural account of corrective justice certainly rules out certain ‘reforms’ to tort law: we have ourselves argued elsewhere that tort cannot function as an adequate compensation scheme because the entire structure of the law (rooted in personal responsibility, and hence enforced ‘bilaterally’ between victim and injurer) works to prevent universal compensation for those in need.94 But once ‘within’ something recognisable as tort (ie people being held responsible, in some way, for some kind of damage sustained by others), corrective justice only raises,   [2005] 2 AC 176, 179. See also Lord Nicholls [52]–[53].   It is the more curious, in the light of his Lordship’s comments in Gregg v Scott, that Lord Hoffmann in Barker tacitly repudiated his earlier explanation of Wilsher (in Fairchild) on NHS budgetary grounds. See above at p 62. 93   cf also Frost 504, 510–11 (Lord Hoffmann). 94   J Morgan, ‘Tort, insurance and incoherence’ (2004) 67 MLR 384. 91 92

70  Jonathan Morgan rather than answers, questions. It is also in great danger of begging them. Reverting to Fairchild, is one wronged by a group when one cannot prove which of them (all admittedly negligent) fired the fatal shot, or let loose the fatal asbestos fibre? Turning to Barker, is there a right not to be exposed to the risk of harm (at least in certain circumstances)? It is submitted that these questions cannot be answered merely by reflecting on the basic structure of corrective justice (personal responsibility, bilaterally enforced). We must go far beyond it, to a full substantive theory of distributive justice, or rights, or responsibility. It is not clear why all concern with ‘policy’ (the effect of legal liability on its regulated subjects’ behaviour) should be ruled out of bounds when choosing between rival conceptions of damage, causation or proof, all apparently compatible with tort law’s ‘basic structure’. Two recent extended corrective justice theories are those of Robert Stevens and Allan Beever.95 For Dr Beever, the policy debate that we now take for granted in tort law would have ‘surprise[d] . . . our juridical ancestors’ for it entails ‘wide ranging debates on issues about which there is nothing even approaching a general consensus’, requiring answers from the courts that will be ‘highly politically controversial’.96 Accordingly, Dr Beever submits: this vision of law contains an invitation to judges to enforce their personal and politically controversial conceptions of the good. The problem with this view of the courts’ role – and I assert this here without argument as I cannot see that one is needed – is that this is just not law. If judges are constrained only by their beliefs as to these and similar issues, then we have the rule of judges, not the rule of law.97

As Underwood J has written, we cannot be surprised if legislators intervene ever more frequently to substitute their own policy preferences for those of the courts.98 Dr Beever accordingly advocates the ‘principled approach’ as a ‘way out of the malaise that is the modern law’.99 Professor Stevens, similarly, argues both that his non-instrumental rights-based approach is a better interpretation of the law as we find it and, moreover, that it is a good thing – at least within our current system of adjudication.100 The focus on policy concerns that can be traced back to the early twentieth century American Legal Realists might make tort law appear more interesting, but the boring truth is that tort is, after all, founded on principles. Judges, in any case, are not well suited to policy reasoning. They do not have the information necessary to make policy, or indeed the training, experience and ‘skills’ to make use of such information. The growth of policy reasoning will make the law much more uncertain, indeed indeterminate. Policies tend to be polycentric and incommensurable; their balancing and application cannot be a matter of logic, in the absence of some overall meta-level ranking of concerns. Finally, of course, courts lack the democratic legitimacy to decide such a policy question as whether trades unions should enjoy defences from suit in the economic torts. The courts should therefore eschew such concerns altogether, at least under our current constitutional settlement. The alternative, if we mean to take ‘policy’ seriously, would be to embrace fully the role of judicial lawmaking – necessitating (at least) ‘Brandeis   Stevens (n 23); Beever (n 24).   Beever (n 24) 5–7. 97   Beever (n 24) 7. Contrast P Cane, ‘Taking disagreement seriously: courts, legislatures and the reform of tort law’ (2005) 25 OJLS 393. 98   Beever (n 24) 7–8 citing P Underwood, ‘Is Ms Donoghue’s snail in mortal peril?’ (2004) 12 Torts Law Journal 39, 59–60. 99   Beever (n 24) 21. 100   Stevens (n 23) ch 14 (‘Policy’). 95 96

The English – and Scottish – Asbestos Saga  71 briefs’ and wide-standing rules to provide the necessary information, prospective overruling, and a politically accountable judiciary.101 It must be admitted that Beever’s and Stevens’s arguments are a powerful challenge for latter-day Legal Realists (if such we are). There are several possible ways to respond. First, we might doubt whether the courts are really comparatively worse than the legislatures at assessing policy arguments.102 Secondly, and more fundamentally, one could question the premise of the ‘pure’ approaches, viz that they enable tort lawyers to avoid engaging in ‘political’ debates about the nature of liability. Both Beever and Stevens insist that tort liability is nothing more than the protection of claimants’ rights by the principles of corrective justice. But as suggested above, this raises a host of further questions: what precisely is corrective justice? What ‘rights’ do we have, and to what extent should they be protected? And who decides on these matters? It is proposed to examine the utility of the rights-based or principled approach to the liability for increased risk that seems to have emerged (rather haphazardly) from the Fairchild – Gregg v Scott – Barker line of cases. Beever and Stevens accord a central place to causation in their theories, in line with the High Priest of corrective justice, Ernest J Weinrib.103 For Professor Stevens, the fact that we require the claimant to pinpoint who caused his injury, and do not simply require those who create risks to pay proportionately into some central compensation fund: is good evidence that the law of torts cannot be solely explained by reference to such goals as deterrence and compensation, nor by a concern to hold the culpable responsible, but is rather concerned with the rights we have one against another.104

It is true that an exclusive focus on tort law as a means of deterring risky conduct has little room for the necessity of causation, as statements by leading lawyer-economists demonstrate.105 Similarly, a rational scheme of compensation would focus on victims’ needs,106 and the needs of a man with a broken leg are identical whether or not it was broken through the fault of another. As Gary Schwartz has observed, the desire to compensate is quite promiscuous and not at all wedded to the causation requirement; individuals can be wonderfully compensated by parties who have no causal connection to their injuries whatsoever.107 101   cf P Cane, ‘Taking disagreement seriously: courts, legislatures and the reform of tort law’ (2005) 25 OJLS 393 (arguing that, while legislatures are in a better position to reform tort law than courts, judges are nevertheless entitled to rely on their personal beliefs and values when deciding novel tort cases, including beliefs in the cogency of policy reasons such as ‘overkill’ concerns – and that judges must ‘state clearly and honestly the real reasons that support their decisions’). 102   See J Morgan, ‘Policy reasoning in tort law: The courts, the Law Commission and the critics’ (2009) 125 LQR 215. 103   EJ Weinrib, The Idea of Private Law (Cambridge (Mass), Harvard University Press, 1995). 104   Stevens (n 23) 129. 105   See eg, G Calabresi, ‘Concerning Cause and the Law of Torts: An Essay for Harry Kalven, Jr’ (1975) 43 University Chicago Law Review 69, 85: individual causation is ‘far from being the . . . almost categorical imperative it is sometimes described to be’; WM Landes and RA Posner, The Economic Structure of Tort Law (Cambridge (Mass), Harvard University Press, 1987) 229: ‘causation can largely be dispensed with in an economic analysis of torts’. 106   And, probably, on the resources available to the funders of the scheme: cf Karl Marx’s slogan ‘From each according to his ability, to each according to his need’ (Critique of the Gotha Program, 1875). cf Acts of the Apostles, Ch 4, vv 34–35: ‘Neither was there any among them that lacked: for as many as were possessors of lands or houses sold them, and brought the prices of the things that were sold, And laid them down at the apostles’ feet: and distribution was made unto every man according as he had need’. 107   CH Schroeder, ‘Corrective Justice and Liability for Increasing Risks’ (1990) 37 UCLA Law Review 439, 467 citing GT Schwartz, ‘The Supreme Court of California, 1977–1978 Term Foreword: Understanding Products Liability’ (1979) 67 California Law Review 435, 445.

72  Jonathan Morgan But all this does not prove that a system of liability mediated through some kind of central fund is necessarily incompatible with corrective justice. Christopher Schroeder has argued that liability based on creation of risk rather than a demonstrable causal link is fully compatible with a non-instrumental theory of corrective justice based on the moral responsibility of actors.108 Indeed, Schroeder maintains that focusing on the level of risk created by the defendant’s conduct is better justified from the ‘action-based responsibility’ angle, since the actual occurrence (or not) of harm as a result of the risks created is purely fortuitous.109 By assessing liability in an amount equal to expected harm, such a system eliminates the lottery characteristics of liability. The agent can know what liability is going to attach, because it attaches on the basis of information and calculations available to the agent at the time of decision. Contingency as to liability is reduced, and an agent can coherently make choices about risky behavior with full knowledge of the cost involved.110

It has been suggested that by severing the ‘bipolar’ linkage of claimant and defendant Schroeder’s approach may not really be a theory of corrective justice.111 Schroeder replies that the obsession with the ‘uniquely just form of bipolar litigation’ which is at the centre of Weinrib’s influential account of corrective justice is no more than an ‘aesthetic preference for symmetry’ and a ‘plea for the supremacy of form over substance’.112 Moreover: Most corrective justice theorists seem to have taken the bipolar traditional litigation form simply for granted. [However] . . . persons can be treated justly in a system that rejects bipolarity. Continuing commitment to the bipolar form sacrifices the moral principles of individual responsibility that corrective justice entails. Once the case for ex ante substantive rules has been made, it becomes incumbent upon corrective justice advocates of causation to explain why form should trump substance.113

Of course, the precise meaning of corrective justice remains doubtful and controversial,114 despite its venerable, Aristotelian heritage.115 The essential point, however, is that it may be possible to sever the link between claimant and defendant without perforce plunging headlong into (Posnerian) deterrence or (Marxist) compensation.116 Moreover, while arguments for recognising creation of risk simpliciter as actionable damage in tort have particularly been associated with instrumental deterrence theorists, by   CH Schroeder, ‘Corrective Justice and Liability for Increasing Risks’ (1990) 37 UCLA Law Review 439.   A pronouncement with which some legal philosophers would disagree. See eg, T Honoré, Responsibility and Fault (Oxford, Hart Publishing, 1999). 110   CH Schroeder, ‘Corrective Justice and Liability for Increasing Risks’ (1990) 37 UCLA Law Review 439, 465. 111   KW Simons, ‘Corrective justice and liability for risk-creation: A comment’ (1991) 38 UCLA Law Review 113. 112   CH Schroeder, ‘Corrective justice, liability for risks, and tort law’ (1991) 38 UCLA Law Review 143, 151. 113   CH Schroeder, ‘Corrective Justice and Liability for Increasing Risks’ (1990) 37 UCLA Law Review 439, 470– 71. 114   See eg, A Beever, ‘Corrective Justice and Personal Responsibility in Tort Law’ (2008) 28 OJLS 475 criticising, inter alios, J Morgan, ‘Tort, insurance and incoherence’ (2004) 67 MLR 384. See further I Englard, Corrective and Distributive Justice: From Aristotle to Modern Times (Oxford, Oxford University Press, 2009). 115   Or perhaps because of it? cf Simons (n 111) 128: ‘One suspects that the startling diversity of scholarly interpretations of corrective justice is due to some scholars’ desire, first, to adopt a noninstrumental theory of tort and, second, to give it a respectable (preferably ancient Greek) intellectual pedigree. (One does not hear much talk of Nietzchian tort theory.)’. 116   Beever (n 24) 414 describes Schroeder’s work as ‘an important argument against the causation requirement . . . that causation is morally arbitrary’, but argues that Schroeder’s argument fails. With respect, Dr Beever’s distinction between the law being ‘ethically arbitrary . . . not morally arbitrary’ (emphasis in original) is elusive to the point of obscurity, and his definition of corrective justice (on the same page), in the present context, simply begs the question. 108 109

The English – and Scottish – Asbestos Saga  73 no means all such arguments can be characterised (and for ‘rights theorists’ dismissed) as ‘instrumental’. A non-instrumental defence of liability for creation of risks per se is readily conceivable. Framing the question in their own terms, why is it that Professor Stevens or Dr Beever deny that we have rights not to be exposed to the risk of injury by the negligence of others? Stevens considers the example of a factory emitting a noxious substance, causing a doubling of the risk in a certain population of developing a certain disease. He does not seek to deny that an increased risk of developing harm is, in fact, a loss.117 However, this is said to be damnum absque injuria: ‘Do we have rights good against the rest of the world not to be exposed to the risk of future harm? No’. Our rights ‘are in relation to the outcome of injury, not its risk of occurring in the future’. The reason given by Stevens is that the law must choose whether it is concerned with outcomes or concerned with risks: making the defendant compensate both for risks and outcomes would make him pay twice over. Thus, we have to opt for one or the other, and the law chooses to focus on outcomes.118 This argument may be questioned. Of course Professor Stevens is right to identify potential double-counting, but it is not clear why the law must make a grand systemic choice in the way that he supposes. Why not let each individual exposed to a toxin choose between an immediate claim for damages to compensate for the (present) increased risk of disease, and a possible subsequent claim for the disease, if and when it develops (but not of course both claims)? It would be wrong to assume that no rational claimant could prefer to take compensation for the risk ‘now’ and thus face apparent undercompensation should the disease eventually result. Quite apart from the whiff of inappropriate paternalism,119 this overlooks the possibility of quantifying the risk in an amount sufficient to purchase an insurance policy (given knowledge of the exposure) against the occurrence of the disease.120 In effect, this would require the defendant to insure the claimant against the possibility of the harm actually occurring. Far from being irrational, it is suggested that a risk-averse claimant would always prefer an insurance policy to ‘waiting and seeing’.121 Allowing claimants such a choice, we submit, is a far better way to deal with the problem of double-counting to which Professor Stevens alludes. It is also likely that most claimants (being risk-averse) would choose the immediate risk claim over the eventual (possible) injury claim – so if the law is to make the choice for them, why should it be the risk claim that is ruled out? It should also be noted that Professor Stevens allows for a major exception. When an increased risk (or lost chance) is consequent on the violation of a voluntarily created right, whether contractual or otherwise, it should be recoverable.122 According to this position, 117   Stevens (n 23) 43. Moreover, Stevens attacks the deterministic view that sees every event as fully determined by earlier events, even though we might lack enough information about the world to apply the relevant causal laws. cf discussion of Hotson at 50: ‘a philosopher may object that upon turning up at the hospital the child either had a 0 per cent or 100 per cent chance of being cured by careful treatment. However, this is irrelevant. That the child lost something valuable can be demonstrated by contemplating what a parent would be prepared to pay to go back in time and have the child given careful treatment. Most would give everything they have’. 118   Stevens (n 23) 44. 119   cf Re Pleural Plaques Litigation [2006] EWCA Civ 27, [139] and [141] per Smith LJ (dissenting). 120   Professors Porat and Stein make the excellent point that if the claimant was already paying life insurance premiums when exposed to the toxin, these premiums will surely be increased in line with the increased risk of disease, showing that possible future physical injury can cause definite present (financial) loss. See ch 11 of this volume. See similarly E Voyiakis (n 3). 121   A fortiori when the claimant may take immediate prophylactic action (once the risk has been apprehended) to prevent the risk ever ripening into actual harm. eg an ‘antidote’ to counteract a drug which is fatally toxic to 20 per cent of the population (cf Voyiakis (n 3) 918–19). 122   Stevens (n 23) 49.

74  Jonathan Morgan Spring v Guardian was correctly decided,123 and Fairchild and Barker could have been more straightforwardly decided on the basis that ‘the defendant had assumed responsibility to the claimant . . . as employers under contracts’.124 On this approach, however, both Wilsher and Hotson were wrongly decided, since in each case the claimant’s increased risk of harm (or reduced chance of recovery) was ‘consequent upon the infringement of a right generated by an assumption of responsibility,’ ie by the hospital ‘as deliverer of treatment’.125 This argument is hard to reconcile with authority. In the Court of Appeal in Hotson, Donaldson MR recognised the axiomatic distinction between breach of contract, actionable per se, and a claim in negligence where damage is the gist of the action.126 Of course this might lead to socially undesirable results in the medical context, since private patients might have a claim (under contract) when NHS patients (suing in tort) would not.127 But this might sometimes work in a claimant’s favour; there has been judicial speculation that the inability to show actionable damage for the purposes of a negligence action in the pleural plaques cases might have been avoided by framing the claim in contract.128 And anyway, the real problem in Hotson was the refusal of the House of Lords to engage with the argument that a lost chance could constitute ‘gist damage’ in negligence.129 More fundamentally, Professor Stevens’s argument is that there should be no difference between contractual and non-contractual assumed responsibilities. This enjoys a certain doctrinal-historical appeal, since the modern English law of contract grew from the action on the case for assumpsit. But ‘voluntary assumption of responsibility’ has had a distinctly chequered recent history as a test for liability in negligence. Lord Bingham collected various sceptical dicta in Customs v Barclays, indicating that the assumption of responsibility had little to do with the ‘the actions and intentions of the actual defendant’ and was rather a ‘notional’ concept, generally being imposed for other reasons.130 The concept seems almost infinitely malleable.131 But there are limits: Stevens accepts that those negligently discharging their shotguns in the famous ‘two hunters’ scenario cannot be said to assume responsibility to the victim. Therefore, on orthodox grounds, the victim’s claim for damages against each hunter must fail.132 Professor Stevens’s argument, then, draws a fundamental distinction between claims for risk and claims for outcomes which is difficult to accept, and allows a major exception to it based on a highly dubious concept (‘voluntary assumption of responsibility’). It is not clear that either stage of the argument does or should exclude all considerations of policy.  ibid. Spring v Guardian Assurance plc [1995] 2 AC 296 (HL).   Stevens (n 23) 148. 125   Stevens (n 23) 50, 148. 126   Hotson v East Berkshire AHA [1987] AC 750 (CA) 760: ‘The distinction between what must be proved in contract and what must be proved in tort in order to establish a cause of action may be regrettable, but it does not lie within the power of this court to do anything about it’. 127   cf McFarlane v Tayside Health Board [2000] 2 AC 59 (HL) 76 (Lord Slynn of Hadley): ‘If a client wants to be able to recover such costs he or she must do so by an appropriate contract’. 128   Rothwell (n 2) [74] (Lord Scott). 129   J Stapleton, ‘The gist of negligence: Part II’ (1988) 104 LQR 389. 130   HM Commissioners of Customs and Excise v Barclays Bank plc [2006] UKHL 28, [2007] 1 AC 181 [5]. 131   Stevens’s example, Spring v Guardian, is by no means easy to explain as a case of assumption of responsibility and reliance thereon. Lord Goff ’s decision in White v Jones [1995] 2 AC 207 (HL) that the defendant solicitor’s assumption of responsibility to the testator must be taken to extend to the plaintiff beneficiaries is perhaps the leading sophistry in this area. Another case, of course, in which ‘practical justice’ drove the result (or: hard cases make bad law, again). 132   Stevens 149–51 (also discussing, but rejecting, the basis for liability suggested by A Porat and A Stein, Tort Liability Under Uncertainty (Oxford, Oxford University Press, 2001)). 123 124

The English – and Scottish – Asbestos Saga  75 Dr Beever’s approach, although working with the same basic assumptions about the nature of tort law, is completely different from Professor Stevens’s. First, Beever would allow recovery in the ‘two hunters’ case.133 The reason seems to be that while only one of the hunters has actually shot the claimant, both of them have violated the claimant’s bodily rights in an extended sense (just as the tort of assault extends from actual battery to the apprehension of harm); and depriving the victim of his claim for damages against the ‘true culprit’ (by making it impossible to show which hunter shot him) is a ‘violation of the claimant’s bodily integrity’:134 ‘both defendants harmed the claimant: one shot the claimant, while the other or both injured the claimant’s personal remedial right to prove liability. Hence, there is no issue of causation, as we know for certain that both defendants violated the claimant’s rights’.135

It would, however, be different if D1 made it impossible for P to sue D2 by destroying the crucial videotape evidence that showed D2 assaulting P, for D1 would not here be violating P’s right to bodily integrity – ‘the claimant would be recovering for violation of a different right, and a right that would have to be manufactured to justify recovery’.136 The discussion here is a little puzzling, but it is clear when Dr Beever turns to discuss Barker that he denies that there is a general right not to be subject to the risk of injury. Although he does not believe that Fairchild can properly be a case of liability for increasing risks,137 Beever discusses the explicit position advanced by Lord Hoffmann in Barker to that effect. He states that it is ‘untenable . . . both because it is immoral and because it would be impossible to realise in practice’.138 The alleged ‘immorality’ seems to flow from the proposition (as Beever puts it) that merely to walk down the street would become an actionable wrong since this (like anything and everything else) places others at risk; but one should not be guilty ‘merely through acting’.139 Moreover: ‘The position is unworkable as a legal rule. We cannot have a situation in which everyone I pass on the street can sue me even if they have suffered no harm’.140 No doubt Dr Beever is here quite right. But could these ‘unworkable’ liabilities not be averted by requiring a significant increase in the risk of injury to others before they can sue? Professor Andy Klein has argued, for example, that recovery should be allowed when exposure to a risk such as toxic pollution has more than doubled the claimant’s risk of contracting disease in the future, noting that this corresponds with the more-probable-than-not standard of proof.141 Professor Michael D Green has suggested a similar doubling of risk 133  See Cook v Lewis [1951] SCR 830 (Supreme Court of Canada). Beever (n 24) 453–63. We must confess that we find Dr Beever’s reasoning at this point difficult to follow: indeed he admits at 462 that ‘at least initially, the argument is not particularly intuitive’. 134   Beever (n 24) 461. 135   Beever (n 24) 459 (emphasis in the original). 136   Beever (n 24) 461–62. 137   Beever’s discussion of Fairchild at 472–83 advances the view that all of the lawyers involved in the Fairchild litigation (counsel and judges alike) misunderstood the medical evidence, and that if the statistical questions had been approached correctly, it was a straightforward situation in which all of the negligent employers had probably contributed to the claimant’s mesothelioma, and thus not a hard case at all. Thus, Beever approves of the Compensation Act 2006, s 3 for returning the law to the position in which, he argues, it should have been all along: 483, 489. 138   Beever (n 24) 484. 139  ibid. 140  ibid. 141   AR Klein, ‘A model for enhanced risk recovery in tort’ (1999) 56 Washington & Lee Law Review 1173, 1194– 96.

76  Jonathan Morgan threshold as a rule of thumb to mitigate the uncertainties of epidemiological evidence.142 If these approaches might seem dangerously pragmatic (or even consequentionalist) to Dr Beever, we might ask whether his own ‘workability’ criterion does not also betray an implicit concern with consequences? It seems that Dr Beever has one final, and fundamental, objection to claims for increased risk and/or lost chances. He subscribes to a deterministic view of both the natural world and even (it appears) of human behaviour, so that there is no real, ultimate or ‘objective’ uncertainty about what will happen in the universe; only an absence of knowledge as to what will happen (‘epistemological’ rather than ‘objective’ probability).143 It is suggested, however, that Professor Stevens’s common-sense response to such chilly logic is greatly to be preferred.144 Presumably, Stevens’s difference with Beever cannot be one of the dreaded policy disagreements. What, however, is the nature of the controversy when avowedly non-instrumental theorists are disagreeing? To what can they appeal to show that their theory of rights or principles (or probability, or harm, or proof) is preferable? While, of course, moral/ rights-based arguments are intellectually reputable, it is less clear that they should take place to the exclusion of all else. To exclude practical (ie consequentialist) arguments from legal reasoning altogether, especially from academic criticism of the law, stands to impoverish legal discourse. The answers to the causal conundrums presented by the asbestos cases cannot simply be deduced from the ‘structural’ axioms of corrective justice (primarily the bilateral nature of private-law litigation). Provided that policy arguments are examined with critical rigour, they may usefully contribute to a solution.145 That is not to say (to return to the start of this section) that courts are necessarily best placed to adjudicate upon questions of policy. Contrary to what many common lawyers feel deeply in their souls,146 judge-made law is not the only game in town. We need to give careful thought also to statutes, and legislators; in the present context, to examine their comparative ability to weigh the relevant policies.147 As we have already argued, legislation is certainly preferable to judge-made law when an ‘exceptional’ category, at odds with ordinary legal principle, is to be created. It is to the response at Westminster to Barker v Corus that we now turn, before examining the Scottish Parliament’s reaction to the pleural plaques litigation.148

142   MD Green, ‘The future of proportional liability: The lessons of toxic substances causation’ in MS Madden (ed), Exploring Tort Law (Cambridge, Cambridge University Press, 2005). cf now Sienkiewicz [2011] UKSC 10. 143   Beever (n 24) 498–504. 144   See n 117 above. See also E Voyiakis (n 3) 915 (fn14). 145   Lord Rodger in Sienkiewicz (n 4) [140] sets out a variety of possible legal approaches to the Fairchild problem. cf Lord Dyson at [219]: ‘Ultimately, questions of burden and standard of proof are policy matters for any system of law’. 146   cf RJC Munday, ‘The common lawyer’s philosophy of legislation’ (1983) 14 Rechtstheorie 193. 147   cf P Cane, ‘Taking disagreement seriously: courts, legislatures and the reform of tort law’ (2005) 25 OJLS 393. 148   Following the substantial completion of this chapter we have had the great benefit of reading in draft James Lee’s ‘ “Inconsiderate Alterations in our Laws”: Legislative Reversal of Supreme Court Decisions’ in J Lee (ed), From House of Lords to Supreme Court: Judges, Jurists and the Process of Judging (Oxford, Hart Publishing, 2010). cf generally RB Ginsburg, ‘Judicial Stimulation of Legislative Change: A View from the United States’, Third Leslie Scarman Lecture, Law Commission, 2010.

The English – and Scottish – Asbestos Saga  77

VI. Causation, Politics and Law I: Barker’s Demise at Westminster

Barker v Corus was a short-lived leading case, being retrospectively reversed by Parliament within three months of the House of Lords’ decision. This shows that tort liability is not purely a matter for judges weaving the seamless web of the common law. Parliament might always intervene to change the incidence of liability, perhaps on nakedly distributive grounds. The matter of note is that Parliament here did so. As will be seen, the ‘reversal’ of Barker was nevertheless incomplete, and perhaps ham-fisted. The result is much more ‘patchwork quilt’ than ‘seamless web’, especially given the rent in the fabric that the courts had already inflicted in Fairchild. At a stroke, Barker had made it very difficult for mesothelioma claimants to recover in full: they would now need to trace all the employers who had exposed them to asbestos over the course of their working lives, and all of those employers’ insurers. This was not an oversight by the House of Lords, but quite deliberate – as Lord Hoffmann put it, Barker was smoothing the rough justice of Fairchild, a decision that was exceptionally generous to claimants, by moving the law back in the favour of defendants (employers and their liability insurers).149 Not surprisingly perhaps, the decision was met with a storm of protest. The Prime Minister (Mr Blair) was asked more than once what Her Majesty’s Government intended to do about the judgment, at his weekly question time in the House of Commons.150 The governing Labour Party’s MPs tended to have most constituents directly affected by Barker, occupying seats in the former industrial heartlands of the UK where asbestos-dependent industries once proliferated – and now asbestos-related diseases. Providentially for the Government, a Bill to reform certain aspects of tort law was at that very moment passing through its parliamentary stages.151 A new clause was introduced (the eventual section 3 of the Compensation Act 2006), briskly to reverse the effect of Barker.152 While mild dissent was expressed by Lord Hunt of Wirral about the ‘haste’ and indeed ‘emotion’ with which Barker’s demise had been brought about,153 most who spoke in the legislative debates welcomed the Government’s rapid action. A few critics questioned the retrospective nature of the change.154 However, as the minister, Bridget Prentice MP, explained to the House of Commons, making the changes retrospective was an ‘exceptional step’ justified because it was of the

  Barker (n 2) [43].   See also Hansard, HC Deb 13 June 2006, col 192WH-216WH.   This became the Compensation Act 2006. Its key provision (apart from reversing Barker) is s 1: which doesn’t change the law in any way at all. The Act’s Explanatory Notes (written by the Government) confirm that s1 was not designed to amend the law. The ‘compensation culture’ problem was thought to stem not from any increase in tort liability, but from people’s fears and misperceptions about liability. Thus s 1 was designed to correct misperceptions rather than curb liability. In effect, a presentational measure. All very Blairite. 152   The speeches in Barker were delivered on 3 May 2006; the relevant amendments were moved in the House of Commons on 17 July 2006, and the Compensation Act received Royal Assent on 25 July 2006. So much (on both sides) for ‘judges who can’t and legislators who won’t’! 153   Hansard, HL Deb, 19 July 2006, col 1318–19. And even Lord Hunt ‘fully accept[ed] the urgency of this matter – this disease [ie mesothelioma] will not indulge us in our deliberations and every day more people are struck down’. 154   Compensation Act 2006, s 16(3): ‘Section 3 shall be treated as having always had effect’. Lord Hunt of Wirral was, again, the most eloquent doubter. 149 150 151

78  Jonathan Morgan utmost importance that all claimants affected by the Barker judgment can secure full compen­sation, including the parties to that case, the parties to the cases conjoined with that case and the parties to cases subsequently settled or determined on the apportionment basis proposed by Barker.155

Since the legislation was being enacted so soon after their Lordships’ decision, Mrs Prentice pointed out, no settled expectations would be defeated by a retrospection of ‘barely seven or eight weeks’.156 In the House of Lords, Lord Goodhart declared that ‘retrospective overruling by Parliament of the decisions of the courts is, in principle, a breach of the rule of law’ but stated that he was satisfied that reversing Barker was justified ‘by the wholly exceptional circumstances of this case’ provided that it was ‘not to be taken as a precedent for the future retrospective overruling of court decisions’.157 We could note that since judicial decisions changing the common law always have retrospective effect, as they apply to facts which by definition lie in the past, legislation intending to reverse the outcome of a case will always have to be retrospective too. Even the former Lord Chancellor Lord Mackay of Clashfern thought that ‘what has been proposed is eminently justified’.158 The only hint of party political opposition came from Conservative MPs who pressed the Government over its involvement in the Barker appeals to the House of Lords (as defendant, and appellant). Oliver Heald MP alleged that while the Government had criticised the role of the insurance industry, ‘the Government funded two of the cases and one was funded privately. Those test cases were pursued because a lot of former Government employees, who worked in the Ministry of Defence and other Departments, were exposed to asbestos’. It was odd, he said, that ‘the Government are now introducing emergency legislation to overturn what they argued for in court as recently as six weeks ago’. Mr Heald asked how much money the Government had spent in legal fees fighting the appeals that it was now striving to reverse, concluding that ‘this is not an advert for joined-up government’.159 Nevertheless, the Official Opposition welcomed the Government’s decision to reverse Barker. David Howarth MP raised an important caveat during the Commons debate. The Law Lords’ decision in Barker had settled two points about Fairchild. The first was whether liability was joint and several; this was what the Government was seeking to reverse. But Barker had also decided that Fairchild liability arose in a situation where some of the exposure to asbestos had been non-tortious (ie by the claimant himself). As Mr Howarth correctly pointed out, the two limbs of the decision were intimately associated – since for Lord Rodger, Lord Walker and Baroness Hale the decision to extend Fairchild was justified by the limited nature of Fairchild liability (ie for proportionate shares rather than in solidum).160 Did the Government appreciate that by reversing the decision in Barker on proportionate liability, it would throw into doubt their Lordships’ (claimant-friendly)   Hansard, HC Deb, 17 July 2006, col 47.   Moreover, ‘It is my understanding that following Barker, the great majority of cases that were under way have been stayed pending clarification of the appropriate method of apportionment or in the light of the Government’s announcement that they intended to introduce these amendments. I am grateful to the judiciary for waiting to hear exactly what we intended to do before they carried those cases through’. Hansard, HC Deb, 17 July 2006, col 48. 157   Hansard, HL Deb, 19 July 2006, col 1319–20. 158   Hansard, HL Deb, 19 July 2006, col 1322. No other judge contributed. Lord Goodhart (col 1319) noted the presence in the chamber of the Lord Hoffmann and wondered aloud what Lord Hoffmann ‘if he chose to intervene in the debate’ would say, but sadly Lord Hoffmann remained mute. 159   Hansard, HC Deb, 17 July 2006, col 50–52. 160   Barker (n 2) [101] (Lord Rodger, considering this point on the basis of the majority decision on liability in solidum, from which he had dissented); [117] (Lord Walker); [128] (Baroness Hale). 155 156

The English – and Scottish – Asbestos Saga  79 decision on the other point?161 The Government’s reply in both Houses was fairly unilluminating, but the ministers promised to write to Mr Howarth setting out the Government’s considered view.162 The Government’s failure to engage fully with Mr Howarth’s point suggests a less-thanfull consideration of the intricate task of reversing a judgment by legislation. The 2006 Act seems in some ways curious, or even misconceived. To be sure, it reverses the outcome of Barker, declaring that when certain conditions are met:163 The responsible person shall be liable – (a) in respect of the whole of the damage caused to the victim by the disease (irrespective of whether the victim was also exposed to asbestos – (i) other than by the responsible person, whether or not in circumstances in which another person has liability in tort, or (ii) by the responsible person in circumstances in which he has no liability in tort), and (b) jointly and severally with any other responsible person.164

But this is achieved without seeking to engage with the reasoning in Barker. Traditionally, Parliament never expresses reasons in the statute itself, and parliamentary draftsmen have generally remained adamant that it should not. Yet the decision in Barker had been reached (at least in part) by Lord Hoffmann’s reinterpretation of Fairchild to put it on an allegedly more principled basis: liability in Fairchild was properly to be seen as liability for increasing the risk of mesothelioma, not for contributing to the mesothelioma itself (since that could not be proved, and it was fictitious to pretend that it had been). Since Fairchild liability was for creating risk (a divisible – indeed ready-divided – ‘harm’) rather than causing mesothelioma (an indivisible harm), liability was for the proportion for which each defendant was responsible, not joint and several. Parliament, however, has now enacted that liability is indeed joint and several – and always has been.165 Has Parliament, by implication, therefore also reversed Lord Hoffmann’s principled reinterpretation of Fairchild? Is the nature of Fairchild liability now after all for ‘causing mesothelioma’ and not ‘increasing risk’? We suggest not. Barker has indeed altered the jurisprudential basis of ‘Fairchild liability’, irrevocably. The Compensation Act 2006 does not purport to disapprove the House of Lords’ reinterpretation of Fairchild: it simply disapplies the conclusion that liability is not joint and several, for mesothelioma claims. While we have spoken in general about the ‘reversal’ of Barker, this refers only to the outcome of the case.166 As noted above, section 3(2) applies only when the conditions in section 3(1) have been met, most pertinently that the victim has contracted mesothelioma as a result of exposure to asbestos[.] 161   This argument is also made by NJ McBride and R Bagshaw, Tort Law, 3rd edn (London, Pearson, 2008) 555. The authors’ prediction that the point ‘may be regarded as too subtle’ receives validation from Sienkiewicz [2011] UKSC 10, where it was not directly discussed. Lord Brown said at [182] that it was ‘quite clear that the preparedness of the majority of the court in Barker to extend the reach of the Fairchild principle this far was specifically dependent upon there being aliquot liability only’ – but apparently did not think that the point was reopened for decision by Parliament’s intervention. 162   David Howarth MP has indicated that that letter, however, was equally unilluminating: communication to Jonathan Morgan, 10 August 2006. 163   S 3(1). 164   S 3(2). 165   S 16(3). 166  See Sienkiewicz (n 4) [131] (Lord Rodger).

80  Jonathan Morgan But there are (presumably) instances of other toxic exposures which might in future satisfy the Fairchild criteria (single causal agent, inherent scientific uncertainty) in addition to asbestos/mesothelioma cases. Barker would still apply to such a Fairchild claim with vigour undiminished.167 This might seem to create another questionable distinction. Mesothelioma victims can rely on Fairchild and section 3 to achieve joint and several liability, but victims of other diseases which satisfy Fairchild will continue to be caught by Barker’s determination that liability was divisible, defendants being liable only for the risk that they created, and not the outcome (the disease). However, because the conditions for the operation of section 3 are laid down in legislation, there can be no plausible argument that section 3 should apply by analogy (to diseases that are ‘like’ mesothelioma, or toxins ‘like’ asbestos). If Parliament had wanted section 3 to apply to similar cases, it could have said so. As it hasn’t, section 3 must be given its limited effect (thus allowing some residual force for the reasoning in Barker). There can be no question of the judges extending the statutory rule to like cases. Unlike Fairchild, the conditions for section 3’s application have not been laid down judicially (although they might be equally indefensible). Parliament has spoken, and the legal position is clear. Some would argue that such legislation is flawed as a matter of political morality, however robust it might be viewed as positive law. Professor Dworkin argues that a society committed to justice and fraternity will place a high value on ‘integrity’ in the political system,168 and will thus abhor ‘checkerboard statutes’ – eg a compromise between pro-life and pro-choice legislators allowing abortions in even months but making them illegal in the odd months.169 Such a statute might be a validly-posited law and thus would be enforced by the courts, but we would condemn its violation of the legal system’s demand for coherent principles – or ‘integrity’. As Dworkin puts it, ‘the state [should] act on a single, coherent set of principles even when its citizens are divided about what the right principles of justice and fairness really are’.170 As suggested above, section 3 of the Compensation Act 2006 creates a distinction between victims of different diseases the normative significance of which is difficult to see. Dworkin would presumably condemn it for violating the legal system’s ‘integrity’. There are a number of possible answers to such charges. First, and most fundamentally, we would argue that Professor Dworkin’s call for all legislation to be informed by ‘a single, coherent set of principles’ is not just unrealistic but undesirable. As Professor Jeremy Waldron argues, legislating takes place against a background of endemic political disagreement, and so it is both necessary and laudable for legislation to be pervaded by compromise.171 From the needs-focused ‘compensation’ viewpoint, no doubt all victims of every misfortune should enjoy restitutio in integrum: but this is unaffordable, and so limited extensions of compensation such as section 3 mark the outer bounds of generosity in the real world. Secondly contra Dworkin, and following on from this, it is notorious that in Britain (and, no doubt, every other country) the compensation available to the victims of   Pace James Lee (n 148): ‘we might expect such a legislative affirmation [ie of Barker] to be more explicit’.   R Dworkin, Law’s Empire (London, Fontana, 1986) ch 6. 169   ibid 278. 170   ibid 166. 171  See generally, J Waldron, The Dignity of Legislation (Cambridge, Cambridge University Press, 1999); J Waldron, Law and Disagreement (Oxford, Clarendon Press, 1999). cf also A Marmor, ‘Should we value legislative integrity?’ in RW Bauman and T Kahana (eds), The Least Examined Branch: The Role of Legislatures in the Constitutional State (Cambridge, Cambridge University Press, 2006). 167 168

The English – and Scottish – Asbestos Saga  81 different kinds of misfortune differs enormously. An egregious example is furnished by the Pneumoconiosis etc. (Workers’ Compensation) Act 1979 – enacted as a sop to the Welsh Nationalist MPs upon whom the minority Callaghan Government was reliant for its continued survival in the House of Commons, even though the Pearson Royal Commission of 1978 had just rejected the Welsh miners’ claim to be treated as a special case. Professor Cane refers to the ‘almost Gilbertian hilarity’ of this episode.172 Maitland famously said that Parliament in the eighteenth century rarely rose to the dignity of a general proposition,173 and it sometimes seems no less true today, not least in the field of compensation. Dworkin’s ‘Law as Integrity’ simply does not fit the practice of legislation (while it does of course describe the postulates of common law reasoning very well). To extend that common law method to statute law, however, seems more Utopian (even deluded) than imperial. Returning to the specific case study, we should recall that quite apart from the ‘principled’ reinterpretation of Fairchild, Lord Hoffmann’s concern in Barker was to ‘smooth the injustice’ that Fairchild had visited upon defendants. But such appeals to judicial intuitions about ‘the just’ invite political disagreement. The instincts of the Law Lords, save Lord Rodger, about where ‘fairness’ lay were clearly at odds with the entire House of Commons and House of Lords in its legislative capacity.174 It may not be coincidental that in the subsequent pleural plaques case, Rothwell, their Lordships were careful to reason in the highest and driest doctrinal fashion (although this has not spared the decision from political criticism).175 So Parliament has reversed the actual outcome of Barker, no doubt in sharp dis­agreement with Lord Hoffmann about the merits of the case, while leaving the main plank of the reasoning in Barker to live on in ghostly fashion, with potential application to non-­ mesothelioma Fairchild cases (should any arise).176 It should be noted that if, as we advocated above, the initial Fairchild ‘problem’ had been resolved by legislation and not by a judicially-created exception, these ragged edges could never have arisen. If (say) Parliament had created a special compensation scheme for mesothelioma victims (which the judges interpreted as creating proportionate liability, an interpretation which Parliament legislated to reverse) there would be no possible category of ‘non-mesothelioma mesotheliomascheme victims’. It is the very fact that Fairchild set an exceptional judicial precedent, of uncertain scope, that the crisp legislative reversal of Barker has left behind some untidy common law remnants.177 172   P Cane, Atiyah’s Accidents, Compensation and the Law, 4th edn (London, Weidenfeld and Nicolson, 1987) 374–75. 173   FW Maitland, ‘English Law (History)’ in Encyclopaedia Britannica (11th edn, 1911). 174   The Scottish Parliament similarly, which on 22 June 2006 gave its consent to s 3, Compensation Act 2006 extending to Scotland (even though it was a ‘devolved’ matter, since this was quicker than introducing a separate bill at Holyrood). 175   J Lee (n 148) warns that it is ‘unwise and constitutionally dangerous for the threat of a legislative reversal to be in the minds of the judiciary when they are deciding cases’. But cf now Baroness Hale in Sienkiewicz [167], quoted at n 58 above. cf Lord Brown: ‘the sole effect of section 3 is to reverse the House’s decision in Barker on the issue of quantum; in no way does it pre-empt or dictate the proper approach of the common law to questions of causation and liability. On the other hand it would be a remarkable thing for this Court now in effect to reverse the decision in Fairchild and revert, in mesothelioma cases as in all others, to the normal, “but for”, rule of causation’. Sienkiewicz [183]. 176   Foskett J has observed of Barker: ‘This decision was reversed in relation to mesothelioma by virtue of section 3 of the Compensation Act 2006. But the issue remains as to the extent to which the decision could have an impact on analogous cases’: AB v Ministry of Defence [2009] EWHC 1225 (QB) [226]. 177   cf James Lee, ‘the law has now been rendered even more incoherent than it was in Barker, as the general approach to liability, of risk as damage, is untouched by the Act. We have a statutory exception to a common law exception’ in ‘Fidelity in interpretation: Lord Hoffmann and the Adventure of the Empty House’ (2008) 28 Legal Studies 1, 12–13.

82  Jonathan Morgan If the most remarkable thing about the reversal of Barker was the speed with which it took place (which may be explained by the political imperatives of the Labour Government and the adventitious availability of the Compensation Bill as a suitable vehicle), the lack of serious scrutiny and contentious debate was almost as surprising. Perhaps the speed of the reversal meant there was no time for interest groups to organise effective parliamentary opposition (presumably liability insurers, despite their apparent non-involvement in the Barker litigation itself, had welcomed the House of Lords’ decision to end joint and several liability).178 All of this was rather different, as will be seen, in the more recent debates about pleural plaques.

VII. Causation, Politics and Law II: Pleural Plaques, the Courts and the Scottish Parliament

Fairchild and Barker concerned mesothelioma. A more recent round of litigation has tested the position of compensation for pleural plaques. The House of Lords determined in Rothwell that plaques do not constitute actionable damage for a claim in negligence. But this was swiftly reversed by statute (again, retrospectively) in Scotland. This legislation has been criticised for giving an unjustified preference to one particular group of asbestos exposure victims, as will be seen. But we suggest that while this might be a valid criticism at the level of (Dworkinian, Utopian) political morality, the enacted special treatment of pleural plaques will at least not cause legal uncertainty comparable to that which Fairchild, being a common law ‘exception’, has wrought in tort law. We briefly examine the litigation and the differing reasoning of the Court of Appeal and House of Lords, before considering in some detail the various arguments unsuccessfully made against the Scottish legislation. Pleural plaques are benign, asymptomatic thickenings of the pleural membrane, which develop as the body’s reaction to inhaled asbestos fibres. In the mid-1980s, a trilogy of cases decided at first instance against the Ministry of Defence had established that plaques were actionable injuries for the purpose of negligence actions, and thereafter damages (at fairly modest levels) had been awarded or agreed for claimants’ anxiety that they would develop asbestos-related cancer in the future.179 More recently, a decision was made to test the legal position again,180 and 10 cases were tried together by Holland J in 2005 who, again, upheld the claims.181 But the Court of Appeal decided, by a majority, that plaques did not constitute an injury upon which an action could be founded.182 On the claimants’ further appeal 178   As Lord Phillips says, ‘The rejoicing with which the insurance industry must have greeted [Barker] was short lived’ Sienkiewicz (n 4) [56]. 179   A ‘commercial decision’ by the insurers, taken while the number and value of such claims appeared low: Petition by Axa General Insurance [2010] CSOH 2 [9]. 180   Medical opinion (which had been uncertain in the 1980s whether or not plaques played a role in the aetiology of cancer) had come to the consensus view that plaques have no causal link to malignant conditions such as mesothelioma. Secondly, the level of payouts by liability insurers had risen sharply (both total numbers of claims, but also rising legal costs). Curiously it seems that, as with Barker, the British Government initiated the test cases to try to curtail its liability, with liability insurers joining in at a later stage. Scots Parliament Justice Committee, Official Report, 2 September 2008, col 1029–30; Scots Parliament Official Record, 5 November 2008, col 12027; cf Petition by Axa General Insurance [2010] CSOH 2 [9] per Lord Emslie: ‘Apparently the prime mover . . . was the Department of Trade and Industry as representing former nationalised industries’. 181   Grieves v FT Everard & Sons [2005] EWHC 88 (QB). 182   [2006] EWCA Civ 27; see J Morgan, ‘Asbestos and Anxiety’ [2006] CLJ 269.

The English – and Scottish – Asbestos Saga  83 to the House of Lords, known as Rothwell, the decision of the Court of Appeal was unanimously upheld, and the claims dismissed.183 It is interesting to note that the reasoning of Lord Phillips of Worth Matravers CJ and Longmore LJ (the majority judgment in the Court of Appeal) was very different from that of the Law Lords. The Court of Appeal stated that they had been persuaded by counsel for the defendants that a number of policy reasons made it undesirable for plaques to con­ stitute actionable damage.184 In dissent, however, Smith LJ subjected these policies to sustained criticism,185 in a judgment to which tribute was paid in the House of Lords.186 Moreover, Lord Hoffmann dismissed the Court of Appeal’s ‘policy arguments of a con­ sequentialist nature, based upon predictions of how people would behave if they could sue for pleural plaques’ as being ‘rather speculative’.187 Their Lordships instead placed their decision for the defendants squarely on established propositions of legal doctrine. For an injury to constitute actionable damage, it had to satisfy the maxim de minimis non curat lex.188 Since (as the claimants’ medical experts accepted) pleural plaques, ‘save in the most exceptional case . . . did not increase the susceptibility of the claimants to other diseases or shorten their expectation of life. They had no effect upon their health at all’.189 If the plaques themselves were not ‘personal injuries’ then there could be no basis for the recovery of damages for anxiety or for the increased risk of asbestos-related cancer in future. These are not, of course, heads of actionable damage in themselves.190 Thus, if none of the alleged injuries (plaques, anxiety, increased risk) was able to sustain an action separ­ ately, aggregating them together could not produce a different result: ‘Nought plus nought plus nought equals nought’.191 As Lord Rodger observed: ‘The logical difficulties of such an approach are self-evident and, in my view, insurmountable’.192 While the two appellate courts ultimately reached the same result, then, the reasoning process was noticeably different. The ready resort to ‘policy reasoning’ in the Court of Appeal was explicitly disapproved in the House of Lords, where great pains were taken to place the decision on the most orthodox doctrinal grounds. Might this reflect an apprehension of Underwood J’s point that when judges engage in policy reasoning, they should not be surprised if the legislature plays at the same game and reverses the judges’ balancing of policy factors?193 But if the judicial straight bat was intended to block out criticism of the plaques decision by politicians it has not been wholly successful. A serious of fast deliveries has been aimed at Rothwell, and in Scotland it has been reversed through (retrospective)

  Rothwell (n 2)   ibid [67]. 185   [2006] EWCA Civ 27 [136]–[146] 186   Rothwell (n 2) [48] Lord Hope. 187   ibid [17]. cf [50] (Lord Hope). 188   Cartledge v E Jopling & Sons Ltd [1963] AC 758 (HL). 189   Rothwell (n 2) [11] (Lord Hoffmann). 190   Hicks v Chief Constable of the South Yorkshire Police [1992] 2 All ER 65 (HL) and Gregg v Scott [2005] 2 AC 176 (HL) respectively. Interestingly, these limits on recovery were also characterised as ones based on ‘policy’ by the Court of Appeal, [62]–[63] (Lord Phillips CJ). Apparently no attempt was made to distinguish Gregg v Scott using the House of Lords’ subsequent recognition of risk as actionable damage in Barker; as K Amirthalingam (2010) 126 LQR 162, 165 observes apropos Rothwell: ‘Curiously, none of the Lords cited Barker’. For the argument that Gregg v Scott itself does not bar claims for expenses (eg insurance) created by an increased risk of future harm see E Voyiakis (n 3). 191   Rothwell (n 2) [73] (Lord Scott). 192   ibid [89]. 193   See n 98 above. 183 184

84  Jonathan Morgan legislation.194 Similarly, in Northern Ireland, the Damages (Asbestos-related Conditions) Bill is about to become law.195 However, the House of Lords’ decision remains intact in England and Wales: after considerable delay, the British Government announced that it would not follow the Scottish lead (save for an ‘exceptional’ ex gratia compensation scheme for ‘individuals who had already begun, but not resolved, a legal claim for compensation for pleural plaques at the time of the Law Lords’ ruling . . . [who] would have had an understandable expectation that their claim would result in compensation’).196 Various Private Members’ Bills have been introduced at Westminster to reverse Rothwell,197 but it is most unlikely that legislation will be enacted against Government opposition. The speeches in Rothwell were delivered on 17 October 2007. The first debate on the decision in the Scottish Parliament was held on 7 November 2007.198 It is fair to say that despite their Lordships’ careful doctrinal opinions, Members of the Scottish Parliament were unconvinced by the reasoning. The decision was variously described as one that ‘contradicts natural justice’ and ‘cannot be justified in any way, shape or form’, and furthermore ‘profoundly unreasonable’, ‘bizarre’, ‘disturbing’, ‘scandalous and unjust’, ‘nonsensical’, ‘certainly not acceptable’, ‘an affront to justice’ and ‘based on a piece of semantic trickery over the definition of the term ‘injury’’.199 Even the Minister for Community Safety (Fergus Ewing MSP) declared that he had read the judgment and thought it ‘not entirely without internal contradictions’ (although without specifying what these were).200 Not surprisingly in the face of such strident cross-party opposition to the plaques judgment, on 29 November 2007 the Scottish Government announced its intention to introduce legislation to reverse the decision (generally known in Scotland as Johnston v NEI International Combustion Ltd, another of the test cases), and engaged in public consultation on its ‘Partial Regulatory Impact Assessment’ between February and April 2008.201 Of the 22 responses, 17 were against the proposal (including 10 from the insurance industry). The points made for and against the proposal were summarised.202 Those who supported legislation argued that it was justified to protect workers diagnosed with plaques, since diagnosis can cause considerable anxiety (as plaques are evidence of exposure to significant levels of asbestos). Moreover, it was suggested, reversal should not cause financial difficulty for insurers since, between the 1980s test cases and the Court of Appeal’s decision in 2006, they should have budgeted for such claims and collected appropriate premiums.   Damages (Asbestos-related Conditions) (Scotland) Act 2009.   The NI Department for Finance and Personnel introduced the Bill into the Northern Ireland Assembly in December 2010, following a consultation initiated by a paper of October 2008. The Bill passed its final legislative scrutiny stage in March 2011. 196   Statement on Pleural Plaques by Secretary of State for Justice and Lord Chancellor, Hansard, HC Deb, 25 February 2010, col 79WS. cf Consultation Paper on Pleural Plaques (London, Ministry of Justice, July 2008); Pleural Plaques: Response to Consultation (London, Ministry of Justice, March 2010). There will be a one-off ex gratia payment of £5,000. The Department for Finance and Personnel in Northern Ireland ruled out a no-fault scheme on the grounds that this would allow insurers and employers to escape their responsibilities: Explanatory and Financial Memorandum accompanying Damages (Asbestos-related Conditions) Bill (Northern Ireland) 2010, para 15–16. 197  Most recently see Damages (Asbestos-related Conditions) Bill [HL] session 2009–10 (Baroness Quin); Damages (Asbestos-Related Conditions) (No 2) Bill [HC] session 2009–10 (Andrew Dinsmore MP) – both were debated before running out of time (Parliament was dissolved on 8 April 2010). 198   Official Report, 7 November 2007, col 3130–50. 199  ibid. 200   col 3149. 201   See ‘Summary of Responses,’ 24 June 2008 (web publication only). 202  ibid. 194 195

The English – and Scottish – Asbestos Saga  85 Points made against the proposed legislation included the following.203 Such a fundamental change to the law of negligence would deny the contemporary medical consensus that plaques have no ill effects on people’s health; historically, claims had been settled when the medical evidence was less clear and the assumption was that plaques affected health. Legislation on plaques would set a dangerous precedent for ‘exposure risk’ claims in other contexts. It would draw an unjustifiable distinction between people with identical risks of developing mesothelioma from identical asbestos exposure, depending on whether they had developed plaques (which have no causal link with mesothelioma). To upset a well-established doctrine in private law through retrospective legislation would violate the rule of law and damage Scotland’s reputation for a stable and principled legal framework upon which businesses and citizens may rely. More specifically, insurers might decide to withdraw from the Scottish market altogether, or raise premiums to higher levels for Scottish businesses, placing them at a competitive disadvantage. Scotland would be out of line with other European countries, which do not compensate for mere exposure. Anxiety would better be allayed by educating people about the benign nature of plaques (and the relatively small 1–5 per cent risk of contracting mesothelioma) than by paying damages. Major bene­ficiaries of the legislation, it was noted, would be the personal injury lawyers who were among its most enthusiastic champions. After considering these responses, the Scottish Government declared its intention to proceed with the legislation in its final Regulatory Impact Assessment.204 The Government replied to some of the criticisms made during consultation as follows.205 The Bill would not ‘set a precedent’ because it applied only to asbestos. It was necessary to legislate retrospectively ‘to fully overrule the effect of that Judgment and in order to maintain the coherence of the law’; retrospective legislation had openly been on the agenda as soon as the House of Lords’ decision was handed down. While those with plaques might not necessarily have a higher risk of mesothelioma, the physical manifestation of asbestos exposure ‘becomes a focus for their anxiety’, which justifies more favourable treatment. The position in other European countries could not affect the Government’s duty to do ‘what is best for the people of Scotland’. The Justice Committee of the Scottish Parliament initiated a further public consultation between June and August 2008, as part of the Bill’s legislative scrutiny. 31 further sub­ missions were received from interest groups and certain individuals.206 No attempt will be made here to describe exhaustively this vast body of information, but some points of interest made against the proposals will be examined. Again, much of the opposition came from the insurance industry. The insurers repeatedly argued that the Scottish Government’s estimates of the financial cost of the measure were overconfident, and that there was a real danger that the total number of plaques claims (and therefore total cost of the measure) for both insurers and governmental bodies (which self-insure) might be much higher than given in the Financial Memorandum accompanying the Bill as introduced.207 Insurers again sought to suggest that education was a better  ibid.   RIA Number 2007/61 (June 2008). 205   ‘Policy Memorandum’ accompanying Bill as introduced in Scottish Parliament on 23 June 2008, Annex A. 206  www.scottish.parliament.uk/s3/committees/justice/inquiries/damages/Damagessubmissions.htm 207   Financial Memorandum para 26 estimated annual cost for non-governmental defenders of £5.45 million. cf Norwich Union’s estimated total liability for plaques liability in Scotland of between £1.1 and £8.6 billion according to [UK] Ministry of Justice figures: para 1.14 of Norwich Union written submission. At para 6.9.5 Norwich Union noted the massive underestimation of claims in the British Coal compensation scheme (592,000 actual claims against 150,000 estimated claims), despite the apparently greater degree of statistical certainty in that case, as the number of workers in the mining industry since nationalisation had been well known 203 204

86  Jonathan Morgan way of dealing with anxiety than litigation – pointing out that there was potential contradiction between doctors assuring those diagnosed with plaques that these were harmless and benign, and legislation classifying plaques as an actionable injury Of particular interest for our purposes was the point made repeatedly in the written submissions that even though the Bill was narrowly drafted to grant compensation for those suffering from asbestos-induced pleural plaques, it would set a worrying precedent. According to Axa Insurance, ‘there is great risk that once having taken this step for the conditions stated in the Bill the Scottish Government will find it impossible to resist calls for other types of condition to be similarly compensated’.208 But there is a danger here, as we suggested above, of confusing the political ‘precedent’ established by legislation with the true precedential force of a judicial decision. A common law principle necessarily applies to analogous cases, hence the difficulties of keeping the explosive, ‘exceptional’ Fairchild decision under control. But binding precedent simply does not operate in the science of legislation. A legislature is perfectly able to draw clear lines which, however arbitrary, cannot be questioned in court.209 We have already observed that there is considerable heterogeneity in the compensation of different kinds of misfortune. This has accumulated over many years of legislation. The untidiness and inevitable inconsistencies seem to be tolerated with equanimity by the legal system. It is not obvious why, against this chaotic background, the Scottish Parliament should uniquely find itself under irresistible pressure to universalise the generous treatment of pleural plaques victims.210 A cynic might dismiss the insurance industry’s arguments against the plaques legislation as so many straws clutched in self-interest. But academic and practising lawyers put forward similar arguments, without the taint of bias.211 Dr Martin Hogg of Edinburgh University gave both written and oral evidence in which he stressed the perils of legal inconsistency. His evidence repays careful reading,212 and space does not allow a full analysis here. A quotation from his oral testimony serves to convey the gist. It would be ‘an incomplete and rather unsatisfactory way of proceeding to simply pluck from the general population one category of people who have inhaled one type of substance’ and compensate them but not other, similarly deserving groups of victims.213

208   cf also DLA Piper Insurance Group submission, point 5.2.1: ‘If Parliament is prepared to overrule established legal principle so as to allow recovery in respect of the anxiety relating to three particular symptomless conditions, it is likely to be pressured to do so in relation to anxiety arising in other circumstances. Even if Parliament does not currently have any appetite for doing this, future Parliaments might, and are likely to come under repeated pressure’. 209   Repeatedly stressed of course by the Scottish Government in response to this line of criticism. See eg, oral evidence to Justice Committee, 9 September 2008, col 1099 (Minister for Community Safety). 210   cf the (inevitable) complaint that those with plaques in England and Wales are being ‘discriminated against’ (‘Can the Justice Secretary, even in his wildest dreams, think of any scheme dafter than the one under which pleural plaques victims in Scotland are compensated by the British taxpayer while English pleural plaques sufferers get nothing?’ Stephen Hepburn MP, Hansard, HC Deb, 9 February 2010, col 753–54). The British Government has tried to forestall arguments to extend its limited, ex gratia plaques scheme (see p 84 above): ‘The Government regard this as a unique situation and as not setting any precedent for any other circumstances where litigants may be disappointed’ (Jack Straw MP, Hansard, HC Deb, 25 February 2010, col 80WS). 211   See written submission of Andrew Smith QC. 212   See also M Hogg, ‘Re-establishing Orthodoxy in the realm of Causation’ (2007) Edinburgh Law Review 8; M Hogg, ‘Asbestos related conditions and the idea of damage in the law of delict’ (2008) 31 Scots Law Times 207. 213   Scots Parliament Justice Committee, Official Report, 2 September 2008, col 1067. See also col 1066: It would be difficult not to extend the proposed legislation to ‘inhalation of a number of substances – coal dust, silica dust, bauxite dust, beryllium, cotton dust and silica and iron mixtures, for example – [that] could produce symptomatic conditions’.

The English – and Scottish – Asbestos Saga  87 Such a concern with general principles is admirable. But as we have repeatedly argued in this essay, legislation can trim, compromise and split the difference in a way in which the common law simply cannot.214 Legislation of this sort should not be taken to lay down any general principles at all. Therefore, speaking generally, Dr Hogg’s admonition to the MSPs that they were the ‘guardians of the whole of the law’,215 and should intervene in the common law only exceptionally and ‘on a principled and sound basis’ seems questionable.216 The principles of the common law can be preserved by the traditional strict construction of legislation. However, as will be seen below, the eventual Act of the Scottish Parliament does have a rather unusual form of drafting, possibly leaving it open to some of Dr Hogg’s criticisms. The Justice Committee’s final Report on the Bill supported the Scottish Government’s proposals.217 After summarising the various criticisms that the proposals were arbitrarily limited in their scope,218 the Committee concluded, first, that the wording of the statute applied only to asbestos (which nobody had sought to deny),219 and secondly that ‘there are compelling grounds to legislate for pleural plaques and the other asbestos related conditions contained within the Bill’220 – although without explaining what these ‘compelling grounds’ were and why they did not extend to inhalation of the ‘coal dust, silica dust, bauxite dust, beryllium, cotton dust and silica and iron mixtures’ mentioned by Dr Martin Hogg.221 Finally, but significantly, on the financial impact of the Bill the Justice Committee concluded that the estimates of the Scottish Government may be too low while the evidence of the insurance industry had ‘probably significantly overestimated the costs’,222 and therefore called for ‘further consideration [of] the figures . . . presented in the Financial Memorandum’.223 The Bill proceeded to be debated on the floor of the Scottish Parliament three times.224 Here, few of the arguments raised against the Bill were even debated, in an atmosphere of cross-party consensus.225 The Conservative Party (alone) did vote against the final enactment of the Bill at Stage Three, but on fairly narrow grounds.226 While supporting the Bill in principle, the Conservatives declared that they could not vote for a measure when its financial implications remained shrouded in mystery. Despite the call made by the Justice Committee for better financial information, the Scottish Government had not been able to bring sufficient clarity to the situation and the Parliament was being ‘asked to sign a cheque that cannot be quantified’.227 As another Tory MSP put it, ‘there remains considerable doubt   On which see generally: JA Weir, ‘All or nothing’ (2004) 78 Tulane Law Review 511.   Scots Parliament Justice Committee, Official Report, 2 September 2008, col 1070.   James Lee (n 148), while praising Dr Hogg, wonders whether ‘arguments for coherence when material [legislative] alterations to the common law are contemplated [are] merely to be viewed as the Canute-like rantings of judicial supremacists’. 217   19th Report, 2008 (Session 3), SP Paper 158. See para 153, 155. 218   ibid 105–10. 219   ibid 111. 220   ibid 112. 221   ibid 107. cf also Lord Brown in Sienkiewicz (n 4) [186], doubting justification for the ‘special treatment’ of mesothelioma. 222   19th Report, 2008 (Session 3), SP Paper 158, 136. 223   ibid 137, see also 154. 224   Stage 1, 5 November 2008; Stage 2, 2 December 2008; Stage 3, 11 March 2009. 225   The nearest thing to dissent at Stage 1 was the Conservative John Lamont’s confession to ‘nervousness’ at the precedent being set by the Bill, but only when ‘playing Devil’s Advocate’: Scots Parliament Official Report, 5 November 2008, col 12051. 226   Scots Parliament Official Report, 11 March 2009, col 15360–2. 227   Scots Parliament Official Report, 11 March 2009, col 15631 (Bill Aitken MSP). 214 215 216

88  Jonathan Morgan about potential liabilities and costs to both the private and the public purse’.228 The reply of the minister was candid, surprising and revealing: The Conservatives’ argument seemed to be that legislation should not proceed unless we can have certainty about what the financial cost will be. If that were the test for legislation, we would not have much legislation, because it is simply not possible to predict with precision what the costs will be . . . If perfection in future estimated costs were a sine qua non of legislation, there would not be any negligence legislation, any compensation for personal injury or any right of recourse to the courts.229

Hence we see that there are serious limits on a well-resourced government’s ability to make accurate predictions about the effects of legal reforms. It may be that this applies a fortori to judicial policy-making. On the other hand, one might argue that judges are not much worse off than politicians when it comes to weighing the scant evidence that might exist on such points.230 We certainly should not assume that legislatures always can and will exhaustively gather empirical evidence when making policy.231 The Scottish minister’s reply makes plain that government legislation is often, perhaps typically, not based on any hard evidence.232 Lord Emslie has accepted that the Scottish Parliament was nonetheless entitled to pass the Bill.233 There was ‘inherent difficulty of producing reliable figures’ in such matters,234 of which the Parliament had been fully aware. While the outcome of the Scottish Government’s attempts at clarification was ‘by no means ideal’, it ‘fell to the Parliament to judge whether the legislation should nevertheless be allowed to proceed’.235 There was, Lord Emslie held, nothing ‘irrational or outrageous about the Parliament’s ultimate reluctance to prolong what had proved to be a fruitless quest for precision in the estimated number and value of future damages claims’236 – particularly not when, by contrast, ‘the cost implications of actionability were not apparently in issue at any stage of the Rothwell test cases’.237 Lord Emslie’s judgment shows judicial awareness and (perhaps inevitably) acceptance of the realities constraining law reform. But this should not relieve the Government of the need to gather empirical evidence, or of the courts to require parties to put essentially empirical arguments to proof.238   Scots Parliament Official Report, 11 March 2009, col 15648 (John Lamont MSP).   Scots Parliament Official Report, 11 March 2009, col 15653. 230   cf J Morgan, ‘Policy reasoning in tort law: The courts, the Law Commission and the critics’ (2009) 125 LQR 215. See also now the Law Commission’s lament concerning the absence of government figures on levels of compensation payouts in Report No 322, ‘Administrative Redress: Public Bodies and the Citizen’ (2010). 231   The Explanatory and Financial Memorandum for the Damages (Asbestos-related Conditions) (Northern Ireland) Bill 2010 is vaguer even than the Scots equivalent: para 18–21. 232   cf ‘Professor Jonathan Shepherd says policy should be scientifically researched’, Mark Henderson, The Times, 20 July 2009 (criticising failure of educational and criminal justice systems to adopt the evidence-based research culture of medicine). 233   Petition by Axa General Insurance [2010] CSOH 2. (Now on appeal to the Supreme Court of the United Kingdom: Axa General Insurance v Lord Advocate (2011).) 234   ibid [219]. 235   ibid [220]. 236   ibid [246]. 237   ibid [220]. 238   cf Buxton LJ, refusing counsel’s invitation to speculate about the likely effect of liability (argument that it would lead to ‘defensive behaviour’) in Perrett v Collins [1998] 2 Lloyd’s Rep 255, 277: ‘wide claims have been made as to the effect of the Court’s decision, particularly on industry or insurance practice. In my respectful view, the Court should be very cautious before reaching or acting on any conclusions that are not argued before it in the way in which technical issues are usually approached, with the assistance of expert evidence’. 228 229

The English – and Scottish – Asbestos Saga  89 Notwithstanding the array of criticism, the Damages (Asbestos-related Conditions) (Scotland) Act received Royal Assent on 17 April 2009.239 Its provisions are somewhat curious.240 Section 1 states: (1) Asbestos-related pleural plaques are a personal injury which is not negligible. (2) Accordingly, they constitute actionable harm for the purposes of an action of damages for personal injuries. (3) Any rule of law the effect of which is that asbestos-related pleural plaques do not constitute actionable harm ceases to apply to the extent it has that effect. (4) But nothing in this section otherwise affects any enactment or rule of law which determines whether and in what circumstances a person may be liable in damages in respect of personal injuries.

Arguably, sections 1(1) and 1(2) are superfluous. Section 1(3) is the operative provision. The preceding subsections are apparently an attempt to reverse not just the result of Rothwell, but also its reasoning.241 The House of Lords held that pleural plaques are not actionable precisely because they are legally ‘negligible’: de minimis non curat lex. The legislation now says that (in Scotland at least), their Lordships were wrong. Does this mean that they applied too high a threshold for actionability? By attempting to give a principled basis for section 1(3), the Scottish Parliament runs the risk of throwing the common law definition of actionable damage into doubt.242 Therefore, the superfluous sections 1(1) and (2) necessitate a further subsection, 1(4), ostensibly to preserve the common law rules. Therefore, certain criticisms of the draft legislation arguably retain their bite. The Forum of Insurance Lawyers identified a lack of ‘transparency’ in the proposed legislation. It required the courts to treat plaques as actionable damage when they are not, ‘in effect to call black “white” ’. The result of such ‘condition specific legislation’ would ‘inevitably’ be ‘inconsistency of our private law’.243 Dr Martin Hogg warned that the legislation would be in clear breach of established Scottish principles of the nature of actionable injury. It would create a legal island within an otherwise uniform and coherent law of damages. . . . Allowing the 239   The insurers’ legal challenge to the Act failed: Petition by Axa General Insurance for Judicial Review [2010] CSOH 2. Lord Emslie held that the human rights in neither Art 6 ECHR nor the first Article of the First Protocol to the ECHR were even engaged. Even if they had been engaged, the legislation was amply justified by considerations of the public interest. On the residual common law challenge, Lord Emslie held that there was ‘nothing intrinsically irrational or outrageous about a legislature deciding to modify or overrule a judicial decision at any level. To make and shape law is a primary function of any legislature, especially where existing rules and principles are perceived as unsatisfactory or unfair’ [236]. The petitioners’ contentions did not come close to establishing the extremes of bad faith or absurdity required for judicial review of primary legislation [229]. Lord Glennie had previously refused ‘interim interdict to prevent the Scottish Ministers from bringing the Act into force before [the petitioners’] challenge to its validity can be heard’: Petition by Axa General Insurance [2009] CSOH 57. 240   Compare cl 1 of the Damages (Asbestos-related Conditions) Bill (Northern Ireland) 2010. 241   The Explanatory Notes to the Bill as introduced (23 June 2008) state that s 1 ‘addresses the central reasoning of the Judgment in [Rothwell]’ (emphasis added). 242   cf Lord Emslie’s decision in Petition by Axa General Insurance [2010] CSOH 2, that the Parliament had simply taken a different view on a matter of degree – ‘on which courts had come to different conclusions since at least 1984’ – as it had been entitled to do, taking into account the social and economic effects of plaques, [234]. ‘sections 1 and 2 of the Act can no more be described as distorting medical fact, or as subverting legal principle, than section 8(6) of the Prevention of Harassment Act 1997 which, again in a limited context, made free-standing anxiety actionable at a time when the common law was moving in the opposite direction’, [235]. ‘In my opinion, a major flaw in the petitioners’ argument is their rigid insistence that pleural plaques claimants “. . . have suffered no harm”. The Scottish Parliament has taken a different view on that matter of mixed fact and law . . .’ [236]. 243   Written evidence to Justice Committee, 8.1. But cf Petition by Axa General Insurance [2010] CSOH 2, [234]: the Parliament ‘was not, as the petitioners insisted, just seeking to turn established fact on its head and declare black to be white’ (Lord Emslie).

90  Jonathan Morgan uninjured to claim damages, by turning their non-injuries into injuries, . . . [would] set a dangerous precedent which might be utilised by other categories of person not currently entitled to claim under the existing rules on damages for worry and upset about their asymptomatic conditions. The floodgates once opened are hard to close again.244

He concluded that the Scottish Parliament should not ‘undermine the fundamental concept of an actionable injury in Scots law merely for the sake of the public appearance of generosity’.245 It is suggested that given the drafting of sections 1(1)–1(2) of the Act these points have some merit, albeit they have not convinced the Court of Session.246 There can be no possible clash between legislation and common law principles (such as the definition of actionable damage in negligence) when the legislature simply creates a different rule for a certain category of case. A self-contained statutory exception leaves the common law untouched. As argued above, section 3 of the Compensation Act 2006 should be interpreted in this way. But the 2009 Act lays down a rule (pleural plaques are ‘actionable harm’) and also an explanation for the rule (plaques ‘are a personal injury which is not negligible’). This apparent attempt to reverse not just the outcome of Rothwell but also its reasoning is, we suggest, wholly misconceived. It creates legal uncertainty which might open the door for the legislation to be applied by analogy to cases in future where injuries which would not have satisfied the common law actionability threshold might arguably be ‘not negligible’ in comparison with pleural plaques. Only the explicit saving provision in section 1(4) curbs this possibility. This seems very convoluted: enacting a reasoned basis for the statute, but then explicitly denying effect to that reasoning for any comparable situation. Should there be any future legislative reversal of Rothwell in England and Wales we suggest that the ‘unreasoned’ drafting of the Compensation Act 2006 should be emulated. This will ensure that the legislation lays down a clear rule (compensation for pleural plaques), rather than a principle of wider, uncertain application.

VIII. Conclusion

As we observed in the introduction above, there is a voluminous and sophisticated literature on the problems of causation, damage and proof raised by the English and Scottish asbestos cases. We have not engaged with the literature in this essay because the courts and legislatures at the heart of our story have not engaged with it either. Such is the prerogative of the legislator. Statutes can lay down political compromises without having to declare their rationale. The oddity of Fairchild is that the judges apparently wished to create such an unprincipled, essentially legislative, exceptional liability to victims of mesothelioma. The House of Lords failed to engage with important debates about proof, and whether risk can amount to actionable damage, at the level of principle appropriate for a leading case. The problems were not resolved by the ‘principled’ reinterpretation of Fairchild in Barker, since the essentially arbitrary conditions for the Fairchild ‘principle’ were reaffirmed. It is now too late to go back.247 The problems of Fairchild and Barker are not that policy concerns   M Hogg, Written submission to Justice Committee, 5.   ibid 6. 246   Petition by Axa General Insurance [2010] CSOH 2. 247   cf discussion of Sienkiewicz at pp 65–66 above. 244 245

The English – and Scottish – Asbestos Saga  91 were considered, but on the contrary that they were not considered carefully enough. ‘Corrective justice’ does not (we suggest) provide the only way properly to analyse these questions. But even a ‘Legal Realist’ defender of policy reasoning can condemn courts attempting to act like legislators, seeking to lay down arbitrary limits on the future reach of their decisions. This is not how the common law works. Only exceptionally have the courts taken the course of designating their decisions as essentially unprincipled ones from which no analogies may properly be drawn. Nervous shock has been one notorious example, to which it seems we must now add Fairchild. This violates every precept of common law reasoning,248 and brings the judicial development of tort law into disrepute. The legislation at Westminster and Edinburgh, by contrast, does little lasting damage to the fabric of the common law. The beauty of limited legislation in the British tradition is that it does not seek to lay down any general principles. It is applicable to the situations enacted, but otherwise has no wider ‘gravitational’ effect.249 Political philosophers may lament legislative inconsistency, but this is to condemn the whole enterprise of politics and legislation in a world of ‘reasonable pluralism’,250 with ‘legislative integrity’ being as Utopian as it is unobtainable. The courts, however, do not have the option of such brutally casuistic reasoning, and it is to be hoped that in future the maxim that ‘hard cases make bad law’ is taken seriously. But equally, ‘policy’ reasons should be examined carefully and critically, with hard empirical data to back them up wherever possible. For legislators with the whole apparatus of the civil service at their disposal, the failure to collect such evidence is far less forgivable than for judges who can decide only on the evidence that parties put before them. The asbestos saga, which is far from over,251 stands as a cautionary tale for both judicial and legislative development of tort law.

248   Admittedly, JR Spencer once praised (ironically?) ‘that manly disregard for vulgar logic which makes the common law so much superior to other systems’ in ‘Motor-cars and the rule in Rylands v Fletcher’ [1983] CLJ 65, 70. 249   J Beatson refers to ‘legislative microsurgery’ (albeit dismissively): [1997] CLJ 291, 301. 250   cf J Rawls, Political Liberalism (New York, Columbia University Press, 1993). 251   The Supreme Court has just heard the insurers’ challenge to the Damages (Asbestos-related Conditions) (Scotland) Act 2009: Axa General Insurance Ltd v Lord Advocate (June 2011).

5 Inferring Cause in Fact and the Search for Legal ‘Truth’ RUSSELL BROWN

I. Introduction

In this chapter, I consider the role that inference-drawing can (and, I will argue, must) serve in a fact-finder’s determination of whether a plaintiff has proven cause-in-fact. This is a subject of peculiar relevance to Canadian torts jurisprudence. Since the Supreme Court of Canada’s 1990 decision in Snell v Farrell,1 findings of cause-in-fact in cases characterised by ‘gaps’ in the plaintiff ’s evidence have been said to rest upon whether a causal link between the defendant’s negligent conduct and the plaintiff ’s harm can be inferred. As I shall explain, this is both a legitimate and necessary standpoint for the Court to have adopted. Admittedly, the Court’s loose conception in Snell of the role of evidence in causal determinations has attracted criticism from Canadian legal academic commentators,2 bringing inference-drawing more generally into disrepute. While I will show that this criticism is too sweeping because it purports to impugn inference-drawing altogether, the critics are correct to emphasise the evidentiary role. A finding of cause-in-fact even when made by the drawing of an inference – requires an account which is oriented to whatever evidence (gaps notwithstanding) is before the fact-finder. A fairer criticism of inference-drawing as it has been articulated in Canadian law, then, might be that judicial preoccupation with theories of and tests for cause-in-fact3 has obscured the essence of the causal enquiry, which is properly directed to what is needed to prove cause-in-fact. By taking seriously that evidentiary role, I seek to achieve something of a rehabilitation of inference-drawing in Canadian tort law discourse. In this chapter, therefore, my case for inference-drawing in the causal enquiry, which first proceeds by identifying and answering the critics’ principal objections, will then be supplemented by discrete consideration of the place of evidence in that inferential process.   Snell v Farrell [1990] 2 SCR 311.   I canvass the criticisms below. 3   I have elsewhere conceptualised the Supreme Court of Canada’s array of definitions of cause-in-fact as being material contribution to harm (pronounced in Athey v Leonati [1996] 3 SCR 458) and material contribution to risk (pronounced in Hanke v Resurfice Corp [2007] 1 SCR 333). (R Brown, ‘Material Contribution’s Expanding Hegemony: Factual Causation after Hanke v Resurfice Corp.’ (2007) 45 Canadian Business Law Journal 432). 1 2

94  Russell Brown

II.  Snell and the Criticisms

The central objections among Canadian legal academics to inference-drawing in causal determinations arise from the Supreme Court of Canada’s pronouncement in Snell. Accordingly, it merits a brief review. The plaintiff, in her 70s by the time of trial, had undergone surgery to remove a cataract. During the surgery, the surgeon noticed a retrobulbar haemorrhage – bleeding behind her eyeball – which was a typical risk of such a procedure, and not the result of the surgeon’s negligence. The surgeon, however, elected to continue, and this was found by the trial judge to be unreasonable. The surgeon ought instead to have aborted the surgery in order to treat the haemorrhage. It was only after several months had passed, when the plaintiff ’s vitreous chamber had cleared, that her optic nerve could be observed to have atrophied to the point of permanent blindness. The obstacle which the law posed for the plaintiff in obtaining recovery from the negligent surgeon was its requirement that she prove cause-in-fact. The aetiology of the optic nerve atrophy was uncertain. It might have resulted from the retrobulbar haemorrhage, or from any of three conditions – diabetes, high blood pressure and glaucoma – which also afflicted the plaintiff. Neither of the party’s experts was able to opine on the cause of the atrophy. There was, therefore, a ‘gap’ in the evidence. For the Court, Sopinka J cited an excerpt from Lord Bridge’s speech in Wilsher v Essex Area Health Authority  4 in which, and with reference McGhee v National Coal Board,5 his Lordship said: The conclusion I draw . . . is that [McGhee] laid down no new principle of law whatever. On the contrary, it affirmed the principle that the onus of proving causation lies on the pursuer or plaintiff. Adopting a robust and pragmatic approach to the undisputed primary facts of the case, the majority concluded that it was a legitimate inference of fact that the defendant’s negligence had materially contributed to the pursuer’s injury.6

Sopinka J then added: I am of the opinion that the dissatisfaction with the traditional approach to causation stems to a large extent from its too rigid application by the courts in many cases. Causation need not be determined by scientific precision. It is, as stated by Lord Salmon in Alphacell Ltd. v Woodward, [1972] 2 All E.R. 497, at 490: [‘essentially a practical question of fact which can best be answered by ordinary common sense rather than abstract metaphysical theory’]. The legal or ultimate burden remains with the plaintiff, but in the absence of evidence to the contrary adduced by the defendant, an inference of causation may be drawn although positive or scientific proof of causation has not been adduced. . . .7

In Canadian law, then, a limited scientific understanding of the aetiology of a plaintiff ’s condition does not, in and of itself, pose an insurmountable hurdle to proving cause-infact. The gap created by what we do not know can be bridged by an inference of cause-infact, drawn from what we do know, where such an inference is supported by ‘common sense’ and by a ‘robust’ and ‘pragmatic’ treatment of evidence.   [1988] 1 All ER 871 (HL).   [1972] 3 All ER 1008 (HL). There, the employer’s failure to provide washing facilities for employees was one of several potential sources of the cause of the plaintiff ’s dermatitis. 6   Snell v Farrell [1990] 2 SCR 311, 324. (Emphasis added). 7   ibid 328, 330. 4 5

Inferring Cause in Fact and the Search for Legal ‘Truth’  95 Of course, the sort of gap – or, more precisely, the sort of factual uncertainty – in Snell was nothing new. The House of Lords had confronted the same problem just three years earlier in Wilsher.8 There, the plaintiff was a prematurely born infant who was negligently oxidised through his umbilical vein instead of an artery, with the result that he was oversupplied with oxygen. He was then negligently monitored, which again exposed him to unsafe oxygen levels. While he subsequently developed retrolental fibroplasia (resulting in blindness), the over-oxidisation was only one possible cause. He may have suffered from as many as four other conditions, any of which on their own might have caused his condition. Although (speaking for the House of Lords) Lord Bridge declined to infer a causal link between the negligent treatment and the plaintiff ’s blindness, he (like Sopinka J) saw McGhee as stating a general rule permitting the drawing of ‘common sense’ inferences ‘where the doctors cannot identify the process of causation scientifically’.9 Despite having come to differing conclusions on their respective facts, then, both Snell and Wilsher affirmed that it is open to a fact-finder, when confronted with negligent conduct and harm falling within the scope of the risk generated by such conduct, to infer a causal link between the two. While this has since been rejected at the House of Lords,10 Justice Sopinka’s statement in Snell remains authoritative in Canadian law. Canadian academic commentary upon Snell has been almost universally critical of the notion that an inference of cause-in-fact can be drawn in the absence of scientifically demonstrated causal linkage.11 While Sopinka J’s reliance in Snell upon evidentiary ‘robust[ness]’ and his echoing of Alphacell’s obscure reference to ‘common sense’ have been particular targets for scorn as supplying fact-finders with a meaningless fact-finding vocabulary,12 other objections have raised fundamental concerns going to the very essence of fact-­ finding. Put generally, their criticism is that allowing fact-finders to ‘infer’ cause-in-fact excuses them from offering clear reasoning for finding a causal link between a defendant’s negligence and a plaintiff ’s harm without evidence proving such a link to a probable standard. The charges levelled by Lewis Klar in his leading Canadian textbook Tort Law13 are generally representative although, as Lara Khoury has pointed out, they are ‘more assertive’14 than most:

 ibid.   Wilsher v Essex Area Health Authority [1988] 1 All ER 871, 880 (HL). There has been substantial debate about the meaning to be ascribed to McGhee (and particularly to Lord Reid’s speech – see eg, the mutually opposing views of J Stapleton, ‘Lords a’Leaping Evidentiary Gaps’ (2002) 10 Tort Law Journal 276, 286–87, and A Beever, Rediscovering the Law of Negligence (Oxford, Hart Publishing, 2007) 466–67). It is unnecessary for me to resolve this here, as I take Sopinka J’s reasons in Snell as my reference point. 10   Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32. 11  V Black, ‘The Transformation of Causation in the Supreme Court: Dilution and “Policyization” ’ in T Archibald and M Cochrane (eds), Annual Review of Civil Litigation 2002 (Toronto, Carswell, 2003) 187, 195; D Cheifetz, ‘The Snell Inference and Material Contribution: Defining the Undefinable and Hunting the Causative Snark’ (2005) 30 The Advocates’ Quarterly 1, 49–54; JG Fleming, ‘Probabilistic Causation in Tort Law’ (1989) 68 Canadian Bar Review 661, 670; L Klar, Tort Law, 3rd edn (Toronto, Thomson Carswell, 2003) 402–03; and SN Pincus, ‘Progress on the Causal Chain Gang: Some Approaches to Causation in Tort Law and Steps Toward a Linguistic Analysis’ (1986) 24 Osgoode Hall Law Journal 961, 983. 12   ‘‘Robust and pragmatic’ are fudge words of the first degree’. (V Black, ‘Review of L Koury, Uncertain Causation in Medical Liability’ (2008) 47 Canadian Business Law Journal 145, 151). See also M MacCrimmon, ‘What is ‘Common’ about Common Sense?: Cautionary Tales for Travelers Crossing Disciplinary Boundaries’ (2001) 22 Cardozo Law Review 1433, 1435: ‘Although frequently relied upon, common sense is seldom defined, nor is reliance on common sense justified’. 13  Klar, Tort Law (n 11). 14   L Khoury, Uncertain Causation in Medical Liability (Portland, Hart Publishing, 2006) 166. 8 9

96  Russell Brown The effect of Snell v Farrell on proving causation in cases where the scientific and expert evidence cannot establish a probable connection between a defendant’s negligence and a plaintiff ’s injury has been significant. To allow an inference of cause to be drawn even where there is no scientific evidence of a probable connection between negligence and injury is in effect to accept the essential principle of McGhee. While this approach may produce a pragmatic solution to a plaintiff ’s dilemma in difficult causation cases, it does depart from the traditional ‘but for’ test and the balance of probability standard.15

I propose to address two general objections that are encapsulated here. The first is that ‘scientific evidence’ is the law’s proper reference point by which cause-in-fact is or is not to be found. The implication here is that such evidence is inherently reliable, or at least more reliable than inference-drawing.16 My answer will show that scientific fact-finding, while of potential assistance to legal fact-finders, must be treated with caution because it fails to account for the cognitive processes that fact-finders necessarily impose upon the evidence. Such processes make inference an unavoidable element of any causal enquiry. A second and related criticism which Klar advances is that no fact-finder can legitimately infer cause-in-fact without the benefit of supportive scientific evidence. Here again, I will suggest in this chapter that the immanent role of inference within fact-finding answers this criticism. I will also cite various criteria that have been discerned by, among others, psychologists and epidemiologists for making reliable inferences of causal association. To a degree, my argument in response to these criticisms of inference-drawing will merely affirm current Canadian tort law doctrine, which allows fact-finders substantial discretion when confronted with factual uncertainty. I assume that this is not, in and of itself, a problem. As Richard Wright has observed, ‘judges and juries . . . consistently have demonstrated an ability to make intuitively plausible factual causal determinations’.17 While the doctrine might be sound, however, my task here cannot be merely to reiterate judicial practice. As I have already mentioned, the critics correctly point out that Snell plays fast and loose with the requirement of supporting evidence (as suggested by its reliance upon Wilsher’s vacuous references to a ‘robust’ and ‘pragmatic’ treatment of the evidence). Hence my coupling in this chapter of consideration of inference-drawing with a concern for restoring the privileged role of evidence in that inferential process.

III. The Role of Scientific Fact-finding

The first criticism at issue here, then, is that Snell’s allowance for inference-drawing in determining cause-in-fact disregards the ostensible requirement for ‘scientific’ evidence that the defendant’s negligence caused the plaintiff ’s injury. There are two reasons for why this criticism is misguided: first, scientific fact-finding is, simply put, an ill-fitting substitute for legal fact-finding. Secondly, scientific fact-finding’s statistical methodology goes only to 15  Klar, Tort Law (n 11) 402–03. ‘[T]he essential principle of McGhee’ being referred to here is Lord Wilberforce’s reversal in McGhee of the burden of proof, by which the defendant was required to prove that his or her negligence did not cause the plaintiff ’s injury. (ibid 397–402. See also Khoury, Uncertain Causation (n 14) 166). 16   This was also the view of Lord Rodger of Earlsferry in Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32, 110. 17   RW Wright, ‘Causation, Responsibility, Risk Probability, Naked Statistics and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (1988) 73 Iowa Law Review 1001, 1018.

Inferring Cause in Fact and the Search for Legal ‘Truth’  97 the possibility of a causal generalisation (that is, the possibility of a cause existing in general) and not to the possibility that such causal generalisation was instantiated in the event under consideration. That is, scientific fact-finding operates at a general level, shedding no insight into the specific causal linkages that might or might not exist between the defendant’s negligence and the plaintiff ’s injury. The first reason implicates, in part, the different standards of proof governing scientific and legal fact-finding. Assuming (for the sake of demonstrating those different standards) the statistical reducibility of burdens of proof, the (civil) legal burden of proof is conventionally understood to be a probability measure of >0.50, signifying that the proposition, while less than certain to be absolutely true, is more likely than not to be true.18 Statistically significant conclusions in science, however, normally fall within a confidence interval such that a scientist would refrain from finding a causal association between a potential cause and effect unless the connection would not occur due to chance or another cause more than 5 per cent of the time.19 The difference is a range of probability measuring ≤0.449, signifying causal propositions whose likelihood of truth would not satisfy a scientist, but within which the plaintiff still ought to succeed. Here I am agreeing with Allen Beever, who has observed that the implications of this gap between legal and scientific fact-finding thresholds mean that a legal fact-finding threshold – which conditions a causal determination upon scientifically established factual linkage – creates a one-sided enquiry favouring the defendant in cases of scientific uncertainty.20 Such dependence of the legal outcome in a tort action upon a scientific enquiry is, quite literally, unjust. These distinct legal and scientific fact-finding thresholds are not as significant, however, as are the distinct methodologies that underlie them. In a situation where we know what has happened but do not know why it has happened, science responds by repeating experiments until the event can be replicated. The emphasis here, then, is upon continual refinement of what we know by eliminating anomalies. Until we reach scientific fact-finding’s confidence interval, causal determinations must always remain contingent or provisional.21 Legal fact-finders, however, do not replicate events in this way. For one thing, as a practical matter law needs finality, not an ongoing process of ongoing refinement of knowledge.22 Indeed, such further investigation may be futile. The very notion that evidence might be ‘complete’ is only a ‘scientific ideal’.23 Because factual uncertainty derives from the unavailability of evidence which by necessity we do not know, nobody can predict what would be proven by the unavailable evidence, were it available. There are, furthermore, simply too many variables, many of them subjective, for which legal fact-finders would have to account. I refer here to trial phenomena such as the manner or demeanour of witnesses, including that of expert witnesses. Assessment of what Laurence Tribe conceived as such 18   RJ Allen and MS Pardo, ‘The Problematic Value of Mathematical Models of Evidence’ (2007) 36 Journal of Legal Studies 107, 111. 19  R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing, 1999) 105; and MA Berger and LM Solan, ‘The Uneasy Relationship between Science and Law: An Essay and Introduction’ (2008) 73 Brooklyn Law Review 847, 852. See also Sienkiewicz v Greif [2011] UKSC 10, [2011] 2 WLR 523 [9]. 20  Beever, Rediscovering the Law of Negligence (n 9) 471. 21   Here I am agreeing with C Miller, ‘Causation in Personal Injury: Legal or Epidemiological Common Sense?’ (2006) 26 Legal Studies 544, 547. 22   A Stein, ‘The Refoundation of Evidence Law’ (1996) 9 Canadian Journal of Law & Jurisprudence 279, 286; and Miller, ‘Causation in Personal Injury’ (n 21) 547. 23   L Dufraimont, ‘Evidence Law and the Jury: A Reassessment’ (2008) 53 McGill Law Journal 199, 206.

98  Russell Brown ‘soft variables’24 entails complex mental activities which resist statistical reduction, thereby undermining science’s claim to a normatively superior fact-finding methodology. Hence my earlier statement that scientific fact-finding and legal fact-finding are ill-­fitting mutual substitutes. For the former to stand in for the latter, it would have to make deductive sense of those ‘soft variables’, with the necessary and sufficient conditions set out in advance. (For example, if the putative expert bites her nails and smiles more than once while describing her credentials, she is uncertain about whether she is qualified to form the opinion which she is about to offer). This is, however, impossible, because the cognitive routes by which such trial phenomena are processed cannot be definitively articulated.25 Legal fact-finding, then, is an instance of what Martha Nussbaum and Hilary Putman have observed as a general phenomenon: [M]ental states are not only compositionally plastic but also computationally plastic, that is, [there are] reasons to believe that physically possible creatures have an indefinite number of different ‘programs’, and that the hypothesis that there are necessary and sufficient conditions for the presence of such a belief in computational or computational-cum-physical, terms is unrealistic in just the way that the theory that there is a necessary and sufficient condition for the presence of a table stateable in phenomenalist terms is unrealistic; such a condition would be infinitely long, and not constructed according to any effective rule.26

Scientific fact-finding’s statistical methodology,27 by which statistically significant correlations or associations are sought to be derived between outcomes and variable exposures, such as the defendant’s creation of risk, furnishes another response to the criticism that inferring cause-in-fact disregards a supposed need for scientific evidence of cause-in-fact. Although legal fact-finders typically do not accept statistical analysis as determinative of the underlying proposition,28 there is without a doubt an appealing quality in statistical ‘evidence’. As Tribe observed, it suggests ‘objectivity and precision’.29 In fact, however, mathematical certainty is itself uncertain,30 as Jonathan Cohen’s ‘gatecrasher paradox’ demonstrates. Cohen posits that where a bare majority – 501 out of 1000 – spectators gatecrashed a rodeo, a statistical analysis would lead to the absurd conclusion that the impresario ought to be able to recover from any and all of the 1000 spectators, solely on the probability that they gatecrashed.31 There is another and more widely recognised normative difficulty with adopting a purely statistical methodology to determining cause-in-fact, going to the distinction between causal generalisations and specific instances of cause-in-fact. Statistical evidence of causal generalisations is seen as detracting from the requirement that we determine whether this

24   LH Tribe, ‘Trial by Mathematics: Precision and Ritual in the Legal Process’ (1971) 84 Harvard Law Review 1329, 1393. 25   RJ Allen, ‘Factual Ambiguity and a Theory of Evidence’ (1993) 88 Northwestern University Law Review 604, 626–27. 26   MC Nussbaum and H Putnam, ‘Changing Aristotle’s Mind’ in MC Nussbaum and A Oksenberg Rorty (eds), Essays on Aristotle’s De Anima (Oxford, Oxford University Press, 1995) 27, 48–49. (Emphasis added). 27   TA Brennan and RF Carter, ‘Legal and Scientific Probability of Causation of Cancer and other Environmental Diseases in Individuals’ (1985) 10 Journal of Health Politics, Policy & Law 33, 38–39. 28   Wright, ‘Pruning the Bramble Bush’ (n 17) 1050–51. See, for a recent and nuanced instance of judicial reluctance to be bound by statistical evidence, Duncalf v Capital Health Authority, 2009 ABQB 80. 29   Tribe, ‘Trial by Mathematics’ (n 24) 1331. 30   Randolph N Jonakait, ‘Stories, Forensic Science, and Improved Verdicts’ (1991) 13 Cardozo Law Review 343, 350. 31   LJ Cohen, The Probable and the Provable (Oxford, Clarendon Press, 1977) 75.

Inferring Cause in Fact and the Search for Legal ‘Truth’  99 defendant caused this injury to this plaintiff.32 The concern here is obviously far from novel, and has been stated by, inter alia, Glanville Williams,33 Wright34 and Khoury35 to the effect that causal determinations in law presuppose a normative ideal that we attempt to discover what probably happened (which we deem to be what actually happened) during a specific, historical interaction between the plaintiff and the defendant. I suggest, however, that this problem runs more deeply than past criticisms within tort law discourse has allowed. That is, statistical evidence is deficient from the standpoint of a legal fact-finder not just as evidence of specific instances of harm, but also as evidence of causal generalisations established by estimating general frequencies. Consider again Cohen’s gatecrasher paradox. We know that 501 spectators gatecrashed, and 499 did not. But this can be modelled in different ways by emphasising different characteristics among the spectators, which in turn generate different outcomes when we ask whether a particular spectator gatecrashed or not. We might, for example, distinguish between those with and without religious faith, on the (admittedly dubious) assumption that the former are more likely to be gatepayers; or between those who have driving-related convictions and those who do not, on the assumption that the former have less respect for the law and therefore are more likely to be gatecrashers. What these subsets of spectators demonstrate is that it is insufficient to know how many spectators there are and how many gatecrashed; we need to know other circumstances that are relevant to the statistical probability that a particular spectator gatecrashed. This problem – typically conceptualised as one of ‘reference classes’36 – implicates causein-fact. In a case like Wilsher, science could generate some provisional or conditional probability of going blind without the defendants’ negligence, as well as contrasting probability of going blind with the defendants’ negligence. From there, science could proceed to generate an equation showing the probability of the causal generalisation – that is, the statistical probability that the blindness was caused by the defendant’s negligence. I have already referred to the criticism that, being a mere probability, this equation is unhelpful in determining specific causation. My point here, however, is that the probability is of questionable value, even as evidence of a causal generalisation, unless we know more about the reference class from which it was derived. Reference classes can be manipulated, after all, to encompass a class as narrow as someone of the same age and medical history as the plaintiff in Wilsher, or as broad as all patients who have ever developed retrolental fibroplasia, or conforming to some midpoint such as subsets of patients (based upon age and gender) who have developed retrolental fibroplasia. 32   As Lord Phillips said in Sienkiewicz v Greif [2011] UKSC 10, [2011] 2 WLR 523, [80]: Epidemiology is the study of the occurrence and distribution of events (such as disease) over human populations. It seeks to determine whether statistical associations between these events and supposed determinants can be demonstrated. Whether those associations if proved demonstrate an underlying biological causal relationship is a further and different question from the question of statistical association on which epidemiology is initially engaged. See also the judgments of Lord Rodger [153]; of Lady Hale [170]; of Lord Mance [190]; and of Lord Kerr [205]. 33   Glanville Williams, ‘Causation in the Law’ (1961) Cambridge Law Journal 62, 66: ‘[T]he scientist is concerned with causal generalisations. But in historical and legal statements this notion of generalisation and reproducibility hardly figures at all’. 34   R Wright, ‘Causation in Tort Law’ (1985) 73 California Law Review 1735, 1822–23. 35  Khoury, Uncertain Causation (n 14) 49–50. 36   JM Keynes, A Treatise on Probability (London, Macmillan, 1921) 94; D Papineau, For Science in the Social Sciences (New York, Palgrave MacMillan, 1978); Wright, ‘Pruning the Bramble Bush’ (n 17) 1046–47; and P Roberts, ‘From Theory into Practice: Introducing the Reference Class Problem’ (2007) 11 International Journal of Evidence & Proof 243.

100  Russell Brown The probative value of statistical evidence, merely as statistical evidence, thus depends upon the assumptions that the data collector made about the group of persons – that is, the reference class – which was selected as the overall population from which probabilities were extracted. Because the probability assessments will vary with variances in those assumptions, the ostensibly objective and rigorous quality of probability assessments is suspect. It might be objected that, while different reference classes yield different probabilities, it is not the case that all reference classes will be seen as equal by the fact-finder, or that parties cannot have a serious debate about whether a given reference class yields a result that is more or less useful than a competing reference class in a given case. In this way, the argument might go, the number of variables can be controlled, and results can be achieved that generate (relatively) more precise connections between causes and effects. In other words, we can account for the reference class problem by assessing the reference classes employed in given studies, comparing them, and preferring the study or studies whose reference classes resemble most closely resemble the circumstances of the plaintiff ’s own claim. This is essentially the point underlying the important distinction drawn in Philip Dawid’s contribution to this volume between experimental studies and observational studies.37 The former accounts for manipulated inputs in clinical trials, thereby also accounting for the reference class problem. The latter, however, does not. Because those studies are merely observational (that is, they adopt the standpoint of an observer), we do not know the inputs, and therefore we do not know if we are comparing true comparables. Where those studies are considered for evidence of cause-in-fact, then, the reference class problem subsists. A distinction between experimental and observational studies does not, however – at least from the standpoint of a legal fact-finder – solve the reference class problem. Unless the fact-finder hears directly from the person who was actually responsible for conducting the study in question – that is, unless that person is able to testify at the trial to the selected reference class(es) – the reference class problem persists, irrespective of whether the study was experimental or observational. This is because information about the reference class adopted in an experimental study that has been cited (but was not conducted) by the expert witness will be hearsay. From the legal fact-finder’s standpoint, then, all studies are observational. In sum, the reference class problem goes to the reliability of the probability assessment: the more or less closely the reference class conforms to the plaintiff ’s peculiar circumstances, the more or less reliable will be the emergent probability assessments. The risk, of course, is that a broad reference class is more likely than a narrow reference class to lead to false causality. This is obviously relevant to statistically expressed causal generalisations derived from epidemiological research. For example, we might want to determine whether the probability of being overweight correlates to having taken a particular drug by finding at random 100 users and 100 non-users. Assume we find there are vastly more overweight users than nonusers. From that, one might conclude that there is a causal correlation between being a user and being overweight. Unless, however, we break down the reference class of ‘users’ (and ‘non-users’) by accounting for age, diet, health (or other potential causal factors such as ethnicity and gender), our chosen reference class leads us into false causality. The failure of researchers to account for the overly general reference class could thus lead a fact-finder in a legal action against the drug manufacturer to misidentify the cause by failing to distinguish a true causal factor from what might well be a spurious covariation.   A Philip Dawid, ‘The Role of Scientific and Statistical Evidence in Assessing Causality’ in ch 7 of this volume.

37

Inferring Cause in Fact and the Search for Legal ‘Truth’  101 My criticism of the use of statistics for determining issues of cause-in-fact raises the question of whether the probabilistic data that scientists can offer in the face of scientific uncertainty should be considered by fact-finders in tort actions for any purpose. At first glance, the disparity to which I have referred between probability theory and human reasoning when confronted with uncertainty suggests that statistics might properly be excluded altogether. Indeed, several evidence theorists have advanced arguments to this effect.38 Others would allow the inclusion of statistics as part of the matrix of evidential sources.39 While I do not purport here to bridge the divide definitively, it seems that statistics may furnish a legitimate basis for confidence (or trepidation) in a fact-finder’s assessment of the competing putative explanations of the evidence relating to the event in question. In Snell, for example, were statistics available to the court to the effect that in 80 per cent of cases involving untreated retrobulbar haemorrhage the optic nerve atrophies to the point of blindness, they might have affirmed for the fact-finder the correctness of an inference that she was already prepared to draw that the surgeon’s negligence caused the patient’s blindness. Conversely, a statistical probability of 20 per cent might force the factfinder to scrutinise more carefully the reason or reasons which had inclined her towards drawing such an inference. Note that statistics are not themselves the basis for the initial factual determination. And, in the converse scenario, while a reconsideration of that determination may be prompted by the statistical data, such reconsideration is then properly directed solely to the evidence. In other words, statistics might influence the fact-finder’s confidence in her determination, but not the determination itself. As such, they serve an affirming or cautionary function, and indeed may have the salutary effect of forcing the fact-finder to express the reasons for drawing (or not drawing) an inference.40

IV. The Role of Inference

I should acknowledge that the reference class problem is not unique to statistical expression of causal probabilities. All evidence has a probabilistic quality,41 because any factual generalisation implies a reference class. The problem is more pronounced, however, where statistics are being used in isolation from a fact-finder’s own reasoning. Unlike instances in which other forms of evidence – whether in documentary, viva voce or physical form – are considered, any interpretation to be drawn from statistics is entirely dependent upon the 38   See eg, CR Callen, ‘Kicking Rocks with Dr. Johnson: A Comment on Professor Allen’s Theory’ (1991) 13 Cardozo Law Review 423; and Tribe, ‘Trial by Mathematics’ (n 24) 1350, 1365. 39   See eg, Allen and Pardo, ‘Mathematical Models’ (n 18) 136–37; Roberts, ‘From Theory into Practice’ (n 36) 251–52; A Stein, ‘An Essay on Uncertainty and Fact-Finding in Civil Litigation, With Special Reference to Contract Cases’ (1998) 48 University of Toronto Law Journal 299, 301; and P Tillers, ‘Mapping Inferential Domains’ (1986) 66 Boston University Law Review 883, 888–89. And, while Robert Rhee has also mistakenly been taken as having argued for exclusion (see MS Pardo, ‘Reference Classes and Legal Evidence’ (2007) 11 International Journal of Evidence & Proof 255, 257), he has expressly stated that statistics are not a question of ‘exclusion, but rather one of appropriate weight’. (See eg, RJ Rhee, ‘Probability, Policy and the Problem of Reference Class’ (2007) 11 International Journal of Evidence & Proof 286, 291). 40   This is not the same role for statistics that was described by Lady Hale in Sienkiewicz v Greif [2011] UK SC 10, [2011] 2 WLR 523[172]. Instead of confining statistics to an affirming or cautionary function, Lady Hale saw them as being no less authoritative than what, on the evidence, the fact-finder thinks probable. 41   United States v Shonubi, 895 F Supp 460 (EDNY 1995) 514.

102  Russell Brown reference class. Statistics purport to speak entirely for themselves. That is, their significance is ostensibly to be found in their own content, irrespective of what our own thinking minds tell us. This is the upshot of the problem to which Wright averred in observing that probabilities provide (at most) evidence supporting the applicability of particular causal generalisations, but ‘do not in themselves indicate which of the possibly applicable causal generalizations actually applies to the particular concrete occurrence’.42 While, therefore (and allowing for the reference class problem), a statistical probability may speak for itself as evidence of a causal generalisation, it says nothing about a specific causal event, because ‘[n]o datum or object has an inferential value standing alone’.43 It remains, as Tribe put it, to be ‘tranform[ed]’ from evidence about the general to evidence about the specific.44 Understood in this way, statistical probabilities are not dissimilar to Tribe’s soft variables such as an expert’s charming turn-of-phrase or his smirks or sideways glances. That is, they are something to which a fact-finder must ascribe meaning through interpretation, and whose probative value depends upon the force by which it supports a particular inference. Indeed, this point is universally applicable – not just to smirks, sideways glances or statistics – but to evidence generally, in all its forms. ‘Evidence’, including circumstantial evidence of temporal proximity between negligent medical treatment and a patient’s blindness, is on its own underdeterminative of the significance which a fact-finder might ascribe to it.45 Such significance is ascribed by a process of inference, by which information furnished through a witness or a document or a physical exhibit become evidence through arguments and the fact-finder’s consideration of what such things might mean. If I am right, then inference-drawing is an unavoidable, indeed immanent characteristic of the causal enquiry. Fact-finding always involves choices between what (or what not) to infer from particular information. I infer (or do not infer) from newspaper reports that the sun shone in Honolulu yesterday that it did. I infer (or do not infer) from history books stating that Sir John A Macdonald was Canada’s first prime minister that he was.46 My point is that adjudicating tort claims entails the same inferential process. There are always ‘gaps’ that need to be bridged in order to construct an historical account of what-caused-what, and those gaps are bridged – if they are to be bridged – by the drawing of inferences. Inference is not merely an unavoidable part, but is the very substance, of causal determination. Indeed, inference-drawing occurs even where there are no competing causal accounts on offer to a fact-finder. Imagine Snell without the factual uncertainty said to have been generated by the plaintiff ’s pre-existing conditions that posed a risk of optic nerve atrophy. Likewise, imagine the plaintiff in Wilsher in a state free of all conditions potentially causative of retrolental fibroplasia. Suppose also that they are both treated in the negligent manner described in the case reports, such that the only known source of risk of retro­ bulbar haemorrhage in Snell is the surgeon’s failure to abort the surgery, and the only known source of risk of retrolental fibroplasia in Wilsher is the negligent over-oxidisation   Wright, ‘Pruning the Bramble Bush’ (n 17) 1047.   CR Callen, ‘Cognitive Science and the Sufficiency of “Sufficiency of the Evidence” Tests’ (1990) 65 Tulane Law Review 1113, 1129. 44   Tribe, ‘Trial by Mathematics’ (n 24) 1346. 45   Wright, ‘Pruning the Bramble Bush’ (n 17) 1047. 46   The historical information analogy is particular compelling, since an historian’s enquiry also entails inferring what was or what really happened. Hence Michael Oakeshott’s observation that ‘historical past’ is a construction composed of ‘what the evidence obliges us to believe’. (M Oakeshott, Experience and Its Modes (Cambridge, Cambridge University Press, 1933) 112). 42 43

Inferring Cause in Fact and the Search for Legal ‘Truth’  103 and monitoring of the plaintiff. In short, there are no ‘gaps’ in the sense that there is no competing set of accounts by which negligence and harm might (or might not) be linked. The only evidence before the court in either case is the temporal proximity of the defendant’s negligence and the plaintiff ’s blindness. In such circumstances, it seems fair to say ex ante that the absence of a competing potential causal source would make the causal determination easy, perhaps even obvious. At the very least, no-one would seriously suggest that scientific expert evidence is required to demonstrate cause-in-fact. Still, the causal linkage is being inferred, not demonstrated. There persists a ‘gap’ between the evidence (being the temporal proximity between the negligence and the blindness) and a settled meaning of such evidence (that is, whether the negligence caused the blindness). Even in the absence of a competing causal account, the fact of the temporal proximity does not speak for itself by demonstrating the linkage between negligence and harm. Rather, its significance is inherently ambiguous until it is ascribed by the drawing of an inference. Here is the essential problem with Klar’s insistence upon plaintiffs demonstrating ‘evidence’ of a ‘probable connection’ between negligence and harm. While Snell’s empty incantations of ‘robust’ and ‘pragmatic’ evidentiary treatment are based upon a too-loose understanding of what sort of evidence is necessary to sustain an inference, his objection presupposes a too-rigid and narrow understanding of the nature of evidence. For purposes of explanation, evidence of cause-in-fact might be usefully distinguished from an evidential source. On its own, the latter carries no weight, but simply is. The document can be read, the witness can be heard, or the physical exhibit can be observed, but as to its implications the evidential source is silent. It is then left to the fact-finder to ponder and then to ascribe significance to an evidential source – to what the document contains, to what the witness says or to how the physical exhibit appears – which significance both gives the evidential source meaning, and becomes part of the evidence.47 How such an ascription is accomplished, and the cogency of such an ascription, represent of course the nub of the problem for Snell’s critics because they are non-demonstrable and inductively uncertain. There is no easy answer (and probably no answer at all) to such an objection, but my point here is that the objection to inferring cause-in-fact, even when the cognitive process of drawing such an inference is elusive and likely impossible to describe with any precision, is itself futile because of the epistemology of fact-finding. And so, even where – as in my revised Snell example (with all competing causes removed) – the cause of the plaintiff ’s injury is ‘obvious’ because the surgeon’s negligence is the only potential cause, the determination that a correlation exists between the surgeon’s risky conduct and the plaintiff ’s injury is a product of inferential reasoning. Absent some observable and objectively verifiable mechanism (which, typically, we do not have), there is no demarcation between the ‘obvious’ case (where the evidence discloses only one possible cause) and cases which pose evidentiary ‘gaps’. The gaps that are said to pose a problem in some cases in fact subsist in every case. This means, of course, that causal determinations can be predicated upon any single evidentiary source containing any amount of information pertinent to cause-in-fact.48 While this risks factually incorrect causal determinations – an 47  I believe this is what Alvin I Goldman referred to as the epistemological principle of ‘total evidence’. AI Goldman, ‘Epistemic Paternalism: Communication Control in Society’ (1991) 88 Journal of Philosophy 113. 48   Because the fact-finder never knows what evidence is not before her, an argument that some evidence which is before her demonstrates the probability of a proposition ‘may be constructed upon virtually any amount of evidence’. (Stein, ‘An Essay on Uncertainty’, 300).

104  Russell Brown unavoidable hazard of what Lord Brown described in Sienkiewicz v Greif as the ‘rough justice about the law of personal injury liability as a whole’49 – that risk is lowered where ‘some credible causal generalization links conditions of that type to results of that type’.50 The subsisting question goes to when and how an inference can be drawn from such causal generalisation, that is, to giving a reasoned structure to the inference of a causal linkage said to be generated by the causal generalisation to the specific instance under consideration.

V. The Role of Evidence

The observation having already been made that the cognitive process of inference-drawing is elusive and resistant to precise description, the search for a reasoned structure of an inference may seem otiose. The problem with treating it as such, however, is that we fail to even consider whether fact-finders can offer better reasoning for drawing causal inferences than empty references to ‘common sense’. Even a general discussion, then, might prove useful. Experimental psychologists Nancy Pennington and Reid Hastie have empirically demonstrated that explanatory considerations guide inferences such that, when multiple propositions might explain the evidence relating to a particular event, fact-finders are naturally inclined to infer as a fact the proposition that best explains it.51 More particularly, explanations of phenomena take the form of stories arising from ‘deductive and inductive reasoning procedures applied to the evidence and world knowledge’, including inferences about ‘events, and causal relations between them.’52 Fact-finders, in other words, attempt in situations of factual uncertainty to make sense of the evidence by cobbling together a plausible account that qualifies as the most compelling of the accounts on offer. While a conception of fact-finding as entailing the construction of narratives is hardly new,53 two points bear emphasising. First, Pennington and Hastie’s work reveal this as an empirically demonstrable phenomenon. Fact-finders infer to the best explanation54 among the array of competing explanations put before them by the parties. Secondly, the process Pennington and Hastie describe – and any legitimate process of inference-drawing in the causal enquiry – must be distinguished from some of the most important recent theorising of the trial as an instance of ‘storytelling’. Inferring cause-in-fact requires a measure of truthseeking rigour that is not discernible, for example, in the argument of Robert Burns that narratives are generally persuasive because of ‘normative considerations that emerge from the narratives themselves.’55 Such an argument is objectionable because it drifts into an emo  [2011] UKSC 10, [2011] 2 WLR 523,[187].   Wright, ‘Pruning the Bramble Bush’ (n 17) 1045.   N Pennington and R Hastie, ‘A Cognitive Theory of Juror Decision Making: The Story Model’ (1991) 13 Cardozo Law Review 519. See, also R Hastie, SD Penrod and N Pennington, Inside the Jury (Cambridge MA, Harvard University Press, 1983); and R Hastie and N Pennington, ‘The O.J Simpson Stories: Behavioral Scientists’ Reflections on The People of the State of California v Orenthal James Simpson’ (1996) 67 University of Colorado Law Review 957. 52   Pennington and Hastie, ‘Juror Decision-Making’ (n 51) 524. 53   HLA Hart and T Honoré, Causation in the Law, 2d edn (Oxford, Clarendon Press, 1985), 28–30; RP Burns, A Theory of the Trial (Princeton, Princeton University Press, 1999); R Lempert, ‘Telling Tales in Court: Trial Procedure and the Story Model’ (1991) 13 Cardozo Law Review 559; and L Hirschman, ‘It Will be Pleasanter to Tell You a Story’ (1991) 13 Cardozo Law Review 445. 54   As I discuss below, ‘Inference to the Best Explanation’, or ‘IBE’, has come become a term of art among various evidence theorists. 55   RP Burns, ‘Some Realism (and Idealism) About the Trial’ (1996) 31 Georgia Law Review 715, 756–57. 49 50 51

Inferring Cause in Fact and the Search for Legal ‘Truth’  105 tive standpoint, privileging the fact-finder’s own preferences that might be satisfied by the narrative, and not the evidence represented by the narrative itself. A plaintiff should win where and only where the fact-finder finds his account of the fact both probable and more plausible than the competing accounts, and not because his story evoked ‘considerations’ that the fact-finder somehow found normatively significant. Otherwise, fact-finding would be guilty of indulging emotivism (or any similar non-cognitive, meta-ethical standpoint, such as the ‘benevolent principle’ which ‘smiles on . . . factual uncertainties and melts them all away’).56 Just how we link the defendant’s risk-creating activity with the defendant’s injury must therefore be reconciled with the law’s truth-seeking aim.57 Hence fact-finding’s focus upon the evidence. Hence also the legitimacy of what Pennington and Hastie have observed, being that fact-finders infer to the explanation that best explains the evidence. Various theorists of (inter alia) evidence in the UK and the US58 have conceptualised this empirical structure as a two-staged process dubbed ‘Inference to the Best Explanation’ (or ‘IBE’). At the first stage, parties offer competing narratives to account for the evidence (the plaintiff offering versions that address the formal doctrinal requirements of her claim, and the defendant offering versions that omit at least one of those elements). The second stage sees the fact-finder subjecting those explanations (along with any explanations generated by the fact-finder herself) to consideration for plausibility in relation to the evidence.59 I have argued elsewhere60 that IBE, at least as it has been elucidated by some evidence theorists, is not without its flaws.61 It does, however, offer the possibility of a basic structure within which inference-drawing can be seen as legitimate, while conforming to the cog­ nitive reasoning processes which Pennington and Hastie have empirically demonstrated are employed by fact-finders. And, because the reference point throughout remains the evidence put before the fact-finder, not only is the risk of intrusion from emotive con­ siderations diminished. As Wright has observed, the risk of undue reliance upon a causal generalisation is also reduced: [Particularistic evidence] can negate the instantiation of one of the abstract elements in a competing causal generalization or lower the ex post probability that it was instantiated, thereby eliminating the competing causal generalization or lowering the ex post causal probability associated with it. On the other hand, particularistic evidence can support the competing causal generalization by establishing that one or more of the abstract elements in the competing causal generalization was actually or probably instantiated.62   Fitzgerald v Lane, [1987] 3 WLR 249 (CA).   In this chapter, I take the law’s truth-seeking aim as a given, being a minimal feature of our legal order. 58   P Lipton, Inference to the Best Explanation (London, Routledge, 1993); A Amaya, ‘Inference to the Best Legal Explanation’ in H Kaptein, H Prakken and B Verheij (eds), Legal Evidence and Proof: Statistics, Stories, Logic (Farnham, Ashgate, 2009); Allen, ‘Factual Ambiguity’ (n 25); RJ Allen and CR Callen, ‘The Juridical Management of Factual Uncertainty’ (2003) 7 International Journal of Evidence & Proof 1; Allen and Pardo, ‘Mathematical Models’ (n 18); RJ Allen and MS Pardo, ‘Probability, Explanation and Inference: A Reply’ (2007) 11 International Journal of Evidence Law & Proof 307;and Pardo, ‘Reference Classes and Legal Evidence’ (n 39). 59   Jonakait, ‘Stories, Forensic Science, and Improved Verdicts’ (n 30) 347. 60   R Brown, ‘The Possibility of “Inference Causation”: Inferring Cause-in-Fact and the Nature of Legal FactFinding’ (2010) 55 McGill Law Journal 1. 61   eg, some of its most prominent supporters have argued that a verdict should be entered for whichever party offers the more plausible account, howsoever implausible the more plausible account is, and irrespective of whether it satisfactorily accounts for the totality of the evidence at trial. (RJ Allen, ‘On the Significance of Batting Averages and Strikeout Totals: A Clarification of the “Naked Statistical Evidence” Debate, The Meaning of “Evidence”, and the Requirement of Proof Beyond a Reasonable Doubt’ (1990) 65 Tulane Law Review 1093; and JA Jolowicz, ‘Civil Litigation: What’s it For?’ (2008) 67 Cambridge Law Journal 508, 518–19). 62   Wright, ‘Pruning the Bramble Bush’ (n 17) 1051. 56 57

106  Russell Brown There is, admittedly, a circularity to all this, in that the evidence justifies the explanation, while the explanation accounts for the evidence. So viewed, the explanation and the evidence are mutually reinforcing.63 I do not claim that IBE is without problems,64 but rather that it accounts for the epistemology of fact-finding. The vexing question remains, however: by what criteria may a fact-finder decide that one particular narrative is more likely to be true than its competitors? This is, of course, impossible to reduce to an algorithm. If cause-in-fact is a ‘complex and subtle’ concept that has long eluded ‘efforts to articulate a precise definition’,65 the decryption of its underlying cognitive processes is likely impossible. I do not, therefore, wish to be taken as proposing that the criteria for inference-drawing are reducible, or even describable.66 There are, however, qualities to narratives constructed from the evidence against which such narratives can be measured for what HLA Hart and Tony Honoré described as ‘special cogency’.67 For example, many of the nine criteria famously listed by A Bradford Hill68 have been emphasised by inter alia evidence theorists, psychologists and computational cognitive scientists, including simplicity,69 coherence or consilience,70 coverage,71 ‘uniqueness’72 and the temporal proximity between risk and harm.73 Certain of these qualities (coherence, coverage, uniqueness and temporal proximity between risk and harm) have been emphasised by Pennington and Hastie,74 while David Danks (whose work involves developing computational models to predict linkages between causal observations and inference) has emphasised inter alia temporal proximity between risk and harm.75 Computer scientist Judea Pearl and engineer TS Verma also stress ‘temporal precedence’ of risk in relation to harm as being significant and possibly even necessary to a causal determination, although not necessarily sufficient to distinguish cause-in-fact from spurious correlation.76

  MS Pardo and RJ Allen, ‘Juridical Proof and the Best Explanation’ (2008) 27 Law and Philosophy 223, 233.   Brown, ‘The Possibility of “Inference Causation” ’ (n 60). 65   Wright, ‘Causation in Tort Law’ (n 34) 826. 66   Here I am as one with MacCrimmon, ‘What is ‘Common’ about ‘Common Sense?’ (n 12) 1436–37. See also M Damaska, ‘Truth in Adjudication’ (1998) 49 Hastings Law Journal 289, 292. 67   Hart and Honoré, Causation in the Law (n 53) 1438. 68   AB Hill, ‘The Environment and Disease: Association or Causation?’ (1965) 58 Proceedings of the Royal Society of Medicine 205. His criteria are: strength (of association); consistency; specificity (of effect and cause); temporality; biological gradient; plausibility; coherence; experiment; and analogy. These are canvassed in Miller, ‘Causation in Personal Injury’ (n 21) 547–53. See also: M Susser, ‘The Logic of Sir Karl Popper and the Practice of Epidemiology’ (1986) 124 American Journal of Epidemiology 711. 69   Allen and Callen, ‘The Juridical Management of Factual Uncertainty’ (n 58) 28 (see in particular the reference to Occam’s Razor); and Allen and Pardo, ‘Probability, Explanation and Inference’ (n 58) 315. 70   ‘Coherence’ here meaning the absence of internal contradictions, whether with elements of the narrative or with the evidence found to be true. See eg, LJ Cohen, ‘Should a Jury Say What it Believes or What it Accepts?’ (1991) 13 Cardozo Law Review 465, 476; and Burns, ‘Some Realism’ (n 55) 753. Stein, ‘The Refoundation of Evidence Law’ (n 22) 312, refers to ‘cogency’, although this is tautological, since a factor is by definition ‘cogent’ as it renders the truth of a conclusion more probable. 71   ‘Coverage’ here meaning the extent to which all the evidence is explained by the narrative). Burns, ‘Some Realism’ (n 55) 753. 72  ibid. ‘Uniqueness’ here meaning ‘the absence of a plausible competing narrative’. 73   J Pearl and TS Verma, ‘A Theory of Inferred Causation’, in Proceedings of the Second International Conference on the Principles of Knowledge Representation and Reasoning (Cambridge MA, April 1991); D Danks, ‘The Psychology of Causal Perception and Reasoning’ in H Beebee, C Hitchcock and P Menzies (eds), Oxford Handbook of Causation (Oxford, Oxford University Press, 2007) 18; and Wright, ‘Pruning the Bramble Bush’ (n 17) 1014. 74   N Pennington and R Hastie, ‘A Cognitive Model of Juror Decision Making: The Story Model’ (1991) 13 Cardozo Law Review 519. 75  Danks, ‘The Psychology of Causal Perception and Reasoning’ (n 73) 18. 76   See Pearl and Verma, ‘A Theory of Inferred Causation’ (n 73). 63 64

Inferring Cause in Fact and the Search for Legal ‘Truth’  107 The common emphasis upon temporal proximity is intriguing in light of Snell and Wilsher, because it suggests a possible reconciliation of the differing outcomes in those cases as to whether it was possible to draw inferences about where the balance of probabil­ ities lay. The respective facts and outcomes, when taken together, suggest that temporal proximity between risk and harm may have been seen as significant. While we are not told in Snell for how long the plaintiff had suffered from her pre-existing sources of risk, given her advanced age it can be assumed that she had suffered longer than the (literally) few hours lived by the plaintiff in Wilsher prior to his negligent treatment. In applying the factor of temporal proximity, a fact-finder might therefore be less reluctant to infer a causal link between risk and harm in cases brought by plaintiffs whose facts resemble Snell than by plaintiffs whose facts resemble Wilsher. This is because, in cases resembling Snell, a factfinder might judge that sufficient time has passed since the onset of the pre-existing sources of risk for us to ascribe significance to their asymptomatic state, and conversely to the active state of the risk generated by the defendant’s negligence. A case like Wilsher, however, offers no such case history of asymptomatology and therefore no basis for a fact-finder to draw conclusions from the contrasts between the pre-existing sources of risk and the negligent treatment. This is not to suggest that the criteria by which cause-in-fact may be inferred are reducible to temporal proximity, or that temporal proximity absolutely supported the finding of a causal link in Snell. It does, however, explain why the facts of one case might be more suggestive than another of a causal link and thus more supportive of an inference. At the same time, consideration of that factor highlights a particular problem with how judges sometimes choose to infer (as in Snell) or not to infer (as in Wilsher) cause-in-fact. The reasoning I have ascribed to the courts in Snell and Wilsher is fictitious: no such reasoning was actually stated in those cases, and in fact the courts’ conclusions (that an inference could be drawn in Snell and that it could not be drawn in Wilsher) were entirely peremptory. As a result, in both cases we are left entirely in the dark as to why cause-in-fact could or could not be inferred. In fairness to critics of inference-drawing, their criticisms – seen in the light of such conclusory statements – are understandable. That is, their driving concern may be that, if we are given no reasoned explanation for why an inference is or is not being drawn, then we are left to the impossible task of divining just how ‘common sense’ led to a particular outcome. In short, where ‘common sense’ is deployed not as the method, but as the explanation, we have no explanation. If, however, this is the concern driving critics of inference-drawing, then their criticisms are too broadly stated inasmuch as their concern implicates not the method, but merely the absence of an accompanying explanation. Vacuous judicial statements that express reliance upon ‘common sense’ ought not to be conflated with inference-drawing, nor should concerns about the former be seen as necessarily implicating the latter. This is, however, an easy (if mistaken) connection for inference-drawing’s Canadian critics to make, given the short shrift given to the evidence in Snell: in addition to his empty rhetoric of treating evidence ‘robust[ly]’ and ‘pragmatic[ally]’, Sopinka J also speculated that the causal inference could be drawn with ‘very little affirmative evidence on the part of the plaintiff ’.77 Snell can thus be fairly described as an authority for the proposition that a barely substantiated invocation of ‘common sense’ is shorthand for determining cause-in-fact. One must also concede in fairness to the critics that there is no reason why lawyers would be particularly nuanced in their critiques of fact-finding generally and of inference-drawing in particular. As law   Snell v Farrell [1990] 2 SCR 311, 328.

77

108  Russell Brown students, lawyers do not take courses in fact-finding, and a typical course in evidence law does not teach credibility assessment or inference-drawing. Fact-finding is not, after all, a uniquely legal phenomenon, but rather is a ubiquitous, quotidian task. The ability to find facts, even in the face of ‘gaps’ or disputed evidence, rests with any person of ordinary intelligence. Employers called upon to resolve employee disputes must exercise it, as must teachers mediating student quarrels. And, of course, parents of young children – who nearly every day hear competing versions of events, and then cobble together the narrative that best explains what the parents have seen and heard (among other senses) – must exercise it. In all these cases, laypersons are finding facts which determine outcomes respecting the status, welfare or even in some cases the rights of others. And, like legal fact-finders, they do so in the same way – by inferring to the best explanation. A good question, then, is why the dearth of causal reasoning in Snell (and Wilsher) should be seen as posing a problem. If there is nothing unique to legal adjudication about indeterminacy in fact-finding generally or in cause-finding specifically, and if laypersons are naturally inclined to assess evidence in order to assess competing narratives as a preliminary step to inference-drawing, perhaps legal fact-finders should be entitled to the same deference. The difference, of course, is that judges perforce exercise judgment. What Snell (and Wilsher) lack is an explicit account of that judgment of the evidence. Absent that account, the judgment that a legal fact-finder ostensibly brings to the causal enquiry is indistinguishable from a layperson’s intuition. This is wrong, because judgment implies a qualitative assessment, not some mere intuitive insight.78 That assessment must in turn depend upon the reasons for supporting it; otherwise, its quality would be a function of its intuitive insight, which (being intuitive insight) cannot be reasoned. For judgment to be seen as such, therefore, it requires a compelling account – that is, an argument. This is hardly a radical proposition. Indeed, the close affinities between how judges bridge ‘gaps’ in evidence and how they bridge gaps in law go to a central ethic of the judicial role. Just as Peter Birks, for example, cautioned that ‘intuitive law-finding . . . frees the judge from the shackles of traditional legal rationality’,79 such rationality – that is, such reasoning – must also bind the judge from engaging in intuitive fact-finding. Judicial reliance upon mere intuition would detract from public confidence in the justice of an outcome, irrespective of whether intuition was applied to find law or to find facts. Similarly, Ronald Dworkin’s view of the law-finding aspect of judging as comprising the practice of legal argumentation also applies mutatis mutandis to the fact-finding aspect of judging which requires factual argumentation. While the former entails the explicit weighing of legal principles in order to discern their net vector as applied to the particular case,80 the latter requires the explicit weighing of evidence in order to discern its implications for the particular case. In short, fact-finding’s ‘argumentative dimension’81 distinguishes it from intuition. In furnishing ‘a reasoned basis for [a] decision’, it falls within what Hart described as ‘characteristic judicial virtues’ which judges employ (or ought to employ) in cases where there is discretion.82 For our purposes, then, the central question is not how Dworkin’s superhuman judge Hercules 78   Here I am agreeing with AT Kronman, ‘Living in the Law’ (1987) 54 University of Chicago Law Review 835, 849–53. 79   P Birks, ‘Three Kinds of Objection to Discretionary Remedialism’ (2000) 29 University of Western Australia Law Review 1, 17. 80   R Dworkin, Law’s Empire (Cambridge, MA, Harvard University Press, 1986) 13: ‘Legal practice, unlike many other social phenomena, is argumentative’. 81   Kronman, ‘Living in the Law’ (n 78) 850. 82   HLA Hart, The Concept of Law (Oxford, Oxford University Press, 1961) 200.

Inferring Cause in Fact and the Search for Legal ‘Truth’  109 would have decided Brown v Board of Education,83 but how he would have reasoned Snell or Wilsher or any case involving multiple sources of risk.

VI. Conclusion

My argument has been, first, that legal fact-finding is distinct from scientific fact-finding. As a consequence, neither an evidentiary gap, nor the absence of scientific evidence bridging the gap, in and of themselves permit a fact-finder to halt the causal enquiry in its tracks. To do so would be, literally, unjust. A fact-finder must, in such circumstances, apply her capacity for judgment that is immanent in the cognitive processes she brings to bear upon the evidence, including the soft variables such as trial phenomena. That judgment is, moreover, as a matter of empirically demonstrated fact naturally inclined to inference-drawing that favours a proposition that best explains a given phenomenon. Such an epistemology of fact-finding should force critics of inference-drawing to reconsider their criticisms. Inference-drawing is not, however, a licence for emotivism in fact-finding. This chapter has, after all, considered inference-drawing as a means of finding (or not finding) cause-infact, not cause-in-feel. Inference-drawing could hardly qualify as a means of factual causal determination were it grounded in emotive considerations, and indifferent to the evidence. For that reason, inference-drawing necessarily entails deliberation and explicit justification of a determination that cause-in-fact is or is not demonstrated on the evidence. It demands of the fact-finder that she give the competing narratives their due, and assess the possibil­ ities that the supporting evidence for each narrative suggests. The fact-finder is literally making a judgment call, but that should not leave us in the dark about why a particular determination on cause-in-fact is (or is not) being made in a given case. Public acceptance of (if not universal agreement with) the factual determination being made requires at a minimum some rational, coherent and deliberative explanation showing that cause-in-fact has not been merely pronounced, but rather sought and found (or not found). To reiterate the earlier discussion of Snell, we might be more likely to accept Sopinka J’s conclusion about cause-in-fact were we told that it was driven by a combination of the temporal proximity between the negligence and the plaintiff ’s blindness and the prior asymptomology of her diabetes, high blood pressure and glaucoma. The rub for critics of inference-drawing is, of course, that we can only assume that throughout the entire process the competing narratives are actually being assessed and compared by the fact-finder with reference to the evidence, and free of all emotive value judgments. Fact-finding is, however, replete with opportunities for disguising emotive fudge as good faith deliberation, and it seems futile to object to that which, given the epistemology of fact-finding, is unavoidable. To object to inference-drawing is to object to factfinding itself, and that is obviously too sweeping an objection for jurists to sustain. That said, fact-finding can and ought to be scrutinised for its reasoning and for the orientation of such reasoning to the evidence. At the end of the day, that is the very least – but also the very most – which we should expect from a fact-finder.

  347 US 483 (1954). See also: Dworkin, Law’s Empire (n 80) 388.

83

6 Causation in French Law: Pragmatism and Policy DUNCAN FAIRGRIEVE AND FLORENCE G’SELL-MACREZ

I. Introduction

Complex theories of causation have long been an enticing intellectual terrain for jurists, attracted by the philosophical, moral as well as legal dimensions to this issue. During the nineteenth century, the pioneering German schools1 were in the vanguard of analysis, whereas French jurists had yet to fully engage with the intricacies of causation. Traditionally resistant to grand systemic theories of causation, French scholars have often been sceptical about the theoretical framework of causality, with commentators asserting that courts decide causation issues intuitively.2 As a leading French author said in 1939: ‘The problems of causation in damages liability were invented to seduce the mind of German jurists’!3 The modern French approach to the issue of causation reveals interesting details about the role and function of the French judiciary, the functioning of the French civil justice system, as well as the underlying policy agenda. This piece will primarily focus on the test of causation in respect of extra-contractual liability, but we will also place this within broader perspective, so as to properly understand how policy factors, civil procedure and the French judicial process have had an influence. An additional complicating factor derives from the well-known jurisdictional divide between public and private law in France. This has meant that the administrative and civil courts have had autonomous approaches to the notion of causation, as we shall examine below. The first part of this chapter will be devoted to the concept of causation itself, in order to show that causation is a basic requirement in all liability regimes, with a very few exceptions, and to illustrate the fact that French courts use the various theories of causation in such a way that we cannot designate the specific theory that prevails. The second part will focus on procedural aspects of the causation issue, especially the law of evidence and the use of presumptive reasoning by French judges. The third part of this chapter will illustrate the way French courts manipulate the concept of causation for policy reasons.   Or, more correctly, the German language schools, as Austrian commentators were also very influential.   See P Esmein, ‘Le nez de Cléopâtre ou les affres de la causalité’ (1964) D 1964, I, 205, fn 30. 3  ‘Le problème causal en matière de responsabilité était fait pour séduire l’esprit des juristes d’Allemagne’: G Marty, ‘La relation de cause à effet comme condition de la responsabilité civile’ (1939) Revue trimestrielle de droit civil 685, 689. 1 2

112  Duncan Fairgrieve and Florence G’Sell-Macrez

II. The Concept of Causation

The concept of causation must at first be analysed through its function as a condition of liability (A) before being considered through the different approaches of causation that have been inspired by German theories (B).

A.  Causation: a Basic Requirement of Liability Causation is an inherent element of the liability equation. In French law, it is however a multi-dimensional issue, with different aspects depending on the basis of liability and the branch of law. We shall therefore address here successively civil liability, criminal law and administrative liability, before turning to special regimes of liability. i.  Causation in Civil Law Causation is a fundamental condition of liability both within the context of tort and contract. In French doctrinal analysis, causation and tortious responsibility are inseparable. Causation is quoted as a ‘logical requirement’ in the words of Dean Carbonnier.4 In substantive law, the notion of causation features prominently in the Civil Code. Causation is mentioned in all the provisions relating to damages liability, which are found in Articles 1382 to 1386 of the Code. However, it is in the famous provision on fault-based liability, Article 1382 C civ, where the requirement is most clearly stated: ‘Any act whatever of man, which causes damage to another, obliges the one by whose fault it occurred, to compensate it’.5 The consecration of a general principle of fault-based liability in Article 1382 of the Civil Code elevated causation to the position of an essential condition for the sanction of any civil wrong. Surprisingly, then, the causal requirement did not feature prominently in the travaux préparatoires of the Civil Code, or in doctrinal commentary immediately thereafter. Indeed, the details of causation were not discussed as such during the drafting of Article 1382 of the Civil Code. Causation was simply assumed as an evident constituent element of the liability equation. Tarrible underlined ‘the requirement that natural law requires every individual to repair the damage he has caused’.6 This tendency prevailed after the enactment of the Napoleonic Code. The early commentators of the Civil Code showed very little interest in defining the causation requirement. In their eyes, causation was a pure question of fact to be solved by the trial judge on a case-by-case basis. Indeed, until the early twentieth century, causation was not even mentioned in French civil law treaties as a condition of liability!7 4   ‘une exigence de raison’: J Carbonnier, Droit civil vol 2, ‘Les obligations’, 22nd edn (Paris, Presses Universitaires de France, 2000, reissued Quadrige, 2004) para 213. 5   Art 1382 Civil Code: ‘Tout fait quelconque de l’homme, qui cause à autrui un dommage, oblige celui par la faute duquel il est arrivé à le réparer’. 6   Cette obligation que le droit naturel impose à tout individu de réparer le dommage qu’il a causé’: J Tarrible in PA Fenet, Travaux préparatoires du Code civil, t XIII, 490. 7   F G’sell-Macrez, Recherches sur la notion de causalité (thesis, Paris I, 2005) para 312. See also C Demolombe, Traité des engagements qui se forment sans convention (Paris, 1882) Bk III, t IV, ch II, para 671 and G BaudryLacantinerie and L-J Barde, Traité théorique et pratique de droit civil, vol IV, ‘Obligations’, 3rd edn (Paris, 1908) para 2853, 536.

Causation in French Law: Pragmatism and Policy  113 On the other hand, causation was discussed within the context of contract law. Article 1151 of the Civil Code explicitly limits compensation for loss caused by breach of contract to the ‘immediate and direct consequences’ of the breach,8 following Pothier in this matter.9 There was speculation about the scope of this requirement, with certain scholars pointing to the lack of an equivalent within tort law. Some even took the view that Article 1151 of the Code Civil expressed a natural condition for the award of damages for a civil wrong, applicable equally to a breach of contract and the commission of a tort.10 Other prominent commentators contended that it was restricted to contract law.11 Ultimately, the lively doctrinal debate on the issue concluded that the requirement in Article 1151 of the Code Civil of ‘direct and immediate consequences’ was applicable to all civil wrongs. This analysis is still generally accepted.12 As a consequence, it is usually stated that causation in general must be shown to be certain and direct. Thus these twin conditions are usually understood as expressing the need for a causal link in itself, not the particular characteristics of that link.13 In other words, requiring a direct link is equivalent to requiring a causal connection. In this respect, it is interesting to note that the draft reform of the Civil Code, the Avant-projet de réforme du droit des obligations et de la prescription (also called Rapport Catala),14 which was compiled by a team of academics, does not even include the terms of Article 1151. The reform instead replaces the requirement of directness by the notion of causality itself, to which the draft devotes Article 1347.15 On this point, substantive law is unlikely to be much changed. In sum, the proof of a causal link is required in French law to compensate the consequences of any breach of contract or any tort. However, a number of observations should 8   Art 1151 Civil Code: ‘Even in the case where the non-performance of the agreement is due to the debtor’s intentional breach, damages may include, with respect to the loss suffered by the creditor and the profit which he has been deprived of, only what is an immediate and direct consequence of the non-performance of the agreement’. ‘Dans le cas même où l’inexécution de la convention résulte du dol du débiteur, les dommages et intérêts ne doivent comprendre à l’égard de la perte éprouvée par le créancier et du gain dont il a été privé, que ce qui est une suite immédiate et directe de l’inexécution de la convention’. 9   Pothier’s example is classical and famous. A livestock dealer made the mistake of selling a plague-stricken cow. The animal died and the buyer’s entire herd perished by contagion. The buyer, deprived of his cattle, couldn’t cultivate his land. Since he lacked resources, he did not pay his creditors. The latter thus seized his goods and sold them at a low price. Among this series of injuries, Pothier for his part considered that only the death of the diseased cow and the herd was a direct damage. He was reluctant to consider the failure to cultivate the land as a cause and believed that the seizure and its implications constituted an indirect injury: ‘Le marchand ne sera pas tenu des dommages que (le paysan) a soufferts par la saisie réelle de ses biens: ce dommage n’est qu’une suite très éloignée et très indirecte de son dol, et il n’y a pas une relation nécessaire : car, quoique la perte de mes bestiaux, que son sol m’a causé, ait influé dans le dérangement de ma fortune, ce dérangement peut avoir eu d’autres causes’. RJ Pothier, Obligations, vol 2, (ed Bugnet, t. II) 80–81, paras 166–67. 10   C Demolombe, Cours de Code Napoléon, t XXI (1882) para 687; M Filippi, Le préjudice indirect (PhD Thesis, Lille, 1933). A Joly, Essai sur la distinction du préjudice direct et du préjudice indirect (Caen, 1938). 11   M Planiol, Traité élémentaire de droit civil, vol II, 1st edn (1900) para 935; L Josserand, Cours de droit civil positif français, vol II, 1st edn (1930) para 440. 12   F Terré, P Simler, Y Lequette, Droit civil, Les obligations, 10th edn (Précis Dalloz, 2009) para 592. G Viney and P Jourdain, Les conditions de la responsabilité civile, 3rd edn (Paris, LGDJ, 2006) para 348. Ph le Tourneau, Responsabilité civile (Dalloz Action, 2008/2009) para 1776; Ph Malaurie, L Aynès, Ph Stoffel-Munck, Droit civil, Les obligations, 4th edn (Defrénois, 2009) para 963. 13  ibid; contra M Fabre-Magnan, Les obligations 2, La responsabilité (Presses Universitaires de France, coll Thémis droit privé, 2007) para 260. 14   P Catala (under the direction of), Avant-projet de réforme du droit des obligations et de la prescription (La documentation française, 2006). An electronic version is accessible on the French Ministry of Justice’s website either in French, English, German, Spanish and Italian versions:www.justice.gouv.fr/index.php?rubrique=10047& ssrubrique=10049&article=11944 15   Art 1347: ‘La responsabilité suppose établi un lien de causalité entre le fait imputé au défendeur et le dommage’.

114  Duncan Fairgrieve and Florence G’Sell-Macrez be made. A first remark must be made regarding the causal requirement under contract law. Whilst the proof of causation is an unquestionable necessity, it should be noted that this requirement is affected by the traditional French distinction between obligation de résultat and obligation de moyen.16 Under an obligation de moyen, the contracting party must strive to achieve the desired result by using reasonable diligence, whereas the obligation de résultat requires the defendant to achieve the promised result at any cost. The breach of an obligation de résultat is ipso facto established when the promised result is not obtained, and in this case, it is often considered that causation does not need to be proved. In other words, when an obligation de résultat is not met, causation is presumed.17 Much does, however, depend on how the court interprets the contract: decisions tend to vary depending on the defendant’s exact circumstances. For example, French judges are particularly severe with the French railways company (SNCF) which is routinely found liable on the basis of an obligation de sécurité de résultat (strict liability safety obligation) for accidents which have occurred in trains, even in the case of injury caused deliberately by a third party or where the victim is obviously responsible for the accident.18 A second remark concerns the fact that under tort law, causation is not only a requirement of fault-based liability (Article 1382 Civil Code), but is also applicable to no-fault liability based on Article 1384 al 1 Civil Code, referred to as liability for the ‘actions of things’.19 Although this provision was originally intended merely as a preface to the following provisions of the Code, French courts have reinterpreted it so as to impose liability for the actions of things within one’s custody.20 Causation is intricately linked to this notion of no-fault liability in the sense that the object in question has played a role in the occurrence of the harm.21 Detailed criteria have been laid down by the courts for determining the extent to which causation is presumed.22 If the object in question was in motion and impacted on the person injured or property damaged, these circumstances give rise to a presumption of causation.23 In other circumstances, the claimant has to show that the thing played an ‘active role’ in the accident, usually by establishing an abnormality in the thing’s position or behaviour.24 The abnormality standard thus introduces a normative analysis of causation. In sum, causation is an essential element of no-fault liability. 16   See generally B Nicholas, The French Law of Contract, 2nd edn (Oxford, Oxford University Press, 2005) 50; J Bell, S Boyron and S Whittaker, Principles of French Law, 2nd edn (Oxford, Oxford University Press, 2008) 342– 43. 17   Cass civ (1) 21 October 1997, D aff, 1419. Cass civ (1) 12 January 1994 and 2 February 1994, JCP G, II, 22294 note Ph Delebecque. Cass civ (1) 20 June 1995, JCP G, IV, 2010. 18   Cass civ (1) 12 December 2000, Bull civ I n°323, D 2001 jur p 1650, note C Paulin. Cass civ (1) 3 July 2002, D 2002, jur p 2631 note J-P Gridel, RTD civ 2002, 821 obs P Jourdain. 19   ‘responsabilité du fait des choses’ Art. 1384 al 1 Code Civil provides that ‘one is liable not only for the harm which one causes by one’s own action but also for that which is caused by the action of persons for whom one is responsible, or of things which one has in one’s keeping’. 20   See generally, J Bell, S Boyron and S Whittaker, Principles of French Law (n 16) 381–88. 21   F G’sell-Macrez, Recherches sur la notion de causalité (n 7) para 362. 22   The decision Cadé established the basic principles in this matter. Cass civ 18 February 1941, Cadé, D C 1941, 85, note J Flour. 23   The courts have appeared to renounce the possibility of reversing the presumption. Cass civ (2) 28 November 1984, JCP G 1985, II, 20477 note N Dejean de la Bâtie (a motorcycle had collided with a car). Cass civ (2) 2 April 1997, RCA com n°255, note F Leduc (two cases, the first concerned an elevator, the second involved an escalator). Contra CA Paris, 25 April 2000, Grillet, Gaz Pal 2001, 1, 478, note F Chabas, (the escalator had had a normal behaviour, the fall of the victim had been caused by a suitcase). 24   eg, in the case where the balustrade of balcony was outdated, Cass civ(2) 17 February 2005, RCA 2005, comm 119. In the case of a glass wall : Cass civ (2)e 15 June 2000, note G Blanc, D 2001, 886; RTD civ 2000, 849 obs P Jourdain; RCA 2000 comm 292 obs H Groutel, or of a block of cement: Cass civ (2) 18 September 2003, D 2004 jur p 25 note N Damas; RTD civ 2004, 108 obs Jourdain RCA 2003 comm 286, obs Groutel, JCP 2004, I, 101 n°18,

Causation in French Law: Pragmatism and Policy  115 ii.  Causation in Criminal Law In criminal law, issues of causation have been particularly controversial for offences based upon negligence. Articles 221-6 and 222-19 of the Criminal Code of 1992 both refer to behaviour which ‘caused’ harm. The French courts initially gave a very broad interpretation of the requisite causal relationship but this led to heavy sentences been handed down for persons whose mens rea was relatively minor.25 Moreover, in actions for compensation accompanying the criminal proceedings,26 the French courts also designated as causes acts which had only a tenuous relationship with the result. For example, judges found liable for negligent homicide the driver who had caused a car accident where the victim had been injured in the head, because the victim had later committed suicide.27 Legislation was adopted to restrict this tendency. The Law of 10 July 200028 laid down that, where the causal link is indirect, criminal liability of individuals was premised upon evidence of wilful or serious misconduct.29 Cases of indirect causation are defined by that provision as those where the defendant ‘has created or helped to create the situation that has resulted in an injury’ or has ‘failed to take measures to avoid it’.30 In the first case, the perpetrator is an ‘indirect’ wrongdoer while, in the second, he is a ‘mediate’ wrongdoer. This reform has thus introduced somewhat confusing terminology about the causation requirement.31 Indeed, for criminal lawyers, direct causation now designates a particularly close relationship whereas it simply expresses the causal link for civil lawyers. iii.  Causation in Administrative Law Mirroring the position of French civil law, the basic principle of causation in French administrative law, applicable to both fault and no-fault liability, is that the claimant’s harm must be direct which, as with civil law, essentially encapsulates the notion that there must be a direct causal relationship between the defendant’s act and the victim’s loss.32 Some obs G Viney, JCP G 2004 II 10013 note C Le Tertre; Defrénois 2004 n°56 obs R Libchaber. Case of a corridor that was slippery because of a recent polishing and which lighting was defective Cass civ (2) 11 December 2003, RCA 2004 comm 61. 25   eg Cass crim 8 July 1971, D 1971, p 625, note E Robert. 26   Under French law, a civil claim can accompany the criminal offence by means of the partie civile procedure. Private parties may both initiate the commencement of criminal proceedings, or alternatively join proceedings which are underway. 27   Crim 14 January 1971, D 1971, p 164, note E Robert. See also Cass crim 19 May 1958, Bull crim n°395, D 1958, somm p 149. Crim 18 October 1995, dr pénal 1996, comm 78, note Véron. Crim 14 February 1996, Bull crim n°78; Rev sc crim 1996, 856, obs Y Mayaud. Crim 23 September 1998, Rev sc crim 1999, 321, obs Y Mayaud. 28   Loi n°2000-647 du 10 juillet 2000 tendant à préciser la définition des délits non-intentionnels. See ‘Aperçu rapide’ par F le Gunehec, JCP G 2000 p 1587 ; Y Mayaud, Retour sur la culpabilité non intentionnelle en droit pénal…í JCP G 2000 chron p 603; J Pradel, ‘De la véritable portée de la loi du 10 juillet 2000 sur la définition des délits non intentionnels’ D. 2000, n°29 Point de vue, p 5; P Mistretta, La responsabilité pénale médicale à ‘Laune de la loi du 10 juillet 2000, évolution ou révolution?’ JCP G 2002 I chron 149; D Commaret, ‘La responsabilité pénale des décideurs en matière de délits non intentionnels depuis la loi du 10 juillet 2000 Gazette du Palais, 10–11 septembre 2004, p 3; D Commaret, ‘La loi Fauchon, cinq ans après’ JCP E 2006, 1528; G Notté, Bilan jurisprudentiel en droit de l’entreprise de la loi n°2000-647 du 10 juillet 2000 sur la définition des délits non intentionnels JCP E 2006, 1527. 29   Now Art 121-3 al 4 Criminal Code. 30   Art 121-3 al 4 Criminal Code. 31   F G’sell-Macrez, Recherches sur la notion de causalité (n 7) para 319. See also Ph Conte, comment under Cass crim 13 November 2002, D 2004, jur 1336. 32   M Paillet, La Faute du Service Public en Droit Administratif Francais (Paris, 1980) para 64; J-P Dubois, La Responsabilité Administrative (Paris, 1996) 39; C Gour, ‘Faute de Service’ in F Gazier and R Drago, (eds) Dalloz Encyclopédie de Droit Public: Répertoire de la Responsabilité de la Puissance Publique, para 255; See eg, CAA Bordeaux 4 April 1991, Francais, [1991] Rec 1241; CE 6 July 1988, Martin, [1988] Rec 1061.

116  Duncan Fairgrieve and Florence G’Sell-Macrez cases similarly refer to the need for the causal link to be certain.33 We will see below that there has been much debate in French civil law about the exact test of causation that has been applied by the courts. Similar questions have been raised in the fora of administrative courts. Taken literally, the rule of causal directness would seem to reflect the direct consequence theory of causation according to which tortfeasors are only liable for the direct consequences of their acts.34 However, as we shall see below, the French case law has engaged with the predominant test of causation, that of adequate causation. iv. Causation and Bespoke Regimes Specific regimes governing damages liability are a familiar feature of French law. Bespoke rules have been implemented in a number of areas, including liability for motor vehicle accidents in 1985,35 liability for defective products in 199836 and medical liability in 2002.37 Whereas liability for defective products and medical liability are both subject to the classic criteria of causation, the notion of causation was ostensibly removed from the so-called ‘liability’ regime for motor vehicle accidents and replaced by the requirement of ‘involvement’ (implication). This regime is sometimes referred to as a simple compensation mechan­ism as no proof of causal link between the defendant’s act and the harm is actually required. Under the relevant legislation,38 the defendant is liable if his/her vehicle is ‘involved’ in a traffic accident. Although French courts retain a broad interpretation of the notion of ‘involvement’, it should be stressed that the claimant must, in any case, prove that his/her individual damage results from the accident. Thus causation is not totally absent from this regime. Compensation for work-related accidents was introduced by virtue of the Act of 9 April 1898 which was amended by Act of 30 October 1946 and included in the Code of Social Security (Articles L 411-1 and following). All medical expenses resulting from a workrelated accident are covered by social security funds. In addition, the victim is entitled to compensation corresponding to the permanent consequences of the accident. In return, the victim cannot sue the employer directly, unless an intentional or inexcusable fault has been committed. Under this regime, the question is whether the damage was caused by work or not. Since 1911, the Cour de Cassation has held that any accident which happened at the time and place of work should be regarded as work-related.39 This presumption now applies to any lesion that appeared at the time and place of work.40 Things are more complicated when the consequences are remote in time from the initial accident. The approach is, in this respect, pragmatic and very close to the solutions adopted in civil liability. 33   CE 4 October 1968, Doukakis, [1968] Rec 1100; CAA Nantes 25 October 1990, Baston, Req 89NT00137. See also comments of J-P Taugourdeau, Le Caractère Certain et Direct du Préjudice en Matière de Responsabilité Extra-Contractuelle de la Puissance PubliqueL (1974) AJDA 508, 517. 34   See T Honoré, ‘Causation and Remoteness of Damage’ in A Tunc (ed), Torts, vol 11, International Encyclopedia of Comparative Law, para 71. 35   Loi n°85-677 of 5 July 1985. 36   Loi n°98-389 of 19 May 1998 creating art 1386-1 to 1386-18 Civil Code. The specific liability regarding defective products requires proof of a causal link between the defect and the harm suffered by the victim (art 1386-9 C civ) L Grynbaum, ‘Le défaut du produit et le lien de causalité’ (2008) Revue de droit sanitaire et social 1026. 37   Loi n°2002-303 of 4 March 2002. The 2002 legislation about medical liability provides that health professionals (doctors, surgeons, nurses) are liable for the harmful consequences of acts of prevention, diagnosis or care only in the case of fault (art L 1142-1 C. santé pub), which implies that a causal link is proven. Art 1142-1 C santé pub also provides no-fault liability in cases of defective products or hospital-acquired infections when it is established that the victim s harm has found its origin in a defect or a stay at the hospital. 38   Loi n°85-677 of 5 July 1985. 39   Cass civ 27 December 1911: S 1911, 1, 383. 40   Cass soc 19 July 1962: Bull civ 1962, IV, n° 670.

Causation in French Law: Pragmatism and Policy  117

B.  Meaning of ‘Causation’ in Civil Law Cases From a historical perspective, there was surprisingly little theoretical reflection on causation in French doctrinal writing during the period immediately following the adoption of the Civil Code. It was only through the influence of German literature that, during the latter part of nineteenth century, commentators gained exposure to theories of causation. A turning point came in 1913, when an avocat from Aix en Provence, Marteau,41 completed a thesis analysing the German theories on causation. Subsequently, during the twentieth century, some commentators, such as Dejean de la Bâtie, have attempted to produce an account of causation based on French case law. Nonetheless, mystery still prevails as to the test of causation applied by the French courts. i.  Imported German Theories In his thesis, Marteau introduced into French academic thinking the two main theories of causation which prevail today in French law: the theory of adequate causation and the theory of the equivalence of conditions. a.  Equivalence of Conditions The theory of the equivalence of the conditions entails that every condition of an injury is a cause. This theory thus encapsulates the sine qua non test, better known to English lawyers as the ‘but for’ test. This theory owes much to German analysis. According to the German theorist, Von Büri,42 since it is impossible to specify the exact share of each condition in the production of the result, every factor should be considered as a cause. Von Büri justified this assertion by the fact that the ultimate result is indivisible. Von Liszt43 also believed that all conditions are necessary to produce the result and are, therefore, inevitably equivalent. In other words, all the conditions without which the effect would not have occurred are necessarily equivalent. As is well known, this test is, however, unsatisfactory in a number of respects. It is often said that the ‘but for’ test is over-inclusive, especially in presence of ‘domino effects’ when events follow and condition each other. In a well-known French case, the victim was injured after a car accident, and consequently became a paraplegic. Ten years later, the person in question died in a fire at his home, from which he could not escape because of his handicap.44 In this case, the court found that there was no causal link between the car accident and the person’s ultimate death. The accident which had caused the handicap was clearly a condition of the ultimate death, but the court denied liability. On the other hand, where victims have been injured in accidents and transported to hospital, and subsequently harmed due to medical malpractice,45 medical risk inherent in the treatment (aléa thérapeutique),46 or transfusion of   P Marteau, La notion de causalité dans la responsabilité civile (thesis, Aix en Provence, 1913).   M von Büri, Die Kausalität und ihre strafrechtlichen Beziehungen (Stuttgart, 1885). 43   F von Liszt, Lehrbuch des deutschen Strafrechts, 13th edn (1903). 44   Cass civ (2) 8 February 1989, JCP G 1990 II 21544 note N Dejean de la Bâtie; RTD civ 1989, 556 obs P Jourdain. 45   Cass crim 14 January 1970, Bull crim 1970, n°23; RTD civ 1970, 574 obs G Durry. 46   Cass civ (1) 16 June 1969, JCP G 1970, II, 16402, note R Savatier. Cass civ (2) 27 January 2000, JCP G 2000, I, chron 247, n°7, obs G Viney ; JCP G II, 10363, note Ph Conte, RCA 2000 comm n°109; RTD civ 2000, 335, obs P Jourdain. 41 42

118  Duncan Fairgrieve and Florence G’Sell-Macrez infected blood,47 the French courts have tended to link the victim’s ultimate harm to the initial accident, namely the driver’s negligent act.48 The ‘but for’ test has also been criticised in French commentary for not offering any method to select the cause among the conditions that have combined to produce the damage.49 b.  Adequate Causation The theory of adequate causation is generally traced back to Von Bar.50 According to Von Bar, a selection of potential causes is required, and the determining cause of an ultimate event is the factor which departs from the ordinary course of events. Von Bar limited the scope of his theory to human actions so that the cause in question is defined as a human action that has altered the normal course of events: A man is in the legal sense the cause of an occurrence to the extent that he may be regarded as the condition by virtue of which what would be otherwise regarded as the regular course of events in human experience is altered.51

Professor Von Kries52 developed Von Bar’s theory further by stating that a given fact can be considered to be the adequate cause of the damage if, and only if, it has significantly increased the objective probability of occurrence of the damage.53 This test of objective probability thereby allows one to distinguish an adequate cause from only incidental or accidental factors of the injury. Subsequently, some German authors expressed the theory of adequate causation in a less rigorous way than Von Kries. For Träger,54 a condition is an adequate cause of a consequence if it tends, according to human experience and in the ordinary course of things, to be followed by the result in question. Träger, however, underlined the fact that for a condition to be an adequate cause of harm, it must have increased significantly the degree of probability of harm. Träger’s formulation of the test has influenced the French approach. French authors have thus considered that a condition is an adequate cause when it is likely to produce the effect according to the ordinary course of things and in light of the experience of everyday occurrences.55 It is also said that necessary antecedents can be described as adequate causes if they normally produce the harm, as opposed to mere occasions that have been followed by the result because of unforeseen and exceptional circumstances.56

47   CA Paris, 7 July 1989, Courtellemont, Gaz Pal 2, 752 concl Pichot. Cass civ (1) 17 February 1993, JCP G 1994, II, 22226 note A Dorsner-Dolivet. Cass civ (1) 2 July 2002, D 2002, IR, 2517. Rappr Cass crim 11 January1961, Bull crim 1961, n°18. 48   Cass crim 5 October 2004, D IR, 2972. contra Cass crim 15 January 1958 JCP G, 1959, II, 11026, note P Esmein. 49   F G’sell-Macrez, Recherches sur la notion de causalité (n 7) paras 142 and 185. 50   L von Bar, Die Lehre vom Kausalzusammenhange im Rechte, besonders in Strafrechte (1871). 51   ibid 14, cited by HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Oxford University Press, 1985) 466. 52  J von Kries ‘Über den Begriff der objektiven Möglichkeit und einiger Anwendungen desselben’ (1888) Vierteljahrsschrift für wissenschaftliche Philosophie 12. 53   See also HLA Hart and T Honoré, Causation in the Law (n 51) 469. 54   L Träger, Der Kausalbegriffe im Straf- und Zivilrecht (Marburg, Elwert, 1904). 55   Ph le Tourneau, Responsabilité civile (n 12) para 1716. Ph Malaurie, L Aynès, Ph Stoffel-Munck, Droit civil, Les obligations (n 122) para 92. F Terré, Ph Simler, Y Lequette, Droit civil, Les obligations (n 12) para 860. M FabreMagnan, Les obligations 2, La responsabilité (n 13) para 46. Ph Brun, Responsabilité civile extra-contractuelle, 2nd edn (Paris, Litec, 2009) para 227. 56   J Flour, J-L Aubert, E Savaux, Les obligations, vol 2, ‘Le fait juridique’, 13th edn (Paris, Armand Colin, 2009) para 158.

Causation in French Law: Pragmatism and Policy  119 The adequacy theory also has as its purpose to establish a hierarchy between different factors based on the increased probability of the outcome. Underlying the concept is the notion that the factor which has increased by, say, 30 per cent the probability of damage (all other things being equal) is more ‘adequate’ than the factor which increased by only 15 per cent the probability of such an outcome. It should be emphasised that the theory involves linking concrete facts to general causal connections in order to assess the likelihood of damage in the absence/presence of the factor considered. Nevertheless, the idea of probability is rarely mentioned in France. For most French lawyers, a factor is an adequate cause when it usually produces the effect by the normal course of things.57 ii.  A French Contribution The only specifically French theory that has gained support among French jurists58 was developed by Professor Dejean de la Bâtie, who suggested that the test of causation should be established in conjunction with the principles governing legal responsibility.59 According to his view, the criterion of causation should be combined with that of unlawfulness as it is only insofar as an act is illegal or wrong that liability is incurred. Therefore, Dejean de la Bâtie defended a theory of causation based on the ideas of unlawfulness or abnormality. Dejean de la Bâtie called this idea ‘the continued spread of evil’ (l’empreinte continue du mal). His theory was intended primarily to resolve the problem posed by long sequences of events. Dejean de la Bâtie believed that when many events occur between the initial act and the damage, each of these facts must contain an element of unlawfulness that explains the unlawfulness of the subsequent fact. Otherwise, the causal chain must be considered as broken. For instance, if an illegal worker is injured at work, it is not possible to assert that the fact of hiring a person without papers is the cause of the accident.60 Some authors think that Dejean de la Bâtie’s analysis has influenced the courts and cite cases that illustrate, in their view, this influence.61 In a famous case, employees of the magazine Paris Match had been allowed to invite guests to the magazine offices in order to watch the traditional Bastille Day parade on 14 July. Firecrackers were thrown into the crowd from the office, and bystanders were injured. The Cour de Cassation ruled that the magazine should have made arrangements to prevent this kind of behaviour from occurring and that there was a causal relationship between this fault and the victim’s injury.62 On the other hand, the negligence of a person who forgot his chequebook in a phone booth has no causal relationship with the fact that a merchant has been paid with bad cheques by the person who stole the chequebook.63 However, it is difficult to see how Dejean de la Bâtie’s approach is embodied in these cases. In particular, the case law cited could be well explained by using the sine qua non criteria implied by the equivalence theory and by taking into account the willingness of the court to assess the wrongful conduct complained of. Whereas Paris Match was perhaps guilty of reckless behaviour, the owner of the chequebook had   See n 51 above.   See Ph Brun, Responsabilité civile extra-contractuelle (n 55) para 238. 59   N Dejean de la Bâtie in Aubry et Rau, Droit civil français, vol VI-2, ‘Responsabilité délictuelle’, 8th edn (1989) para 74, p 137. 60   Cass soc 7 May 1943, DH 1943, 51. Ph Brun, Responsabilité civile extra-contractuelle (n 55) para 239. 61   Ph Brun, Responsabilité civile extra-contractuelle (n 55) n°242 in fine. Ph le Tourneau, Responsabilité civile (n 12) para 1777 62   Cass civ (2) 4 March 1966, D 1966 somm 110. 63   Cass civ (2) 7 December 1988, Bull civ II, n°246; RTD civ 1989, 557, obs P Jourdain. 57 58

120  Duncan Fairgrieve and Florence G’Sell-Macrez only committed minor negligence. Indeed, these examples show that this approach fails to distinguish clearly between causation and unlawfulness. iii  Case law: pragmatic approach a.  Approach of the Civil Courts Despite the aforementioned criticism, the theory of the equivalence of conditions has commonly been applied in French civil law,64 so that a factor must be a ‘but for’ condition of the damage to qualify as a cause. This means that causation is not established unless it is shown that the damage would have not occurred absent the factor in question.65 Indeed, in one recent decision of the civil law courts, the sine qua non condition has even been referred to as a principle.66 In general, judges do not require that the factor under consideration played a predominant role in the accident or had significantly increased the likelihood of injury. For instance, in the case of a bar tender who had continued to serve drinks to a clearly intoxicated customer who took part in a brawl, had his nose broken, then died of asphyxiation by regurgitation, the court retained a causal link between the bar tender’s negligence and the death.67 Similarly, in cases of multiple causes, when several events have combined to produce the damage, the defendant is liable when his act was a condition sine qua non of the harm.68 Furthermore, in French civil law it is accepted that the presence of other causes does not reduce contribution: each defendant is responsible for everything, even if he has only contributed to the damage. However, the principle of equivalence is not always applied by the courts. The notions of direct link69 or adequate causation are also invoked as well. We have already seen that the death in a house fire of the paraplegic victim of a traffic accident was not held to be linked to the initial accident, even though it was clearly his disability that prevented the victim to escape from the fire.70 Likewise, despite the earlier cases, the Cour de Cassation71 has decided to ignore the victim’s contributory negligence to accidents in cases of blood contamination after a traffic accident.72 Similarly, the Criminal Division of the Cour de Cassation has refused to link the death of the victim of a hospital-acquired infection to the   Viney and Jourdain, Les conditions de la responsabilité civile (n 12) para 355.   eg, there is no causal link if it is proved that, had the victim been fully informed of the risks of surgery, he or she would still have chosen to undergo it. CA, Paris, 20 November 2003, RCA 2004 comm n°76 obs Ch Radé. This applies even when it is established that a notary s compliance with his or her obligation to counsel would not have changed the behaviour of his/her client Cass civ (1) 25 November 1971, Bull civ I, n°296. In a recent article, Mr. Jourdain highlights ‘the principle that any necessary condition of the harm is a legal cause, even if the relationship with the damage is indirect’. According to him, any rejection of the equivalence of the conditions is an exception to the principle though it must be acknowledged that the exceptions to the principle of equivalence of conditions are difficult to determine. P Jourdain, obs under Cass civ (3) 19 February 2003, RTD civ 2003, 508. 66   Le principe de l‘équivalence des causes dans la production du dommage en matière de responsabilité délictuelleé. Cass civ (2) 27 March 2003, Bull civ II, n°76; JCP G 2004, I, 101, n°13, obs G Viney. 67   Cass civ (2) 20 June 2002, RCA 2002 comm n°279. In contrast, a druggist who sold a firecracker to a child in contravention of statute was not convicted because it was not established that the crackers would not have served the same purpose if they had been sold in compliance with the law. Cass (civ) 1 8 April 1986, RTD civ 1987, p 557, obs J Huet. 68   Cass civ (1) 13 December 1988, Defrénois 1989, art 34554, n°56 obs J-L Aubert. 69   Cass civ (3)19 February 2003, n°00-13253, RCA 2003, comm n°125, RTD civ 2003, 508. 70   Cass civ (2) 8 February 1989, JCP G 1990, II, 21544 note N Dejean de la Bâtie; RTD civ 1989, 556, obs P Jourdain. 71   French Supreme Court in civil and criminal matters. 72   Cass civ (2) 20 October 2005, Bull civ II n°274, RTD civ 2006, p 122, obs P Jourdain; D 2006 pan 1930 obs P Jourdain. 64 65

Causation in French Law: Pragmatism and Policy  121 accident which had made hospitalisation necessary.73 Moreover, in medical malpractice cases, the Civil Chambers of the Cour de Cassation have sometimes judged that the victim’s harm cannot be considered as resulting from a previous accident74 or from earlier medical malpractice75 even when such events had made the surgery necessary. Consequently, it is not possible to speak of a uniform application of the equivalence principle: it varies on a case-by-case basis. As to the test of adequate causation, there is some debate about the exact influence of the theory on case law. For leading commentators Viney and Jourdain, adequate causation is applied only in ‘exceptional’ cases that correspond to exceptions to the principle of equivalence.76 However, some other authors have on the contrary opined that the adequate causation theory is the dominant theory in French positive law.77 Indeed, in a recent decision, the Avocat-Général presented this theory as the causal test most frequently applied by courts.78 b.  Approach of the Administrative Law Courts In French administrative law, most commentators have asserted that the stipulation of causal directness is expressed by the courts through the test of la théorie de la causalité adéquate.79 There is a good deal of consensus on the use of this test by the administrative judiciary,80 and indeed there is even one decision in which the courts have referred to the test explicitly.81 However, identifying the exact nature of this test before the administrative law courts has, as with the civil law courts above, not been an easy task. The French public law doctrine has been left with the Herculean task of trawling through the mass of case law in order to identify some guiding principles.82 Some commentators have rejected the idea of any coherent test, concluding that the courts proceed on an ‘empirical’ basis.83 And indeed, one member of the French administrative judiciary has admitted that ‘this topic is not amenable to any systematisation’.84 73   Cass crim 5 October 2004, D IR p 2972. It should be noted that this case applied the new provisions of the law of 10 July 2000, which require the proof of serious misconduct if the causal link is indirect. 74   Cass civ (2) 4 February 1987, Bull civ II, n°38. 75   Cass civ (1) 30 September 1997, Bull civ I n°259; RCA 1997 comm 373. 76   Viney and Jourdain, Les conditions de la responsabilité civile (n 12) para 356. 77   F Terré, Ph Simler, Y Lequette, Droit civil, Les obligations (n 12) para 860. 78   Concl R Kessous, Cass civ (2) 20 November 2003, D 2003, 2902. 79  ‘Adequacy theory’ or ‘adequate cause theory’. See M Deguergue, ‘Responsabilité du Fait des Travaux et Ouvrages Publics’ (1997) Fascicule 930, vol 8, Juris-Classeur Administratif para 149; C Guettier, La Responsabilité Administrative (Paris, LGDJ, 1996) 127; S Rials, Le Juge Administratif Francais et la Technique du Standard (Paris, 1980) 166; Paillet (n 32) para 78; J Moreau, Droit Public: Droit Administratif, vol 2, 3rd edn (Paris, 1995) 598. See also W Van Gerven, J Lever and P Larouche, Tort Law (Oxford, Hart Publishing, 2000) 421. But note P Vialle’s view that the impugned act must first be a necessary condition of the loss and then it must satisfy a variant of the theory of immediate consequence: ‘Lien de Causalité et Dommage Direct dans La Responsabilité Administrative’ (1974) RDP 1243, 1268. 80   See also CG Galmont’s conclusions in Conseil d’Etat (14 October 1966), Marais, (1966) Dalloz Jurisprudence 636; CG Arrighi de Casanova’s conclusions in CE (9 June 1995), Lesprit (1995) AJDA 745. 81   CAA Marseille 3 May 2006, SOCIETE AXA COURTAGE, Req 03MA00948. 82  J-P Taugourdeau, ‘Le Caractère Certain et Direct du Préjudice en Matière de Responsabilité ExtraContractuelle de la Puissance Publique’ (1974) AJDA 508, 518. 83   C Gour, ‘Faute de Service’ in F Gazier and R Drago (eds), Dalloz Encyclopédie de Droit Public: Répertoire de la Responsabilité de la Puissance Publique, para 383; M Deguergue, ‘Responsabilité du Fait des Travaux et Ouvrages Publics’ (n 79) para 147. 84   CG Courtial argued in his conclusions in CE (16 June 1999), Tripot Req 177075 (unreported) that in respect of causation ‘an attempt at theorising such a mode of reasoning doesn’t provide much assistance . . . this topic is unamenable to all systematisation’.

122  Duncan Fairgrieve and Florence G’Sell-Macrez Other commentators have identified a test, with the common reference point as that of normality, as expressed in the stipulation that the harm must have followed the defendant’s act ‘in the normal course of events’.85 Chapus thus asserts that the correct enquiry is whether the defendant’s act could ‘in the normal run of things’ be considered as having played a ‘particular’ role in causing the damage.86 Paillet opines that ‘the judge will undertake a selection of the various [potential causes] of the loss in order to accept those which, in the ordinary course of events, must logically have caused [the loss]’.87 The looseness of the test affords a margin of discretion to the judiciary.88 Chapus himself recognises that the judges combine one part objective reasoning with at least one part intuition: ‘[t]he judicial assessment is not mechanical. It is undertaken with a good deal of freedom, and is influenced by considerations of common sense and subjectivity’.89 Startlingly, Deguergue has described this approach as guided by ‘emotion’ and ‘intuition’: ‘The judiciary, both administrative and ordinary, decide the question whether the occurrence of the loss is a direct consequence of an act by emotion and intuition and not by logical reasoning’.90 It should also be noted that in some administrative law cases, the ‘but for’ test has also been given preference. Indeed, some authors have made a distinction between two stages of the causal enquiry, drawing a dichotomy between the threshold and extent of causal responsibility. The adequacy theory is applied to determine the ‘threshold situation’ where the court is establishing whether the defendant is liable at all. The extent of liability of one who has caused tortious damage is determined by the test of directness, so that the courts hold that too indirect consequences are not recoverable.91

III. Procedure and Causation

Causal theories are not of course applied in a vacuum. Rather than the laboratory conditions of a scientific examination of causation, the legal approach is marked and shaped by a number of extraneous factors, including environmental, cultural, socio-legal and otherwise. In understanding the different tests of causation applied by legal systems, it is import­ ant to place this in a wider context. Beyond the penumbra of substantive law, a whole series of external factors impact on the operation of the rules of causation, and help to explain the content of the French law rules. We will refer to the broader legal architecture in France in order to make good this point, pointing to the rules relating to evidence, as well as other elements in French civil and criminal procedure, such as the role of experts. 85   See CG Galmont’s conclusions in CE (14 October 1966), D Marais (1966) Jurisprudence 636. S Rials opines that causalité adéquate is ‘often known as the theory of normality’ (Le Juge Administratif Francais et la Technique du Standard (Paris, 1980) 166). 86   R Chapus, Droit Administratif Général, vol 1, 15th edn (Paris, 2001) para 1413. 87   M Paillet, La Faute du Service Public en Droit Administratif Francais (Paris, 1980) para 78. 88   CE (1 February 1995), de Bray, (1995) Rec 60. 89  R Chapus, Droit Administratif Général, vol 1 (n 86) para 1413. This is reflected in the conclusions of CG Cazin d’Honincthun in CE (27 March 1985), Henry (1985) Revue française de droit administrative 575, 576, who refers to the ‘subjectivity’ of the causal enquiry. 90   M Deguergue, ‘Causalité et Imputabilité’ (2000) Fascicule 830, vol 8, Juris-Classeur Administratif, para 3. 91   See L Richer, ‘Préjudice Réparable’ in F Gazier and R Drago (eds), Dalloz Encyclopédie de Droit Public: Répertoire de la Responsabilité de la Puissance Publique, para 106; Paillet, La Faute du Service Public en Droit Administratif Francais (Paris, 1980) paras 63 and 439; F-P Benoit, ‘Essai sur les Conditions de la Responsabilité en Droit Public et Privé’ (1957) JCP .I.1351.

Causation in French Law: Pragmatism and Policy  123

A.  Proving Causation in Civil Cases In principle, the burden in French law is upon the claimant to prove that the defendant’s wrongful act has generated his or her loss. In this context, causation is considered as a legal fact (fait juridique) that can be proved by all means (par tous moyens). This implies that all types of evidence are admissible. Moreover, the assessment of the evidence submitted by the claimant falls within the sovereign appreciation of lower courts (appreciation souveraine des juges du fond). However, the French Cour de Cassation has the ability to review the grounds given by trial judges to justify their decisions. In many cases, judges resort to the use of presumptive evidence by basing their decision on Article 1349 of the Civil Code which defines presumptions as ‘the consequences that a statute or the court draws from a known fact to an unknown fact’.92 Specifically, Article 1353 of the Civil Code provides that presumptions ‘are left to the insight and carefulness of the judges, who shall only admit serious, precise and concurrent presumptions’.93 In this context, the Cour de Cassation controls the arguments used by judges to justify the admission or rejection of a causal link. Recently, the Cour de Cassation has made several important decisions to guide lower judges on the use of presumptive reasoning. Before addressing these, we should emphasise that French lawyers are not accustomed to draw a distinction between general causation and specific causation.94 Therefore, the same rules and reasoning apply whatever the type of causal relationship in question. For example, even if the proving of general causation involves relying upon scientific or epidemiological data, French judges will not refer to any probabilistic reasoning but rather base their decision on ‘serious, precise and concurrent presumptions’, as illustrated by the recent case concerning the vaccine against Hepatitis B.95 Reference to presumptions as outlined in the aforementioned Article 1353 of the Civil Code is also used when proving individual or specific causation. Making out individual causation may, in many circumstances, be difficult, even if a general link is established between a product and a given disease. Once the toxicity of a given substance is admitted, it must be shown that the victim was exposed to the toxic substance and that the victim’s damage resulted specifically from exposure. Indeed, the victim’s harm does not necessarily result from the toxic exposure even if this exposure has increased the risk of such harm, in particular, when other factors may explain the disease. For instance, if the victim suffers from lung cancer and has been exposed to asbestos, then one can conclude that exposure has generated the disease. But if the victim has also been smoking for years, it becomes difficult to assert that asbestos is the origin of the disease. In such a case, the appreciation of the facts is left to lower judges’ discretion and the courts may use presumptions provided that they are ‘serious, precise and concurrent’.96 It must also be noted that, for certain kind of injuries, the French legislator has introduced mandatory presumptions, called ‘presumptions of law’ (presomptions de droit). A ‘presumption of law’ requires the judge to assume a certain fact once another fact is established. Under such a mechanism, judicial discretion   ‘des conséquences que la loi ou le magistrat tire d’un fait connu à un fait inconnu’.   ‘sont abandonnées aux lumières et à la prudence du magistrat, qui ne doit admettre que des présomptions graves, précises et concordantes’. 94   See Joseph V Rodricks, ‘Evaluating disease causation in humans exposed to toxic substances’ (2006) 14 Journal of Law and Policy 39. 95   See above. 96   Art 1353 of the Civil Code. 92 93

124  Duncan Fairgrieve and Florence G’Sell-Macrez disappears. Presumptions of law are prescribed by law for diseases resulting from contamin­ ated blood transfusions, such as HIV97 or hepatitis C.98 In such cases, causation is assumed once the victim has proved that a transfusion took place and that there was a subsequent contamination in the absence of other risk factors. The Cour de Cassation has very recently adopted important decisions regarding cases where the defendant cannot be specifically identified because several producers have marketed the toxic product. In the DES litigation, victims are, in most cases, unable to identify the producer who marketed the specific pills absorbed by their mothers during pregnancy. Therefore, the Cour de cassation decided to reverse the burden of proof for the benefit of victims.99 Once the victim has established that her injury results from exposure to DES, she may obtain compensation from any DES producer or even from all DES producers, jointly and severally (responsibility in solidum). To avoid being responsible, producers must establish that their own pills were not those which caused the harm. To date, courts have not yet decided how the burden of compensating the victims will be spread among producers when they are jointly liable. Finally, it should be emphasised that French courts tend to compensate the uncertainty of causal judgments by using widely the concept of loss of chance,100 especially in medical malpractice cases.101 Although the Cour de cassation does not admit officially that compensation for loss of chance is a way to relax the causation requirement,102 it accepts, however, the principle of measuring compensation in reference to the extent of the lost opportunity. When the impact of the defendant’s action (or forbearance) is difficult to measure, the defendant’s liability is modulated according to the probability that he has caused the damage: thus the entire injury is not compensated.103 Khoury has recently pointed out the extreme flexibility of French law on the subject and the proximity of the mechanism of compensation for missed opportunities with liability based on increased risk.104

B.  Systemic Differences In this section, a word will be said about the broader systemic differences of French law which have also impacted upon the causal enquiry in French cases. Amongst the specific

  Art L 3122-2 Code de la santé publique.   Art 102 al 1st, de la loi n° 2002-303, 4 March 2002.   Civ 1 24 September 2009, n°08-10.081 and n°08–16.305, D 2009 AJ 2342, obs Gallmeister; ibid 2010. Pan 49, obs Brun; JCP 2009. 304 obs Mistretta; ibid 383, note Hocquet-Berg; RLDC 2009/65, n° 3605, obs Bugnicourt; RLDC 2010/01, n°67–10, note B Parance. Civ 1 28 January 2010, n°08–18837. 100   F G’sell-Macrez, Recherches sur la notion de causalité (n 7) para 455–457. See also J Boré, ‘L’indemnisation pour les chances perdues, une forme d’appréciation quantitative de la causalité d’un fait dommageable’, (1974) JCP G, I, 2620. G Mémeteau, ‘Perte de chances et responsabilité médicale’ (1997) Gazette du Palais 7, II, 1367. 101  In a seminal case, a doctor had committed an error of diagnosis on a child who remained crippled. The judges condemned the doctor to pay damages for the loss of ‘chances of cure’ Cass civ (1)14 December 1965, JCP G, 1966, II, 14753, note R Savatier. See also, Cass civ (1) 27 January 1970, Bull I n°37, 31, JCP G 1970, II, 16422, note Rabut. Viney and Jourdain, Les conditions de la responsabilité civile (n 12) para 370. 102   Cass civ (1).17 November 1982, Bull civ I, n°133; JCP 1983, II, 20056, note M Saluden; D 1984, 305, note A Dorsner-Dolivet. Comp CA Versailles 28 November 2003 D 2004 jur p 2814, note P Hennion-Jacquet. 103   Cass (civ) 1 14 December 1965, Cass civ (1) 27 March 1973, Bull civ I, n° 115. Cass civ (1) 8 July 1997, Bull Civ I n° 239. Cass civ (1)10 July 2002, Bull Civ I, n°197. See in administrative law: CE, s. 21 December 2007, Centre Hospitalier de Vienne c/ M Joncart, req n° 289328, concl T Olson, RFDA 2008, 348. 104   L Khoury, ‘Causation and Risk in the Highest Courts of Canada, England and France’ (January 2008) LQR, 124, 103–31. 97 98 99

Causation in French Law: Pragmatism and Policy  125 characteristics of French system of civil justice, two features will be underlined here: the predominance of written, documentary evidence and the role of the expertise.105 Traditionally, French civil procedure is focused on a written procedure with, at its centre, the judicial dossier composed of the parties’ respective written pleadings, supplemented by documentary evidence.106 Whilst the courts may indeed hear witnesses (as part of an enquête), this in practice rarely actually occurs. Unlike the common law trial, the French civil justice system is characterised by a distrust of testimonial over documentary evid­ ence.107 This poses challenges for the court when deciding upon technical issues (which may often arise in respect of causation). As a consequence of the document-based approach, the court does not hear evidence on technical matters and instead relies upon outside experts to complete the picture, as we shall now see.108 Whilst the litigating parties have primary responsibility in France for assembling the relevant evidence,109 the parties may nonetheless request the judge to undertake measures for collecting evidence,110 known as mesures d’instruction. Indeed, a decree of 13 October 13111 reformed the Code of civil procedure by creating a ‘juge de la mise en état’ whose task is to prepare the trial and manage the inquiry through various measures. The most promin­ ent of these is the expertise whereby, upon request of the parties, an expert is formally instructed by the court to produce an expert report on specific issues relating to the dispute.112 This court-appointed expert plays an important role in civil procedure, and is often appointed at a pre-trial stage as a conservatory measure by means of the expedited référé procedure for urgent matters.113 In principle, the court is not bound by his or her resultant opinion: Article 246 of the Code of Civil Procedure provides that the judge is not obliged to follow the findings or conclusions of the expert.114 In other words, the judge must always exercise his discretionary right to assess the implications of the findings and opinions of the expert – otherwise, he incurs the censure of the Cour de Cassation.115 However, the views of scientific experts seem to have a significant impact on the judiciary. This is all the more so in technical cases, including of course causal uncertainty in areas such as environmental litigation, pharmaceutical cases116 or, as Khoury has shown,117 in the context of medical liability. 105   See generally, J Bell, S Boyron and S Whittaker, Principles of French Law (n 16) 85–126; L Cadiet, J Normand, S Amrani-Mekki, Théorie générale du procès, (Presses universitaires de France, coll Thémis, 2010). S Guinchard, F Ferrand, C Chainais, Procédure civile, Droit interne et droit communautaire, 29th edn (Précis Dalloz, 2008). 106   See generally, J Bell, S Boyron and S Whittaker, Principles of French Law (n 16) 90. 107  See generally, J Beardsley, ‘Proof of Fact in French Civil Procedure’ (1986) (34) American Journal of Comparative Law 459, 483. 108   See J Bell, S Boyron and S Whittaker, Principles of French Law (n 16) 106. 109   See J Beardsley, ‘Proof of Fact in French Civil Procedure’ (n 107) 459. 110   Note also here that the judge (juge de la mise en état) who decides upon the procedural issues of ordering measures for collecting evidence is not the same judge who will decide on the ultimate substantive issues after the final hearing. 111   H Motulsky, ‘La réforme du code de procédure civile par le décret du 13 octobre 1965 et les principes directeurs du procès’ (1966) JCP, I, 1996 112   The court chooses an expert for the purposes adumbrated in the court order, which additionally defines the task to be undertaken. Under Art 234 of the Civil Procedure Code, parties may challenge the judge’s appointment of an expert. See F Ferrand, V°Preuve, Répertoire civil (Dalloz) n°472 and following. 113   Before the juge des référés, see Art 145 Civil Procedure Code. 114   Le juge n’est pas lié par les constatations ou les conclusions du technicien. 115   Cass civ (3) 23 February 1972, Bull civ III, no 130; 27 June 1972, Bull civ III, no 427; Cass. (com) 26 May 1964, Bull civ III, no 271. F Ferrand, n°475. 116   See generally, R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing, 1999). 117   L Khoury, Uncertain Causation in Medical Liability (Oxford, Hart Publishing, 2006).

126  Duncan Fairgrieve and Florence G’Sell-Macrez In fact, the distribution of roles between the court and the expert is not always crystal clear. Indeed, despite the injunction found in Code of Civil Procedure that ‘[the expert] must never make any assessments of a legal nature’,118 one comparatist has observed that ‘in many cases, the expertise is a kind of mini trial’.119 It may thus be wondered if the deference accorded by the judge to the expert as a ‘professional on technical issues of fact’, ‘reduces the juge de l’audience to a very passive role in respect of the facts’.120 As Khoury has illustrated in her seminal work in this sphere,121 whilst courts generally try to assert their independence vis-à-vis the views of expert opinions, experience in fact shows that the impact of science on the judicial determination of causation is significant. The French system is an example par excellence of this phenomenon, as the French system of civil justice is particularly marked by the judicial delegation of technical matters to a sage, the court-appointed expert. Nevertheless, the Cour de Cassation has recently illustrated the willingness of the French judiciary not to follow systematically the view of appointed court-experts. Litigation concerning the Hepatitis B vaccine has given rise to important decisions of the Cour de Cassation in a context where there is no tangible scientific evidence of the vaccine’s toxicity.122 Confronted with victims of neurological disorders such as multiple sclerosis, French judges have gradually changed their minds. In 2003, the Cour de Cassation refused to accept the existence of a causal link between vaccination against Hepatitis B and multiple sclerosis because of the scientific uncertainty of that link as highlighted by experts.123 However, in 2008, the Cour de Cassation handed down six important decisions124 in which it was accepted that such a causal link could be established by ‘serious, precise and concurrent’ presumptions, notwithstanding scientific uncertainty and lack of conclusive statistic data. Since then, the Cour de Cassation has had the opportunity to reaffirm its position on several occasions in cases concerning multiple sclerosis125 as well as other neurological dis-

  Art 238(3) Civil Procedure Code.   J Beardsley, ‘Proof of Fact in French Civil Procedure’ (n 107) 483. 120   J Bell, French Legal Cultures (London, Butterworths, 2001) 91. See also, L Khoury, Uncertain Causation in Medical Liability (n 117) 63–64. Compare with E Jeuland, ‘V° Expertise’ in L Cadiet (dir) Dictionnaire de la Justice, (Presses universitaires de France, 2004) 503. 121   See in particular, L Khoury, Uncertain Causation in Medical Liability (n 117). 122   See, further, Richard Goldberg, ch 8, pp 000. 123   Cass civ (1) 23 September 2003; Resp civ et assur 2003, chron 28, Ch Radé; D 2004, 898, note Y-M Sérinet and R Mislawski; JCP G 2003, II, 10179; RLDC 2004, 11, chron S Hocquet-Berg; RTD civ 2004, obs P Jourdain. 124  Cass civ (1) 22 May 2008, 6 judgments, n° 05–20.317, n° 06–14.952, n° 05–10.593, n° 06–10.967, n° 06–18.848, n° 07–17.200; Resp civ et assur 2008, étude 8, Ch Radé; RDSS, 2008, 578, J Peigné; JCP G 2008, II, 10131, L Grynbaum; RTD civ, 2008, 492, P Jourdain; Gaz Pal 9 October 2008, n° 283, 49, S Hocquet-Berg; RDC 2008, 1186, J-S Borghetti; JCP G 2008, I, 186, P Stoffel-Munck. See further, Richard Goldberg, ch 8, pp 000. 125   ‘si les études scientifiques versées aux débats par la société Sanofi Pasteur MSD n’ont pas permis de mettre en évidence une augmentation statistiquement significative du risque relatif de sclérose en plaque ou de démyélinisation après vaccination contre l’hépatite B, elles n’excluent pas, pour autant, un lien possible entre cette vaccination et la survenance d’une démyélinisation de type sclérose en plaque ; qu’ayant, ensuite, relevé que les premières manifestations de la sclérose en plaque avaient eu lieu moins de deux mois après la dernière injection du produit ; que ni Mme X . . . ni aucun membre de sa famille n’avaient souffert d’antécédents neurologiques, et que dès lors aucune autre cause ne pouvait expliquer cette maladie, dont le lien avec la vaccination relevait de l’évidence selon le médecin traitant de Mme X . . ., la cour d’appel, qui a souverainement estimé que ces faits constituaient des présomptions graves, précises et concordantes, a pu en déduire un lien causal entre la vaccination de Mme X . . ., et le préjudice subi par elle’: Cass civ (1) 9 July 2009, n° 08-11.073; Gaz Pal 13 August 2009, n° 225, 9, opinion A Legoux; Resp civ et assur 2009, study 13, Ch Radé. 118 119

Causation in French Law: Pragmatism and Policy  127 orders.126 In such cases, two considerations are taken into account by judges: first, the fact that no other factor could explain the disease (victim in good health, no medical history); second, the temporal proximity between the injection of the vaccine and the appearance of the first symptoms. It should also be mentioned here that the Conseil d’Etat has adopted a similar position and criteria in litigation arising from the compulsory vaccination against Hepatitis B.127 However, in other recent vaccine cases, French judges have denied the existence of a causal relationship when other factors could explain the disease128 or when the evidence brought by the claimants does not appear sufficient to constitute ‘serious, precise and concurrent presumptions’.129 In light of this case law, French authors have started to distinguish between scientific causation and legal causation.130 Here it should be highlighted that French courts have never defined any criteria for scientific or epidemiological studies to be eligible: there is no Daubert131 decision in France.132 Moreover, the constant reference to ‘serious, precise and concurrent presumptions’ seems somehow to prevent French courts from adopting probabilistic reasoning regarding causation.133 Judges may simply decide that proof of causal connection is provided on the basis of presumptions following from circumstances specific to the victim, even though the ‘expertise’ is not conclusive and even if such a general causal link is not established on a scientific and/or statistic point of view.

IV. Policy Factors in Determining Causation

The orthodox doctrinal view has generally been that for constitutional and institutional reasons, the French judicial decision-making is not marked by the policy discourse which is such a familiar feature of the reasoning of common law judges.134 Certainly, the formalistic style of French judgments leaves little scope for the intricate discussion of competing policy 126   ‘Qu’en exigeant une preuve scientifique certaine quand le rôle causal peut résulter de simples présomptions, pourvu qu’elles soient graves, précises et concordantes, la cour d’appel a violé les textes susvisés’: Cass civ (1) 25 June 2009, n° 08–12.781; Bull civ 2009, I, n°141; JCP G 2009, n°41, 308, note P Sargos; Resp civ et assur 2009, comm 293. 127   CE, 9 March 2007, n° 267635; JCP A 2007, 2108, note D Jean-Pierre; 2277, note S Carpi-Petit, act 297; JCP G 2007, II, 10142, note A Laude; AJDA 2007, 861 concl T Olson; A Rouyère, ‘Variations jurisprudentielles à propos du lien de causalité entre vaccination contre l hépatite B et sclérose en plaque. Questions de méthode’ (2008) Revue française de droit administrative 1011. CE, 18 February 2009, Gaz Pal 19–20 June 2009, 20, note D Cristol. 128   Cass civ (1) 22 January 2009, n°07–16.449, Bull civ 2009, I, n° 11 (maladie de Guillain-Barré); RDC 2009, n°3, 1028, obs O Deshayes. 129   Cass civ (1) 24 September 2009, n° 08–16.097, Resp civ et Assur 2009, comm 328, obs Ch Radé. 130   Ph Brun, ‘Causalité juridique et causalité scientifique’ (July/August 2007) RLDC n°40 (supp) 2630. 131   Daubert v Merrell Dow Pharmaceuticals 509 US 579 (1993). 132  R Encinas de Munagorri, ‘La recevabilité dLune expertise scientifique aux Etats-Unis’ (1999) Revue International de Droit Comparé, vol 51, n°3, 621–32. M-A Hermitte, ‘V°Science’ in L Cadiet (dir), Dictionnaire de la Justice (Presses universitaires de France, 2004) 1204. O Leclerc, Le juge et l’expert-contribution à l’étude des rapports entre le droit et la science, préf A Lyon-Caen, (LGDG Bibl droit privé t. 443, 2005). 133   See KM Clermont and E Sherwin, ‘A comparative view of standards of proof ” (spring 2002) American Journal of Comparative Law vol 50 n°2 243–75. 134   For the traditionalist view, see the judicial style guide of P Mimin, Le Style des Jugements, 4th edn (Paris, 1978). Counterpoints to this account can be seen in comparative law studies: J Bell, ‘Reflections on the Procedure of the Conseil d’Etat’ in G Hand and J McBride, Droit Sans Frontières (Birmingham, 1991) and more generally in J Bell, Judiciaries within Europe: a Comparative Review (Cambridge University Press, 2006), as well as M Lasser, Judicial Deliberations: A Comparative Analysis of Judicial Transparence and Legitimacy (Oxford, Oxford University Press, 2004). See also Ch Jamin and Ph Jestaz, La doctrine (Dalloz Coll Méthode du droit, Paris, 2004).

128  Duncan Fairgrieve and Florence G’Sell-Macrez factors.135 However, recent studies have shown that behind the ‘terse’ judgments,136 it is clear from other sources such as the conclusions or doctrinal commentary, that in reality many similarities can be drawn with the common law judicial discourse. The issue of causation provides an example of the way in which policy influences judicial decision-making. Causation cannot of course be applied in a formulaic or mechanistic way, and despite the abstract theories, issues of policy inevitably colour the judicial appreciation of the causal link. This is particularly true in French law since no clear theory of causation has been privileged, thereby allowing for the introduction of policy factors.137 Different purposes may be served by judges in the making of causation judgments but courts usually try to favour the goal of compensating the victims.138 The well-known wrongful life case of Perruche is the classic example of this tendency,139 as is clear from the conclusions of the Avocat Général,140 as well as from the ample literature following the decision and the subsequent legislative amendment.141 Policy concerns have also shaped French civil courts approach to car accidents which have led to loss due to medical treatment: if the victim is taken to the hospital with a minor injury from the accident and then is victim of a medical error or of a contaminated blood transfusion,142 judges will accept the link of the final damage to the accident in order to allow the victim to seek compensation from the motorist. For policy reasons, judges decide, in such cases, to transfer the burden of medical risk onto motor insurance. Judges may also choose, because of a certain conception of social justice,143 to condemn those who are best able to assume the cost, either because they are insured, or because they have some financial strength. For example, judges are very strict with the French National railway company (SNCF) which is responsible for assaults committed in trains144 or when the victim is obviously at fault.145 Administrative law’s approach to causation similarly illustrates this phenomenon of policy factors. There is evidence that the notion of causation is used by the French administrative courts in a very specific manner, so as to circumvent the scope and extent of liability.146 The use of causation as a control mechanism is illustrated by the governmental liability cases, where a study undertaken by one of the present authors has shown that the causal enquiry plays an extremely important role in circumventing the liability of the

135   See however the pioneering work of, in particular, G Canivet and N Molfessis, ‘La politique jurisprudentielle’ in La création du droit jurisprudentiel Mélanges en l’honneur de Jacques Boré (Dalloz, 2007) 79–97; G Canivet, ‘Variations sur la politique jurisprudentielle: Les juges ont-ils une âme?’ in M Andenas and D Fairgrieve (eds), Tom Bingham and the Transformation of the Law (Oxford, Oxford University Press, 2009). 136   M Wells, ‘French and American Judicial Opinions’ (1994) 19 Yale Journal of International Law 81, 82; J Bell, ‘Principles and Methods of Judicial Selection in France’ (1988) 61 Southern California Law Review 1757, 1764. 137   M Fabre-Magnan, Les obligations 2, La responsabilité (n 13) para 49. 138   F G’sell-Macrez, Recherches sur la notion de causalité (n 7) para 588 and following. para 595 and following. 139   Cass Ass Plen 17 November 2000 Perruche, JCP G 2000, II, 10438, rapp P Sargos, concl J. Sainte-Rose, note F Chabas Gazette du Palais, 24–25 January 2001; D 2001 Jurisprudence 316. 140   Cass Ass Plen 17 November 2000, Perruche, Gazette du Palais, 24–25 January 2001; D 2001 Jurisprudence 316. 141   Loi n° 2002-303, 4 March 2002. 142   See eg, Cass civ (2) 27 January 2000, JCP G 2000, II, 10363 note Ph Conte; Cass (civ) 1 17 February 1993, JCP G. 1994 II, 22226, note A Dorsner-Dolivet. 143   G Viney, Introduction à la responsabilité, 3rd edn (LGDJ, Paris, 2008) para 38. 144   Cass civ (1)12 December 2000, D 2001, jur p 1650 note C Paulin. Cass civ (1) July 2002, D 2002 jur, 2631 note J P Gridel; RTD civ 2002, 821 obs P Jourdain. 145   Cass civ (2) 23 January 2003, D 2003, 2465, note V Depadt-Sebag. 146   Particularly in cases of pure economic loss, see D Fairgrieve, State Liability in Tort: A Comparative Law Study (2003, Oxford University Press) 170–71 and 200.

Causation in French Law: Pragmatism and Policy  129 administration.147 This research has been borne out by French writers. Deguergue in particular has commented that: ‘[I]n the investigation for the determinant causal factor, the administrative judge pursues policy imperatives deriving from his dual desire to compensate the victims of loss caused by the admin­istration whilst [at the same time] adjusting the liability of the latter in order to avoid a dispro­portionate sanction and not to overburden public finances.’148

From this perspective, the test of causation in French administrative law may well be seen as a counterbalance to the extensive interpretation of the notion of fault (as well as the many heads of no-fault liability), thereby balancing the liability equation with an effective control mechanism of causation. One final point to make is that the French examination of causation is by no means impervious to outside influences. Indeed, it might be thought that the complexity of causal analysis makes it attractive for the judiciary to examine how courts elsewhere have resolved similar solutions.149 In the controversial area of wrongful life, in two actions brought independently before the Conseil d’Etat and Cour de Cassation, both courts were referred to comparative law solutions respectively in the conclusions of CG Pécresse150 and Avocat Général Sainte-Rose.151 Moreover, very recently, the decisions made by the Cour de Cassation in DES cases have led French authors to look at the American decision Sindell v Abbott Laboratories152 and to question the desirability of apportioning the burden of compensating the plaintiff according to the market shares held by each defendant.153 This last point may well be raised soon before French courts. However, French traditions and the very complexity of the concept of causation may pose barriers to such a transplant.

V. Conclusion

The requirement of causation in French law is structured ultimately around the basic requirement of the sine qua non test. However, over and above this initial requirement, we have seen that pragmatic and policy considerations have had a significant impact on the case law, allowing the French courts to relax the exigencies of the ‘but for’ test when evidential obstacles are problematic, either by a deft use of the concept of loss of chance, or by the development of presumption of causality in certain cases. It is thus evident that beyond the technical debate of the application of intricate tests of causation, the modern French approach to causation provides revealing glimpses about the role and function of the French judiciary, the French civil justice system, as well as the policy issues underlying the substantive law.

 ibid.   M Deguergue, ‘Causalité et Imputabilité’ (n 90) para 45. See also C Guettier, La Responsabilité Administrative (Paris, 1996) 17, 126–27. 149   See eg, Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32 [32] (Lord Bingham). 150   CE 14 February 1997, Epoux Quarez, RFDA 1997.375, 379–80. 151   Cass Ass Plen 17 November 2000, Perruche, Gazette du Palais, 24–25 January 2001; D 2001 Jurisprudence 316. 152   607 P.2d 924 (Cal 1980). 153   S Hocquet-Berg, JCP G 2009, 383. 147 148

7 The Role of Scientific and Statistical Evidence in Assessing Causality A PHILIP DAWID*

I. Introduction

‘Causality’ is a big topic in Statistics and cognate fields such as econometrics and epidemiology, but there is remarkably little overlap between the frameworks, assumptions and approaches common in these disciplines and those commonly taken by lawyers and legal philosophers. One of the most obvious differences is that, as a matter of course, statisticians think of causality as an uncertain relation that needs to be described and qualified probabilistically, while lawyers may be more naturally inclined to a fundamentally deterministic view. It is my aim here to introduce lawyers to statistical ways of thinking about causality, while at the same time probing just how statistical evidence might or might not be used in resolving legal problems of causality. We shall identify many subtleties and snakes in the grass, and I hope this will also serve as a caution to be alert to the dangers.

II.  Varieties of Causal Questions

A.  Effects of Causes (EoC) and Causes of Effects (CoE) We shall deal with causal queries where there is no ambiguity about the events at issue, but there is uncertainty about their relationship: in particular, whether or not that relationship can be regarded as ‘causal’. So, important though they are, we shall not here be considering questions such as ‘Who killed [caused the death of] cock robin?’, where there is no dispute about the causal relation, but there is uncertainty about the identity of the causal agent.1 Nor shall we address such legally important questions as allocation of responsibility, or identification of the ‘actual cause’, among several coexisting causal agents. *  I have benefited from comments of Sander Greenland, Judea Pearl, William Twining and Carlo Berzuini on an earlier draft. 1   See A Philip Dawid and Julia Mortera, ‘Coherent Analysis of Forensic Identification Evidence’ Journal of the Royal Statistical Society Series B (1996) 58, 425–43 and A Philip Dawid, ‘Statistics and the Law’ in Andrew Bell, John Swenson-Wright and Karin Tybjerg (eds), Evidence (Cambridge, Cambridge University Press, 2008) 119–48, for statistical approaches to that topic.

134  A Philip Dawid Within this limited ambit it is still important to distinguish two types of causal query which, although not entirely unconnected, are nevertheless very different, both in form and in the type of answer they require: Effects of Causes (EoC). I have a headache. I am wondering whether to take aspirin. Will that cause my headache to disappear? Causes of Effects (CoE). I had a headache and took aspirin. My headache went away. Was that caused by the aspirin?

EoC type questions have the form ‘Does [will] A cause B?’. They are also known as problems of type or general causation. CoE questions have the form ‘Did A cause B?’, and are also known as problems of token or individual causation. Thus the questions ‘Does smoking cause lung cancer?’, ‘Joe is a smoker, will that cause him to develop lung cancer?’, and ‘If I take up smoking, will that cause me to develop lung cancer?’ are all of EoC type; while the question ‘Joe is a smoker who developed lung cancer. Was that caused by his smoking?’ is of CoE type. From one point of view the difference between EoC and CoE questions is one of tense – EoC looks forwards from present A to future B, while CoE looks backwards from present B to past A. A more helpful distinction is epistemological, in terms of what is known when the question is posed: when we ask the CoE question, we already have knowledge of the outcome B, as well as the putative cause A; for a EoC question, we do not yet know the outcome B (and even the putative cause A may be yet to be realised). This apparently slight difference in the knowledge base leads to very significant differences in how the two questions are to be formulated and addressed. As a very loose generalisation, Science is more typically concerned with general, EoC type, queries, and Law with individual, CoE type queries. Because of this mismatch we must be particularly careful when we try to bring scientific evidence and reasoning to bear on questions of legal causality.

B.  Contrastive Causality Much philosophical discussion of legal causation centres on the difficulty of identifying the ‘true’ or ‘actual’ cause of some event, from among the many antecedent events that might have some claim to this title. While by no means belittling the importance of addressing such conundra, we shall here take them as settled, and concentrate on assessing the evidence for a clearly enunciated causal claim.2 Such a claim is most usefully phrased as ‘contrastive’, having the pattern ‘Is [was] it A1 rather than A2 that will [did] lead to B?’. In all cases we must take care that the ingredients A1, A2, B are defined in a way suitable to our purpose – again, an important and highly non-trivial issue that, for our purposes, we shall take as settled. For CoE analysis we will know A1 (eg Joe is a smoker) and B (Joe developed lung cancer) as facts, but need to specify an alternative ‘foil’ A2 (eg Joe never took up smoking), as a – necessarily counterfactual 3 – alternative to A1. For EoC analysis we do not yet know the outcome B, which can be regarded as a random variable (with a clearly specified set of potential realisations). The putative cause A1 might already have been activated (Joe is   Of course, if several competing claims are in play, we can conduct this investigation for each of them.   Because counter to known facts!

2 3

Statistical Evidence for Causality  135 already a smoker), in which case we again need to specify an appropriate foil A2 (Joe did not in fact take up smoking); or both A1 and A2 could be alternative future hypotheticals (I will/ will not take up smoking). Although the term ‘counterfactual’ is often used to describe such future hypotheticals, it is not really appropriate as there is no incompatibility with known facts. I believe there is considerable value in making a clear distinction between counterfactuals and hypotheticals.4

III.  Effects of Causes: Experimentation

A. Introduction Data, even ‘scientific’ data, never speak for themselves: before we can sensibly interpret the data, it is vital to know, and properly take into account, the nature, protocol and properties of the study leading to the generation and/or collection of the data. Interpretation is most straightforward when the data arise from a properly designed and conducted experimental study. The statistical theory of experimental design5 is one of the crowning glories of twentieth-century science, just as its special case, the randomised controlled clinical trial,6 is the most important medical advance of that century. How many people know that? At its most basic, careful experimentation to investigate the effects of various treatments or actions involves the following steps: Treatment protocol, describing, in adequate detail, the treatments and procedures that are to be compared. Assembly of a suitable set of ‘experimental units’. Stratification (possibly) of these units according to various pre-existing attributes. Randomisation (taking appropriate account of the stratification) of the assignment of treatments to units.

When it comes to analysis of the data, the fact that those data were generated from a well-designed and well-conducted experiment will support robust conclusions, unaffected by disputable assumptions7 about the relationships between the variables measured (and unmeasured). i.  Example 1 A simple experiment is conducted to investigate the effectiveness of aspirin as a cure for headache. A test group of 200 individuals is assembled, and, by a process of physical randomisation, 100 of these are assigned to the treatment group, and given aspirin tablets, and 100 to the control group, who are given chalk tablets. In order to eliminate subjective biases, the two kinds of tablets are indistinguishable, and no-one apart from the statistician conducting the randomisation knows which patient was assigned which treatment (a ‘double-blind’ trial). 4   A Philip Dawid, ‘Counterfactuals, Hypotheticals and Potential Responses: A Philosophical Examination of Statistical Causality’ in Federica Russo and Jon Williamson (eds), Causality and Probability in the Sciences (London, College Publications, 2007) 503–32. 5   Ronald A Fisher, The Design of Experiments (Edinburgh, Oliver and Boyd, 1935). 6   Austin Bradford Hill, ‘The Clinical Trial’ (1951) British Medical Bulletin 7 278–82. 7   Like all broad generalisations this is an oversimplification, but it will serve, for our purposes, as a contrast with the case of observational studies.

136  A Philip Dawid The patients take their assigned tablets the next time they get a headache, and record how long it is until the headache has gone. For the sake of simplicity, attention concentrates on the outcome ‘Headache disappears within 30 minutes’, coded as ‘recovery’. After the experiment is over the results are as displayed in Table 1. Table 1: Results of aspirin trial Recover

Not recover

Total

Chalk

12

88

100

Aspirin

30

70

100

Assuming the trial is large enough for these results to be meaningful,8 we can regard the contrast between the 30 per cent recovery rate for aspirin and the 12 per cent rate for the selected foil, chalk, as a quantitative measure of the ‘causal effect’ of aspirin in bringing about recovery from headache. The purpose of the randomisation is to ensure that, when we form the contrast between the aspirin and the chalk groups, we are ‘comparing like with like’, and that there are no systematic pre-existing differences between the two groups that might be alternative explanations of the observed difference in response. For example, if we formed the treatment group solely of migraine sufferers (whose headaches are generally very long-lasting) and the control group solely of non-sufferers, we might well see the treatment group performing worse than the controls – though this would simply be a reflection of the different constitutions of the groups, and not causally ascribable to the treatment received. When we suspect that certain measured factors (such as migraine-sufferer status, sex, age, . . .) may affect response to treatment, we might take this into account in the design, by stratification. Thus we might divide our 200 patients into four groups: male migraine sufferers, female migraine sufferers, male non-sufferers, female non-sufferers; and (to guard against possible imbalances in other, unmeasured, disturbing factors) randomise within each group. Whether or not we have conducted such ‘pre-stratification’ at the design stage, when the data are in it would be sensible to conduct a ‘post-stratified’ analysis, looking at the results within each of these groups, rather than overall. This would obviously be valuable if we wanted to understand the ‘causal effect’ of aspirin for selected kinds of patients (male non-sufferers, . . .). In a non-prestratified study it can also correct for any (randomly generated) imbalance between the treatment and control groups for the factors considered. (An important caveat here is that any one data set can only take so much detailed probing before its apparent message becomes more an artifact of random noise than a reliable measure of an underlying signal.)

B.  What Kind of Science is This? Some may think of the clinical trial as a very poor example of ‘science’. It appears to be a very blunt instrument, making no enquiry into the pharmacologicial or psychological processes whereby aspirin might influence the sensation of headache – effectively, it is a ‘black-box’ approach, perhaps more akin to engineering than science. Well, let us not fight   We shall not here be dealing with issues of ‘statistical significance’.

8

Statistical Evidence for Causality  137 over the nomenclature. What is important is that this empirical black-box approach does deliver the desired results, allowing us to predict what might happen in future administrations of aspirin. There is of course nothing to stop us peering into the box, and performing experiments on its component systems. But, because we may be uncertain or mistaken about their completeness and interrelations, trying to piece together the results of such sub-experiments can often lead to overall conclusions that are less reliable than those grounded in the simple (but far from simplistic) experimental study of the behaviour of the system as a whole.

IV.  Effects of Causes: Observational Studies

While experimentation is the scientific ideal, there are many cases where it is simply not available, be it for pragmatic, economic or ethical reasons, and recourse to pure observation becomes a pragmatic necessity. However, making defensible inferences from evidence is problematic when that evidence is obtained from observational studies. Suppose we wanted to to apply the trial design of Example 1 to the problem of assessing the effect of smoking on lung cancer. We would aim to assemble a group of 200 youngsters, and randomly assign half of them to take up smoking, half to refrain. After suitably many years have passed, we would count how many in each group developed lung cancer, and compare the percentages to get an estimate of the causal effect of smoking on developing lung cancer. The manifold pragmatic and ethical obstacles to such a study should be self-evident. In such cases we may have no alternative but to rely on purely observational studies of various kinds. For example, we could conduct a ‘retrospective study’, collecting a group of individuals who have contracted lung cancer (‘cases’) and another group who have not (‘controls’), and identifying each individual as a smoker or non-smoker. However, in an observational study, because we have had no control over the application of the treatment, we can not be sure that we are ‘comparing like with like’ – failing which, an observed association might be an artifact of the differing constitutions of the treatment groups, rather than ascribable to the different treatments applied. In addition, a particular problem of the retrospective study design, where we ourselves have fixed the numbers of cases and controls, is that (even if we are indeed comparing like with like) we cannot obtain meaningful estimates of the proportions of smokers or of non-smokers who go on to develop cancer. The best we can do is turn this on its head, estimating the proportions of cases or of controls who had smoked. While these are of no intrinsic interest in themselves, a comparison between them can give some indirect and imperfect information about whether there is a positive association (which might or might not have a causal interpretation) between smoking and lung cancer.

138  A Philip Dawid

V. Cautionary Tales

A.  Some Examples i.  Example 2 Berkeley Graduate Admissions9 Table 2 summarises the overall acceptances of male and female candidates for admission to the Graduate School of the University of California at Berkeley in 1973. The large difference between the admission rates for men and for women might be taken as prima facie evidence of a bias against admitting women. One simple way of quantifying this imbalance is to note that, if the overall acceptance rate of 41 per cent applied equally to men and women, we would have expected 1771 female admissions. So the shortfall is 1771 − 1494 = 277 fewer female admissions than would be expected under a non-discriminatory regime. Table 2: Graduate admissions, Berkeley, 1973 Accept

Reject

Total

% admitted

Male

3738

4704

8442

44

Female

1494

2827

4321

35

Total

5232

7531

12763

41

In an attempt to uncover the sources of this bias, the admission figures were examined separately for the 85 academic departments to which application was made. For each department individually, the shortfall or excess in female admissions was computed, as described above. On aggregation across departments, it turned out there was now no shortfall, but rather an excess of 60 women above what would be expected in the absence of sex discrimination; and in no department was there a shortfall of any significant degree. What are we to make of these competing and confusing messages from the same data? ii.  Example 3 Hormone Replacement Therapy and Coronary Artery Disease10 A substantial body of observational research on postmenopausal hormone replacement therapy, over a number of years, suggested that there could be a 40−50 per cent reduction in coronary heart disease incidence among women using these preparations. To investigate this further, a large-scale clinical trial of oestrogen plus progestin (the Women’s Health Initiative trial, following up 16,608 postmenopausal women aged 50−79 years) was conducted – in parallel with a new observational study of 53,054 women. The clinical trial found an elevated incidence of CHD among those taking HRT. The age-adjusted estimates of the incidence of coronary heart disease, stroke and venous thromboembolism were 40−50 per cent lower in the observational study than in the clinical trial. 9   Peter J Bickel, Eugene A Hammel and J William O’Connell ‘Sex Bias in Graduate Admissions: Data from Berkeley’ (1975) Science 187, 398–404. 10   Ross L Prentice, Robert Langer, Marcia L Stefanick, Barbara V Howard, Mary Pettinger, Garnet Anderson, David Barad, J David Curb, Jane Kotchen, Lewis Kuller, Marian Limacher and Jean Wactawski-Wende ‘Combined Postmenopausal Hormone Therapy and Cardiovascular Disease: Toward Resolving the Discrepancy between Observational Studies and the Women’s Health Initiative Clinical Trial’ (2005) American Journal of Epidemiology 162, 404–14.

Statistical Evidence for Causality  139

B. What Is Going On? The above examples indicate the problems of taking observational data at face value. These problems are due to the existence of ‘confounding’ variables, that disturb the relationship between a putative causal factor and its measured effects. Table 3: Confounding by department Department

% admitted

% female applications

A

64

12

B

63

4

C

35

65

D

34

47

E

25

67

F

6

48

Table 3 goes at least part way to explain the odd behaviour in Example 2. For each of the six largest admitting departments, coded A to F, it lists their overall admission rate, and the percentage of their applicants who are female. We see that there is a strong tendency for the women to apply to those departments with a low admission rate, while the men go for those that are easier to get into. Reflecting on this, the authors of this study opine: The graduate departments that are easier to enter tend to be those that require more mathematics in the undergraduate preparatory curriculum. The bias in the aggregated data stems not from any pattern of discrimination on the part of admissions committees, which seem quite fair on the whole, but apparently from prior screening at earlier levels of the educational system. Women are shunted by their socialization and education toward fields of graduate study that are generally more crowded, less productive of completed degrees, and less well funded, and that frequently offer poorer professional employment prospects.

In this example, the department to which application is made operates as a confounding variable, and the overall analysis, which does not account for this, is likely to be highly misleading. But one should not immediately conclude that the analysis disaggregated by department can be relied on either, since there might be further confounding variables. One such might be residency. The admission procedures favour California residents over out-of-state applicants. If for some reason the proportions of women in these two groups were markedly different, then a further, more detailed investigation of admissions by sex, within each department, for each of the two residence categories, could well suggest yet another pattern of sexual discrimination. While arguments may be made for or against the appropriateness of adjusting for particular potential confounders, or the completeness of such adjustments, there is nothing in the data themselves that can support such arguments: so that any purported interpretation of observational data is beset by vagueness and subjectivity. These interpretative difficulties are multiplied when we realise that there may well be important confounding factors that we are not even in a position to adjust for, because we have not measured them.

140  A Philip Dawid The differing messages of observational and experimental studies in Example 3 might also be due to confounding by unmeasured attributes: women who choose to take HRT are largely middle class and generally healthier than those who do not, and the difference between the outcomes of the two groups seen in the observational study might have been distorted by such unmeasured differences between the groups, reflecting the betweengroup differences in overall health, irrespective of treatment taken. (Note that here we also had an observed potential confounder, age. The women in the experimental study tended to be older, and the effect of HRT is very age-dependent. But, while an analysis making a suitable adjustment for age brought the two kinds of study into closer agreement, this was not sufficient to resolve the differences.)

VI. Causes of Effects

A. Introduction Any attempt to bring statistical and scientific evidence to bear on cases at law must be fully cognizant of the problematic issues of data interpretation discussed above. But even when these can be regarded as fully solved, the implications of such ‘effects of causes’ evidence for the ‘causes of effects’ task of assigning causality in an individual case are far from straightforward. The problem is as much philosophical as technical. Suppose that I wish to know whether it was the aspirin I took 30 minutes ago that caused my headache to disappear. We assume that there is no uncertainty about the generic causal effect of aspirin, as described in Example 1. But how to apply that knowledge to my particular case? In our contrastive understanding of causality, we would need to address a question such as ‘What would have happened to my headache if I had not taken the aspirin?’.11 But this is ‘counterfactual’ – contradicting known facts – in two ways: first, by taking seriously the possibility, counter to fact, that I did not take the aspirin; and secondly, by considering the outcome as uncertain and possibly different from the known fact that I did recover. How can we meaningfully address such a nonsensical question? Let us code the act of taking aspirin as A = 1, and of not taking aspirin as A = 0. The response will be coded as R = 1 if the headache goes away, or R = 0 if it does not. My factual knowledge includes the information that A = 1 and R = 1; nevertheless, I want to consider the possibility A = 0, with R left unspecified. One approach to this that has gained much popularity in statistical circles is based on the idea of ‘potential responses’. We conceive that the response variable R comes in two flavours: R1, the ‘potential response’ to taking aspirin (A = 1); and R0, the ‘potential response’ to not taking aspirin (A = 0). And we must imagine that the values of these quantities have been determined even before I decide whether or not to take the aspirin. Then if I do take aspirin, I will uncover the pre-existing value R1, ie in the case A = 1 the response will be R = R1; whereas if A = 0 then R = R0. What we want to know is the probability of causation, PC, which can be expressed in the language of potential responses as the conditional probability:

  We might want to specify this foil more clearly: eg, by supposing that I had taken a chalk tablet instead.

11

Statistical Evidence for Causality  141 PC = Pr(R0 = 0 | A = 1, R = 1) (1) Here Pr(X|Y), read as ‘the probability of X, given Y’, is to be understood as quantifying uncertainty as to whether X is true, taking into account the (known or hypothesised) information Y. So PC addresses the question: knowing that I did take aspirin (A = 1), and the actual response was recovery (R = 1), what is the probability that the (necessarily unobserved) potential response R0, that would have been observed had I not taken aspirin, would have been different (R0 = 0)? When the response R is not a simple all-or-nothing event, but a variable, such as the actual time it takes for my headache to go away, we might need to reassess the interpretation of ‘probability of causation’ – one reasonable choice in this case might be (R0 > r | A = 1, R = r), ie, the probability, given that I did take aspirin and observed the headache went away after a certain time r, that it would have lasted longer if I had not taken the aspirin. Such cases are both very common and very important, but here we shall concentrate on the simple case of a binary (all-or-nothing) response, since this will suffice to make our principal logical points. The specific details will of course differ when we move to more complex cases.12 One simplification of formula (1) is immediate. Since I did take aspirin, the observed response was R = R1. The information A = 1, R = 1 is thus equivalent to A = 1, R1 = 1. So PC = Pr(R0 = 0 | A = 1, R1 = 1). We shall now make a (questionable) assumption: that my decision to take aspirin was unrelated to the (then hidden) values of my potential responses (R0, R1).13 In that case, the information does not affect uncertainty about the joint behaviour of (R0, R1), so we can ignore it, obtaining PC = Pr(R0 = 0 | R1 = 1). (2)

B.  Estimating the Probability of Causation Expressing the desired target of our interest in mathematical terms is all very well. But the real question is: how are we to estimate this probability, on the basis of available statistical evidence? What, in principle, can we learn from a well-designed and well-conducted statistical experiment along the lines of Example 1? We will have data on the observed response, R = R1, for the treatment group, who did take the aspirin. The proportion recovering (30 per cent in Table 1) is thus an estimate of the marginal14 recovery probability if treated with aspirin: Pr(R1 = 1). Similarly, from the data on the control group15 we can estimate Pr(R0 = 1), the marginal recovery probability if untreated. 12   See James M Robins and Sander Greenland ‘The Probability of Causation Under a Stochastic Model for Individual Risk’ (1989) Biometrics 45 1125–38 for some relevant considerations. 13   This is an important and often unappreciated requirement, equivalent to the assumption that there are no confounding variables bringing about dependence between my treatment decision and my possible outcomes. A contrary case might be when I learn, from the very fact that I was minded to take aspirin, that I am a hypochondriac with no serious medical problem, and will probably recover whatever I decide. 14   ie, not conditional on any other information. 15   Assuming the foil in the experiment is regarded as a relevant foil for the case at hand.

142  A Philip Dawid Now, by simple probability rules, we have Pr(R0 = 0 and R1 = 1) . PC =   Pr(R1 = 1) (3) Since we can estimate the denominator, Pr(R1 = 1), it is enough to estimate the numerator, Pr(R0 = 1 and R1 = 1) . But in no individual can we ever observe both potential responses, R0 and R1 – which makes estimation of this joint probability highly problematic. Different assumptions about the relationship between R0 and R1 might be made. One such is independence, ie there is no dependence whatsoever between how you would respond to aspirin or no aspirin. This is equivalent to assuming the conditioning in (2) can be ignored, so that PC reduces to the estimable quantity Pr(R0 = 0) (88 per cent if taken from Table 1). However, it might appear more realistic to consider both R0 and R1 as related to the individual’s underlying health status, which would make them positively dependent, and reduce the above figure. Another possible assumption is monotonicity, which states that if you would recover without aspirin then you would recover with aspirin, and can be expressed as Pr(R0 = 1, R1 = 0) = 0. In this case, using the figures of Table 1 we would find PC = 60%. The trouble is, such assumptions are largely arbitrary. In particular, so long as we do not measure any further variables, nothing in the data collected, or even in any data that could be collected, can shed any light on whether or not they hold. Since our evaluation of PC is sensitive to what assumptions we make, we appear to have reached the limits of justifiable inference before being able to address the question we are interested in. However, as we shall see in section VI.D below, it is possible to be more precise when we can measure additional variables.

C.  Bounding the Probability of Causation Table 4 displays the possible patterns of dependence between the two potential responses R0 and R1, as would be seen in 100 individuals. We have filled in the marginal figures based on the results in Table 1 – which is as far as that evidence can take us. What we cannot know (since we can never observe both responses on the same individual) is x, the number of individuals having R0 = 0 and R1 = 1. However, once we have assigned a value for x, the remaining figures in the table are determined, as shown. Table 4: Joint distribution of potential responses R0 and R1.

R0 R1

0

1

Total

0

88 – x

x – 18

70

1

x

30 – x

30

Total

88

12

100

The only thing we know for sure is that every one of the entries in Table 4 must be nonnegative. In particular, x must be 18 or more to ensure that the entry for R0 = 1, R1 = 0 is

Statistical Evidence for Causality  143 non-negative. Likewise, x cannot exceed 30 in order that the entry for R1 = 1, R0 = 1 be nonnegative. Now from (3) we see PC = x/30. We can thus infer that, whatever be the unknown (and unknowable) relationship between R0 and R1, PC must lie between 18/30 = 60% and (uninformatively) 30/30= 100%. That is, the result we found under the additional assumption of monotonicity supplies a lower bound for PC. This argument holds in general. Suppose that the response rate among the treated, Pr(R1 = 1), is at least as large as that among the untreated, Pr(R0 = 1). Then the (experimental) risk ratio, defined as Pr(R1 = 1) , RR = Pr(R0 = 1) (4) is at least 1. Working through the above argument yields the lower bound: (5) PC ≥ 1 – 1 . RR In our example, RR = 30/12 = 2.5, and (5) indeed reduces to PC ≥ 60%. In particular, whenever RR exceeds 2, we can deduce from (5), without making any further assumptions, that PC must exceed 50 per cent. In a civil case (and absent other evidence), this might be considered as meeting the civil standard of proof on the balance of probabilities. As a final caveat, we emphasise that it is the experimental risk ratio RR defined by (4) that enters (5). This should be contrasted with the observational risk ratio, Pr(R = 1 | A = 1)/ Pr(R = 1 | A = 0), which can be estimated from data on A and R. In a properly conducted experimental study, it might be appropriate to equate these two risk ratios, so getting an empirical purchase on RR; but, as we have seen above, this is very problematic in an observational study, in which both A and R are observed but there is no manipulation or control over the application of the treatment.

D.  The Role of Science If we know more about the internal workings of the black box that converts treatment into outcome, we can sometimes improve the bound (5) on PC. Robins and Greenland16 show how sharper bounds on PC can be obtained when certain assumptions about the biological process can be made. i.  Example 4 Suppose that scientific research has identified a gene, with two variants, G and G, each present in 50 per cent of the population. It has been found that, among those having variant G, the headache recovery rate if untreated is 24 per cent; but because of an interaction with the action of aspirin, such an individual will never recover from a headache (within 30 minutes) if treated with aspirin. As for those with variant G, they never recover if left untreated, 16  James M Robins and Sander Greenland, ‘The Probability of Causation Under a Stochastic Model for Individual Risk’ (1989) Biometrics 45, 1125–38.

144  A Philip Dawid but have a 60 per cent chance of recovery if they take aspirin. Over the whole population, ignoring genetic status, the recovery rate is 30 per cent with aspirin and 12 per cent without aspirin – as observed in Table 1. We now come to the case at hand: did taking the aspirin cause my headache to disappear? We do not need to suppose that my own genetic status, G or G, is known. Nevertheless, given that I did recover after taking aspirin, it can be inferred that I can not be of type G, so must be G; and it then follows that I would certainly not have recovered if I had not taken the aspirin. So with this extra scientific knowledge we can deduce PC = 1 – a major improvement on (although of course consistent with) the ‘black-box’ conclusion PC ≥ 60%. The above argument was essentially logical. But even when the more detailed science is less certain, we can often still improve the bound (5) on PC. Thus suppose we have identified an additional attribute, S, which can take on a number of values s (in the above example, S was the gene, with values G and G); and, in a good experimental study, we have measured the response rates, for both treated and untreated individuals, within each category s of S: thus we have estimates of Pr(R=1 | A = 1, S = s) = Pr(R1 = 1 | S = s), and Pr(R=1 | A = 0, S = s) = Pr(R0 = 1 | S = s), for each value of s. We also know the distribution of S in the population, Pr(S = s) for each s. If, in the case in issue, we knew the value s of S – eg, we knew which gene variant, G or G, I possessed – we could simply apply formula (5), using the response rates relevant to that subgroup. But what if we do not know S? We can compute: Δ= Γ=

Σs  Pr(S = s) × max {0, Pr(R=1 | A = 1, S = s) – Pr(R = 1 | A = 0, S = s)}

Σs  Pr(S = s) × max {0, Pr(R=1 | A = 1, S = s) – Pr(R = 0 | A = 0, S = s)}.

It can now be shown,17 without any further assumptions, that Δ Γ . ≤ PC ≤ 1 – Pr (R = 1 | A = 1) Pr(R = 1 | A = 1)

(6)

The lower bound will be an improvement on the ‘black-box’ lower bound (5), so long as, as we vary the level s, the level-specific treated response rate, Pr(R = 1 | A = 1, S = s), is sometimes greater and sometimes less than the corresponding level-specific untreated response rate, Pr(R = 1 | A = 0, S = s). ii.  Example 5 We reconsider Example 4, with the following modification: we now assume Pr(R = 1 | A = 1,G) and Pr(R = 1 | A = 0, G) are each 12%, rather than 0 as previously. The overall recovery rates are then 36 per cent with aspirin, and 18 per cent without, for a risk ratio RR = 2, and a corresponding ‘black-box’ lower bound on PC of 50 per cent – just insufficient to prove causation on the balance of probabilities. Taking the gene into account, if it is known that I am of type G, then the relevant risk ratio becomes 60/12 = 5, and we can deduce PC ≥ 80%; whereas if I am known to be of type G, the risk ratio is less than 1, and nothing non-trivial can be said about PC.   Those interested in the details are referred to the Appendix.

17

Statistical Evidence for Causality  145 For the case that my genetic status is unknown, we compute Δ = 24%. We can then infer PC ≥ 24/36 = 67% – which should swing the case. It is interesting to contrast the roles of science18 as applied to CoE and to EoC types cases. For CoE, as we have just seen, we can use a more detailed understanding of the system to improve inference in an individual case, even when we are unable to look inside the black box in that case. For EoC, on the other hand, the benefits of a more detailed understanding of internal mechanisms are much less direct. Clearly if we can see what is inside the box for the new case, such understanding might help us focus and so improve our predictions for that case, by taking into account additional (eg genetic) information observed on it. In the absence of such case-specific information, however, what we can gain from a more detailed scientific understanding is much more limited. If the new case arises from a regime with different features from those in our data, we might be able to apply a more detailed understanding to focus on the component mechanisms that we can assume are stable, and ignore (or find other ways of handling) those that may have changed. But even when dealing with an unchanging environment, while more detailed modelling might improve the accuracy of our predictions for new cases, if it introduces too many additional quantities requiring (necessarily imperfect) estimation, it is more likely to degrade it.

VII. Conclusions

The application of scientific evidence to inform issues of causality in cases at law is a very subtle issue. When the evidence is based on observational rather than experimental studies, there is plenty of scope for drawing a wide range of conflicting conclusions from the same data. And even with the best experimental evidence, its implications for a particular case at hand are far from clear-cut, and can usually only support bounds on the ‘probability of causation’, at best. But it may be possible to sharpen those bounds by taking into account more detailed scientific understanding of the cause-effect process.

Further Reading

For general accounts of the relationships between Statistics and Law, see Dawid (2005, 2008).19 The relationship (5) between probability of causation and experimental relative risk was first developed by Robins and Greenland (1989)20 although, as pointed out by Greenland

18   We caution that the ‘science’ should be soundly based. As discussed in section III.B, this may not be easy to achieve. 19   A Philip Dawid, Probability and Proof (on-line appendix to Terence J Anderson, David A Schum and William L Twining, Analysis of Evidence 2nd edn (Cambridge, Cambridge University Press, 2005) http://tinyurl.com/tz85o; A Philip Dawid, ‘Statistics and the Law’ in Andrew Bell, John Swenson-Wright and Karin Tybjerg (eds), Evidence (Cambridge, Cambridge University Press, 2008) 119–48. 20  James M Robins and Sander Greenland, ‘The Probability of Causation Under a Stochastic Model for Individual Risk’ (1989) Biometrics 45, 1125–38.

146  A Philip Dawid (1999),21 and in a specifically legal context by Greenland and Robins (2000),22 there is a common misconception that this is an equality rather than a lower bound. I am not aware of any prior discussion of the material in section VI.D on improving this bound. The bible of statistical causality is the book by Pearl (2009).23 Pearl’s framework is based on graphical representations of essentially deterministic causal processes. Tian and Pearl (2000)24 consider a variety of concepts of individual causation: they term PC the ‘probability of necessity’, and show the bound (5). They also show that it is sometimes possible to improve on this bound when possibly confounded observational data, as well as experimental data, are available. The language and calculus of potential outcomes was developed by Rubin (1974, 1978)25 although much of that theory is aimed at EoC type queries, rather than CoE, as here. Detailed philosophical investigations into the nature and scope of arguments based on potential responses may be found in Dawid (2000, 2007).26 An alternative approach for handling EoC problems without needing to consider potential responses was developed by Dawid (2002).27 The importance of deeper scientific knowledge in helping to refine the probability of causation, in cases with responses that may be more complex than a simple binary outcome, has been discussed by Beyea and Greenland (1999).28

Appendix: Improving the Bounds for PC

We first note that, for each value of s, (7) Pr(R0 = 0, R1 = 1, S = s) ≥ max {0, Pr(R1 = 1, S = s) – Pr(R0 = 1, S = s)} (8) Pr(R0 = 0, R1 = 1, S = s) ≤ min {Pr(R1 = 1, S = s), Pr(R0 = 0, S = s)}. (These properties are easy consequences of the non-negativity of Pr(R0 = i, R1 = j, S = s) for all i, j = 0,1). On summing (7) and (8) over s, realising that Pr(R1 = 1, S = s) = Pr(S = s) × Pr(R1 = 1 | S = s) = Pr(S = s) × Pr(R = 1 | A = 1, S = s), etc, we obtain Pr(R0 = 0, R1 = 1) ≥ Δ 21   Sander Greenland, ‘Relation of Probability of Causation to Relative Risk and Doubling Dose: A Methodologic Error That Has Become a Social Problem’ (1999) American Journal of Public Health 89, 1166–69. 22   Sander Greenland and James M Robins, ‘Epidemiology, Justice, and the Probability of Causation’ (2000) Jurimetrics 40, 321–40. 23   Judea Pearl, Causality: Models, Reasoning and Inference 2nd edn (Cambridge, Cambridge University Press, 2009). 24   Jin Tian and Judea Pearl, ‘Probabilities of Causation: Bounds and Identification’ (2000) Annals of Mathematics and Artificial Intelligence 28, 287–313. 25   Donald B Rubin, ‘Estimating Causal Effects of Treatments in Randomized and Nonrandomized Studies’ (1974) Journal of Educational Psychology 66, 688–701; Donald B Rubin, ‘Bayesian Inference for Causal Effects: the Role of Randomization’ (1978) Annals of Statistics 6, 34–68. 26  A Philip Dawid, ‘Causal Inference Without Counterfactuals’ (with Discussion) Journal of The American Statistical Association (2000) 95, 407–48; A Philip Dawid, ‘Counterfactuals, Hypotheticals and Potential Responses: A Philosophical Examination of Statistical Causality’ in Federica Russo and Jon Williamson (eds), Causality and Probability in the Sciences (London, College Publications, 2007) 503–32. 27   A Philip Dawid, ‘Influence Diagrams for Causal Modelling and Inference’ International Statistical Review (2002) 70, 161–89. (Corrigenda, ibid 437.) 28   Jan Beyea and Sander Greenland, ‘The Importance of Specifying the Underlying Biologic Model in Estimating the Probability of Causation’ Health Physics (1999) 76, 269–74.

Statistical Evidence for Causality  147 and Pr(R0 = 0, R1 = 1) ≤ Pr(R1 = 1) – Γ. Since PC = Pr(R0 = 0, R1 = 1)/Pr(R1 = 1), and Pr(R1 = 1) = Pr(R = 1 | A = 1), (6) follows.

8 Using Scientific Evidence to Resolve Causation Problems in Product Liability: UK, US and French Experiences RICHARD GOLDBERG*

I. Introduction

Proof of causation in product liability litigation is an inherently difficult problem, which regularly requires time-consuming analysis of complex scientific evidence.1 By no means has this been merely a recent phenomenon. For instance, almost 40 years ago, in Ashington Piggeries Ltd v Christopher Hill Ltd 2 the defendants were successful in proving, but only after a 50-day trial in the High Court, that the death of their mink was caused by the presence of the contaminating chemical dimethylnitrosamine (DMNA) in the herring meal of the plaintiffs’ mink feed.3 However, in recent years, the cases have become even more complex, demanding much from lawyers and scientific experts on both sides and the judges themselves.4 *  I wish to thank Professor David Goldberg, Professor Mike Green and Professor Joseph Sanders for helpful discussions. 1   CJ Miller and RS Goldberg, Product Liability, 2nd edn (Oxford, Oxford University Press, 2004) para 17.05. 2   Ashington Piggeries Ltd v Christopher Hill Ltd [1972] AC 441. 3   [1972] AC 441, 466–88; also Wright v Dunlop Rubber Co Ltd [1972] 13 KIR 255 (cancer from industrial chemicals). cf the US mass tort litigation concerning Bendectin (Debendox), the anti-nausea drug used in pregnancy. In contrast to thalidomide, where many lines of evidence showed that the drug caused phocomelia malformations (see H Sjöström and R Nilsson, Thalidomide and the Power of Drug Companies (Harmondsworth, Penguin, 1972) 156–59), no causal link has ever been scientifically established between Bendectin and birth defects. The Bendectin litigation demonstrated a persistent failure of plaintiffs’ lawyers to prove causation: see the seminal paper on this topic, J Sanders, ‘The Bendectin Litigation: A Case Study in the Life Cycle of Mass Torts’ (1992) 43 Hastings Law Journal 301. The litigation spawned two formative monographs (M Green, Bendectin and Birth Defects: The Challenges of Mass Toxic Substances Litigation (Philadelphia, Penn, University of Pennsylvania Press, 1996); J Sanders, Bendectin on Trial: A Study of Mass Tort Litigation (Ann Arbor, MI, The University of Michigan Press, 1998)), which influenced the current writer in his work: R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing, 1999) 102–31. 4   See eg, Loveday v Renton [1990] 1 Med LR 117, where Stuart-Smith LJ, after examining complex scientific evidence and arguments, held that the plaintiff had failed to prove on a balance of probabilities that pertussis (whooping cough) vaccine could cause permanent brain damage in young children: [1990] 1 Med LR 117, 185 (discussed in Goldberg (n 3) 137–43); and Reay v British Nuclear Fuels; Hope v BNFL [1994] 5 Med LR 1, 53 (plaintiffs failed to establish that paternal preconception irradiation was a material contributory cause of childhood leukaemia of Dorothy Reay or non-Hodgkin’s Lymphoma of Vivien Hope); Davis & Doherty v Balfour Kilpatrick [2002] EWCA Civ 736 (claimants failed to establish that their alleged radiation sickness was caused by radio frequency emissions).

150  Richard Goldberg This chapter examines the recent trends in such litigation in the UK, US and in France. In essence, it seeks to determine the extent to which the courts in these countries in the highlighted cases have been pragmatic and fair in their interpretation and utilisation of epidemiological evidence from the perspective of both consumers and producers. The first section explores the difference between evidence of causation for purposes of science and for the law, and the difficulties in reconciling the standards of proof in law and science, including the theory that causation can be proved on the balance of probabilities by reference to the doubling of risk of injury. The distinction between association and causation and the difficulty in proving general and specific causation between a product and damage using epidemiological evidence is reviewed in the context of the controversial Scottish case of McTear v Imperial Tobacco Limited,5 which is subject to criticism of the way that the epidemiological evidence was received by the trial judge. The problem of utilising statistics deriving from trends in general populations to prove causation in an individual case is highlighted in McTear, and a possible solution in the form of utilisation of the Bayes’ Theorem is discussed. A developing issue in the United States has been the assessment of the sufficiency of scientific evidence admitted in the determination of proof of causation in complex product liability cases, which often involve prescription drugs and vaccines. In particular, of recent interest is the assessment of the value of scientific evidence in six test cases of the Omnibus Autism Proceeding (OAP) under the National Childhood Vaccine Injury Act of 1986 (NVIA), decided in February 2009 and March 2010. These cases essentially explored two causation theories, viz that MMR vaccines and thimerosal containing vaccines could combine to cause autism, and that thimerosal vaccines alone can cause autism. The implications of these complex and lengthy judgments are explored. The final section of the chapter explores the much more liberal approach to causation established in France by the Cour de Cassation for medicinal product liability cases in the context of injury allegedly caused by the Hepatitis B vaccine through the use of presumptions of causation.

II. Reconciling the Standards of Proof in Law and Science in the UK

A.  Evidence of Causation for Purposes of Science and for Purposes of Law There is considerable difference between evidence of causation for purposes of science and for the law. For the law of negligence, it is sufficient to show that the balance of probabilities – meaning more than 50 per cent, or on a preponderance of the evidence – indicates a causal connection. For medical science, on the other hand, rules of epidemiology require evidential proof on a balance of probabilities of at least 95 per cent to establish causation. The most pertinent issue6 is the lack of clarity in being able to determine at what point the balance of probabilities standard (legal) and the standard for epidemiology (science) intersect.7 While   McTear v Imperial Tobacco Limited 2005 2 SC 1.   See R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (n 3) 105. 7   L Lasagna and SR Shulman, ‘Bendectin and the Language of Causation’ in KR Foster, DE Bernstein and PW Huber (eds), Phantom Risk: Scientific Inference and the Law (Cambridge, Mass, MIT Press, 1993) 112. 5 6

Causation Problems in Product Liability  151 neither the claimant nor the defendant is required to apply scientific standards of proof when determining causation on a balance of probabilities,8 courts must be alert to the problem that may be faced by an expert ‘in readjusting his focus from the 95 per cent confidence limits approach to the balance of probabilities test’.9 However, in Vadera v Shaw10 the Court of Appeal reconciled the legal standard of proof on a balance of probabilities with the scientific standard of statistical significance in showing that a failure to establish a statistically sig­ nificant connection between the oral contraceptive Logynon and strokes was fatal to the establishment of proof of causation on a balance of probabilities. Henry LJ stated: The judge concluded, and in our respectful view was right on the evidence to conclude, that the studies carried out and referred to by Dr Lidegaard [for the plaintiff] did not establish a statistically significant connection between Logynon and strokes. Such evidence cannot be ignored by a judge. It is as common sense a conclusion as one could wish to say that if the connection between ‘a’ and ‘b’ cannot be shown with confidence to be other than a coincidence, then it cannot be held on a balance of probabilities that ‘a’ caused ‘b’. This is not to allow scientists or statisticians to usurp the judge’s function, but rather to permit him to use their skills to discern a connection, or lack of connection between the two phenomena.11

B.  Doubling of Risk Theory A significant attempt to reconcile the balance of probabilities standard and the standard for epidemiology has emerged with the theory that causation can be proved on the balance of probabilities by reference to the doubling of risk of injury. That theory has long been recognised in the United States,12 where it has been said that ‘[t]he use of scientifically reliable epidemiological studies and the requirement of more than a doubling of the risk strikes a balance between the needs of our legal system and the limits of science’.13 However, the theory has also been subject to trenchant criticism.14 In particular, it has been submitted 8   Carter v Basildon and Thurrock University Hospitals NHS Foundation Trust [2007] EWHC 1882 (QB), [2007] LS Law Medical 657, [92]. 9   ibid [97]. A confidence interval or confidence limit is a range within which a study parameter lies 95 per cent of the time: see R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (n 3) 137; American Law Institute Reporters’ Study, Enterprise Responsibility for Personal Injury Vol II: Approaches to Legal and Institutional Change (Philadelphia, Penn, American Law Institute, 1991) 324–28. 10   Vadera v Shaw (1998) 45 BMLR 162 (CA). 11   ibid 164. However, it is suggested that there was a failure by the trial judge and the Court of Appeal to scrutinise adequately the scientific evidence in respect of causation in this case: R Goldberg, ‘The Contraceptive Pill, Negligence and Causation: Views on Vadera v Shaw’ (2000) 8 Medical Law Review 316, 331–35. 12   See, in particular, Daubert v Merrell Dow Pharmaceuticals Inc 43 F3d 1311 (9th Cir 1995), cert.denied, 116 SCt 189 (1995) [Daubert II]. In that case, the court of appeals, on remand from the Supreme Court of the United States, held that the plaintiffs had to show not merely that Bendectin increased the likelihood of injury but more likely than not caused their injuries. In terms of statistical proof, it had to be shown that their mothers’ ingestion of Bendectin more than doubled the likelihood of birth defects: ibid 1320. This was reaffirmed by the Supreme Court of Texas in Merrell Dow Pharmaceuticals Inc v Havner 953 SW 2d 706, 716–18 (Tex 1997). 13   Merrell Dow Pharmaceuticals Inc v Havner 953 SW 2d 706, 718 (Tex 1997) (echoing the views of the court of appeals in Daubert II (n 12)). 14   See eg, LM Finley, ‘Guarding the Gate to the Courthouse: How Trial Judges Are Using their Evidentiary Screening Role to Remake Tort Causation Rules’ (1999) 49 DePaul Law Review 335, 336, 348 (criticism of the doubling-in-risk evidentiary requirement for epidemiological proof, describing the trend as ‘seriously scientifically and legally misguided’); MA Berger, ‘Upsetting the Balance Between Adverse Interests: The Impact of the Supreme Court’s trilogy on Expert Testimony in Toxic Tort Litigation’ (2001) 64 Law & Contemporary Problems 289, 304–06 (criticism of the doubling of the risk rule as ‘a legal invention that creates a hard and fast rule that disposes of cases efficiently but rests on assumptions that cannot be scientifically validated at this time’); S Greenland and JM Robins, ‘Epidemiology, Justice, and the Probability of Causation’ (2000) 40 Jurimetrics

152  Richard Goldberg that judges have adopted ‘substantive changes in causation law through the rubric of evid­ entiary admissibility decisions’,15 and have frequently conflated admissibility decisions and sufficiency of evidence decisions.16 Those courts which require plaintiffs to produce epidemiological studies with a relative risk of two are making a ‘legal policy determination to equate epidemiology, relative risk, general causation, and the burden of proof on individual causation’. Moreover, while the frequency of judicial opinions referring to relative risk greater than two is increasing, the US courts disagree as to the proper role of the doubling of risk theory in deciding questions of both sufficiency and admissibility of scientific evidence of causation in toxic tort cases. They do not agree on whether to adopt the doubling of risk as a threshold, nor do they agree on the meaning of such a threshold.17 As the reporters for the American Law Institute’s Restatement Third of Torts have noted:18 Many courts accept the doubling of the incidence of disease in group studies; some courts insist on doubling of risk as a minimum threshold for establishing specific causation. Others have recognised that if other known causes can be identified and eliminated, something less than a doubling would still be sufficient to find specific causation

Accordingly, the requirement of a relative risk of two for the admissibility or sufficiency of epidemiological evidence has been subject to much scepticism:19 and the reporters for the Restatement Third of Torts, in discussing the considerations which affect the appropriateness of determining the probability of specific causation based on the outcome of group studies, have concluded that a judicial requirement that plaintiffs must show a threshold increase in risk or a doubling of incidence in a group study to satisfy the burden of proof of specific causation is ‘usually inappropriate’.20 However, despite the problems with the doubling of risk theory in the US, its existence appears to be gaining ground in the UK. In XYZ v Schering Health Care Ltd,21 a trial of seven lead cases involving third generation oral contraceptives, Mackay J stated that the claimant could prove that an exposure to risk caused injury if that exposure had more than doubled the risk of the injury occurring.22 This method of proving causation has been applied in a case of bladder cancer where the claimant had been tortiously exposed to carcinogens and nontortiously exposed to cigarette smoke, both of which are potent causes of the condition.23 Journal 321, 325. cf M Geistfeld, ‘Scientific Uncertainty and Causation in Tort Law’ (2001) 54 Vanderbilt Law Review 1011, 1015, 1018, 1020. 15   ibid 336, 347. 16   ibid 336; supported by JM Eggen, ‘Clinical Medical Evidence of Causation in Toxic Tort Cases: Into the Crucible of Daubert’ (2001) 38 Houston Law Review 369, 378–79; M Green, ‘The Future of Proportional Liability: The Lessons of Toxic Substances Causation’ in S Madden (ed), Exploring Tort Law (New York, Cambridge University Press, 2005) 368–69. 17   RS Carruth and RD Goldstein, ‘Relative Risk Greater than Two in Proof of Causation in Toxic Tort Litigation’ (2001) 41 Jurimetrics Journal 195, 199, 202–03. 18   Restatement of the Law Third, Torts: Liability for Physical and Emotional Harm [Restatement Third] (St Paul, American Law Institute, 2010) §28(a), Reporters’ Notes to comment c, 450–52. 19   C Cranor, Toxic Torts: Science, Law, and the Possibility of Justice (Cambridge, Cambridge University Press, 2007) 234–38, 281. 20   Restatement Third §28(a) comment c (4), 409. 21   XYZ v Schering Health Care Ltd [2002] EWHC 1420, (2002) 70 BMLR 88 (QB). 22   ibid [21] per Mackay J. See further, CJ Miller and RS Goldberg, Product Liability, 2nd edn (Oxford, Oxford University Press, 2004) paras 17.06–17.08. 23   Novartis Grimsby Ltd v Cookson [2007] EWCA Civ 1261, [74] per Smith LJ; AB v Ministry of Defence [2010] EWCA Civ 1317, (2011) 117 BMLR 101, [151] (doubling of risk theory relevant in the context of examining strength of claimants’ cases on causation in determining the exercise of discretion under the Limitation Act 1980, s 33) (on appeal to Supreme Court).

Causation Problems in Product Liability  153 The utilisation and value of epidemiological or statistical evidence alone in determining causation on a balance of probabilities was subject to some interesting debate in the Supreme Court in Sienkiewicz v Greif.24 The reason for this discussion, as pointed out by Baroness Hale,25 was the presence of an obiter observation of Smith LJ in her judgment in Sienkiewicz that ‘in a case of multiple potential causes, a claimant can demonstrate causation by showing that the tortious exposure has at least doubled the risk arising from the non-tortious cause or causes’. 26 While there was unanimity in their Lordships holding that there was no room for introducing the doubling of risk approach to ‘single exposure’27 mesothelioma cases or multiple defendant mesothelioma cases,28 the differences in view emerged with the obiter discussion of the general applicability of the doubling of risk theory using epidemiological evidence to determine proof of causation in personal injury cases. Lord Phillips discussed the decision of XYZ v Schering Health Care Ltd,29 and took the view that, while the case contained ‘a detailed and illuminating discussion of epidemiology’, it did not afford any direct assistance to the question of whether the ‘doubles the risk’ test, as he called it, was an appropriate test for determining causation in a case of multiple potential causes.30 His reasoning was somewhat obscured by his misclassification of the contraceptives in this case. He stated that the issue: was whether a second generation of oral contraceptives more than doubled the risk of causing deep vein thrombosis (DVT) that was created by the first generation of contraceptives . . . It was not whether the DVT suffered by the claimants had been caused by the second generation of oral contraceptives.

In fact, the issue was whether the claimants had proved that third generation combined oral contraceptives caused a true excess risk of Venous Thromboembolism (VTE), which was more than twice that caused by second generation combined oral contraceptives. This had been suggested in three unpublished studies highlighted in a ‘Dear Doctor’ letter sent to prescribers by the Committee on Safety of Medicines (CSM) on 18 October 1995. Both sides agreed that if the claimants failed to prove this, the action would go no further as it could not succeed. The reason for this utilisation of doubling of risk theory in fact related to proving that the third generation contraceptives were defective. Both parties had agreed that if the claimants could prove a true excess risk of VTE, they would also succeed on the second issue, which was whether the relevant products were defective within the meaning of section 3 of the Consumer Protection Act 1987, ie that their safety would not be such as persons generally were entitled to expect. The test of defectiveness under section 3 includes consideration of instructions or warnings associated with the product. The reasoning   Sienkiewicz v Greif [2011] UK SC 10, [2011] 2 WLR 523.   ibid [169]. 26   Sienkiewicz v Greif (UK) Ltd [2009] EWCA Civ 1159, [2010] QB 370, [23] per Smith LJ. 27   Sienkiewicz v Greif [2011] UK SC 10, [67]. As Lord Kerr has noted, ‘[t]he use of the expression “single exposure” may be misleading in this context’: ibid [199]. It is probably better expressed as ‘single tortious exposure cases’: ibid [173] per Baroness Hale. These are cases where only one defendant exposed the victim to asbestos and there was only one possible tortious source for the exposure, and the only other exposure creating a risk of developing mesothelioma was environmental exposure to low level asbestos dust in the general atmosphere: ibid [113] per Lord Rodger; [199] per Lord Kerr, [207] per Lord Dyson. 28   ibid [106] per Lord Phillips, [160] per Lord Rodger, [169] per Baroness Hale, [188] per Lord Mance, [2003] per Lord Kerr, [220] per Lord Dyson. 29   XYZ v Schering Health Care Ltd [2002] EWHC 1420, (2002) 70 BMLR 88 (QB). 30   Sienkiewicz v Greif [2011] UK SC 10, [74] per Lord Phillips. 24 25

154  Richard Goldberg behind the doubling of risk theory’s relevance to establishing that the third generation contraceptives were defective was that if the Court ruled that the true risk of VTE was more than doubled with third generation combined oral contraceptives, women and their prescribers were entitled to be told this before making their decisions or giving their advice, and they were not.31 However, Lord Phillips’ reasoning seems to ignore the fact that causation was inherently behind the court’s approach. As Mackay J explained in XYZ: The reason why the Claimants accept through Lord Brennan QC that this first issue is capable of disposing of the claims should be set out. It is not because an increase of less than 2 would fail to render the product defective within the meaning of the Act, though the Defendants would so argue if they had to. It is for reasons of causation that he accepts this burden, correctly in my view. If factor X increases the risk of condition Y by more than 2 when compared with factor Z it can then be said, of a group of say 100 with both exposure to factor X and the condition, that as a matter of probability more than 50 would not have suffered Y without being exposed to X. If medical science cannot identify the members of the group who would and who would not have suffered Y, it can nevertheless be said of each member that she was more likely than not to have avoided Y had she not been exposed to X.32

While Lord Phillips concluded33 that there was no scope for the doubling the risk test in cases where two agents operated cumulatively and simultaneously in causing the onset of the disease, since in such cases the material contribution rule in Bonnington Castings v Wardlaw34 would apply, he submitted35 that there was no reason in principle why the ‘doubles the risk test’ should not be applied where the initiation of a disease was dose related and there had been consecutive exposures to an agent or agents that cause the disease, eg McGhee v National Coal Board.36 He also considered37 that there was no reason in principle why the ‘doubles the risk’ test should not apply to competing alternative, rather than cumulative potential causes of a disease or injury, such as Hotson v East Berks Area Health Authority.38 However, neither Lord Rodger nor Baroness Hale took such a view, and both thought that a doubling of risk approach was not an appropriate test of causation.39 Lord Rodger stressed that where statistical evidence established that exposure to a substance more than doubled of the risk of a disease, this would not amount to proof, on the balance of probability, that the exposure actually caused the disease.40 While Lord Dyson felt it ‘unnecessary to decide whether there [were] any circumstances in which, as a matter of English law, causation [could] be proved on the basis of epidemiological evidence alone’, 41 he expressed the view that ‘there [was] no a priori reason why, if the epidemiological evidence [was] cogent enough, it should not be sufficient to enable a claimant to prove his case without more’.42 By contrast, Lord Kerr stressed the need to treat the use of epidemiological 31   XYZ v Schering Health Care Ltd [2002] EWHC 1420, (2002) 70 BMLR 88 (QB), [20]–[21]. See, further, CJ Miller and RS Goldberg, Product Liability, 2nd edn (Oxford, Oxford University Press, 2004), para 17.06–17.08. 32   XYZ v Schering Health Care Ltd [2002] EWHC 1420, (2002) 70 BMLR 88 (QB), 21 (emphasis added). 33   Sienkiewicz v Greif [2011] UK SC 10, [91] per Lord Phillips. 34   Bonnington Castings v Wardlaw [1956] AC 613, 620 per Lord Reid. 35   Sienkiewicz v Greif [2011] UK SC 10, [92] per Lord Phillips. 36   McGhee v National Coal Board [1973] 1 WLR 1. 37   Sienkiewicz v Greif [2011] UK SC 10, [93] per Lord Phillips. 38   Hotson v East Berks Area Health Authority [1987] AC 750. 39   Sienkiewicz v Greif [2011] UK SC 10, [156], [158], [162] per Lord Rodger, [170], [172]–[173] per Baroness Hale. 40   ibid [156], [158] per Lord Rodger. 41   Sienkiewicz v Greif [2011] UK SC 10, [221] per Lord Dyson. 42   ibid [222] per Lord Dyson.

Causation Problems in Product Liability  155 evidence to seek to establish any specific proposition in an individual case with great caution.43 He felt that there was a real danger that ‘so-called “epidemiological evidence” [would] carry a false air of authority’.44 Lord Mance felt that whether and if so when epidemiological evidence could, per se, prove a case, was ‘a question best considered not in the abstract but in a particular case, when and if that question [arose]’.45 If it could, he would hope and expect that this would only occur in the rarest of cases.46

C.  Association Versus Causation It is suggested that it would be an oversimplification to think that the views of Lord Phillips in Sienkiewicz will help to signal a green light to the establishment of proof of causation on a balance of probabilities by a mere doubling of relative risk. The matter was addressed in Merrell Dow Pharmaceuticals, Inc v Havner47 where, having stated that the balance between the needs of the legal system and the limits of science could be shown by the use of scientifically reliable epidemiological studies and the requirement of more than doubling the risk, the Supreme Court of Texas added the caveat: We do not hold, however that a relative risk of more than 2.0 is a litmus test or that a single epidemiological test is legally sufficient evidence of causation. Other factors must be considered. As already noted, epidemiological studies only show an association. There may in fact be no causal relationship even if the relative risk is high.48

The latter sentence is of particular importance, and while Lord Phillips referred to the remainder of this caveat in his speech in Sienkiewicz,49 he omitted the last sentence, and ignored its import in his final analysis. Unlike Lord Phillips, Lord Rodger stressed the importance of the distinction between association and causation. Lord Rodger’s speech is more compelling in that it shows a greater understanding of both the significance and limitations of epidemiological evidence and demonstrates a reluctance to support the general applicability of the doubling of risk theory using epidemiological evidence to determine proof of causation in personal injury cases.50 Lord Rodger accepted that epidemiological and statistical evidence may form an important element in proof of causation, and he supported the utilisation and value of epidemiological evidence where a claimant was required to prove his case on a balance of probabilities.51 However, he emphasised that, since by its very nature the statistical evidence does not deal with the individual case, the court should not proceed to find a causal relationship without further, non-statistical evidence, eg temporality.52 In so doing, he cited Phipson on Evidence, which states that ‘[w]here there is   ibid [205] per Lord Kerr.   ibid [206]. 45   ibid [192]. 46  ibid. 47   Merrell Dow Pharmaceuticals Inc v Havner 953 SW 2d 706 (Tex 1997). 48   ibid 718. 49   Sienkiewicz v Greif [2011] UK SC 10, [88] per Lord Phillips. 50   ibid [162], per Lord Rodger, [170], [173] per Baroness Hale. 51   ibid [163] per Lord Rodger. 52   ibid; see also, Baroness Hale, who opined that ‘the existence of a statistically significant association between factor X and disease Y does not prove that in the individual case it is more likely than not that factor X caused disease Y’: ibid [170]. Lord Mance accepted that epidemiological evidence, used with proper caution, could be admissible and relevant in conjunction with specific evidence related to the individual circumstances and parties. 43 44

156  Richard Goldberg epidemiological evidence of association, the court should not proceed to find a causal relationship without further, non-statistical evidence’.53 Lord Rodger illustrated his example of evidence of temporality in the context of a medicinal product and an adverse effect, where there was ‘a strong epidemiological association between a drug and some condition that could have been caused in some other way’.54 He submitted that that epidemiological evidence, ‘along with evidence that the claimant developed the condition immediately after taking the drug’ could be sufficient to allow the judge to conclude that the drug caused the condition on the balance of probability.55 In the light of Lord Rodger’s observations in Sienkiewicz, the mere existence of a statistic­ ally significant association is insufficient to establish a causal relationship without the presence of further non-statistical evidence, and, that, in order to do so, factors such as those adumbrated by Sir Austin Bradford Hill would need to be utilised to determine whether a reported association is causal.56 This point was emphasised by the Scottish Court of Session in McTear v Imperial Tobacco Limited,57 a decision which takes a cautious approach to the use of epidemiological evidence, and stresses the impossibility of applying epidemiological studies to determine causation in individual cases. UK developments in this area have often focused on the difficulty in proving general and specific causation between a product and damage using epidemiological evidence58 derived from trends in general populations, and this was graphically illustrated by McTear. In that case, the pursuer, the widow of a smoker, sought to recover damages from the defenders, manufacturers of John Player brand cigarettes, which were smoked by her late husband. Her husband had contracted squamous cell carcinoma of the lung, and the pursuer averred that cigarette smoking could cause lung cancer (an issue of general causation), and that her husband’s lung cancer was caused by his smoking (an issue of individual or specific causation). The problem of establishing a general causal link between cigarette smoking and cancer was exacerbated by the fact that, unlike all the United States cigarette companies, and all the other UK cigarette companies, Imperial Tobacco Ltd had not accepted that there was a The significance a court might attach to it depended ‘on the nature of the epidemiological evidence, and of the particular factual issues before the court’: ibid [191]. Lord Kerr considered that ‘[i]t is an essential and minimum requirement . . . that there be evidence connecting avowedly relevant statistical information produced by the epidemiological studies to the facts of the case’: ibid [205]. Lord Dyson also stressed the association/causation dichotomy, stating that ‘epidemiology . . . seeks to establish associations between alleged causes and effects . . . However, in an individual case, epidemiology alone cannot conclusively prove causation’: ibid [218]. 53   HM Malek (gen ed), Phipson on Evidence, 17th edn (London, Sweet & Maxwell, 2010) 34–27. 54   [2011] UK SC 10, [163] per Lord Rodger. 55  ibid. 56   See AB Hill, ‘The Environment and Disease: Association or Causation’ [1965] 58 Proceedings of the Royal Society of Medicine 295–300. These aspects of association (viz strength of association, consistency, specificity, temporality, biological gradient, plausibility, coherence, experiment and analogy) are utilised to determine whether a reported association is causal or non-genuine. 57   McTear v Imperial Tobacco Limited 2005 2 SC 1, [6.158]. However, the presentation of the list of factors in textbooks as ‘criteria’ for inferring causality or associations in a way as to imply that all the conditions are necessary has been described as ‘unfortunate’: S Greenland (ed), The Evolution of Epidemiological Ideas: Annotated Readings on Concepts and Methods (Newton Lower Falls, MA, Epidemiology Resources Inc, 1987) 14. As Greenland correctly observes, Hill expressly stated that he did not intend to lay down ‘hard-and-fast rules of evidence that must be obeyed before we accept cause and effect’: AB Hill, ‘The Environment and Disease: Association or Causation’ (1965) Proceedings of the Royal Society of Medicine 295, 299 (emphasis in original). He added that ‘[n]one of [his] nine viewpoints can bring indisputable evidence for and against the cause and effect hypothesis and none can be required as a sine qua non’: ibid. See further, C Cranor, Toxic Torts: Science, Law, and the Possibility of Justice (n 20) 102–05. 58   Epidemiology has been defined as ‘the study of patterns of disease occurring in human populations and the factors that influence these patterns’: McTear v Imperial Tobacco Limited 2005 2 SC 1, [6.157].

Causation Problems in Product Liability  157 causal link between smoking and disease, especially lung cancer.59 Lord Nimmo Smith concluded that, in the absence of such an admission, and indeed of any evidence that this was an inference which should be drawn, the defenders were entitled to put the pursuer to proof of her averment that cigarette smoking could cause lung cancer (ie the general causation issue).60 In the absence of support from animal experiments, proof of causation between cigarette smoking and lung cancer depended on what was proved before the court about epidemiological studies. He held that, in accordance with the Scots law of expert evidence, it was necessary to consider whether the evidence of any expert witness had imparted to him special knowledge of the subject matter of epidemiology, including published material lying within the witness’s field of expertise, so as to enable the court to form its own judgment about the subject matter and the conclusions to be drawn from it. Accordingly, it was not open to the court to form its judgment on the evidence without being taught how to do the epidemiology to a sufficient extent, and without it being provided with sufficient factual material to enable it to be proved on the balance of probabilities, not only that there was an association between cigarette smoking and lung cancer, but also that the proper conclusion to be drawn from this was that there was a causal connection between them.61 This distinction between association and causation lay at the heart of Lord Nimmo Smith’s conclusions. In his view, while an association between an exposure and a condition was judged to be statistically significant, that in itself did not constitute a judgment that there was a causal connection between an exposure and a condition.62 He explained: The finding of an association between an exposure and a condition or disease, even if judged to be statistically significant, does not of itself connote that a causal connection between the two is established. This is a matter for further exercise of judgment, taking account of such criteria as the consistency, the strength, the specificity, the temporal relationship and the coherence of the association. This must, I think be especially so, when, in the view of Sir Richard Doll . . . cigarette smoking is not a necessary cause nor a sufficient cause of lung cancer.63

It has been suggested that, while Lord Nimmo Smith’s ‘dogmatic aversion to statistical evidence’ means that epidemiology alone will never secure recovery in such cases,64 utilisation of epidemiological evidence that satisfies the Bradford Hill criteria would seem to be hard to gainsay.65 Thus it has been argued by Chris Miller that if an individual had been one of the cases in a case control study that yields strength of association (relative risk) then, in the light of such strength of association and other Bradford Hill criteria ‘it seems perverse 59   2005 2 SC 1, [2.58], [2.76], [6.30]. This was notwithstanding the generally accepted view for over 50 years that cigarette smoking could cause lung cancer: R Doll and AB Hill, ‘Smoking and carcinoma of the lung: Preliminary report’ (1950) 2 British Medical Journal 739–48; R Doll and AB Hill, ‘The mortality of doctors in relation to their smoking habits: A preliminary report’ (1954) 1 British Medical Journal 1451–55; 2005 2 SC 1, [5.208] (evidence of Sir Richard Doll). The defenders admitted that the World Health Organization, the United Kingdom Government and the United States Government had accepted for many years that cigarette smoking can cause lung cancer: 2005 2 SC 1, [2.7]. However, they averred that ‘[c]igarette smoking has not been scientifically established as a cause of lung cancer and, although various theories have been advanced, the cause or causes of lung cancer are unknown and the mechanism or mechanisms whereby lung cancer develops are unknown’: ibid. 60   2005 2 SC 1, [2.78], [2.80] (OH). 61   2005 2 SC 1, [6.155]. 62   2005 2 SC 1, [6.158]. 63  ibid. 64   C Miller, ‘Causation in personal injury: legal or epidemiological common sense’ (2006) 26 Legal Studies 544, 566. 65   ibid 566.

158  Richard Goldberg to hold that it is less probable than not that the exposure caused that individual’s condition’.66 It is submitted that Miller is correct in concluding that an association would exist in such circumstances. Indeed, Sir Austin Bradford Hill emphasised that ‘[n]one of [his] nine viewpoints can bring indisputable evidence for and against the cause and effect hypothesis and none can be required as a sine qua non’,67 and this has been judicially approved in the US.68 Bradford Hill specifically cautioned against overemphasis of the importance of specificity at the expense of strength of association, and specifically referred to smoking and lung cancer smoking in such circumstances.69 In doing so, he provided a particularly apt example: Coming to modern times the prospective investigations of smoking and cancer of the lung have been criticized for not showing specificity-in other words the death rate of smokers is higher than the death rate of non-smokers from many causes of death . . . But here surely one must return to my first characteristic, the strength of association. If other causes of death are raised 10, 20, or even 50% in smokers whereas cancer of the lung is raised 900–1,000% we have specificity- a specificity in the magnitude of the association . . .

We must also keep in mind that diseases may have more than one cause. In short, if specificity exists we may be able to draw conclusions without hesitation; if it is not apparent, we are not thereby necessarily left sitting irresolutely on the fence.70

It is suggested that the importance of such a widely accepted magnitude of strength of association between cigarette smoking and cancer was underplayed by Lord Nimmo Smith in McTear, and that he was wrong to treat all the Bradford Hill factors as being criteria that needed to be satisfied before such an association could amount to a causal connection between smoking and lung cancer. However, as we shall now see, this was not the only problem that the pursuer had in establishing general causation: the court also had to be taught the relevant epidemiology.

D.  Teaching the Court the Epidemiology In respect of general causation, Lord Nimmo Smith held that the pursuer had failed to prove, in accordance with the requirements of the Scots law of evidence relating to expert witnesses, that cigarette smoking could cause lung cancer.71 This was because the pursuer had failed to lead evidence about the primary literature sufficient to impart to the court special knowledge of the subject matter (epidemiological evidence of a causal connection between cigarette smoking and lung cancer, within the expert’s field of expertise) so as to enable the court to form its own judgment about it and the conclusions to be drawn from it.72 Lord Nimmo Smith stated that ‘a fundamental defect in the presentation of the pursuer’s case’ was the failure to take the court to any of the primary literature in which it had

 ibid.   AB Hill, ‘The Environment and Disease: Association or Causation’ (1965) Proceedings of the Royal Society of Medicine 295, 299 (emphasis in original). 68   Cook v Rockwell Intern Corp 580 F Supp2d 1071, 1098 (D Colo 2006). 69   AB Hill, ‘The Environment and Disease: Association or Causation’ (1965) Proceedings of the Royal Society of Medicine 295, 297. 70  ibid. 71   2005 2 SC 1, [6.170]–[6.171]. 72   ibid [6.155], [6.162]–[6.163]. 66 67

Causation Problems in Product Liability  159 been concluded that there was a causal connection between cigarette smoking and lung cancer.73 In his view, this was a missed opportunity: This could have been done: it is clear that the survey of British doctors, on which Sir Richard Doll and colleagues have worked for many years, is regarded as a classic of its kind, both because of the pioneering nature of the research, a preliminary report of which was published as Doll and Hill in 1950, and because this had been followed up with subsequent papers over several decades. I could at least have been shown these papers, which I assume disclosed the data, the statistical techniques and all the other considerations which led to the authors’ conclusions so that I could see for myself whether these conclusions were soundly based. The opportunity was there with Sir Richard Doll in the witness box, and indeed Professor Friend for one thought that evidence would be given about this survey. Warning had been given on behalf of [Imperial Tobacco Ltd] . . . that Sir Richard Doll’s data were of potential interest to the court. But in the event no attempt was made to show me the data.74

Recent Scots law in the decision of Smith v McNair 75 reaffirms the very same cautious approach to the interpretation of epidemiological evidence established in previous cases.76 It stresses the need for experts to teach the court how to do the epidemiology before it can form its reasoned judgment on interpreting the epidemiological evidence. While acknow­ ledging that medical witnesses are entitled to refer to medical literature, in particular to refer to published papers by epidemiologists, even though they themselves are not epidemiologists,77 Lord McEwan in Smith has stressed the need to look at such evidence critically because the writers of it could not be cross-examined themselves. Such scientific evidence only became a factor for consideration if it was ‘intelligible, convincing and tested’.78 Accordingly, in Scotland the cases are at one in emphasising that where a pursuer seeks to rely on epidemiological evidence of disease through an expert witness in proving causation, the pursuer must import to the court special knowledge of the subject matter of epidemiology, so that the court can form its reasoned judgment on the epidemiological evidence.79 Such a cautious approach to epidemiological evidence was central to the decision in Smith. While sympathetic to the experts who were ‘outwith their chosen discipline and abroad in the field of epidemiology’,80 Lord McEwan concluded nonetheless that they were unable to explain the studies, which seemed to him to ‘raise more questions than answers’.81 Unlike McTear, however, Smith shows less of an impression of what Chris Miller has described as a ‘dogmatic aversion’82 to statistical evidence. Lord McEwan felt that many of   ibid [6.163].   ibid [6.162]. 75   Smith v McNair [2008] CSOH 154; see R Goldberg, ‘Causation, Idiopathic Conditions and the Limits of Epidemiology’ (2009) 13 Edinburgh Law Review 282. 76  See Dingley v The Chief Constable of Strathclyde 1998 SC 548, at 555 per Lord President (Rodger), at 604 per Lord Prosser, aff ’d 2000 SC (HL) 77; McTear v Imperial Tobacco 2005 2 SC 1 (OH) [5.11] per Lord Nimmo Smith. 77   Main v McAndrew Wormald Ltd 1988 SLT 141, 142 per Lord Justice-Clerk (Ross). 78   Smith v McNair [2008] CSOH 154, [18], citing Davie v Magistrates of Edinburgh 1953 SC 34 at 40. 79   cf in the United States, the Advisory Committee Note (2000 Amendment) to Fed R Evidence 702. The Amendment not only stresses that the expert conducts the application of principles and methods to the facts of cases reliably, but also reiterates the ‘verifiable practice of using expert testimony to educate the factfinder on general principles’. Fed R Evidence 702 Advisory Committee’s Note (2000 amendment). It notes that it might be important in some cases for an expert to educate the fact-finder about general principles, without ever attempting to apply these principles to the specific facts of the case: ibid. 80   [2008] CSOH 154, [81]. 81  ibid. 82   C Miller, ‘Causation in Personal Injury; Legal or Epidemiological Common Sense’ (2006) 26 Legal Studies 544, 566. 73 74

160  Richard Goldberg the problems with the evidence might have been helped if the authors of the reports had been called and there had been some statistical evidence. Without such assistance, the judge was ‘at once disabled from being able properly to evaluate the worth of the study or to draw on the proper conclusions’.83 In his view, therefore, this was an appropriate case for epidemiologists to give evidence and for experts to explain their studies. He did not, however, believe that this was always the case, and suggested that reliance on doctors and epidemiologists ‘can almost lead the court unwittingly into a kind of satellite litigation on issues away from the pursuer’s case’.84 He seemed to regard McTear and another Scottish decision, Dingley, as two recent examples of this,85 yet it is submitted that the use of statistics in determining causation is hardly satellite litigation: in both cases it was a primary issue which required resolution in the face of scientific uncertainty. It is submitted that the concern with such a cautious approach in Scots law to epidemiological evidence is that it may make it harder to even discern that there is any reconciliation of the legal standard of proof on a balance of probabilities with the scientific standard of statistical significance. Even more importantly, there is also concern that the mere placing of an obligation on the pursuer to teach the epidemiology to the court suggests that the court can act passively in this process. This is surely an unhelpful approach in cases such as McTear, where there is a clear societal function of a judge to resolve these matters to the satisfaction of both parties. As a leading American judge has observed in cases where judges preside over non-jury trials: Passivity of the court is no virtue when serious scientific questions of more than passing import­ ance are involved. The court owes an obligation to the parties, to society, and to itself to assist in obtaining the best possible answers to the scientific questions before it. That will mean forcing the parties to gather and present evidence effectively, calling upon other experts as necessary, and studying to obtain the understanding needed to maintain effective control.86

Had the pursuer explained the epidemiologic evidence properly, and had Lord Nimmo Smith been more receptive to the strength of evidence of relative risk, as well as taken a more active role in forcing the pursuers to present their evidence effectively, it would seem that general causation could have been established in this case. Moreover, the fact that the defenders admitted that the World Health Organization, the United Kingdom Government and the United States Government had accepted for many years that cigarette smoking can cause lung cancer87 was surely important generally accepted scientific evidence to which Lord Nimmo Smith should have been given adequate weight.88 Irrespective of the con­ clusions on general causation, the problem of course remained of establishing individual causation, in the context of naked statistical evidence. It is to this that we now turn.

  [2008] CSOH 154, [80].   ibid [16]. 85   ibid [16], [29]. 86   J Weinstein, ‘Improving Expert Testimony’ (1986) 20 University of Richmond Law Review 473, 495–96. Judge Weinstein also encourages judges presiding over non-jury trials ‘to become familiar with the scientific background by reading about the issues and discussing them with the experts’: ibid 494; noted in Snyder v Sec’y of Health & Human Services, 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *2. 87   2005 2 SC 1, [2.7], [6.30]. 88   Counsel for the pursuer had submitted (unsuccessfully) that considerable weight be placed on the fact that this proposition had come to be generally accepted: ibid [6.41]. 83 84

Causation Problems in Product Liability  161

E.  The Statistical Chance/Personal Chance Dichotomy It has been argued that there is a dichotomy between two kinds of chances – one ‘statistical’ and the other ‘personal’. A statistical chance is a figure collected from ‘previous unconnected outcomes, giving a probability of that outcome in any non-individual case’, whereas a personal chance is ‘peculiar to a particular individual’.89 A statistical chance has no compensatory value, until the data is ‘personalised.’90 The impossibility of applying statistics derived from epidemiological studies to determine causation in individual cases was cited as the principal reason for the pursuer’s failure to prove individual causation in McTear, since such evidence could not prove that it was more likely than not that but for her husband’s smoking of cigarettes he would not have contacted lung cancer.91 As Lord Nimmo Smith put it:92 The information provided in an observational epidemiology is generally such that it can neither confirm nor refute a causal relationship, particularly when the exposure in question is not specific­ally associated with a certain condition (ie the exposure is always associated with the condition, and vice versa). Epidemiology cannot provide information on the likelihood that an exposure produced an individual’s condition. The population attributable risk is a measure for populations only and does not imply a likelihood of disease occurrence within an individual, contingent upon that individual’s exposure. The fact that cases and non-cases can emerge both from the unexposed and the exposed groups shows that the likelihood of the individual occurrence cannot be reliably predicted from his or her exposure group membership alone. The group estimates obscure the underlying heterogeneity of the population, so that it is entirely possible that other group memberships besides exposure, like genetic profile, socio-economic status, workplace, diet and other exposures make a major contribution to disease occurrence. The question of using epidemiological data for individual causation raises the problem of identifying a particular individual who was harmed by the exposure. While models such as the assigned share concept, derived from attributable fractions, have attempted to deal with this, they suffer from the limitations mentioned by Dr Lewis. The attempt to identify exposure as the sole cause of disease in an individual produces a statement counter to fact in that it implies that the individual would have remained healthy if the exposure had not occurred. This, as Dr Lewis said, is not provable and cannot be derived from epidemiological data.

He concluded that given there were other possible causes of lung cancer other than cigarette smoking, and that lung cancer could occur in a non-smoker, it was not possible to determine in any individual case whether but for an individual’s cigarette smoking he probably would not have contracted lung cancer.93 In doing so, he referred to ‘[t]he fallacy of applying statistical probability to individual causation’.94 89   T Hill, ‘A Lost Chance for Compensation in the Tort of Negligence by the House of Lords’ (1991) 54 MLR 511, 512. 90   ibid 518. See Hotson v East Berkshire AHA [1987] 2 WLR 287, 303 (CA) per Croom-Johnson LJ. 91   2005 2 SC 1, [6.180], [6.184–6.185]. See also the above discussion in Sienkiewicz v Greif [2011] UK SC 10, [163] per Lord Rodger, [170] per Baroness Hale, [191], per Lord Mance, [205] per Lord Kerr, [218] per Lord Dyson (n 49–52). As in the UK, US courts often bifurcate both specific and general causation elements: see J Sanders, ‘The Controversial Comment C: Factual Causation in Toxic-Substance and Disease Cases’ (2009) 44 Wake Forest Law Review 1029, 1032. 92   ibid [6.180]. 93   2005 2 SC 1, [6.184]–[6.185]. 94   2005 2 SC 1, [6.184]. For the need to exercise caution in the use of general statistics in establishing causation, and the importance of looking at the claimant’s individual circumstances, see the observations of Brooke LJ in Wardlaw v Farrar [2003] 4 All ER 1358, [2004] Lloyd’s Rep Med 98, [2004] PIQR 19 at [ 35]–[36]. See, further, Sienkiewicz v Greif [2011] UK SC 10, [152], [163] per Lord Rodger, [170], [172] per Baroness Hale, [190]–[192] per Lord Mance, [204]–[206] per Lord Kerr.

162  Richard Goldberg However, his dicta require closer scrutiny. While Lord Nimmo Smith was correct to observe that there are limitations of epidemiological evidence, his description of such limitations of epidemiology is somewhat inaccurate. In stating that ‘group estimates obscure the underlying heterogeneity of the population so that it is entirely possible that other group memberships besides exposure, like genetic profile, socio-economic status, workplace, diet and other exposures make a major contribution to disease occurrence’,95 he fails to appreciate that epidemiologists can and do adjust for these potentially confounding factors through logistic regression statistical techniques.96

F.  Overcoming the Statistical Chance/Personal Chance Dichotomy: Bayes’ Theorem The problem of utilising statistics deriving from trends in general populations to prove causation in an individual case has been recognised judicially by the House of Lords in Hotson v East Berkshire Area Health Authority97 and, more recently, in Gregg v Scott.98 Yet it is arguable that epidemiological evidence can be refined to draw conclusions about the cause of disease in an individual using specific risk factors such as were present in Mr McTear’s case. A ‘statistical chance’ could be refined and personalised into a ‘personal chance’99 using the Bayes’ Theorem, which can modify evaluations of probability based on initial assumptions in the light of more data and thus can help refine statistical chances into personal chances.100 Bayes’ Theorem expresses the relationship between the probability of a proposition (A) evaluated before the utilisation of new data (B) (prior probability) and the probability of the same proposition evaluated after the utilisation of the new data (posterior probability). Thus: Posterior Probability = Prior Probability of A × Probability of B given A of A given B 1 Unconditional Probability of B i.e. P(A/B) = P(A) × P(B/A) P(B) Thus, applied to McTear, and assuming that general causation was established, we would be looking to determine the following:

  2005 2 SC 1, [6.180].  See, further, KJ Rothman, S. Greenland, TL Lash, Modern Epidemiology, 3rd edn (Philadelphia, PA, Lippincott Williams & Wilkins, 2008) 394–95. 97   Hotson v East Berkshire Area Health Authority [1987] AC 750, 789 per Lord Mackay. 98   Gregg v Scott [2005] 2 AC 176, 26–33 per Lord Nicholls and 153 per Lord Phillips. 99   T Hill, ‘A Lost Chance for Compensation in the Tort of Negligence by the House of Lords’ (1991) 54 MLR 511, 518. 100   See, for further discussion, R Goldberg Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (n 3) 39–40, the principles of which are adapted to meet the circumstances of McTear, below. 95 96

Causation Problems in Product Liability  163 The probability of a cigarette-induced cancer given McTear has cancer

The probability of cigarette-induced = cancer

The probability of cancer given McTear has × cigar­ette induced cancer The probability of cancer from any cause

A statistical chance could be refined and personalised into a personal chance using specific factors which are embodied in the likelihood ratio.101 The probabilities in the likelihood ratio (LR) can be decomposed into factors in the light of specific case information in respect of patient history. Such factors would include the risk factors in McTear’s case, viz his personality traits (LR (Pt)), family history of lung cancer (LR (Fh)), stressful lifestyle (LR (Sl)), viral infections of the respiratory tract (LR (ViRt)), alcohol abuse (LR (Aa)), vitamin A deficiency (LR(VAd)), low socio-economic status (LR (Ls-eS)) and residence in an urban area in the west of Scotland (LR (RuWS)).102 The likelihood ratio is found by obtaining the product of all the individual likelihood ratio factors. Diagrammatically this can be expressed by: LR = LR (Pt)× LR (Fh)× LR(Sl) × LR(ViRt), × LR(Aa) × LR(VAd) × LR (Ls-eS) × LR (RuWS) (Caveat = components, ie Pt etc must be statistically independent).

The use of all these factors is dependent on the specific case information available. If all specific case information in respect of the factors is available, the posterior odds103 is calculated as follows: Posterior Odds = Prior Odds × LR (Pt) × LR (Fh) × LR(Sl) × LR(ViRt), × LR(Aa) × LR(VAd) × LR (Ls-eS) × LR (RuWS) Thus the posterior odds can be further refined by combining the prior odds, based on background information and the likelihood ratios, based on case-specific information, to produce as accurate a posterior probability as possible.104 This would seem to be a possible tool in improving probabilistic precision in the McTear type of case, and in so doing, helping to overcome the difficulties of the statistical chance – personal chance dichotomy.

  The likelihood ratio (LR) of a test can be defined as: The true positive incidence of the test The false positive incidence of the test 102   See 2005 2 SC 1, [6.181]. 103   The relationship between odds and probability is: Odds = Probability 1-Probability Thus the probability of 0.9 = odds of 9:1. 104   B Donatini et al, ‘Causality Assessment of Spontaneous Reporting: Correlation Between Bayesian and Other Approaches’ (1993) 7 Pharmaceutical Medicine 255, 256. 101

164  Richard Goldberg

III. Developments in the US: Omnibus Autism Proceeding Test Cases

A.  Assessing the Significance of Admitted Scientific Testimony of Proof of Causation in Product Liability Cases The role of scientific evidence in determining causation in product liability suits is well-established in the US. Here the emphasis is often on the relevance and reliability of evidence since the landmark Supreme Court decision of Daubert v Merrill Dow Pharmaceuticals, Inc.105 The Daubert framework has been utilised in exercising the gatekeeping role to the admissibility of scientific evidence – in some cases in order to reduce the use of so-called ‘junk science’ by juries. However, a further issue developing in the case law has been the assessment of the sufficiency of scientific evidence admitted in the determination of proof of causation in complex product liability cases, which often involve prescription drugs and vaccines. Of recent interest is the assessment of the value of scientific evidence in three test cases of the Omnibus Autism Proceeding (OAP) under the National Childhood Vaccine Injury Act of 1986 (NVIA), which were decided in February 2009, as well as a further three test cases in March 2010.106 The Petitioners’ Steering Committee (PSC) presented two different theories of ‘general causation’ in the OAP, designating three ‘test cases’ for each of the two theories. The long-awaited test cases in these proceedings are of considerable importance, since they have irrefutably rejected the petitioners’ first and second general causation theories. The special masters in these proceedings, having considered all the available scientific evidence, concluded in the first three test cases that there was no merit in the petitioners’ first general causation theory that MMR vaccines and thimerosal containing vaccines could combine to cause autism, and in the second three test cases that there was no merit in the petitioners’ second causation theory that thimerosal vaccines alone can cause autism.

B.  The First Three Test Cases and the Petitioners’ First Theory i.  The First Three Test Cases The United States Vaccine Court Omnibus Autism Proceeding under the NVIA gave three rulings in the three test cases where the petitioners claimed that measles-mumps-rubella vaccines combined with thimerosal-containing vaccines administered to three children 105   Daubert v Merrill Dow Pharmaceuticals Inc 509 US 579, 589–95 (1993); and, further, General Electric Co v Joiner, 522 US 136, 139, 141, 143, 146 (1997); Kumho Tire Co v Carmichael, 526 US 137 (1999), 147–49, 152–53 (1999); Weisgram v Marley Co 528 US 440, 456–57 (2000); Fed R Evidence Rule 702, 28 USCA; Fed R Evidence 702 Advisory Committee’s Note (2000 Amendment); see D Owen, Products Liability Law, 2nd edn (St Paul, Thomson West, 2008) § 6.3. For criticism that Daubert and its progeny have in some cases hindered the search for justice in product liability law, see generally, C Cranor, Toxic Torts: Science, Law, and the Possibility of Justice (n 20) and, in particular, ibid 337, 340–41, 348, 353, 356, 368. 106   The US test cases can be contrasted with the UK MMR/MR Vaccine Litigation, where, with only two claimants continuing with claims (those two having had their public funding restored, the rest having had their funding withdrawn by the Legal Services Commission), and there being no realistic prospect of any new claims being progressed in the light of the unavailability of public funding, the status of the litigation as group litigation was dissolved: see Re MMR and MR Vaccine Litigation; Sayers and others v Smithkline Beecham plc and others [2007] EWHC 1335, QB, [2007] All ER (D) 67 (June), [35], [37].

Causation Problems in Product Liability  165 had caused several conditions, including autism107 and chronic gastrointestinal symptoms. The key question under the National Vaccine Programme is the establishment of a causal link between the vaccination and the injury. In some cases the petitioner may simply demonstrate the occurrence of a so-called Table Injury, ie that the vaccine recipient was administered a vaccine and suffered an injury covered by the NVIA and occurring within an applicable time period following the vaccination specified in the Vaccine Injury Table.108 If so, the Table Injury is presumed to have been caused by the vaccination.109 However, in the Omnibus Autism Proceeding, each of the petitioners’ test cases was based on an exception to the Table. Here, the petitioners claimed that they suffered injuries not of the type covered in the Table, but that they could show by a preponderance of evidence that their injuries were ‘caused- in- fact’ by the vaccination in question.110 This is known as an offTable injury or causation-in- fact claim. In contrast to the relaxation of the burden of proving causation for injuries satisfying the Table, the burden of proof on the petitioner in a causation in fact claim is a heavy one.111 Essentially, the three test cases, Cedillo,112 Snyder113 and Hazlehurst,114 were three of more than 5000 cases filed with the National Childhood Vaccine Injury Compensation Program, in which it has been alleged that autism or a similar disorder was caused by one or more vaccines. The evidentiary record was described by the special master in Cedillo as ‘massive’,115 and one which dwarfed, by far, any evidentiary record in any Program case. The amount of medical literature filed into records of the three cases was noted as being ‘staggering’.116 During the evidentiary hearings, a total of 28 expert witnesses testified. A total of 939 different items of medical literature were filed into the three cases, the complexity of the material involving many different specialities of biology and medicine, including neurology, 107   ‘Autism’ is the term used to describe ‘a complex and severe set of developmental disorders characterised by sustained impairments in social interaction, impairments in verbal and nonverbal communication, and stereotypically restricted or repetitive patterns of behaviours and interests’: Institute of Medicine, Immunization Safety Review: Vaccines and Autism (Washington DC, The National Academies Press, 2004), Executive Summary, 3. In the Omnibus Proceeding, it was noted that the terms ‘autism’, ‘autistic’ and ‘autism spectrum disorder’ would be used interchangeably to refer to the entire group of disorders within the category of ‘pervasive developmental disorder’ (PDD). 108   42 USC §300aa-11(c)(1)(C)(i). 109   42 USC §300aa-11(c)(1)(C)(i); §300aa-13(a)(1)(A). 110   42 USC §300aa-11(c)(1)(C)(ii)(I); §300aa-13(a)(1)(A); Moberly v Sec’y of Health & Human Services 592 F 3d 1315, 1322 (Fed Cir 2010); Shyface v Sec’y of Health & Human Services 165 F 3d 1344, 1352–53 (Fed Cir 1999); Hines v Sec’y of Health & Human Services 940 F 2d 1518, 1525 (Fed Cir 1991). 111   Grant v Sec’y of Dep’t of Health & Human Services 956 F 2d 1144, 1148 (Fed Cir 1992); Hodges v Sec’y of Dep’t of Health & Human Services 9 F3d 958, 961 (Fed Cir 1993). Nonetheless, it has been judicially observed that Congress ‘clearly intended’ that its goal of rendering expeditious, certain and generous determinations should apply equally to Table and off-Table claims: Stevens v Sec’y of HHS, No 99–594 V, 2001 WL 387418 (Fed Cl 30 March 2001) *7 per Chief Special Master Golkiewicz (noting the difficulties associated with causation in fact cases under the National Childhood Vaccine Injury Compensation Program); HR Rep No 99–908, 13; see, further, K E Strong, Note, ‘Proving Causation Under the Vaccine Injury Act: A New Approach for a New Day’ (2007) 75 George Washington Law Review 426, 442–46 (submitting that the medical and scientific uncertainties surrounding vaccine injuries, as well as the lack of a uniform standard for causation in fact cases, has meant that the goals of Congress have not been met for petitioners who require to prove off-Table claims). 112   Cedillo v Sec’y of Health & Human Services, 98–916V, 2009 WL 331968 (Fed Cl 12 February 2009), aff ’d 89 Fed Cl 158, 164, 184 (2009), aff ’d 617 F3d 1328, 1334, 1349–1350 (Fed Cir 2010). 113   Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009), aff ’d, 88 Fed Cl 706, 708, 748 (2009). 114   Hazlehurst v Sec’y of Health & Human Services 03-654V, 2009 WL 332306 (Fed Cl 12 February 2009), aff ’d 88 Fed Cl 473, 475, 490 (2009), aff ’d 604 F 3d 1343, 1345, 1354 (Fed Cir 2010). 115   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *14. 116   ibid *14–15.

166  Richard Goldberg gastroenterology, virology, immunology, molecular biology, toxicology, genetics and epidemiology.117 ii.  The First General Causation Theory The petitioners advanced a causation theory which had several parts, including three main contentions, viz: (1) that thimerosal-containing vaccines can cause immune dysfunction; (2) that the MMR vaccine can cause autism; and (3) that the MMR vaccine can cause gastrointestinal dysfunction.118 It was agreed that the Petitioners’ Steering Committee (PSC) would present its general causation evidence concerning the first theory, along with all the evidence specific to the Cedillo case.119 As to each of the general causation theory elements, Special Master Hastings concluded that ‘the evidence was overwhelmingly contrary to the petitioners’ contentions’.120 Considerable emphasis was placed on the respondent’s expert witnesses, who were ‘far better qualified, far more experienced and far more persuasive than the petitioners’ experts concerning most of the key points’.121 The numerous medical studies came down strongly against the petitioners’ contentions. Having considered all the evidence, the special master found that the petitioners had failed to demonstrate that thimerosal containing vaccines in general could contribute to causing immune dysfunction or that the MMR vaccine could contribute to causing either autism or gastrointestinal dysfunction.122 The petitioners’ general causation theory concerning the causation of autism was contingent on a weakening of the immune system by thimerosal containing vaccines which allowed the measles virus contained in the MMR vaccine to persist within the child’s body.123 However, the determination by the special masters in all three cases that the testing 117   ibid. For a comprehensive discussion of the early stages of the Omnibus Autism Proceeding prior to the decisions in the test cases being decided, as well as discussion of potential problems in the aftermath of the Proceeding, see G Shemin, ‘Mercury Rising: The Omnibus Autism Proceeding and What Families Should Know Before Rushing Out of Vaccine Court’ (2008) 58 American University Law Review 459. 118   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *1. 119   ibid *10. Then, over the following months, the PSC would present its case specific evidence concerning the two additional test cases, viz Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) and Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009): ibid. 120   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *1. See also, Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *1 per Special Master Vowell. 121   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *1. See also Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *1 per Special Master Vowell (stating that ‘it was abundantly clear that petitioners’ theories of causation were speculative and unpersuasive’); and Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *13. 122   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *1. See also Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *1, *76, *104, *137; Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *85, *150, *171. 123   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *15; Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *28; Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *86. The special master in Hazlehurst summarised the theory of causation as follows: Specifically, petitioners assert that the measles component of the MMR vaccine causes an immune dysfunction that impairs the vaccinee’s ability to clear the measles virus. Unable to clear properly clear [sic] the measles virus from the body, the vaccinee experiences measles virus persistence which leads to chronic inflammation in the gastrointestinal system and, in turn, chronic inflammation in the brain. Petitioners argue that the inflammation in the brain causes neurological damage that manifests as autism. It is also the position of petitioners

Causation Problems in Product Liability  167 for the presence of the measles virus in the intestinal tissue of Cedillo, Snyder and Hazlehurst and other autistic children was unreliable124 was fatal to all three decisions. The petitioners’ general theory concerning the causation of autism was rejected on the basis of nine grounds, viz: (1) the general theory depended upon the existence of reliable laboratory test findings of persisting measles virus, but such a reliable test did not exist;125 (2) the available evidence did not demonstrate any substantial likelihood that measles virus persistence in the brain would cause autism;126 (3) the evidence indicated that the wild measles virus had never been shown to cause autism which made it quite unlikely that the vaccine strain form of the measles virus could cause autism;127 (4) the petitioners’ theory seemed unlikely in the light of several accepted understandings concerning the causation of autism, in particular that there was a very strong genetic component to the causation of autism;128 (5) there were contradictions and inconsistencies in the testimony concerning the appropriate time period between the MMR vaccination and the onset of autism symptoms;129 (6) the testimony of three other experts failed to provide substantial support to the causation theory of the petitioners’ expert Dr Kinsbourne;130 (7)  the qualifications of the respondent’s experts concerning this issue substantially exceeded the qualifications of the petitioners’ expert witnesses;131 that the viral persistence is facilitated by the vaccinee’s receipt of thimerosal containing vaccines that suppress the immune system of the vaccinee and impair the immune system’s ability to respond properly to the viral presence: ibid *86. 124   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *29–59, aff ’d 89 Fed Cl 158, 171–72 (2009), aff ’d 617 F3d 1328, 1345 (Fed Cir 2010); and, further, Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *116; Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *150. The studies purported to find the presence of the measles virus in the biological material of autistic children and primarily derived from two sources: the work of Dr Andrew Wakefield of the Royal Free Hospital in London (see, in particular, his article, AJ Wakefield et al, ‘Ileallymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children’ [retracted] (1998) 351 Lancet 637–41), and his colleagues John O’Leary and Orla Sheils at the for profit, non-accredited Unigenetics laboratory in Dublin; and the research of Dr Stephen Walker of Wake Forest University School of Medicine, North Carolina. Dr Wakefield and his colleagues were ‘the principle proponents of the hypothesis that the receipt of the MMR vaccine results in the development of autism spectrum disorders and gastrointestinal problems in certain children’: Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *87, *126. The special master found that the work of Dr Wakefield had been largely discredited and that none of the studies indicating the presence of the measles virus in autistic children had been successfully replicated independently of Wakefield or Unigenetics: ibid *90, 124. The testimony of a government expert, Professor Stephen Bustin (who had also been an expert for the vaccine manufacturers in the UK MMR litigation) helped to discredit the reliability of the testing conducted at Unigenetics: ibid *129–32. It was held on appeal that this testimony had been properly admitted: Hazlehurst v Sec’y of Health & Human Services 88 Fed Cl 473, 480–83 (2009), aff ’d 604 F 3d 1343, 1349–50 (Fed Cir 2010). Shortly before the first test case, Cedillo, was due to begin, the Secretary of Health and Human Services was granted permission to obtain from the records of the High Court, copies of expert witness reports of Professors Bustin, Simmonds and Rima, filed by the defendants in the UK MMR/MR Vaccine Litigation, so as to use these documents in the Omnibus Autism Proceedings in the US: Sayers v Smithkline Beecham Plc, Smith Kline & French Laboratories Limited, Merck & Co Inc, Sanofi Pasteur MSD Limited [2007] EWHC 1346 (QB). 125   ibid *67–68. 126   ibid *67–69. 127   ibid *67, *69–71. 128   ibid *67, * 71–77. 129   ibid *67, *77–79. 130   ibid *67, *79–83. 131   ibid *67, *83–84.

168  Richard Goldberg (8) the epidemiologic evidence consisting of numerous studies by qualified medical researchers around the world132 added another reason to reject the petitioners’ theory that vaccines could contribute to the causation of autism;133 and (9) Two reports of well qualified experts published by the Institute of Medicine in 2001 and 2004 studied the general MMR/autism causation issue and concluded that the evidence favoured rejection of the proposition that the MMR vaccine could cause autism.134 Taken together, all this evidence was irrefutable.

C.  The Second Three Test Cases and the Petitioners’ Second Theory i.  The Second Three Test Cases The Petitioners Steering Committee’s second causation theory was that thimerosal-­ containing vaccines alone can cause autism.135 The same three special masters who had been tasked with hearing the first three test cases concerning the first theory of general causation were also tasked with hearing the second three test cases concerning the second 132   See B Taylor et al, ‘Autism and Measles, Mumps, and Rubella Vaccine: No Epidemiological Evidence for a Causal Association’ (1999) 353 Lancet 2026–29; F DeStefano et al, ‘Age at First Measles-Mumps-Rubella Vaccination in Children With Autism and School-Matched Control Subjects: A Population-Based Study in Metropolitan Atlanta’ (2004) 113 Pediatrics 259–66; L Smeeth et al, ‘MMR Vaccination and Pervasive Developmental Disorders: A Case-Control Study’ (2004) 364 Lancet 963–69; B Taylor et al, ‘Measles, Mumps, and Rubella Vaccination and Bowel Problems or Developmental Regression in Children with Autism: Population Study’ British Medical Journal 2002; 324: 393–96; CP Farrington et al, ‘MMR and Autism: Further Evidence Against a Causal Association’ (2001) 19 Vaccine 3632–35; E Fombonne and S Chakrabarti, ‘No Evidence for a New Variant of Measles-Mumps-Rubella-Induced Autism’ (2001) 108 Pediatrics e58; S Dewilde, ‘Do Children Who Become Autistic Consult More Often After MMR Vaccination?’ (2001) British Journal of General Practice 51, 226– 27; A Mäkelä, et al, ‘Neurologic Disorders After Measles-Mumps-rubella Vaccination’ (2002) 110 Pediatrics 957– 63; H Takahashi, et al, ‘An Epidemiological Study on Japanese Autism Concerning Routine Childhood Immunization History’ (2003) 56 Japanese Journal of Infectious Diseases 114–17; W Chen et al, ‘No Evidence for Links Between Autism, MMR and Measles Virus’ (2004) 34 Psychological Medicine 543–53; H Honda et al, ‘No Effect of MMR Withdrawal on the Incidence of Autism: A Total Population Study’ (2005) 46 Journal of Child Psychology and Psychiatry 572–79. 133   ibid *67, *92–93. Special Master Hastings effectively destroyed the sufficiency of the epidemiologic evidence proffered by the petitioners in the following two paragraphs: The numerous epidemiologic studies done over the past ten years, when taken together, make it very unlikely that the MMR vaccination has played any significant role in the overall causation of autism. It is true, as the petitioners argue, that the available epidemiologic studies do not completely rule out the possibility that the MMR vaccine might be associated with some small subset of autism, such as regressive autism. However, there are three reasons why the epidemiologic evidence still must be said to provide significant evidence against the petitioners’ general causation theory set forth in this case. First, none of the numerous competent studies has yielded the slightest bit of evidence in the petitioners’ favor. Second, the failure of so many studies to find any association between MMR vaccine and autism, while not completely ruling out a possible causal role with respect to a subset of autism, at least casts considerable doubt upon the proposition that the MMR vaccine ever plays a role in causing any kind of autism, including regressive autism. And, third, five studies provide evidence that is directly relevant to the petitioners’ ‘regressive autism only’ argument, supplying significant evidence against the theory that the MMR vaccine plays a causal role even in the subset of autism known as regressive autism. Accordingly, my conclusion is that the epidemiologic evidence does provide yet another strong reason to reject the petitioners’ general causation theory presented in this case: ibid * 92–93. 134   ibid *68, *93–94; see Institute of Medicine, Immunization Safety Review: Measles-Mumps-Rubella Vaccine and Autism (Washington DC, The National Academies Press, 2001) 60; Institute of Medicine, Immunization Safety Review: Vaccines and Autism (Washington DC, The National Academies Press, 2004) 7, 16, 126, 151–52. 135   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *2.

Causation Problems in Product Liability  169 theory of general causation advanced by the petitioners.136 The evidentiary record was described as ‘massive’,137 and one which exceeded any evidentiary record in any Program case, with the exception of the record in the first three test cases. During the evidentiary hearings, a total of 26 expert witnesses testified. The amount of medical literature filed into the records of the three cases was a ‘staggering’ figure of more than 1200 different items.138 In March 2010, each of the three special masters issued a decision in the test case assigned to them, ie respectively in Mead,139 Dwyer140 and King.141 All three special masters found that the parents had failed to prove that their children’s autism was caused by the thimerosalcontaining vaccines that they received.142 ii.  The Second General Causation Theory The petitioners’ medical theory contended that ‘the thimerosal component of the received childhood vaccines dissociates into organomercurial ethylmercury once in the body’.143 That ethylmercury ‘then courses through the blood stream to diffuse across the bloodbrain barrier to reach the brain’.144 On reaching the brain, ‘the ethylmercury is de-ethylated to become inorganic mercury-a form of mercury that is not quickly removed from the brain-and once deposited, provokes a series of detrimental responses that ultimately manifest as autism’.145 It was found that the underpinnings for the opinions of the petitioners’ experts concerning the second theory were ‘scientifically flawed’, and in the absence of a sound basis for the offered opinions of causation, these opinions ‘[could not] be credited’.146 The theory that the thimerosal content of the vaccines contributed to the development of autism was ‘scientifically unsupportable’.147 Several epidemiological studies148 were examined and it was found that they showed no association between thimerosal-containing vaccines and the development of autistic   ibid *4.   King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *12. 138  ibid. 139   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010). 140   Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010). 141   King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010). 142   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *1, 13, 113; Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *1-2, 201; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *1, 90–91. 143   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *106; and, further, at *17. 144  ibid. 145  ibid. 146   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *109, citing Perreira v Sec’y of Health & Human Services, 33 F 3d 1375, 1377 fn 6 (Fed Cir 1994) (‘An expert opinion is no better than the soundness of the reasons supporting it’). 147   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *113; see, further, Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *165, 198–199; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *34–35. 148   See A Hviid et al, ‘Association Between Thimerosal-Containing Vaccine and Autism’ (2003) Journal of the American Medical Association 290 (13) 1763–66; K Madsen et al, ‘Thimerosal and the Occurrence of Autism: Negative Ecological Evidence From Danish Population-Based Data’ (2003) 112 Pediatrics 604–06; T Verstraeten et al, ‘Safety of Thimersoal-Containing Vaccines: A Two Phased Study of Computerized Health Maintenance Organization Databases’ Pediatrics (2003) 112, 1039–48; P Stehr-Green et al, ‘Autism and Thimerosal-Containing Vaccines’ (2003) 25 American Journal of Preventative Medicine 101–06; N Andrews et al, ‘Thimerosal Exposure in Infants and Developmental Disorders: A Retrospective Cohort Study in the United Kingdom Does Not Support a Causal Association’ (2004) 114 Pediatrics 584–91; H Jick and J Kaye, ‘Autism and DPT Vaccination in the United Kingdom’ (2004) 350 New England Journal of Medicine 2722–23; E Fombonne et al, ‘Pervasive Development Disorders in Montreal, Quebec, Canada: Prevalence and Links with Immunizations’ (2006) 118 Pediatrics 118 (1): 136 137

170  Richard Goldberg spectrum disorders.149 Reference was made to the evidence given by the eminent paediatric psychiatrist, Professor Sir Michael Rutter who, having examined the limitations of the epidemiological studies, had concluded that ‘taken as whole, the studies were all “unsupportive of a causal association” ’.150

D.  Implications of the Test Cases i.  Epidemiological Evidence Should be Given Appropriate Weight Some of the most significant evidence used to reject the general causation theory were the numerous epidemiologic studies performed over the previous 10 years which, when taken together, made it very unlikely that the MMR vaccination played a significant role in the overall causation of autism.151 Both the Cedillo and King test cases, determined by Special Master Hastings, clarify the position surrounding the use of such epidemiological evidence supporting a causation in fact claim under the National Childhood Vaccine Injury Program (the Program). They reaffirm the settled legal position that while there is no requirement that epidemiological evidence supports a causation in fact claim under the Program,152 in the relatively rare instance in which general causation has been the subject of published e139–50; R Schechter and J Grether, ‘Continuing Increases in Autism Reported to California’s Developmental Services System’ (2008) 65 Archives of General Psychiatry 19–24. Two other studies did not directly address the question of an association between thiomersal and autism, but provided relevant information: (see, J Heron and J Golding, ‘Thimerosal Exposure in Infants and Developmental Disorders: A Prospective Cohort Study in the United Kingdom Does Not Support a Causal Association’ (2004) 114 Pediatrics 577–83; and WW Thompson et al, ‘Early Thimerosal Exposure and Neuropsychological Outcomes at 7 to 10 Years’ (2007) 357 New England Journal of Medicine 1281–92). In the light of the strength of the epidemiological evidence of no association, and given the absence of any direct evidence for a biological mechanism, the Institute of Medicine concluded that the evidence favoured rejection of a causal association between thimerosal-containing vaccines and autism: Institute of Medicine, Immunization Safety Review: Vaccines and Autism (Washington DC, The National Academies Press, 2004) 16, 151–52. 149   Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *39. See, further, Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *77 (‘In this case, the epidemiological studies furnish powerful evidence refuting a causal association between TCVs [thimerosal containing vaccines] and ASD’); and King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) * 66–67. While the petitioners conceded through their expert, Professor Sander Greenland, that the epidemiologic literature to date had not detected an association of mercury-containing vaccines and autism in general or autistic spectrum disorders, Dr Greenland claimed that the performed epidemiological studies lacked the requisite specificity to detect an association between the receipt of thimerosal-containing vaccines and regressive autism: Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *41–45. This position presumed that regressive autism was a distinct phenotype of autism. However, the Special Master found that studies of the developmental patterns in children described as having early onset autism and in children described as having regressive autism, militated against a finding that regression in autism constituted a separate phenotype of autistic disorder: ibid *45, 112; and, further, Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *62–63 (petitioners failed to demonstrate the exist­ ence of ‘clearly regressive autism’ as a separate phenotype; Dr Greenland’s opinion that the existing epidemiologic studies could not rule out a substantial causal rule for thimerosal-containing vaccines in one form of autism was ‘not relevant or persuasive’); and King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *39, 70–72. When the results of the epidemiological studies were viewed as a whole, they were found to reach the consistent conclusion that there was no association between thimerosal-containing vaccines and autism: Mead v Sec’y of Health & Human Services, 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *45; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *75. 150   ibid *40. 151   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *123. 152   Capizzano v Sec’y of Health & Human Services 440 F 3d 1317, 1325–26 (Fed Cir 2006). Indeed, causation can be demonstrated under the Program without any support from medical literature: Althen v Sec’y of Health & Human Services 418 F 3d 1274, 1281 (Fed Cir 2005).

Causation Problems in Product Liability  171 epidemiological studies, such evidence should be given appropriate weight, along with the other evidence of the record.153 ii.  Reliability of Expert Testimony Crucial to the determination of these test cases are the factors that a special master is required to consider in evaluating the reliability of expert testimony and other scientific evidence relating to causation. Even though the Federal Rules of Evidence do not apply in Program cases,154 the test cases reaffirm that it is appropriate to use the Daubert factors as a tool or framework for conducting the enquiry into the reliability of causation in fact theor­ ies.155 In particular, two of the important factors listed in Daubert156 and utilised by the special masters in evaluating these theories157 were whether the scientific theory had been subject to peer review or publication and also whether the theory or technique enjoyed general acceptance.158 Such epidemiological evidence, while not dispositive, should be considered in evaluating scientific theories, such as the general causation theory in issue in the test cases.159 iii.  The Causation-in-Fact Standard: General and Specific Causation and Temporal Proximity One of the most interesting aspects of the test cases is their utilisation of the causation in fact standard. The cases emphasise the importance of establishing both general and specific 153   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *92; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *74. See, further, Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *39; Terran v Sec’y of Health & Human Services 195 F 3d 1302, 1315–17 (Fed Cir 1999); Grant v Sec’y of Health & Human Services 956 F 2d 1144, 1149 (Fed Cir 1992). Epidemiological evidence should be considered in evaluating scientific theories: Scott v Sec’y of Health & Human Services 03–2211V, 2006 WL 2559776, *21; Garcia v Sec’y of Health & Human Services 05–720V, 2008 WL 5068934, *3, *10. The reliance of a special master on epidemiologic evidence has been subject to express approval: see eg, Moberly v Sec’y of Health & Human Services 85 Fed Cl 571, 596 (2009), aff ’d 592 F 3d 1315, 1325 (Fed Cir 2010); Estep v Sec’y of Health & Human Services, 28 Fed Cl 664, 668 (1993); Sharpnack v Sec’y of Health & Human Services 27 Fed Cl 457, 459 (1993); Sumrall v Sec’y of Health & Human Services 23 Cl Ct 1, 8 (1991); Hennessey v Sec’y of Health & Human Services 01–190V, 2010 WL 94560, *6-7, *11–13. 154  42 USC §300aa-12(d)(2)(B): Vaccine Rules ‘shall include flexible informal studies of admissibility of evidence’. 155   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *3; aff ’d, 89 Fed Cl 158, 182 (2009), aff ’d 617 F 3d 1328, 1338–39 (Fed Cir 2010), (applying Terran v Sec’y of Health & Human Services 195 F 3d 1302, 1316 (Fed Cir 1999)); Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *30, *138, *194, aff ’d 88 Fed Cl 706, 736, 744–45 (2009); Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *16–17; aff ’d 88 Fed Cl 473, 483 (2009), aff ’d 604 F3d 1343, 1353 (Fed Cir 2010); Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *13–15; Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *7, 25–26; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *3, 73. For further approval of the utilisation of the Daubert factors in evaluating the reliability of scientific evidence in cases under the Program, see Moberly v Sec’y of Health & Human Services 592 F 3d 1315, 1324 (Fed Cir 2010); Andreu v Sec’y of Health & Human Services 569 F 3d 1367, 1379 (Fed Cir 2009); Knudsen v Sec’y of Health & Human Services 35 F 3d 543, 548 (Fed Cir 1994); Perreira v Sec’y of Health & Human Services 33 F 3d 1375, 1377 fn 6 (Fed Cir 1994). 156   Daubert v Merrill Dow Pharmaceuticals Inc 509 US 579, 593–94 (1993). 157   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *3. 158   Daubert v Merrill Dow Pharmaceuticals Inc 509 US 579, 593-94 (1993). 159   Terran v Sec’y of Health & Human Services 195 F 3d 1302, 1315–17 (Fed Cir 1999); Grant v Sec’y of Health & Human Services 956 F 2d 1144, 1149 (Fed Cir 1992); Scott v Sec’y of Health & Human Services, 03–2211V, 2006 WL 2559776, *21; Garcia v Sec’y of Health & Human Services 05–720V, 2008 WL 5068934, *3, *10; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *74.

172  Richard Goldberg causation in vaccine damage cases, as well as the need for temporal proximity between the vaccine and the damage in each case. This legal standard of proof for causation in fact under the Program was elaborated on in the leading case of Althen v Sec’y of Health & Human Services.160 There, the Federal Circuit established three factors which had to be satisfied to overcome the burden of proof, viz (1) a medical theory causally connecting the vaccination and the injury; (2) a logical sequence of cause and effect showing that the vaccination was the reason for the injury; and (3) a showing of a proximate temporal relationship between the vaccination and the injury.161 In all six test cases in the Omnibus Autism Proceeding, the special masters were able to explain how their analyses of the petitioners’ contentions on the scientific evidence fitted within the three prongs of the test and how in each case none of the requirements of the three factors were satisfied.162 The principal test case of the first general causation theory, Cedillo, provides an import­ ant explanation of the three prongs of the Althen test. The first prong, viz the requirement of a medical theory causally connecting the vaccination and the injury is explained as a general causation requirement, ie that the type of vaccination in question can cause the type of injury in question. The second prong, a logical sequence of cause and effect showing that the vaccination was the reason for the injury, is explained as a specific causation requirement, ie that the particular vaccination received by the specific vaccinee did cause the vaccinee’s own injury. Cedillo affirms the ‘can/did cause’ test, as being equivalent to the first two prongs of Althen.163 Applying the available scientific evidence, the special master held that the petitioners’ arguments fell far short of demonstrating that the MMR vaccination could contribute in general to the causation of either autism or chronic gastrointestinal dysfunction, or that the MMR vaccination did contribute to the causation of Cedillo’s own autism and gastrointestinal symptoms.164 Moreover, there was no doubt that the Althen test required that as an overall matter, a petitioner had to demonstrate that it was more probable than not that the particular vaccine was a substantial contributing factor in causing the particular injury in question.165 This was clear from the ‘preponderance of evidence’ standard in the Vaccine Act.166 Regardless of the precise meanings of Althen, the overall evidence fell far short of demonstrating that it was ‘more probable than not’ that the MMR vaccine contributed to the causation of either Cedillo’s autism or gastrointestinal   418 F 3d 1274 (Fed Cir 2005).   ibid 1278. 162   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *132–33; aff ’d 89 Fed Cl 158, 182–83 (2009); For discussion of the application of the Althen test in the other two test cases of the first general causation theory, see Snyder v Sec’y of Health & Human Services 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009) *29, *192–98, aff ’d 88 Fed Cl 706, 745–46 (2009) and Hazlehurst v Sec’y of Health & Human Services 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009) *15–19, *83–86. For discussion of the application of the Althen test in the three test cases of the second general causation theory, see Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010) *15–16, 106–13; Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March, 2010) *23–24, 196–201; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *87–89. 163   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *131, affirming Pafford v Sec’y of Health & Human Services 451 F 3d 1352, 1355–56 (Fed Cir 2006); and, further, in respect of the second general causation theory, Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *197; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *87. 164   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *132; aff ’d, 89 Fed Cl 158, 182–83 (2009); aff ’d 617 F 3d 1328, 1338 (Fed Cir 2010). 165  ibid. 166   ibid, citing §300aa-13(a)(I)(A). 160 161

Causation Problems in Product Liability  173 symptoms.167 The petitioners also failed to satisfy the third element of Althen, viz the need to show a ‘proximate temporal relationship between vaccination and injury’.168 They were unable to establish that the first symptom of autism and/or the first symptoms of the chronic gastrointestinal problems occurred within a time-frame consistent with causation by the MMR vaccination in question.169 iv.  Looking Beyond the Epidemiology: the Overall Evidence A strength of these Omnibus Autism Proceeding test cases is that in determining whether the petitioners have demonstrated causation by a preponderance of evidence, the special masters have looked beyond the epidemiologic evidence to determine whether the overall evidence – ie medical opinion and circumstantial evidence and other evidence considered as a whole – tipped the balance even slightly in favour of a causation showing.170 Ultimately, in each case, the overall weight of the evidence was overwhelmingly contrary to the petitioners’ causation theories. In respect of general causation, the evidence advanced by the petitioners had fallen far short of demonstrating a causal link.171 v.  On the Side of Science Thus we can conclude that in these important test cases, the special masters have come down clearly on the side of science, and in doing so have considered the evidence overall. Indeed, one master, Special Master George Hastings in Cedillo severely criticised those physicians who are supporting a link between MMR and autism. He stated that: ‘Unfortunately, the Cedillos have been misled by physicians who are guilty . . . of gross medical misjudgment’.172 All of the special masters concluded that the petitioners had failed to demonstrate that the vaccinations played any role in causing autism.

IV. The Liberal French Approach to Medicinal Product Liability Cases: Presumptions of Causation

In France, a much more liberal approach to causation appeared to be established by the Cour de Cassation for medicinal product liability cases. On 22 May 2008, the Cour de 167   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *132; aff ’d, 89 Fed Cl 158, 182–83 (2009); and, further, in respect of the second general causation theory, King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *88. 168   Althen v Sec’y of Health & Human Services 418 F 3d 1274, 1278 (Fed Cir 2005). 169   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *132–33, aff ’d, 89 Fed Cl 158, 182–83 (2009); Pafford v Sec’y of Health & Human Services 451 F 3d 1352, 1358 (Fed Cir 2006) (need for evidence demonstrating petitioner’s injury within medically accepted time-frame). 170   In determining if a petitioner is entitled to compensation, the special master is not bound by any diagnosis, conclusion, judgment, test result, report, or summary, and in evaluating the weight to be afforded to such matters, ‘shall consider the entire record . . .’: 42 USC §300aa-13(b)(1) (emphasis added). 171   Cedillo v Sec’y of Health & Human Services, 2009 WL 331968 (Fed Cl 12 February 2009) *134–35; and, further, in respect of the second general causal theory, King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *91. 172   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 2009) *135; and, further, Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *201 (‘Unfortunately the Dwyers (and uncounted other parents of children with ASD) have relied upon practitioners who peddled hope, not opinions grounded in science and medicine’).

174  Richard Goldberg Cassation acknowledged in a series of five cases concerning hepatitis B which it was alleged to have caused neurological disorders,173 and one case concerning two medications that were alleged to have caused Lyell’s syndrome,174 that causal link could be established by the presence of ‘serious, precise and concordant’ presumptions of causation. Such presumptions had to be supported by specific causation-related data submitted by each specific claimant on a case-by-case basis relating to the claimant’s medical history, but not through generalised statistical or probabilistic studies. As a result, despite the absence of any scientific and statistical data showing a causal link between hepatitis B vaccine and multiple sclerosis or other neurological illnesses, the Cour de Cassation quashed two175 out of five judgments concerning the hepatitis C vaccine which had previously dismissed claims for compensation. The decisions were quashed on the grounds that the appellate courts had ruled without investigating specific causation-related data submitted by each claimant and whether this constituted serious, precise and concordant presumptions of causation. In one appellate judgment, the court had relied on general studies and statistics to determine that there was no causal link between hepatitis B and multiple sclerosis. Accordingly, a claim against a pharmaceutical producer could not be rejected on the sole basis of the absence of any scientific and statistical data showing a causal link between a medicinal product and an illness. This decision to allow the claimants to prove a causal link on the basis of serious, precise and concordant presumptions of causation was confirmed by the Cour de Cassation by a judgment of 25 June 2009.176 This led to considerable concern from the pharmaceutical industry since the existence of a causal link could no longer be excluded on the basis of an absence of general statistical evidence of a causal link between drug and damage. The industry became worried that this position had opened the door to compensation for the alleged side effects of medicinal products generally, especially when the Cour de Cassation’s position conflicted with legal certainty and fairness in the absence of conclusive epidemiology. It also appeared unclear in what circumstances trial judges would be able to demonstrate the necessary presumption of causation, in cases where there was an absence of scientific evidence of general causation.177 The opportunity to confirm what type of facts could potentially give rise to serious, precise and concordant presumptions quickly arose with the judgment of the Cour de Cassation of 9 July 2009.178 In an extremely controversial judgment, the court went beyond its previous decisions of 22 May 2008 and 25 June 2009, and upheld a judgment by the Court of Appeal of Lyon, granting a patient’s claim against the manufacturer of the Hepatitis B vaccine, by finding that causation had been proven even in the absence of general causation, but where such a causal link could not be excluded. The Court of Appeal of Lyon had utilised two factual criteria to establish a presumption of a causal link between the vaccination and the development of multiple sclerosis, viz (1), a temporal proximity 173   Appeals to the Cour de Cassation, n° M 06-18.848 n° V 06-10.967, n° B 06-14.952, n° T 05-10.593, and n° N 05-20.317, 22 May 2008 (I am grateful to Cécile Derycke and Agnès Roman-Amat, Lovells, Paris, for the provision of copies of several of the French cases utilised in this chapter). 174   Appeal to the Cour de Cassation, n° R 07-17.200, 22 May 2008. 175   n° V 06-10.967 and n° N 05-20.317. 176   Cour de Cassation Civ 1, n° 08-12781, 25 June 2009. 177   C Derycke and A Roman-Amat, ‘Law against science: French Civil Supreme Court opens the door to compensation in product liability cases involving the hepatitis B vaccine’ (June 2008) Lovells’ European Product Liability Review, 14, 17–18 178   Civ 1, n° 08-11.073, 9 July 2009; P Brun and O Gout, Recueil, Dalloz 2010, 49, 2.

Causation Problems in Product Liability  175 between the vaccine injection and the development of the illness; and (2) the absence of other personal risk factors. The Cour de Cassation held that while scientific evidence had failed to establish a statistically significant increase in relative risk of multiple sclerosis following vaccination against Hepatitis B, but nevertheless could not exclude such a possible link, and there existed proximity between the injection and the development of the disease and an absence of other individual risk factors, such facts could constitute serious precise and concordant presumptions. From this, a causal link would be inferred between the vaccine and the damage.179 It is strongly suspected that the purpose of the Cour de Cassation’s judgment was to adopt the same position as the Counseil d’État in actions brought by individuals subject to compulsory vaccination to prevent Hepatitis B against the State or other employers.180 However, certain French Courts of Appeal have resisted this controversial approach adopted by the Cour de Cassation and have continued to refuse to hold manufacturers liable where there is an absence of scientific evidence of general causation.181 In particular, the Paris Court of Appeal stressed the need to base the decisions on specific personal data of the claimant, whilst at the same time reaffirming the absence of any scientific consensus between hepatitis B vaccine and neurological disorders, the fact that that the aetiology is unknown and that multiple sclerosis can be caused by various genetic factors.182 Moreover, in its judgment of 19 June 2009, the Paris Court of Appeal held that temporal proximity and the absence of personal risk factors did not constitute serious, precise and concurrent presumptions.183 This position has now been upheld by the Cour de Cassation in its two most recent decisions of 24 September 2009, and 25 November 2010.184 The Cour de Cassation now appears to be retreating somewhat from its position on 9 July 2009. In upholding the decision of the Court of Appeal of Paris of 19 June 2009, it has refused to establish causation in the absence of scientific consensus in favour of a causal link between the Hepatitis B vaccine and multiple sclerosis, despite the existence of proximity between the vaccine injection and the disease (two weeks) and an absence of other individual risk factors. Such evidence, without a scientific consensus in favour of causation, was insufficient to constitute serious, precise and concordant presumptions.185 The inconsistency of these decisions has been unhelpful in generating uncertainty for both claimant and defendant. However, it is submitted that, without scientific evidence of   Civ 1, n° 08-11.073, 9 July 2009.   CE, n° 267635, 9 March 2007. See, further, D Fairgrieve and F G’sell-Macrez, Ch 6, p 127; C Derycke and A Roman-Amat, ‘The adverse side effects of the French Supreme Court’s judgments in relation to the Hepatitis B vaccine’ (September 2009) Lovells’ European Product Liability Review, 13, 14. 181   Cour d’Appel de Paris, n° 04/19067, n°04/19068, 9 January 2009, Cour d’Appel de Paris, n° 06/13741, 19 June 2009 (upheld by Cour de Cassation: Civ. 1, n° 09-16.556, 25 November 2010). See also, Cour d’Appel de Bourges, 22 January 2009, n° 07/01489. 182   Cour d’Appel de Paris, 9 January 2009, RG n°08/1407, 6; RG n° 04/19067, 8. 183   Cour d’Appel de Paris, n° 06/13741, 19 June 2009 (upheld by Cour de Cassation: Civ 1, n° 09-16.556, 25 November 2010). 184   Civ 1, n° 08-16.097, 24 September 2009 (evidence of claimant (seven month period between vaccination and outbreak of multiple sclerosis, and fact that claimant presented no personal or family history in relation to multiple sclerosis), did not constitute serious, precise and concordant presumptions of a causal link); Civ 1, n° 09-16.556, 25 November 2010. See also the decision of the Cour de Cassation prior to its judgment of 9 July 2009: Civ 1, n° Z 07-16.449, 22 January 2009, which confirmed that appellate courts could dismiss patient claims concerning the hepatitis B vaccine, provided that they assess with care not only general causation, but also specific causation, and are able to conclude that the factual evidence submitted to them does not amount to serious, precise and concordant presumptions of causation; C Derycke and A Roman-Amat, ‘The judge and science: a new episode in the hepatitis B vaccine saga’ (March 2009) Lovells’ European Product Liability Review, Issue 34, 21, 22. 185   Civ 1, n° 09-16.556, 25 November 2010. 179 180

176  Richard Goldberg general causation, there should be no question of overcoming the burden of proof of causation in such cases. The Cour de Cassation would be wise to study the factors required to overcome that burden as established in the National Vaccine Injury Compensation Program.186 While the current decisions of the French Courts appear to accept prongs two and three of the Althen187 test, viz (2) a logical sequence of cause and effect showing that the vaccination was the reason for the injury; and (3) a showing of a proximate temporal relationship between the vaccination and the injury, the uncertainty seems to stem from whether there should be an acceptance of prong (1), ie a medical theory causally connecting the vaccination and the injury, which is a general causation requirement that the type of vaccine can cause the type of injury in question.188 Were the French courts to adopt an Althen type approach, which gives primacy to the general causation issue, this would help create more consistency in its decisions, in line with the Cour de Cassation’s objective laid down in its Annual Report for 2008 to harmonise case law on the Hepatitis B vaccine.189

V. Conclusions

We can make the following observations about recent cases involving the role of scientific evidence in assessing causation in product liability in the UK, US and France. There remain considerable difficulties in reconciling the standards of proof in law and science. Despite the trenchant criticisms of the doubling of risk theory in the US, its support appears to be gaining ground in the UK. However, the majority of the Supreme Court in Sienkiewicz appears to be sceptical of introducing a threshold for the use of epidemiological evidence and remain of the view that epidemiological evidence can be useful but must be viewed with caution; without further non-statistical evidence there is reluctance for courts to proceed to find the existence of a causal relationship. A danger otherwise is that the Legal Services Commission, in assessing whether to fund multi-party product liability litigation, may regard this doubling of risk theory as the sole basis on which to allow or prevent cases from going forward to trial, even in cases where epidemiological evidence is lacking. This could potentially prejudice access to justice in future cases. It is submitted that in such cases where there is a dearth of epidemiological evidence, courts and, for that matter, funding bodies, should learn from the US experience and should avoid insisting on epidemiological studies which have a relative risk of greater than two, allowing all evidence which falls ‘within a zone of reasonable scientific disagreement’190 to be considered.  See Althen v Sec’y of Health & Human Services 418 F 3d 1274, 1278 (Fed Cir 2005).  ibid. 188   Cedillo v Sec’y of Health & Human Services 2009 WL 331968 (Fed Cl 12 February 12009) *131, affirming Pafford v Sec’y of Health & Human Services 451 F 3d 1352, 1355–56 (Fed Cir 2006); and, further, in respect of the second general causation theory, Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010) *197; King v Sec’y of Health & Human Services 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010) *87. 189  See: www.courdecassation.fr/publications_cour_26/rapport_annuel_36/rapport_2008_2903/quatrieme_ partie_jurisprudence_cour_2922/responsabilite_civile_assurances_2953/droit_responsabilite_2954/produits_ defectueux_12240.html 190   C Cranor, Toxic Torts: Science, Law, and the Possibility of Justice (n 20) 366, 289–90, 335. Courts should not exclude causal opinions based on non-epidemiological evidence where a body of epidemiological data does not exist: DL Faigman, DH Kaye, MJ Saks, J Sanders, ‘How Good is Good Enough?: Expert Evidence Under Daubert and Kumho’ (2000) 50 Case Western Reserve Law Review 645, 663. 186 187

Causation Problems in Product Liability  177 While it seems the UK is becoming more receptive to the need for epidemiologists to come to court to speak to their evidence and for it to be taught to the fact-finder, there has recently developed an overly cautious approach to the use of epidemiological evidence, particularly in Scots law. It has been submitted that the concern with such a cautious approach to epidemiological evidence is that it may make it harder to even discern that there is any reconciliation of the legal standard of proof on a balance of probabilities with the scientific standard of statistical significance. Moreover, there is also concern that the mere placing of an obligation on the pursuer to teach the epidemiology to the court suggests that the court can act passively in this process. This is surely an unhelpful approach in cases such as McTear, where there is a clear societal function of a judge to resolve these matters to the satisfaction of both parties. Had the pursuer explained the epidemiologic evidence properly, and had Lord Nimmo Smith been more receptive to evidence of relative risk, taken a more active role in forcing the pursuers to present their evidence effectively, and given adequate weight to the generally accepted scientific evidence since 1950 that cigarette smoking can cause lung cancer, it would seem that general causation should have been established in this case. There also remains a lack of clarity to the extent to which generalised epidemiological evidence can be useful in determining individual causation. I have suggested previously that in complex pharmaceutical cases, where non-numerical solutions have proved elusive, statistical refining processes, including Bayes’ Theorem using specific factors embodied in the likelihood ratio are a way forward in this area.191 It is arguable that epidemiological evidence can be refined to draw conclusions about the cause of disease in an individual using specific risk factors such as were present in Mr McTear’s case. Such probabilities in the likelihood ratio could be decomposed into factors in the light of specific case information, in respect of patient history, timing, characteristics of the adverse event, the response to dechallenge, ie discontinuing the drug, and the response to rechallenge, ie restarting the drug.192 Indeed, the Interdisciplinary Vaccine Safety Committee of the Institute of Medicine adopted such an ‘informal Bayesian approach’ to assessing case reports in its review of scientific and medical literature on specific risks to children associated with vaccines.193 It is my view that all of these courts could utilise this information to refine these generalised statistics to produce as accurate a posterior probability as possible, especially in the pharmaceutical field. This, however, would need epidemiologists and physicians to assist the court in this exercise, and clearly without existing prior probabilities the process would be limited. Nevertheless, in the absence of epidemiologic evidence, the Institute of Medicine in its report on Adverse Events Associated with Childhood Vaccines has supported the reliance upon individual case reports and case series, providing the nature and timing of the adverse event following vaccination and the absence of likely alternative aetiologic candidates were such that a reasonable certainty of causality could be inferred from one or more case reports.194 And as Carl Cranor has astutely reported, while many US judges reject case reports in traditional tort cases as inadmissible on the Daubert criteria, the special masters under the National Vaccine Injury Compensation Program, while following Daubert under 191   R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (n 3) 38–43. 192   ibid 40. 193   KR Stratton, CJ Howe and RB Johnston, Institute of Medicine, Adverse Events Associated with Childhood Vaccines: Evidence bearing on Causality (National Academy Press, 1994) 3, 25. 194   ibid 31–33.

178  Richard Goldberg their less formal procedures for introducing evidence, are much more receptive to the use of case reports supporting causation than most judges in traditional torts cases.195 The US experience in the test cases in autism is in many ways a paradigm of how to address such controversial issues. Unburdened by the emotions of a jury and the usual restrictions imposed by the Federal Rules of Evidence, a single trier of fact has been able to look at all the available evidence and come to a reasoned decision. In these autism test cases, issues of general and specific causation have been addressed and factors personal to the individual children have been taken into account. While the Daubert factors have been utilised, they have not prevented evidence being made admissible in these proceedings through an overly strenuous evidentiary threshold. They have been relevant to the assessment of weight at the adjudication stage, which has allowed the evidence as a whole to ventilate in the proceedings. It suggests that this more flexible approach to scientific evid­ ence, albeit with high standards at the adjudication stage, is welcome and may counter some of the criticisms196 of Daubert that it has in some cases hindered the search for justice in product liability law. No doubt the most radical approach would be to build on the template of the National Childhood Vaccine injury programme, and extend it to one involving all prescription drugs generally. This is unlikely to take place in the short term. But what should be possible is a greater flexibility in the use of the gatekeeping role for scientific evidence in these types of cases in the future. In France, a liberal approach to causation appears to be established by the Cour de Cassation for medicinal product liability cases. The Cour de Cassation has acknowledged that causal link can be established by the presence of serious, precise and concordant presumptions of causation. However, the inconsistency of the decisions has been unhelpful in generating uncertainty for both claimant and defendant. It is submitted that, without scientific evidence of general causation, there should be no question of overcoming the burden of proof of causation in such cases. The Cour de Cassation would be wise to study the factors required to overcome that burden as established in the National Vaccine Injury Compensation Program, as adumbrated in the Althen case, and utilised by the OAP test cases, which, it is submitted, generate more clarity and consistency in approach.

  C Cranor, Toxic Torts: Science, Law, and the Possibility of Justice (n 20) 256–59.   See eg, RL Cupp Jr, ‘Believing in Products Liability: Reflections on Daubert, Doctrinal Evolution and David Owen’s Products Liability Law’ (2006) 40 University of California Davis Law Review 511, 528–30; and, generally, C Cranor, Toxic Torts: Science, Law, and the Possibility of Justice (n 20). 195 196

9 Biostatistics and Causation in Medicinal Product Liability Suits PETER FELDSCHREIBER, LEIGH-ANN MULCAHY AND SIMON DAY

I. Introduction

This chapter discusses the evaluation and attribution of causality of injury from biologically active agents with particular reference to therapeutic medicinal products. We will review the biostatistical and the epidemiological principles which are applied to the analysis of association and causation of such medical events and the degrees of confidence they can give us in drawing conclusions on the safety and efficacy of medicines and other potentially injurious exposures. We will also identify the potential for divergence of approach in the evaluation and attribution of causality of injury between scientists, regulators and the courts. The title of this chapter implies that there are hard and fast rules by which medical science can use biostatistical and epidemiological data to draw definitive conclusions as to how medicines can and do cause harm and that these conclusions can be used to support rigorous and robust judgements on the therapeutic use or withdrawal from the marketplace. However, this is an oversimplification of the true position which we will illustrate by examining some of the high profile drug-induced disasters of the past few decades where our scientific analyses and explanations of causation of serious adverse drug reactions events have been wrong. The attribution of causality in medical science can be said to be both an art and science.

II. Use of Epidemiology and Biostatistics in the Assessment of Risk Benefit

A. Thalidomide The use of biostatistics and epidemiology in the regulation of medicines is in fact relatively recent. In 1951, Dr WG McBride, a general practitioner obstetrician in New South Wales, reported his anecdotal observations of the appearance of phocomelia in his practice. Although congenital abnormalities had been reported to be present in approximately

180  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day 1.5 per cent of babies, he had personally observed in his practice that the incidence of multiple severe abnormalities delivered of women who had been given Distival (thalidomide) as an antiemetic during pregnancy was almost 20 per cent. Rather than explicitly attributing causality, Dr McBride simply invited any other readers to report similar anecdotes!1 The manufacturers also reported two ‘overseas cases’ of potential harmful effects on the foetus and the product was withdrawn in the UK by the manufacturer. The epidemic of some 10,000 cases of phocomelia ended as abruptly as it had started some nine months later2. It is noteworthy that even in the absence of robust epidemiology and clinical trial data to support the evaluation of risk: benefit, prompt action was taken by governments in Europe and elsewhere and the manufacturer to withdraw a majorly defective medicine. Thalidomide was the iconic drug disaster of the twentieth century and its withdrawal led directly to the regulatory system we now enjoy. Prior to thalidomide, there were no statutory requirements for implementation of the biostatistical/epidemiologic methodology that we now employ. In particular, there had been no requirement for pre-clinical testing of new medicines to determine, for example, their teratogenic potential. Similarly, there was no requirement for clinical and post-marketing trials which, together with adverse event recording, could form the basis for epidemiologic databases which could allow interrogation as to association and causal relationships between medicines and adverse events. This was in spite of the fact that knowledge about the potential for epidemiology to examine these relationships had existed since the middle of the nineteenth century.

B.  The Introduction of Scientific Method Principles of modern day epidemiology had in fact been used to great effect in protecting public health in Victorian times. In particular the study described below contains all the fundamental principles of epidemiology that are used today in medical regulatory science. They are illustrated in a mid-nineteenth century study by Dr John Snow, a London physician. In 1849, Snow published on the transmission of cholera, a water-borne disease well known to the residents of Victorian London3. Cholera was endemic in London and frequent cyclical epidemics of the disease occurred. Snow investigated one such epidemic in 1853 and published his paper ‘On the mode of communication of cholera’ in 1855 predating by some 30 years Robert Koch’s postulates on the transmission of microbial disease. Cholera is primarily a disease of the gastrointestinal tract and Snow used his clinical observations of patients to test his theories of transmission. He concluded that cholera ‘was a local affectation of the mucosal membrane of the alimentary tract passed from one patient to another by swallowing’. Further, ‘as the disease grows in the community it was clear that the cholera poison must multiply itself by a kind of growth’. Thus, Snow was the first medical scientist to postulate self-propagating agents which spread from person to person through contaminated water and food. In his investigation of the 1853 cholera epidemic, Snow described three distinct epidemiological studies:   WG McBride, ‘Thalidomide and congenital abnormalities’ (1961) 278 The Lancet 1358.   RR Shah, ‘Thalidomide, drug safety and early drug regulation in the UK’ (2001) 20 Adverse Drug Reactions and Toxicological Reviews 199–255. 3   J Snow, ‘On the pathology and mode of communication of cholera’ (1849) 44 London Medical Gazette 745– 52, 923–29. 1 2

Causality in Medicinal Product Liability  181 •  an ecological study which compared mortality data by region; •  a cohort study comparing water quality and distribution by region; and •  a case control study comparing the water sources used by cholera cases and non-cases during the outbreak4. He found that during the 1853 outbreak the Southwark and Vauxhall Water Company provided water which had been contaminated downstream with sewage. His cohort study was able to calculate mortality rates by supplier by studying where water had travelled down parallel distribution pipes along streets, indiscriminately supplying water to both rich and poor households. The results showed water supplied by Southwark and Vauxhall caused 315 deaths per 10,000 houses compared to 37 deaths per 10,000 houses from water supplied by Lambeth Water Co and 59 deaths per 10,000 houses throughout the rest of London. Furthermore, his case control study in Golden Square showed the distribution of fatalities: 61 patients who died had used water from the Broad Street pump, six patients had definitely not drawn water from the pump and six were indeterminate. By contrast, there were no deaths in men drinking beer from a nearby brewery who had never obtained water from the pump! Snow also made another crucial observation from hearsay evidence. When investigating the case of a widow from Hampstead he stated: I was informed by this lady’s son that she had not been in the neighbourhood of Broad Street for many months. A cart went from Broad Street to the West End every day and it was the custom to take a large bottle from the pump in Broad Street, as she preferred it. The water was taken on Thursday the 31st of August and she drank of it in the evening and also on the Friday. She was seized with cholera on the evening of the latter day and died on Saturday . . . A niece who was on a visit to this lady also drank of the water; she returned to her residence in a high and healthy part of Islington, was attacked with cholera and died also. There was no cholera at the time either at West End or in the neighbourhood where the niece died.5

He ultimately concluded that cholera patients were more likely than non-cases to have consumed sewage contaminated water and he persuaded the parish to remove the handle of the Broad Street pump and the epidemic ceased! Snow’s research embodied virtually all of the scientific principles and methodologies of the state of the art of medical epidemiology.

C.  The Bradford Hill ‘Criteria’ In 1965, Bradford Hill6 described the criteria to be taken into account to determine the relationship between evidence of association and causality. These were: •  strength of association, ie the magnitude of the risk ratio; •  consistency of the epidemiological and clinical data showing similar findings in different populations in diverse studies;

4   W Winklestein, Jr ‘A New Perspective on John Snow’s Communicable Disease Theory’ (1995) 142 American Journal of Epidemiology (9 Suppl) S3-9. 5   J Snow, On the mode of Communication of Cholera, 2nd edn (London, John Churchill, 1855); WH Frost (ed), Snow on Cholera, (New York, the Commonwealth Fund, 1936). 6   AB Hill, ‘The Environment and Disease: Association or Causation’ (1965) 58 Proceedings of the Royal Society of Medicine 295–300.

182  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day •  specificity of the effect of exposure, ie that the causal factor should lead to only one disease; •  the temporality of the appearance of results of exposure, ie that exposure precedes the onset of disease; •  biological gradient of the increase in intensity, increase in level and duration of effects of exposure, ie a dose relationship; •  plausibility; (whether there is an association with known biologic facts about the pathophysiology of the disease or reaction. This latter criterion is of course dependent on the state of scientific knowledge at the time the data is being analysed and collated. It may be inadequate to explain associations that may in fact be causal and, conversely, may be unable to provide a realistic estimate of the probability of such an association not being causally related.) •  whether there is general coherence of theories and evidence of the natural history of the material disease and associated exposure; •  are there any objective experimental models to investigate the link between exposure, association and subsequent causality? •  is there any evidence or observations on analogous cases? (For example, is there an analogous scenario that implies similarities between things that are otherwise different such as different drugs causing birth defects? If thalidomide can, so might other drugs with similarities in structure and pharmacologic properties.) It is important to note, however, that Bradford Hill did not claim that satisfying all or even the majority of the criteria on a tick-box approach would automatically conclude inference of a causal relationship. He observed: ‘All scientific work is incomplete, whether it is observational or experimental’. Moreover, he specifically disclaimed a tick-box approach. He advised epidemiologists to avoid overemphasising statistical significance testing, given the observation that systematic error is often greater than random error.7 He wrote: ‘Epidemiology deals with the characteristics of human populations and therefore is more an observational than an experimental discipline’.8 In a particularly useful review of the Bradford Hill criteria, Morabio puts it thus: [W]hen proof is not available pragmatic epidemiologists simply acknowledge that there is as yet no alternative to the causal criteria logic that says in substance: before inferring causation it is imperative to check for illogicalities and rule out gross contradictions between what has been found and what we think we know.9

Epidemiological observational data are therefore necessary but not sufficient evidence to conclude causation. It can be indicative of the overall strength of evidence when concluding causation but a great deal of judgement concerning any conclusion of causality is required.

7  CV Philips and KJ Goodman, ‘The missed lessons of Sir Austin Bradford Hill’ (2004) 1 Epidemiologic Perspectives and Innovations 3. 8   AB Hill, ‘Observation and Experiment’ (1953) 248 New England Journal of Medicine 995–1001. 9   A Morabio, ‘On the origin of Hill’s Causal Criteria’ (1991) 2 Epidemiology 367–69.

Causality in Medicinal Product Liability  183

D.  Current Biostatistical and Epidemiological Principles Relevant to Causation We now turn to a review of some of the biostatistical and epidemiological concepts and principles that are used by scientists to evidence the strength of measurement of causal relationships.

i. Risk In epidemiological terms, risk is defined as the number of incident cases occurring during a period of observation, divided by the size of the population at risk. The denominator excludes people who already have the disease (by virtue of already having the disease, these people are no longer at risk of getting it) and others who are incapable of contracting the disease.10 Comparisons of risk can be expressed arithmetically as a simple ratio: per cent relative to baseline or as a per cent difference relative to baseline. The absolute comparison is the risk difference (RD) and this provides an absolute measure of difference between risk when exposed (R1) and risk when not exposed (R0) (RD = R1–R0). Similarly, the risk ratio provides a relative measure of association between exposure and disease (RR = R1/R0). The classic example is that of smoking and lung cancer. Suppose smokers have a lifetime lung cancer risk (R1) of 1 per 100 (1 per cent) while non-smokers have a lifetime lung cancer risk (R0) of 1 per 1000 (0.1 per cent). The risk difference provides an absolute measure of association between exposure and disease equal to 1 per cent – 0.1 per cent (0.9 per cent). On the other hand, the risk ratio is a relative measure of association between exposure (smoking) and disease (lung cancer) and is equal to 1 per cent divided by 0.1 per cent (10). These arithmetic methods can be used to infer the strength of an association between exposure and disease such that an estimate of causality can be inferred. At their most basic, risk ratios indicate the direction of an association between an exposure and disease. A risk ratio (RR) of 1 indicates no association between exposure and disease whereas a RR of greater than 1 indicates a positive association and a RR of less than 1 indicates a negative association (ie a protective effect). Ratio measures quantify the strength of an association, ie a RR is effectively a multiplier relative to baseline risk. In spite of being able to quantify the relative strength of the association between an exposure and disease, this does not automatically mean we can draw conclusions that a statistical association is necessarily causal and association must not be confused with causation. ii.  Error, Bias and Confounding A major issue in analysing pharmaco-epidemiologic data for the purposes of evaluating potential causal relationships is the incidence of random and systematic error in databases collated from epidemiologic research. For instance, when anybody calculates a risk they are seeking the absolutely correct value of that incidence in the population under study. Similarly, when calculating a risk ratio they are seeking an absolutely correct value which 10   BB Gerstman, Epidemiology Kept Simple: An Introduction to Traditional and Modern Epidemiology, 2nd edn (New York, Wiley-Liss, 2003).

184  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day describes the exposure−disease relation in the population being studied. This ‘absolutely correct’ value quantifies the relative effect of the exposure and could be called the risk ratio causal-effect parameter. Although this is an objective characteristic of the population being studied, it is not possible to calculate it precisely. Instead, the biostatistician will calculate an imperfect estimate which will be prone to both random and systematic error. As the size of a sample increases, its inherent random error decreases, but any systematic error may not. Systematic error is also called bias. For instance, measures can be distorted by adopting methods of selecting the population for study in such a way as make the sample atypical of the population to which inferences will be made. In studies of causal effects, systematic bias occurs when the association between exposure and disease differs between those who participate in the study and those who do not. Another important source of systematic error is the distortion between study exposure and disease brought about by the influence of extraneous factors. This is known as confounding. Confounders are associated with the exposure in the source population but they are independent factors for the disease. In addition, the confounder is not intermediate in the causal pathway between the exposure and disease and is not a consequence of the disease. iii.  Estimation and Significance Testing The statistical methods used to address random error in quantitative research are estimation and significance testing. Aside from point estimation (eg 25 per cent of the population smoke), estimation may be described in terms of an interval (eg between 20 per cent and 30 per cent of the population smoke). Such ‘interval estimation’ surrounds the point estimate with a margin of error, thus creating a ‘confidence interval’. The width of the confidence interval is a measure of the estimate’s precision. Wide intervals indicate low precision, and vice versa. Statistical testing is an inferential method used to determine how likely an observed characteristic of a sample has occurred given a set of assumptions about the population from which the sample was derived. These tests can be used in conjunction with confidence intervals. Significance testing starts with the adoption of an assumption about the population (the null hypothesis). This is a general statement of no association between the exposure variable and disease variable, or no difference in expected disease frequencies between exposed and unexposed individuals. Thus, the null hypothesis is given the benefit of the doubt until the data convince us otherwise. Scientists then use probability theory to derive a probability answer to the question ‘ought I to take any notice of that?’ The p value, when calculated at a fixed arbitrary level for biological purposes (normally set at 0.05) allows a determination of the level of confidence one should have in ascribing an observed association as being a causal relationship. When the p value falls below this fixed threshold the investigator can reject the null hypothesis; eg there is a less than 1 in 20 chance that the link between exposure and disease is a random association. A common misinterpretation of the p value is that small p values provide unassailable support for a causal relationship. This is not so. Without making certain further tests for causality, such as the application of all or some of the Bradford Hill criteria, p values cannot be used to indicate evidentiary support for a hypothesis. The p value only tells us about random sampling errors, it tells us nothing about possible systematic errors, (ie bias).

Causality in Medicinal Product Liability  185

E.  Fallibility of Scientific Approach We now turn to some further examples of drug-induced injury where epidemiologic and clinical trial/pharmacology data were not only available but had also been subject to robust scrutiny by the regulator. Despite this scrutiny, both manufacturers and regulators failed to make an appropriate evaluation of risk: benefit. i. Vioxx The first example is that of Vioxx (rofecoxib). Vioxx is a non-steroidal anti-inflammatory agent/drug (NSAID) which focuses its activity on the biochemical pathway leading to inflammation without affecting the stomach. Vioxx was an example of a new class of drugs – the ‘COX 2 inhibitors’. The drug had been marketed in the US and Europe since 2001. In September 2004, the manufacturer, Merck and Co, voluntarily withdrew the drug worldwide because evidence appeared from placebo-controlled clinical trials that suggested that the drug caused fatal heart attacks and stroke. It is instructive to examine this story in a little more detail. The cyclogenases are intracellular enzymes that inhibit the production of compounds that inhibit the production of chemical mediators responsible for pain, inflammation and platelet function. The terms cycloxygenase 1 and 2 refer to individual forms of the enzyme in the family of these biological catalysts. The COX 2 inhibitors such as Vioxx specifically inhibit cyclogenase 2, the enzyme involved in the synthesis of inflammatory mediators in, for example, arthritis. However, the specific inhibition of this pathway leaves the other cycloxygenase pathway, the COX 1 pathway unhindered. This ‘COX 1’ pathway is involved in tissue protection including protection of the gastric mucosa. One of the most important side effects of conventional non-steroidal anti-inflammatory drugs is damage to the gastric mucosa causing stomach ulceration. Vioxx appeared to avoid this problem. However, by allowing unhindered activity of COX 1 it may have allowed enhanced platelet aggregation which in turn might cause thrombosis particularly in the coronary circulation. Therefore drugs such as Vioxx which block cyclo-oxygenase 2 predictably increase the risk of thrombotic events such as myocardial infarction.11 Such potential was apparent from the pharmacology of these drugs, particularly their ability to inhibit prostacyclin biosynthesis (a biochemical mediator which relaxes smooth muscle and inhibits platelet aggregation). The early studies of Vioxx hinted at such problems and this in fact was what was observed about a year after the drug had been authorised by the US Food and Drug Administration (FDA). Why had this safety issue not been observed and evaluated before the drug was marketed? The answer, we believe, is in the way the pre-marketing pivotal trials were designed and conducted. The pivotal trial of clinical evidence of efficacy in Vioxx was known as ‘VIGOR’. This was a clinical study examining the efficacy of Vioxx in its primary indication, rheumatoid arthritis, compared to Naproxen, a very commonly used non-selective NSAID. The trial was large (8076 patients) and robustly statistically powered but there was no placebo arm. The trial found a fivefold increase in thrombotic events with Vioxx compared to Naproxen but also showed that Vioxx had 2 per cent gastro-intestinal events compared to 11   HP Rang, MM Dale, JM Ritter, RJ Flower, Pharmacology, 6th edn (Oxford, Elsevier Science 2008) 236; GA FitzGerald and C Patrono, ‘The coxibs, Selective Inhibitors of Cyclooxygenase 2’ (2001) 345 New England Journal of Medicine 433–42.

186  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day 4 per cent with Naproxen. So this trial did suggest an increased cardiovascular risk for Vioxx. The FDA did express concern in its communications with the manufacturer, but the company responded with the hypothesis that Naproxen was ‘cardioprotective’ and that the result simply highlighted the lack of an intrinsic cardioprotective effect for Vioxx. This explanation was accepted by the FDA12, as indeed it was by the European regulators. In September 2004, there became available results of the first placebo-controlled study of Vioxx, the APPROVE study. This trial in 2585 patients with benign colon polyps investigated the effect of treatment with Vioxx in the prevention of such polyps becoming malignant. Although the study demonstrated that Vioxx did prevent the malignant transformation of polyps, it caused approximately a twofold increase in the risk of cardio­ vascular thrombotic events and mortality. The company voluntarily withdrew the drug worldwide. This incident has lead to huge litigation in the US and during the course of court proceedings allegations have been made that during the submission of its application to FDA to approve its use in US, the company misrepresented the cardiovascular morbidity and mortality results of the pivotal clinical trials.13 ii.  Seroxat (Paroxetine) The second example of the difficulties the regulator may have in determining causation and risk: benefit of a new medicinal product is that of the anti-depressant Seroxat (paroxetine). During the clinical development of this drug the manufacturer (at the time SmithKlineBeecham, now GlaxoSmithKline) undertook a series of pivotal trials on the safety and efficacy of Seroxat’s anti-depressant activity. The manufacturer concluded from Phase 1, 2 and 3 studies a positive risk: benefit ratio for all age ranges, but in its submission for the marketing authorisation, the company failed to disclose data from studies in children.14 During post-marketing surveillance, there were a number of spontaneous anecdotal reports of children and adolescents prescribed Seroxat describing signs of akisthesia (extreme agitation) and suicidal ideation. Because it was difficult to distinguish between underlying depressive illness and drug-induced adverse events (ie there were potential confounding factors) the company and the regulators concluded that the epidemiological data were insufficient to demonstrate a safety signal strong enough to establish a causal relationship with Seroxat sufficient to withdraw the drug from the population at large. However, eventually sufficient evidence did accrue to justify contraindication in children and adolescents together with label warnings on withdrawal effects, suicidal feelings and akisthesia. The decision to impose these changes on the label had taken several years and it was sub­ sequently discovered that the manufacturer had failed to disclose studies of the drug’s effect in children. It has now been claimed that the company also misrepresented data in one of these studies.15

12   EJ Topol, ‘Failing the public health – Rofecoxib, Merck, and the FDA’ (2004) 351 New England Journal of Medicine 1707–09. 13  ibid. 14   L McGoey, and E Jackson, ‘Seroxat and the suppression of clinical trial data: regulatory failure and the convenience of legal ambiguity’ (2009) 35 Journal of Medical Ethics 107–12. 15   M Keller et al, ‘Efficacy of Paroxetine in the Treatment of Adolescent Major Depression: A randomised Clinical Trial’ (2001) 40 Journal of the American Academy of Child & Adolescent Psychiatry 762–72; J Jureidini et al, ‘Clinical trials and drug promotion’ (2008) 20 International Journal of Risk and Safety in Medicine 73–81.

Causality in Medicinal Product Liability  187 iii.  The TeGenero Incident Our final example of the difficulty posed in predicting potential causal relationships in new drug development is the TeGenero incident or Northwick Park clinical trial disaster16. Although not strictly an example of identifying the cause of a drug safety event, TGN 1412 is an example of the care needed in identifying safety signals and potential causes of druginduced injury as early as possible before a drug is given to man for the first time. TGN 1412 was a new monoclonal antibody, an example of advanced molecular biologic methods of drug development, designed to mitigate autoimmune and immunodeficiency diseases such as leukaemia and, at the other end of the scale, rheumatoid arthritis. When this drug was administered to six healthy volunteers they all developed catastrophic multiple organ failure because of the drug-induced exaggerated immunological effect known as ‘cytokine storm’. The ‘first in man’ study design (protocol and dosage regime) had been accepted by the regulators in both the UK and Germany. The UK authorities instituted an investigation of this episode and the way in which first in man trials were designed and performed. One of the key recommendations was on the way in which the dose for first in man studies was calculated from the available pre-clinical animal data. This directly questioned the scientific basis on which such extrapolations had previously been made.17 It subsequently transpired also that the original trial protocol had already referred to the possibility of cytokine storm and that there had been previously unpublished reports of similar adverse events with similar molecules. These examples of difficult and ultimately erroneous decisions regarding causation and risk: benefit of drug-induced adverse events, whilst not exhaustive, illustrate in our view a common theme. The statistical and epidemiologic methodology for attributing causation and measuring the strength of associations and causal relationships are highly useful and practical in achieving robust and practical conclusions. However, the methodology is only as good as the integrity, comprehensiveness and quality of the data analysed. A common theme running through the Vioxx and Seroxat examples discussed above is that in each case allegations have been made that the manufacturer may have failed to report material data to the regulator. If that is right, whether this was the result of a deliberate attempt to mislead or inadvertent errors matters not. The fact that the regulator had to work with erroneous and incomplete databases led to false assumptions as to causation of injury, adverse events and risk: benefit.

III. The Relevance of Biostatistics to the Determination of Causation by the Court

A. Introduction Having reviewed at a very basic level the principles and components of the methodologies the regulatory/medical scientist uses to evaluate causation when assessing the safety and efficacy of drugs, it is important to assess the relevance and practicality of such thinking to   M Goodyear, ‘Learning from the TGN1412 Trial’ (2006) 332 British Medical Journal 677–78.   www.dh.gov.uk/Consultations/LiveConsultations/fs/en 7 December 2006; Expert Scientific Group on Phase One Clinical Trials Final Report, 30 November 2006. 16 17

188  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day the way in which the courts determine causation, in particular in alleged cases of druginduced injury.

B.  General Principles In any product liability case in which an agent is alleged to have caused injury, including drug-induced injury claims, and to adopt the terminology used in McTear v Imperial Tobacco,18 a claimant must prove three matters: •  general causation (ie the agent can cause the injury in question); •  individual causation (ie the agent did cause the injury in question); and •  fault or proximate causation (ie that the exposure to the agent as a result of the defendant’s breach of duty (in the case of liability in negligence) as opposed to exposure from another source caused the injury in question).19 Biostatistics and epidemiology are particularly relevant to proof of causation. It is sometimes sought to prove general causation by reference to epidemiology or statistical evidence alone. In this situation, the ‘balance of probabilities’ test has been translated into a requirement to establish a relative risk (RR) of more than two of contracting the condition as a result of the exposure to the product in question.20 An RR of more than two would demonstrate a statistical probability that a particular agent or drug or exposure was capable of causing a condition. However, even if an RR of more than two is established, this is not necessarily enough to discharge the burden of proof of general causation on the balance of probabilities and the court may still need to have regard to additional evidence in order to be satisfied that a particular agent or drug is capable of causing the condition at issue.21 Many associations turn out in the end not to have been causative.22 A statistical probability, particularly a finely balanced one, may not be a sufficient basis to conclude that there a causal probability in fact. Further, relative risk takes no account of absolute risk. If there is an incidence of disease in an unexposed population of one in a million cases and in an exposed population of two in a million cases, the RR is two but the absolute risk is very low. Accordingly, even if statistically significant, the extent to which conclusions can be drawn as to general causation needs to be tested by reference to non-statistical evidence. Individual causation cannot be proved by epidemiology or statistical probability alone.23 Accordingly, non-statistical evidence (which it is suggested should take the form of an

  2005 2 SC 1, [6.1].   Note that fault causation would not require to be established under the strict liability provisions of the Consumer Protection Act 1987, s 2(1) of which merely requires that the person injured by a defective product must prove the damage, the defect, and the causal relationship between them: see CJ Miller and RS Goldberg, Product Liability, 2nd edn (Oxford, Oxford University Press, 2004) para 17.87–17.95. 20   XYZ v Schering Health Care Ltd & ors [2002] EWHC 1420, (2002) 70 BMLR 88 (QB). See also Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523, [26]. 21   McTear v Imperial Tobacco [2005] 2 SC 1, 6.158. See also Loveday v Renton [1990] 1 Med LR 117 (where the court considered the following factors to be relevant in addition to epidemiological evidence: a distinct and specific clinical syndrome, a specific pathology, temporal association, plausible mechanisms and animal experimentation). 22   Hope and Reay v British Nuclear Fuels [1994] 5 Med LR 1, 23. 23   Hotson v East Berkshire Area Health Authority [1987] AC 750, 769 and 789, McTear v Imperial Tobacco [2005] 2 SC 1, [6.28] and Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523, [95–96], [153], [156], [170–71], [191], [205], [216] and [218]. 18 19

Causality in Medicinal Product Liability  189 application of the applicable Bradford Hill criteria24) will be required to prove individual causation. In the recent case of Sienkiewicz v Greif (UK) Ltd25 the Supreme Court made a series of obiter observations about the use of epidemiological and statistical evidence to prove causation which are relevant to medicinal product liability suits. They expressed the view that a statement by Smith LJ in the Court of Appeal26 that a claimant can demonstrate causation by showing that the tortious exposure has at least doubled the risk arising from the nontortious exposure was wrong and should not be followed.27 Lord Phillips considered that an RR of just over two was a tenuous basis for concluding that the statistical probable cause of a disease was also the probable biological cause or cause in fact.28 This perhaps goes no further than disapproving an approach which suggested it might be possible to prove individual causation on the basis of statistical evidence alone. However, the possibility that it might ever be possible to prove causation on statistical evidence alone was not completely ruled out.29 It was recognised that epidemiological and statistical evidence may form an important element in proof of causation (at least outside of the mesothelioma context).30 However, apart from stressing the need for additional non-statistical evidence in order to find a causal relationship,31 the Justices emphasised the need to ensure the adequacy and reliability of any statistical evidence relied on to show the statistical probability of the cause or causes of disease.32 Lord Kerr warned against the real danger that epidemiological evidence will carry ‘a false air of authority’.33

C.  Differences Between the Medical Science and Legal Approaches We now highlight some identifiable differences between the approach which is taken by medical science and law in the attribution of causality. i.  Prospective Versus Retrospective Approaches The intensive analyses of epidemiological and prospective clinical trial data are used by healthcare professionals and regulators in the making of practical policy-based decisions and judgements in protecting public health. The attribution of causality in this context is generally an exercise in prospective forecasting of the magnitude of the safety issues and risks inherent in allowing the widespread use of medicine in the population at large. Causality is a key component of the decision on the risk: benefit ratio of a drug and the ongoing monitoring of risk: benefit depends on ongoing re-evaluation of pharmaco-epidemiological data and pharmacological hypotheses of the mechanisms of potential drug-induced injury. By contrast, the attribution of causation in personal injury, clinical negligence and product   As in Hope and Reay v British Nuclear Fuels [1994] 5 Med LR 1, 13 and 49–53.   [2011] UKSC 10, [2011] 2 WLR 523. 26   Sienkiewicz v Greif (UK) Ltd [2009] EWCA Civ 1159, [2010] QB 370, [23] per Smith LJ. 27   [2011] UKSC 10, [2011] 2 WLR 523, [162] per Lord Rodger. 28   ibid [83]. 29   See ibid, Lord Mance at [192] and Lord Dyson at [221–22]. 30   ibid [163], [172] and [191]. 31   See n 23 above. 32   See eg, [2011] UKSC 10, [91] [98] per Lord Phillips and [164] per Lord Rodger. 33   ibid [206]. 24 25

190  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day liability legal actions require a retrospective assessment of clinical and biological data by the court in deciding whether an association predicted by the scientific body of epidemiological and pharmacology data was sufficiently strong and robust to satisfy the applicable legal tests for causation. In Sienkiewicz, Baroness Hale commented on the prospective nature of risk contrasting it with the retrospective nature of causation: ‘Risk is a forward looking concept – what are the chances that I will get a particular disease in the future? Causation usually looks backwards – what is the possible cause of the disease which I now have?’34 The difference of approach means that it cannot be assumed that because regulatory action has been taken, causation can be proved. A regulator will make a decision about whether to take action on the basis of a far narrower knowledge base than is usually possible at the retrospective analysis stage and by employing (in practice if not necessarily consciously) a precautionary approach. The regulatory action may turn out to have been unnecessary if in fact there was not a sufficient causative link between a particular medicine and the alleged drug-induced injury. A good example was the regulatory action taken by the Committee on the Safety of Medicines in respect of the combined oral contraceptive pill where a civil action followed but failed to establish that the third generation of contraceptive pills created a significantly greater risk of thromboembolism than the second generation and therefore failed to even begin to establish causation to the required standard.35 ii.  Tests for Causation – Differences in Level of Proof Required Another key difference between the approaches of the scientist and the lawyer is that the medical scientist/regulator is concerned with the attribution of causality primarily as a component of the risk: benefit decision. It is sometimes assumed by lawyers that scientists treat an association as causal if it is 95 per cent certain. However, this is a misinterpretation of the p-value which, as explained above, is merely the level of statistical significance often used to exclude the possibility that when something occurs in a cohort of cases studied it does so by chance rather than by reference to the issue under investigation. There is no generally accepted scientific standard of proof for causality.36 What is clear, however, is that the standard is not one of marginal probability. In Sienkiewicz, Lord Phillips commented that, ‘When a scientific expert gives an opinion on causation, he is likely to do so in terms of certainty or uncertainty, rather than probability’.37 He quoted Lord Prosser’s observations in Dingley v Chief Constable, Strathclyde Police:38 Whether one uses the word ‘scientific’ or not, no hypothesis or proposition would be seen as ‘proved’ or ‘established’ by anyone with any form of medical expertise merely upon the basis that he had come to regard it as probably sound. (Indeed, I think even the word ‘probable’ would be reserved for situations where the likelihood is thought to be much more than marginal.) And even if, in relation to any possible proposition or hypothesis, such an expert even troubled to notice that he had come to the point of regarding it as not merely probable but on balance ‘probable’, then I think he would regard that point as one from which he must set off on further inquiry, and by no means as being (as it is in the courts) a point of arrival. Mere marginal probability will not much interest him. But it must satisfy a court.   Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10; [2011] 2 WLR 523, [170].   XYZ v Schering Health Care Ltd & ors [2002] EWHC 1420, (2002) 70 BMLR 88 (QB). 36   Loveday v Renton & anr [1990] 1 Med LR 117, 124 per Stuart-Smith LJ. 37   [2011] UKSC 10, [2011] 2 WLR 523, [9]. 38   1998 SC 548, 603. 34 35

Causality in Medicinal Product Liability  191 For the medical scientist as regulator, absolute determination of causality is not required; simply identifying the strength of an association in the direction of ‘plausible causality’ is sufficient for judgements to be made on therapeutic use for individual patients and the public at large. Plausible causality sufficient to ground regulatory action may be determined at a level of mere possibility rather than high probability. In the legal context, the courts are seeking to determine causation but again absolute determination (or proof) is not required. The traditional test for causation is the balance of probabilities approach or ‘but for’ test. Once it has been established that a causative link is more likely than not (ie more than 50 per cent likely), the defendant is deemed in law to have caused the injury. Conversely, if this level of probability is not established, the claimant’s claim fails in toto even though a causative link may be a strong possibility. There is therefore a degree of (inevitable) arbitrariness in the courts’ historic approach to attribution of causality.39 iii.  Divergent Approaches to Risk as a Test of Causation There is arguably an increasing difference in the relevance of ‘risk’ and its significance/ relevance in attributing causality between science and law. As stated above, scientists define ‘risk’ as the incidence of disease in the population exposed to injurious agents or pathogens. It is a quantifiable parameter which may suggest an association between exposure and disease. In itself, it is not indicative of causation. The strength and direction of the association may be a measure of the probability of a causal relationship. The quantification of this probability, ie the calculation of relative risk and the application of confidence intervals to determine the robustness of the calculated estimates, is a good starting point for the evaluation of causation. As discussed above, the full review of causation is facilitated by the application of the Bradford Hill criteria, subject to the important caveat that the criteria were never meant to be used as a checklist, but rather to provide the framework for defining the circumstances in which causation can be inferred from non-experimental observation. In other words, risk does not equate to cause or even a potential cause. It is at best evidence of a possible association. In the legal arena, the courts have developed three different strands of principle relating to causation of injury in the product liability context. These are: 1. The orthodox balance of probabilities/‘but for’ approach taken in multiple agent cases (Wilsher v Essex Area Health Authority40) where the possible causes could have caused the injury either independently or in combination.41 If it cannot be proved that the tortious cause was the actual cause on the balance of probabilities, the claimant fails in full. If it can be so proved, the claimant succeeds in full. 2. The ‘material contribution to injury’ approach which is applicable in cumulative injury cases, such as asbestosis, where it is proved that the defendant’s negligence actually contributed to the injury in the sense of making it worse or more severe than it would otherwise have been (Bonnington Castings Ltd v Wardlaw42). In this situation, the defendant is liable to the extent of his contribution to the injury.43  See Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523, [26], per Lord Phillips.   [1988] AC 1074, HL. 41  See Wilsher v Essex Area Health Authority [1987] 1 QB 730, CA per Glidewell LJ, 775G–776A and Lord Hoffmann in Gregg v Scott [2005] 2 AC 176, [77]. 42   [1956] AC 613. 43   Holtby v Brigham & Cowan (Hull) Ltd [2000] 3 All ER 421; Performance Cars Ltd v Abraham [1962] 1 QB 33; Allen v British Rail Engineering Ltd [2001] PIQR Q101; Thompson v Smiths Shiprepairers (North Shields) Ltd [1984] 1 All ER 881. 39 40

192  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day 3. The ‘material contribution to the risk of injury’ approach (otherwise known as ‘the material increase in risk’ approach) which is taken in an exceptional category of cases where it is not possible to prove that the defendant actually caused or contributed to the injury but only to prove that he materially increased the risk of suffering the injury (McGhee v National Coal Board;44 Fairchild v Glenhaven Funeral Services Ltd45). In relation to mesothelioma, the defendant is jointly and severally liable for all of the loss,46 subject to apportionment as between tortfeasors. In relation to other conditions or injuries, he is liable only to the extent of his contribution to the risk.47 The cases of McGhee v National Coal Board48 and Fairchild v Glenhaven Funeral Services Ltd49 support a departure from the need for claimants to prove causation of injury or material contribution to an injury (ie contributing a component element of the injury or making it worse) by substituting, in circumstances that are said to be ‘exceptional’, a test of material contribution to the risk of injury. In effect, a ‘cause’ is treated by lawyers as an actual cause of the whole or part of an injury (assessed as a matter of probability) and a ‘risk’ is treated as a possible or potential cause. The material increase in risk principle in the Fairchild exception to the ‘but for’ test of causation has only been applied to date at the stage of determining fault causation, when general and individual causation have already been established on the balance of probabil­ ities. For example, in McGhee, it had already been proved that brick dust was capable of causing dermatitis and that it had caused the pursuer’s dermatitis. The issue was whether it was the (innocent) brick dust to which the pursuer was exposed in the course of his working day, or the (guilty) brick dust to which he remained exposed whilst cycling home due to the negligent failure to provide showers, which was responsible for the dermatitis. In Fairchild, it was proved that asbestos was capable of causing mesothelioma (indeed it is the near-exclusive cause of pleural mesothelioma) and that it had in fact caused the claimants’ mesotheliomas. Accordingly, the Fairchild exception only obviates the claimant’s requirement to prove which of two or more sources (whether all tortious or some tortious and others not) was the origin of the agent which caused his injury and does not affect the need to prove the anterior issues of general and individual causation on the conventional balance of probabilities basis. Unfortunately, the existence of the Fairchild exception and the reference to being able to establish causation by reference to a material contribution to risk has sometimes been misunderstood in the legal context and, in our experience, has led claimants to claim in respect of a whole range of other injuries that they can come within the exception and succeed on the requirement to prove all the elements of causation if they can merely prove a statistical association, even one which falls far short of the requirement that the relative risk be two or above required as the starting point for demonstrating a risk on the balance of probabilities. This has recently been seen to be an unfruitful tactic in the Atomic Veterans Litigation, where the Court of Appeal accepted the submission of the Ministry of Defence that the cases were not ones to which the Fairchild exception could foreseeably be made to apply. It   [1973] 1 WLR 1.   [2002] UKHL 22, [2003] 1 AC 32.   S 3 Compensation Act 2006, effectively reversed Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572 but only in relation to mesothelioma. 47   Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572. 48   McGhee v National Coal Board [1973] 1 WLR 1. 49   Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32. 44 45 46

Causality in Medicinal Product Liability  193 reaffirmed that the exception would only apply where the two or more potential causes act either through the same agent or possibly through different agents which act on the body in the same way.50 In Sienkiewicz, Lord Brown sounded an express warning to claimants seeking to extend any relaxations to the traditional ‘but for’ test of causation that ‘[s]ave only for mesothelioma cases, claimants should henceforth expect little flexibility from the courts in their approach to causation’,51 warning further that ‘the law tampers with the ‘but for’ test for causation at its peril’.52 Since the issue of determining risk involves the use of epidemiology and biostatistics, it is ever more important that lawyers and judges understand the statistical methodologies, and what they can or cannot prove in terms of causation. The observations of the Supreme Court in Sienkiewicz on the limits of epidemiological and statistical evidence to prove causation in legal proceedings no doubt assist in this regard. iv.  Distinction Between Divisible/Indivisible Injuries A further distinction which has been brought into focus in the law by the Fairchild decision and which may influence which legal causation test is applicable is the distinction between ‘divisible’ and ‘indivisible’ injuries but it is a distinction which is not known to science and which we would predict is likely to prove increasingly problematic. The Court of Appeal in Fairchild v Glenhaven Funeral Services Ltd  53 noted that a ‘divisible’ disease was one where the injury can be divided into component parts or degrees of injury resulting from different causes whereas an ‘indivisible’ disease was ‘all or nothing’ in the sense that the attribution of different parts of the injury or condition to different causes is not possible. A distinction was made between the aetiology of asbestosis or silicosis (divisible) versus the aetiology of mesothelioma (indivisible).54 The reasoning in Fairchild and their Lordships’ express approval55 of Wilsher v Essex Area Health Authority56 indicates that the Fairchild exception is limited to single agent/cause57 cases of indivisible injury (where it cannot be proved which of multiple sources of the said agent was actually responsible for causing the ‘all or nothing’ injury) and not to multiple agent cases of indivisible injury (where it cannot be proved which of several possible agents or causes actually caused the injury). The problem from a scientific perspective is that the Fairchild exception is not expressly limited to the mesothelioma context, where almost all tumours are caused by exposure to asbestos, a complete carcinogen. It clearly also covers dermatitis as a result of exposure to brick dust since it is an application of the principle established in McGhee.58 What is unclear is which other conditions might be governed by the exception and the extent to which the precise aetiology of such conditions fit within the logic for the distinction. It is, however, 50   AB & ors v Ministry of Defence [2010] EWCA Civ 1317, [152]; Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572, [24]. 51   Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523, [187]. 52   ibid [186]; see also Baroness Hale, [167]. 53   Fairchild v Glenhaven Funeral Services Ltd [2001] EWCA Civ 1881, [2002] 1 WLR 1052, [21], [27]. 54   ibid [21]–[28]. In Sienkiewicz, the UK Supreme Court confirmed that mesothelioma is an indivisible disease: Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523, [15]. 55   [2002] UKHL 22, [2003] 1 AC 32 [22], [70], [118], [149], [170]. 56   [1988] 1 AC 1074. 57   Or possibly different agents which act on the body in the same way: Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572, [24] per Lord Hoffmann, applied in AB & ors v Ministry of Defence [2010] EWCA Civ 1317, [152]. 58   McGhee v National Coal Board [1973] 1 WLR 1.

194  Peter Feldschreiber, Leigh-Ann Mulcahy and Simon Day clear that they must involve the same agent or possibly different agents which act on the body in the same way.59 The distinction between divisible and indivisible injuries was perhaps unduly influenced by the fact that their Lordships in Fairchild understood that mesothelioma was or could be caused by a single fibre of asbestos. However, the ‘single fibre theory’ is now discredited60 and it is now accepted that an increased dose leads to increased risk of contracting the disease and that a critical mass of asbestos fibres is almost certainly needed. If in fact there is something more akin to a dose-response relationship for cases of indivisible injury such as mesothelioma, as is typical of cases of divisible injuries like silicosis and asbestosis, the theoretical distinction between divisible and indivisible injuries would appear to begin to break down. In any event, it is not yet known how the divisible/indivisible distinction will be applied in relation to the aetiology of other conditions and an obvious difficulty is that this is a distinction invented by lawyers and unknown to the scientific experts who will be asked to opine on this for the purpose of litigation.

IV. Conclusion

This chapter has reviewed the biostatistical and epidemiological methodologies adopted by medical/regulatory scientists to evaluate risk and assign causality for the mainly prospective purpose of evaluating risk: benefit and compared this approach with that taken by lawyers. It must be emphasised that such an approach is by no means foolproof and that the methodologies are only as good as the integrity and quality of the data being analysed. At the end of the day the attribution of causality by the medical scientist requires both quantitative calculation and qualitative judgement. In summary however, there are some similarities between the approach taken by scientists/regulators and lawyers in using biostatistics to determine causation, and in particular in neither situation is absolute proof of causation required. However, there are more divergences of approach. Whilst science/regulation is generally prospective in its approach to assessing causality, litigation is retrospective. There are very significant differences in the standard of proof required to establish causation in each context. There is no apparent consensus as to what is the scientific standard of proof. This is of significance in circumstances where litigation is encouraged or instituted based upon, say, a controversial scientific study (such as the MMR litigation,61 which was ultimately discontinued) or upon precautionary regulatory action (such as the oral contraceptive litigation,62 where the claimants failed to prove causation). Scientific or regulatory standards of proof in relation to causation may not always be clearly articulated (or understood by lawyers) and will not necessarily bear any relation to the legal standard of proof for causation. Further, there are divergent approaches to ‘risk’ as a test for causality by scientists, regulators and the courts.

59   AB & ors v Ministry of Defence [2010] EWCA Civ 1317, [152]; Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572, [24]. 60   See the Employers’ Liability Policy Trigger Litigation; Durham v BAI (in run off) & ors [2009] 2 All ER 26, [31](i); Sienkiewicz [2011] UKSC 10, [2011] 2 WLR 523, [102]. 61   Sayers & ors v Smithkline Beecham Plc & ors [2006] EWHC 3179, [2007] EWHC 1346. 62   XYZ v Schering Health Care Ltd & ors [2002] EWHC 1420, (2002) 70 BMLR 88 (QB).

10 Proving Causation: Probability versus Belief  1 RICHARD W WRIGHT

I. Introduction

One of the frequently assumed major differences between civil law and common law systems is the standard of persuasion applied by each in civil (non-criminal) cases. In most civil law jurisdictions, it is commonly assumed that the standard of persuasion is the same for criminal and civil proceedings. The plaintiff in a civil case, as well as the prosecutor in a criminal case, must provide sufficient proof to convince the trier of fact of the truth of the facts at issue on the particular occasion. This is phrased in France as a requirement that the trier of fact must have an intime conviction, an inner, personal, subjective conviction or belief in the truth of the facts at issue. The German courts have stated that a degree of conviction is required that silences doubts for practical purposes without completely eliminating them. Both of these standards are sometimes interpreted as being equivalent to the common law standard of proof beyond a reasonable doubt. Although it is recognised that absolute certainty is impossible to achieve, the required degree of belief is often expressed in terms of a virtual certainty, or at least a very high probability. However, a mere statistical probability, no matter how high, will not suffice in the absence of the required conviction or belief in the truth of the facts at issue.2 1   This chapter includes substantial material from my prior papers on this topic. See RW Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts [Pruning] (1988) 73 Iowa Law Review 1001, 1044–67; RW Wright, ‘Liability for Possible Wrongs: Causation, Statistical Probability, and the Burden of Proof ’ [Possible Wrongs] (2008) 41 Loyola Los Angeles Law Review 1295; RW Wright, ‘Proving Facts: Belief versus Probability’ [Proving Facts] in H Koziol and BC Steininger (eds), European Tort Law 2008 (Vienna, Springer-Verlag, 2009) 79. 2   eg BGH, 17 February 1970 (‘Anastasia Decision’), 53 BGHZ 245, 256 (German Federal Court of Justice) (interpreting s 286 of the German Code of Civil Procedure as requiring ‘full judicial conviction in the form of a degree of certainty that silences doubt for practical purposes, even if it does not eliminate them entirely’); FHS Bridge, The Council of Europe French-English Legal Dictionary (Strasbourg, Council of Europe, 1998) 173 (defining intime conviction as ‘reasonable conviction; reasonable certainty; state of being satisfied beyond reasonable doubt (personally convinced); personal conviction of the court (after considering all the evidence)’); KM Clermont and E Sherwin, ‘A Comparative View of Standards of Proof ’ (2002) 50 American Journal of Comparative Law 243, 245–46, 250, 256, 271; D Demougin and C Fluet, ‘Preponderance of Evidence’ (2006) 50 European Economic Review 963, 964; C Engel, ‘Preponderance of the Evidence versus Intime Conviction: A Behavioral Perspective on a Conflict Between American and Continental European Law’ (2009) 33 Vermont Law Review 435; J Kokott, The Burden of Proof in Comparative and International Human Rights Law (Dordrecht, Kluwer Law International, 1998) 18–19; H Nagel, ‘Evidence, Law of ’ in 7 Encyclopædia Britannica, Macropædia,

196  Richard W Wright In common law jurisdictions, there is an explicit distinction between the standard of persuasion in criminal and civil proceedings. In criminal proceedings, the standard is very high: the prosecutor must prove the defendant’s guilt ‘beyond a reasonable doubt’. In civil proceedings, however, the plaintiff generally only needs to prove his case by the much lower standard of a ‘preponderance of the evidence’ (the usual formulation in the United States) or a ‘balance of probability’ (the usual formulation in the British Commonwealth and Scandinavia). Both the ‘preponderance of the evidence’ standard, which is commonly rephrased or interpreted as a ‘more likely than not’ standard, and the ‘balance of probability’ standard are often interpreted by academics, and sometimes by judges, as merely requiring a 50+ per cent probability.3 For both standards, as so interpreted, it would be better to employ the term ‘standard of proof ’ rather than ‘standard of persuasion’, since the latter implies an element of conviction or belief that is lacking when all that is involved is a class-based statistical probability. Common law lawyers find it hard to understand why the very high standard of per­ suasion required for a criminal conviction should also apply in a civil action.4 Some of them also believe that the civil law’s rejection of proof by statistical probabilities and its insistence that the trier of fact instead be convinced of the truth of the facts at issue in the particular case is naïve, irrational, ‘strange’ and ‘very odd,’ especially when such conviction is interpreted as requiring virtual certainty. They note that nothing is certain, that all evid­ ence merely gives rise to probabilities regarding the facts at issue, and that the 50+ per cent probability standard best serves the assumed goal of minimising errors.5 They also question the focus on conviction, belief, and truth in civil law jurisdictions given the limited means for acquiring relevant evidence in those jurisdictions.6 Conversely, many civil law lawyers believe that it is naïve to believe that aggregate statistical probabilities can establish what actually happened in a particular case, and they find it hard to understand why a defendant should be held civilly liable based merely on such aggregate class-based probabilities, in the absence of evidence sufficient to convince the trier of fact of what actually happened in the particular case. Believing that the common law’s ‘preponderance of the evidence’ and ‘balance of probability’ standards allow such proof, they reject both standards, at least as so interpreted.7 I have argued that the supposed major differences between the standards of persuasion in common law and civil law jurisdictions are greatly overstated.8 I retrace that argument in section II below. I conclude that, in general, the common law as well as the civil law con­ tinues to view the applicable standards of persuasion in civil as well as criminal actions as requiring the formation of a belief in the truth of the facts at issue in the particular case, rather than a mere class-based statistical probability, and that concrete ‘particularistic 15th edn (Chicago, Encyclopædia Britannica, 1974) 1a, 2b; R Zo˝ ller, Zivilprozessordnung (Cologne, Verlag Dr Otto Schmidt, 1999) (‘Less than the conviction of truth is not sufficient for a proof . . . Standards that are based on a measure of likelihood do not find any support in the law.’). 3   eg Clermont and Sherwin (n 2) 243, 251–52 and fn 39, 257, 261 fn 86; Demougin and Fluet (n 2) 964; Kokott (n 2) 18–20. 4   eg Clermont and Sherwin (n 2); Engel (n 2). 5  eg E Adeney, ‘The Challenge of Medical Uncertainty: Factual Causation in Anglo-Australian Toxic Tort Litigation’ (1993) Monash University Law Review 23, 57, 59–62; Clermont and Sherwin (n 2) 243–44, 251–52, 258–59, 267, 271, 273–74. 6   Clermont and Sherwin (n 2) 249–50; see Wright, Proving Facts (n 1) 80. 7   eg M Taruffo, ‘Rethinking the Standards of Proof ’ (2003) 51 American Journal of Comparative Law 659, 663– 64, 667–71. 8   Wright, Proving Facts (n 1) 82–90.

Proving Causation: Probability versus Belief  197 evidence’ specific to the particular case is necessary to support such a belief. Although the available evidence is slim, I also conclude that in the civil law as well as the common law the strength of the required belief is lower in civil actions than in criminal actions. The distinctions between belief and probability and between particularistic evidence and statistical evidence have been questioned. As the proponents of proof by mere statistical probability argue, given our limited knowledge of causal laws and the circumstances of particular situations, nothing is certain. All evidence, including concrete evidence specific to the particular occasion – for example, fingerprint evidence and eyewitness testimony – merely gives rise to a probability regarding the facts at issue.9 Does it then make any sense to distinguish between probability and belief? I address these points and questions in section III below, in which I argue that the holding of a belief regarding what actually happened in a particular situation is very different from being willing to place a bet on what happened, and that while class-based statistics are very useful for the placing of the bet, they are insufficient and generally unhelpful for the formation of the belief, for which instead particularistic evidence is essential because only it is capable of converting possibly applicable causal generalisations (with their associated statistical frequencies) into actually instantiated causal laws. Finally, in section IV, I apply the lessons of sections II and III to the various types of problematic causal situations that courts around the world have struggled to deal with in recent decades. Clear recognition of the issues in these cases and their proper resolution has often been hindered by an unanalysed assumption that the standard of persuasion in civil actions is a mere statistical probability standard, which however is not consistently applied, since doing so would generate significant problems and paradoxes. The considerable confusion that now exists could and should be greatly reduced simply by replacing the highly misleading phrases ‘more likely than not’ and ‘balance of probability’, and even the less misleading phrase ‘preponderance of the evidence’, with ‘a minimal belief ’.

II. The Standards of Persuasion

A.  The Civil Law Italian law professor Michele Taruffo notes that ‘no rule, in any civil law system, requires the courts to apply in civil cases the same standard of proof that is applied in criminal cases.’10 Instead, the relevant code provisions merely state the need for the triers of fact to be convinced regarding the truth of the facts at issue. They do not provide any standard for reaching that conviction or require that whatever standard is employed be the same in criminal and civil proceedings. To the contrary, they emphasise the judge’s discretion in reaching that conviction based on his or her ‘free evaluation’ of all the available evidence, unconstrained by rigid rules of legal proof, such as the hearsay rules in the common law and the weighting of different types of evidence and mathematical calculation of ‘full 9   The difficulties of attributing a specific piece of evidence to a particular source are examined in National Research Council, Strengthening Forensic Science in the United States: A Path Forward (Washington DC, National Academies Press, 2009). 10   Taruffo (n 7) 665.

198  Richard W Wright proof ’ that existed under the French regime of preuve légale prior to the adoption of the civil codes.11 Changes to the French Code of Criminal Procedure, beginning in 1983, explicitly state that a civil action can proceed, in the criminal court or the civil court as appropriate, despite the acquittal of the defendant in the criminal action.12 This makes little sense unless the civil standard, as in the common law, is lower than the criminal standard. While the difference could be in the degree of fault required rather than the standard of persuasion, the relevant changes to the Code of Criminal Procedure do not refer to any change in the required degree of fault, which previously was assumed to be the same in criminal and civil actions: an acquittal in the criminal action precluded any civil action. In Italy, the Supreme Court of Cassation has explicitly adopted the ‘beyond a reasonable doubt’ standard for criminal actions and the ‘more probable than not’ standard for civil actions: As this Court has previously stated, the main difference [between the penal and civil processes] is in the standards of proof that each system requires (Cass. Pen., S.U., 11.09.2002, n. 30328). The Penal Code requires proof ‘beyond a reasonable doubt’ while the Civil Code merely requires ‘more probable than not.’ The different standards correspond to the different values at stake in each system (Cass. 16.10.2007, n. 21619; Cass. 18.04.2007, n. 9238; Cass. 05.09.2006, n. 19047; Cass. 04.03.2004, n. 4400; Cass. 21.01.2000, n. 632).13

In many civil law jurisdictions, there is little discussion by the courts, or even by academics, of the content of the standards of persuasion. However, anecdotal evidence indicates that the standards of persuasion in France, even in criminal proceedings,14 vary depending on the discretion of the judge. This is also said to be the case in Germany.15 Kevin Clermont and Emily Sherwin, at the end of an article generally assuming otherwise, eventually conclude that the supposed identity of the (presumed very high) criminal and civil standards of persuasion in civil law jurisdictions is a myth, which they argue is purposely maintained by the courts to shore up their legitimacy.16 They state that ‘civil-law judges likely apply a haphazardly variable civil standard of proof ’,17 and they cite scholars who believe that the civil standard is closer to the ‘more likely than not’ standard than the ‘beyond a reasonable doubt’ standard.18 11   ibid 666–67; Nagel (n 2) 1b, 3a; see Wright, Proving Facts (n 1) 83–84; n 20 below. The judge’s ‘free evaluation’ is not completely unconstrained. In addition to the substantial limitations on the power of the parties or the judge to obtain access to relevant evidence, some legal proof rules continue to exist, especially in France. See Clermont and Sherwin (n 2) 249; R Vouin, ‘The Exclusionary Rule: France’ (1961) 52 Journal of Criminal Law, Criminology, and Police Science 275. But cf Taruffo (n 7) 661, 674–75. 12   eg Code of Criminal Procedure arts 4-1, 371, 372, 470–71. 13   Cass Civ Sez Un 581 [2008] s 3.9 (translated by Claudia DiMarzo, University of Palermo). The concept of ‘probabilistic certainty’ is discussed in Wright, Proving Facts (n 1) 95–96. 14   See eg, M Foucault, Abnormal: Lectures at the Collège de France 1974–1975 (G Burchell transl) (New York, Picador, 2003) 8–11. 15  PL Murray and R Stürner, German Civil Justice (Durham, Carolina Academic Press, 2004) 310–11; P Gottwald, ‘Fact Finding: A German Perspective’ in DL Carey Miller and PR Beaumont (eds), The Option of Litigating in Europe (London, British Institute of International and Comparative Law, 1993) 67, 77 (stating that German courts apply a preponderance standard ‘with regard to prima facie cases, to causation, to negligence and to assessment of damages’). 16   Clermont and Sherwin (n 2) 258–59, 269–73. 17   ibid 273. 18   ibid 261, citing, among other sources, ALI/UNIDROIT Principles and Rules of Transnational Civil Procedure, Discussion Draft No 3 (St Paul, American Law Institute, 2002) P-18A, R-31E. See also Murray and Stürner (n 15) 310–11.

Proving Causation: Probability versus Belief  199 However, contrary to Clermont and Sherwin’s identification of the ‘preponderance of the evidence’ and ‘more likely than not’ standards with a mere 50+ per cent statistical probability,19 civil law judges and lawyers, as noted above, generally reject such identification and instead insist that the trier of fact must be convinced (believe) that the alleged causal connection actually existed in the particular case. Although the codes do not specify any specific standard of persuasion in (or criminal) cases, the references to the judge’s ‘conviction’ in the French intime conviction standard and other civil code provisions, for example, the German criminal and civil code provisions,20 provide a minimum standard of persuasion: the judge is required to have a conviction – a belief – regarding the truth of the facts at issue. This is the core of the civil law approach to proof, which is thought to be absent in the common law’s ‘preponderance of the evidence’, ‘more likely than not’, and ‘balance of probability’ standards.21 Civil law courts generally insist that mere statistical probabilities are insufficient evidence to support a belief in what actually happened in a particular case. Instead, evidence specific to the particular case is necessary.22 For example, in a continuation of the language quoted above from the recent decision of the Italian Supreme Court of Cassation, the Court stated: The [European] Court of Justice CE has recently stated that causation cannot be based on probabilities (CGCE 13.07.2006, n. 295; CGCE 15.02.2005, n. 12). The concept of ‘probabilistic certainty’ [in Italian law] is a standard that is necessary in all civil cases. The mere statistical likelihood that one act or omission caused certain harm is not enough to impose liability. Probabilistic certainty also requires evidence from the specific case to support that statistical likelihood.23

B.  The Common Law No-one, other than die-hard mathematical probabilists, argues that the ‘beyond a reasonable doubt’ standard of persuasion for criminal liability should be, or is, interpreted or applied as a mere (very high) statistical probability. Less commonly appreciated nowadays is that the ‘preponderance of the evidence’, ‘more likely than not’, and ‘balance of probability’ standards of persuasion for civil liability traditionally also have been understood by judges and presented to juries as requiring the formation of a minimal belief regarding the   Clermont and Sherwin (n 2) 265.   Section 261 of the German Code of Criminal Procedure states: ‘The court shall decide on the result of the evidence taken according to its free conviction gained from the hearing as a whole’. CJM Safferling, ‘Terror and Law – Is the German Legal System able to deal with Terrorism? – The Bundesgerichtshof (Federal Court of Justice) decision in the case against El Motassadeq’ (2004) 5 German Law Journal 515, 520. Subsection 1 of section 286 of the Code of Civil Procedure states: ‘The court shall decide at its free discretion, by taking into account the whole substance of the proceedings and the results of any evidence taking, whether a factual allegation should be regarded as true or untrue. The grounds which prompted the court’s conviction shall be stated in the judgment’. SL Goren, The Code of Civil Procedure Rules of the Federal Republic of Germany of January 30, 1877 and the Introductory Act for the Code of Civil Procedure Rules of January 30, 1877 (Littleton, Fred B Rothman & Co, 1990) 73; see Wright, Proving Facts (n 1) 82–83, 94. 21   Taruffo (n 7) 675. 22   See eg, BGH, 6 October 1998, 1999 NJW 860 (German Federal Court of Justice) (upholding the lower courts’ decision that a 70 per cent statistical probability that prompt diagnosis and treatment of the patient’s illness would have prevented the patient’s injury was insufficient proof of causation of the injury by negligent non-diagnosis); Taruffo (n 7) 659, 663–64, 667–71; n 2 above (quote from Zoller). 23   Cass Civ Sez Un 581 [2008] s 3.9 (translated by Claudia DiMarzo, University of Palermo). The Italian court’s concept of ‘probabilistic certainty’ is discussed in Wright, Proving Facts (n 1) 95–96. 19 20

200  Richard W Wright truth of the fact(s) at issue, rather than a mere 50+ per cent statistical probability. My arguments in support of this conclusion are based primarily on an analysis of relevant materials in the United States. However, the situation appears to be the same in the other common law countries, as is indicated especially by the recent judgments of the United Kingdom Supreme Court in Sienkiewicz v Greif (UK) Ltd.24 A highly respected source of jury instructions in the United States defines the ‘preponderance of the evidence’ standard of persuasion as follows: To ‘establish by a preponderance of the evidence’ means to prove that something is more likely so than not so. In other words, a preponderance of the evidence in the case means such evidence as, when considered and compared with that opposed to it, has more convincing force, and produces in your minds belief that what is sought to be proved is more likely true than not true. This rule does not, of course, require proof to an absolute certainty, since proof to an absolute certainty is seldom possible in any case.25

The core of this instruction is essentially identical to the standards of persuasion that are stated in the German codes of criminal and civil procedure.26 While the American instruction does not include the ‘free evaluation of the evidence’ principle that is stressed in the German code provisions, it has the same focus on the required formation of a conviction (belief) in the truth of the facts at issue. Moreover, unlike the German code provisions, code provisions and jury instructions in the United States generally specify the required degree of belief. For example, section 115 of the California Evidence Code states: ‘Burden of proof ’ means the obligation of a party to establish by evidence a requisite degree of belief concerning a fact in the mind of the trier of fact or the court. . . . The burden of proof may require a party to raise a reasonable doubt concerning the existence or nonexistence of a fact or that he establish the existence or nonexistence of a fact by a preponderance of the evidence, by clear and convincing proof, or by proof beyond a reasonable doubt. Except as otherwise provided by law, the burden of proof requires proof by a preponderance of the evidence.27

As this code provision states, the default required degree of belief in a civil case in the United States is by a mere ‘preponderance of the evidence’, that is, the slightest degree of belief,28 rather than the much stronger degrees of belief required under the ‘clear and convincing evidence’ or ‘beyond a reasonable doubt’ standards, which are more literally worded belief standards. Civil jury instructions in the Unites States often refer to proof that the disputed fact is ‘more probably true than not true,’ rather than simply ‘more likely than not’ as a matter of

24   [2011] UKSC 10 [2011] 2 WLR 523; see text to nn 41–48 below; Adeney (n 5) 56–59 (discussing Australian cases). Adeney incorrectly equates my insistence on the formation of a minimal belief with insistence on absolute certainty. Adeney (n 5) 59–60. 25   EH Devitt et al, 3 Federal Jury Practice and Instructions (Civil), 4th edn (St Paul, West Publishing Co, 1987) s 72.01 at 32 (emphasis added); see State Bar of Nevada, Nevada Pattern Jury Instructions – Civil (Charlottesville, Michie, 1986) s 3.00.1 (‘The term “preponderance of the evidence” means such evidence as, when weighed with that opposed to it, has more convincing force, and from which it appears that the greater probability of truth lies therein’.) (emphasis added); Wright, Pruning (n 1) 1065 and fns 337–39. 26   See n 20 above. 27   California Evidence Code (St Paul, West Publishing Co, 2010) s 115. 28   See eg, Livanovitch v Livanovitch (1926) 131 A 799, 800 (Supreme Court of Vermont) (‘If . . . you are more inclined to believe from the evidence that he did so deliver the bonds to the defendant, even though your belief is only the slightest degree greater than that he did not, your verdict should be for the plaintiff ’. (quoting the trial court’s jury instructions)).

Proving Causation: Probability versus Belief  201 mere statistical probability.29 When ‘more likely than not’ or some similar phrase is employed, it is usually clear from the surrounding language that the phrase is not being used to refer to a mere 50+ per cent statistical probability, but rather to refer to the formation of a minimal belief in the truth of what actually happened on the particular occasion. In a famous case involving an unidentified bus that forced an oncoming automobile off the road, in which the statistical probability that it was the defendant’s bus was very high, the Supreme Court of Massachusetts stated: It has been held not enough that mathematically the chances somewhat favor a proposition to be proved; for example, the fact that colored automobiles made in the current year outnumber black ones would not warrant a finding that an undescribed automobile of the current year is colored and not black, nor would the fact that only a minority of men die of cancer warrant a finding that a particular man did not die of cancer. The weight or ponderance of evidence is its power to convince the tribunal which has the determination of the fact, of the actual truth of the proposition to be proved. After the evidence has been weighed, that proposition is proved by a preponderance of the evidence if it is made to appear more likely or probable in the sense that actual belief in its truth, derived from the evidence, exists in the mind or minds of the tribunal notwithstanding any doubts that may still linger there.30

Similarly, in the more recent, also well-known Cipillone case involving cancer allegedly caused by cigarettes, the US Court of Appeals for the Third Circuit questioned attempts by some courts and commentators to use mathematical probability to define and prove ‘but for’ causation: We are not convinced that when a jury determines that ‘but for’ a defendant’s conduct, the injury would not have occurred, it is determining that the chances of that injury being the result of defendant’s conduct are 50% or greater. Traditionally, jury instructions have been in words, not numbers.31

When asked to do so by researchers, many judges similarly object to interpreting standards of persuasion in terms of quantitative probabilities.32 In one survey, 80 out of 255 judges refused to specify a probability sufficient for a ‘preponderance of the evidence’ finding.33 Of the judges who were willing to do so, only about three-fifths chose a probab­ ility of 50 to 55 per cent; about two-fifths chose a probability of 60 per cent or greater, almost one-fifth a probability of 70 per cent or greater, one-tenth a probability of 80 per cent or greater, and one-twentieth a probability of 90 to 100 per cent.34 The distribution of 29  eg Judicial Council of California Civil Jury Instructions (Matthew Bender, 2010) No 200 (‘more likely to be true than not true’); Illinois Supreme Court Committee on Pattern Jury Instructions in Civil Cases, Illinois Pattern Jury Instructions: Civil (St Paul, West Publishing Co, 2006) s 21.01 (‘more probably true than not true’); LB Sand et al, 4 Modern Federal Jury Instructions (Newark, LexisNexis, 2007) s 73.01, Instruction 73–2 (stating that ‘by a preponderance of the evidence’ means ‘more likely true than not true,’ considering the ‘weight’ and ‘quality and persuasiveness’ of the evidence). 30   Sargent v Massachusetts Accident Co (1940) 29 NE 2d 825, 827 (Supreme Court of Massachusetts) (citations omitted). 31   Cipollone v Liggett Group Inc (1990) 893 F 2d 541, 561 fn 17 (United States Court of Appeals, Third Circuit). I am grateful to my colleague, Kathy Baker, for bringing this footnote to my attention. 32   CMA McCauliff, ‘Burdens of Proof: Degrees of Belief, Quanta of Evidence, or Constitutional Guarantees?’ (1982) 35 Vanderbilt Law Review 1293, 1332; RJ Simon and L Mahan, ‘Quantifying Burdens of Proof: A View from the Bench, the Jury, and the Classroom’ (1971) 5 Law and Society Review 319, 329 (quoting judges as stating that ‘[p]ercentages or probabilities simply cannot encompass all the factors, tangible and intangible, in determining guilt – evidence cannot be evaluated in such terms’). 33   McCauliff (n 32) 1325 fn 184, 1330. 34   ibid 1331; Simon and Mahan (n 32) 324–25, 327 table 7.

202  Richard W Wright probabilities was about the same for the ‘more probable than not’ standard.35 Laypersons – jurors and students – were even less willing to interpret the preponderance standard as a mere 50+ per cent probability. About four-fifths of the laypersons chose a probability of 70 per cent or greater, half a probability of 80 per cent or greater, and more than one-tenth a probability of 95 to 100 per cent.36 Over 90 per cent of the judges and about two-thirds of the laypersons were opposed to having jurors simply make a probability finding, which the judge would then use to determine liability.37 Trial consultants advise American plaintiffs’ lawyers that ‘[m]any jurors will not agree to decide on the basis of 80 percent or 70 percent or 60 percent certainty’, but rather ‘expect you to prove your case beyond a reasonable doubt, and you won’t change their minds by explaining preponderance’. Instead, the trial consultants advise, repeatedly get witnesses to testify that something is ‘more likely right than wrong’ and, ‘beyond that,’ that they are ‘certain’ of the truth of the fact at issue.38 Courts in the United States also usually agree with the civil law courts that, to prove what actually happened in a particular case – to establish what the facts actually were in that case – the party with the burden of persuasion regarding those facts must employ evidence specific to that particular case, rather than mere statistical probabilities. In Day v Boston & Maine Railroad, the Supreme Court of Maine famously stated: Quantitative probability, however, is only the greater chance. It is not proof, nor even probative evidence, of the proposition to be proved. That in one throw of the dice there is a quantitative probability, or greater chance, that a less number of spots than sixes will fall uppermost [on both die] is no evidence whatever that in a given throw such was the actual result. Without something more, the actual result of the throw would still be utterly unknown. The slightest real evidence that sixes did in fact fall uppermost would outweigh all the probability otherwise.39

Jury instructions often refer to the ‘weight’ of the evidence.40 Abstract statistics do not have ‘weight’. Concrete evidence specific to the particular case does have ‘weight’. The recent judgments of the United Kingdom Supreme Court in Sienkiewicz similarly warn against literally interpreting the ‘balance of probability’ standard of persuasion as merely requiring a statistical 50+ per cent probability. Lord Rodger’s judgment is the clearest and most emphatic. Discussing a hypothetical situation in which reliable statistical evidence established that exposure to a substance more than doubled the risk of a specific injury, he stated: It is important to recognize that in such a case the claimant would not have proved, on the balance of probability, that his exposure to the [substance] actually caused [the injury]. . . . [B]y leading the epidemiological evidence, the only “fact” that the plaintiff can prove and offers to prove, on the balance of probability, is that in most cases the [injury] would have been related to the [exposure].   McCauliff (n 32) 1331.   Simon and Mahan (n 32) 327 table 7; see also DK Kagehiro and WC Stanton, ‘Legal vs. Quantified Definitions of Standards of Proof ’ (1985) 9 Law and Human Behavior 159, 164, 169 (discussing an empirical study demonstrating a divergence between subjects’ findings under the preponderance standard and a quantified 51 per cent standard, with results closer to those obtained under the preponderance standard even when the two standards were combined in the same instruction). 37  Simon and Mahan (n 32) 329, 330 fn 8. See also GL Wells, ‘Naked Statistical Evidence of Liability: Is Subjective Probability Enough?’ (1992) 62 Journal of Personality and Social Psychology 739 (finding that laypersons and trial judges are unwilling to base liability on mere statistical probability, requiring instead evidence specific to the particular situation that gives rise to a belief in the truth of the asserted facts). 38   D Ball, ‘Making Preponderance Work’ (March 2008) Trial 38–40. 39   Day v Boston & Maine Railroad (1902) 52 A 771, 774 (Supreme Court of Maine). 40   See eg, nn 25 and 29 and text to n 30 above. 35 36

Proving Causation: Probability versus Belief  203 So, if the judge accepts the evidence, it may legitimately satisfy him, on the balance of probability, not that the plaintiff ’s [injury] was caused by the [exposure], but that, in the absence of any evidence that the claimant is atypical, it is more probable than not that his [injury] was caused by the [exposure]. In short, the chances are that it was. Whether, in any particular case, the claimant’s [injury] was actually caused by the [exposure] is a matter of fact – and one that remains unknown, if the only available evidence is statistical. . . . Of course, it is possible to conceive of a legal system which chose, as a matter of policy, to make defendants liable for all the damages which a court was satisfied, on the balance of probability, they had probably caused. But only the legislature could alter English or Scots law so as to introduce a general rule to that effect, which would change the very nature of the system and completely alter its balance. . . . [I]n civil proceedings for damages the role of the judge is to decide, on the balance of probabilities, what actually happened.41

Baroness Hale stated: I do agree with Lord Rodger that doubling the risk is not an appropriate test of causation. . . . Risk is a forward looking concept – what are the chances that I will get a particular disease in the future? . . . But if the disease materialises, the existence of a statistically significant association between factor X and disease Y does not prove that in the individual case it is more likely than not that factor X caused disease Y. . . . But as a fact-finder, how can one ignore these statistical associations? Fact-finding judges are told that they must judge a conflict of oral evidence against “the overall probabilities” coupled with objective facts and contemporaneous documentation. . . . Yet judges do not define what they mean by “the overall probabilities” other than their own particular hunches about human behaviour. Surely statistical associations are at least as valuable as hunches about human behaviour. . . . Most judges will put everything into the mix before deciding which account is more likely than not. As long as they correctly direct themselves that statistical probabilities do not prove a case, any more than their own views about the overall probabilities will do so, their findings will be safe.42

Lord Mance also cautioned against sole reliance on statistical evidence to prove what actually happened in a particular case: I share a reluctance to place too much weight on such evidence. . . . [T]he law is concerned with the rights or wrongs of an individual situation, and should not treat people and even companies as statistics. . . . [A]n attribution of liability based substantially on statistical evidence, that, viewing the relevant population or group as a whole, it is more likely than not that the particular defendant was negligent or causatively responsible, appears to me most undesirable. That epidemiological evidence used with proper caution, can be admissible and relevant in conjunction with specific evidence related to the individual circumstances and parties is, however, common ground and clearly right. What significance a court may attach to it must depend on the nature of the epidemiological evidence, and of the particular factual issues before the court. Whether and if so when epidemiological evidence can by itself prove a case is a question best considered not in the abstract but in a particular case, when and if that question arises. If it can, then, I would hope and expect that this would occur only in the rarest of cases.43

Lord Kerr stated that he ‘shared the misgivings expressed’ by the other judges and that ‘[i]t is an essential and minimum requirement . . . that there be evidence connecting avowedly relevant statistical information produced by the epidemiological studies to the facts of the case’.44   [2011] UKSC 10 [156] (citation omitted) and [158]; see ibid [143], [153]–[160].   ibid [170] and [172]; see ibid [171]. 43   ibid [190]–[192]. 44   ibid [204] and [205]; see ibid [206]. 41 42

204  Richard W Wright Lord Dyson acknowledged the distinction between ‘fact probability’ based on a mere 50+ per cent statistical probability and ‘belief probability’ based on an actual belief in what actually occurred in a particular case: In my view, this is an important distinction and it is of particular relevance in relation to causation in toxic torts. It is often the basic impossibility of proving individual causation which distinguishes toxic tort cases from ordinary personal injury cases. . . . [E]pidemiology is based on the study of populations, not individuals. . . . [I]n an individual case, epidemiology alone cannot conclusively prove causation. At best, it can establish only a certain [statistical] probability that a randomly selected case of disease was one that would not have occurred absent exposure. Ultimately, questions of burden and standard of proof are policy matters for any system of law. It is trite law that our system requires a civil claim to be proved by a claimant on the balance of probability. It is a matter of policy choice whether, and, if so, in what circumstances the courts are willing to find causation proved on the balance of probability on the basis of epidemiological evidence alone. . . . [I]t is not necessary for the resolution of the present appeal to decide whether epidemiological evidence alone suffices. . . . It seems to me, however, that there is no a priori reason why, if the epidemiological evidence is cogent enough, it should not be sufficient to enable a claimant to prove his case without more. Our civil law does not deal in scientific or logical certainties. The statistical evidence may be so compelling that . . . the court may be able to infer belief probability from fact probability. To permit the drawing of such an inference is not to collapse the distinction between fact probability and belief probability. It merely recognises that, in a particular case, the fact probability may be so strong that the court is satisfied as to belief probability.45

Lord Brown did not address the statistical evidence issue.46 Only Lord Phillips expressed a willingness to treat adequate and reliable epidemiological evidence (which he and the other judges agreed was lacking in Sienkiwicz) as being generally sufficient, by itself, to satisfy the ‘balance of probability’ standard of persuasion, although he acknowledged, as did all the other judges, that doing so would be supportable only as a matter of judicial policy, since such statistical evidence is insufficient to establish actual causation in a particular case as a matter of scientific fact,47 and he stated that a statistical probability only minimally greater than 50 per cent would be ‘a tenuous basis for concluding that the statistical probable cause of a disease was also the probable biological cause, or cause in fact’, since ‘the balance of that probability is a very fine one’.48 In sum, contrary to what is commonly thought, I believe that there is strong agreement between civil law and common law jurisdictions regarding the standards of persuasion. In both systems, the plaintiff generally must provide evidence sufficient to convince the trier of fact of the truth of the facts at issue in the particular situation; a mere statistical probability, no matter how high, generally is insufficient. Moreover, in both systems, the required degree of belief varies in criminal actions and civil actions, given the different interests at stake in each action. For criminal actions, a very high degree of belief is required: no reasonable doubt can remain. For most issues in civil actions, however, the standard of persua  ibid [217]–[222].   The closest Lord Brown came was his observation that ‘[t]here is rough justice about the law of personal injury liability as a whole. To compensate a claimant in full for a lost finger because there was a 60:40 chance that he would have worn protective gloves had they been made available to him may be regarded as rough justice for the defendants. But it is balanced by the denial of compensation to a plaintiff who cannot establish that he probably would have worn the gloves’. ibid [187]. 47   ibid [6], [9]–[10], [26], [52]–[54], [78]–[106]. 48   ibid [83]. 45 46

Proving Causation: Probability versus Belief  205 sion is much lower: all that is required is the formation of a minimal personal belief (intime conviction) by the trier of fact in the truth of the facts at issue.

III.  Proving Causation

Proof of a singular instance of causation requires proof of (1) a scientifically valid causal generalisation that contains in its antecedent an abstract condition that is at least partially instantiated by the putative cause, the so-called ‘general causation’ or causal capacity issue, and (2) complete instantiation of the allegedly relevant causal generalisation and its underlying causal laws on the particular occasion, the so-called ‘specific causation’ issue. Causation is an empirical relation between concrete conditions. A singular instance of causation consists of the complete instantiation of one or more causal laws by concrete conditions on a particular occasion. A causal law is a law of nature; it describes an empirically based, invariable, nonprobabilistic relation49 between some minimal set of abstractly described antecedent conditions and some abstractly described consequent condition, such that the concrete instantiation of all the antecedent conditions will always immediately result in the concrete instantiation of the consequent condition. Any concrete condition that is part of the instantiation of the completely instantiated antecedent of the causal law is a cause of (contributed to) the instantiation of the consequent.50 Our knowledge of causal laws generally is incomplete, and even when it is complete we rarely refer to completely specified causal laws, since such complete specification would be extremely burdensome and unnecessarily detailed and lengthy. We rather employ causal generalisations, which refer to only some of the antecedent conditions in the relevant causal laws and have only as much specificity as is possible and needed in the particular situation. Moreover, the generalisations that we employ usually refer elliptically to a large number of simultaneously or successively operative causal laws. However, when we make an assertion regarding a singular instance of causation, we are implicitly asserting that all the unstated as well as the stated conditions in the relevant causal generalisations and all the unknown as well as known conditions in the causal laws underlying the causal generalisations were instantiated on the particular occasion.51 In the bash/crash/slash physical trauma situations that once dominated tort litigation and are still common, causal capacity usually is assumed and need not be alleged or proven.52 However, modern tort litigation often involves complex and poorly understood biological and chemical processes, so that the issue of causal capacity is contested and must be proved through expert scientific evidence. To establish causal capacity, scientists typically rely on controlled experiments (eg toxicology) or observation and statistical analysis of uncontrolled events (epidemiology). 49   In a radically indeterministic world, in which nothing was even weakly necessary or sufficient for anything else, the concept of causation likely would not exist. RW Wright, ‘The NESS Account of Natural Causation: A Response to Criticisms’, chapter 14 in this volume, III.G. 50   ibid II. 51  ibid. 52   But see Vosburg v Putney (1890) 47 NW 99, (1891) 50 NW 403 (Supreme Court of Wisconsin), frequently used to teach liability for intentional injury in the United States, in which the capacity of a traumatic blow to cause osteomyelitis was the major contested issue. See ZL Zile, ‘Vosburg v Putney: A Centennial Story’ 1992 Wisconsin Law Review 877, 910–14, 933–42, 956–64.

206  Richard W Wright Toxicological experimentation is limited by ethical restrictions on research on humans and other animals, by the frequent need to use doses of the substance being tested that are much higher than the actual or projected exposure levels in order to get meaningful results within reasonable time-frames using a reasonable number of tests or samples, and by the difficulty of extrapolating from effects on nonhuman species to effects on humans and from effects at high doses to effects at much lower doses. Scientists therefore often rely on the uncontrolled ‘natural experiments’ that occur in ordinary life without their intervention. Assessment of these natural experiments relies on epidemiology: the use of statistical methods to attempt to discern statistically significant correlations. However, this method often requires very large sample populations which either do not exist or cannot practically be evaluated. Moreover, as scientists recognise, a statistically significant correlation between A and B is insufficient by itself to establish a causal relationship between A and B, since the correlation may be spurious – that is, due to some antecedent common cause of both A and B, with no causal relation existing between A and B themselves. For example, my neighbour may take the train to work every morning exactly one hour after I take the train. The perfect correlation does not prove or demonstrate that my catching the train is a cause of his catching the train one hour later; rather, both of our actions are independently caused by other factors, including the occurrence of certain times of the day.53 To avoid attributing a causal relation based on such spurious correlations, scientists generally require that A occur before B, that other possible explanations for the correlation be ruled out, and that there be some explanation or theory indicating how A causes B.54 Since a causal generalisation is not a complete specification of the underlying causal laws, instantiation of all of the antecedent conditions in the causal generalisation provides only a probability, rather than a certainty, that the underlying causal laws were completely instantiated and, thus, that the condition at issue actually was a cause of (contributed to) the relevant consequence. This probability is what I have called an ex ante causal probability. The statistical probability provided by an epidemiological study will provide only a lower bound, and perhaps even a negative value, for the ex ante probability of causation, since it will not take into account situations involving causal overdetermination, in which the condition at issue contributed to the occurrence of the consequence but the consequence would have occurred anyway owing to other duplicative or preempted conditions. Thus, a requirement that exposure to a substance more than double, or even increase, the risk of a certain injury is too strict when applied to the issue of general causation or causal capacity (the existence of a relevant causal generalisation linking such exposures to such injuries).55 More significantly for our purposes, the ex ante causal probability is a non-individualised, class-based probability that describes, on average or in the aggregate, the frequency of actual causation by the causal process at issue in the totality of situations in which the antecedent of the causal generalisation is completely instantiated. As was recognised by the United Kingdom   See Wright (n 49) III.C.   AB Hill, ‘The Environment and Disease: Association or Causation?’ in (1965) 58 Proceedings of the Royal Society of Medicine 295; ‘Symposium on Legal and Scientific Perspectives on Causation in Mass Tort Litigation’ (1991) 1 Courts, Health Science & the Law pt 3. 55  See KJ Rothman, S Greenland, and TL Lash, Modern Epidemiology, 3rd edn (Philadelphia, Lippincott, Williams and Wilkins, 2008) 62–64; S Greenland and J Robins, ‘Epidemiology, Justice, and the Probability of Causation’ (2000) 40 Jurimetrics 321, 326–27. I am grateful to Alex Broadbent for bringing Greenland’s work to my attention. For discussion of duplicative and preemptive causation, see Wright (n 49) III.D and III.E. 53 54

Proving Causation: Probability versus Belief  207 Supreme Court in Sienkiewicz,56 this ex ante causal probability does not describe an individualised probability of instantiation for the specific situation at issue, which will vary widely depending on the state of the antecedent conditions in the under­lying causal laws that are not included in the causal generalisation. For example, the individualised probability that smoking cigarettes or exposure to some other carcinogen or toxin will cause (or has caused) the death of a specific individual will vary widely depending on that individual’s genetics and lifestyle. The aggregate statistical probability provided by an epidemiological study of a certain exposed population over a certain period of time will not constitute an accurate individualised probability even for a person in the studied population, much less for a person with a different exposure during a different period of time.57 Nevertheless, as recognised by the judges in Sienkiewicz,58 ex ante causal probabilities are useful, indeed necessary, for causal prediction or ‘postdiction’ – predicting, in the absence of further evidence, what is likely to happen in the future or the most likely cause of an already occurred event. However, as all the judges in Sienkiewicz also recognised,59 ex ante causal probabilities are insufficient for establishing what actually happened in a particular situation, that is, which of the possibly applicable causal generalisations actually applied in the particular situation. Indeed, as only Lord Rodger perhaps realised,60 apart from its possible use in implying (but being neither necessary nor sufficient for proving) a possibly applicable causal general­ isation, the ex ante causal probability is neither useful nor relevant in establishing what actually happened in a particular situation, since it provides no information on whether the abstract elements in the underlying causal laws that are not included in the causal generalisation actually were instantiated on that occasion. It merely states that, on average or in the aggregate, X per cent of the time that the abstract elements specified in the causal generalisation are instantiated, the unspecified abstract elements required to complete the underlying causal laws are also instantiated. It does not help us determine whether this particular occasion is one of the X per cent in which the underlying causal laws were fully instantiated, or instead is one of the 100–X per cent in which they were not. The ex ante causal probabilities can be used to place a bet on what happened, but they cannot be used to resolve the bet. If a horse wins 90 per cent of its races or the odds are 90 per cent that a spin of a roulette wheel will not result in the ball’s landing on a certain number, no-one who placed a bet either way in either situation will consider themselves to have won or lost the bet in the absence of specific evidence of the actual outcome of the particular race or spin of the wheel. Moreover, on the issue of what actually happened such specific evidence of the actual outcome is all that counts; the ex ante causal probabilities are irrelevant (apart from their possible usefulness in implying general causal capacity). This is the point made 56   See [2011] UKSC 10 [83], [96]–[103] (Lord Phillips); ibid [152] (Lord Rodger); ibid [170] (Baroness Hale): ibid [205] (Lord Kerr); ibid [223] (Lord Dyson); text to n 47 above. 57   See eg, R Goldberg, ‘The Contraceptive Pill, Negligence and Causation: Views on Vadera v. Shaw’ (2000) 8 Medical Law Review 316, 331–35 (noting the court’s failure to particularise evidence to the individual patient, a 22-year-old Asian woman with no history of health problems, who suffered a stroke within six weeks of administration of the ‘second generation’ oral contraceptive, Logynon). 58   See, eg [2011] UKSC 10 [7] (Lord Phillips); ibid [163] (Lord Rodger); text to nn 41–48 above. 59   See ibid [6]–[10] (Lord Phillips); text to nn 41–48 above. Although Lord Dyson correctly observes that a very high statistical probability may cause the trier of fact to form a belief in actual causation in the particular instance, see ibid [222], it would not be a soundly based belief, for the reasons I state in the following text. 60   Compare Lord Rodger’s comments at ibid [153]–[156] with his comments at ibid [163]; see text to nn 41–48 above.

208  Richard W Wright in the previously quoted language from the Supreme Court of Maine’s opinion in Day v Boston & Maine Railroad: Quantitative [ex ante] probability, however, is only the greater chance. It is not proof, nor even probative evidence, of the proposition to be proved. That in one throw of the dice there is a quantitative [ex ante] probability, or greater chance, that a less number of spots than sixes will fall uppermost [on both die] is no evidence whatever that in a given throw such was the actual result. Without something more, the actual result of the throw would still be utterly unknown. The slightest real [particularistic] evidence that sixes did in fact fall uppermost would outweigh all the [ex ante] probability otherwise.61

In the dice throwing hypothetical, there are six competing causal generalisations – ‘throwing a die causes six (five/four/three/two/one) spots to fall uppermost (on average or in the aggregate, given a sufficient number of throws) one-sixth of the time’ – that, for heuristic purposes only,62 can be rephrased when a pair of dice are thrown as two competing causal generalisations: (1) ‘throwing a pair of dice causes a pair of sixes to fall uppermost 1/36th of the time’ and (2) ‘throwing a pair of dice causes something other than a pair of sixes to fall uppermost 35/36th of the time’. The ex ante causal probabilities associated with these two causal generalisations – 3 per cent and 97 per cent, respectively – provide a strong basis for a causal prediction, ‘postdiction’, or bet that the second causal generalisation was much more likely to be instantiated on this particular occasion and every other occasion considered separately. Nevertheless, they provide no information at all on which of the two causal generalisations actually was instantiated on the particular occasion. Legal fact-finders are not told that they merely need to place a bet on the existence of some fact, but rather are instructed that they must determine whether the fact actually existed. As we have seen, both civil law and common law courts generally insist that the finders of fact be convinced, to the required degree of belief, of the truth of the asserted facts in a case, and they further insist that such belief must be supported by concrete evidence that is specific to the particular case, rather than mere ex ante statistical probabilities. To prove what actually happened in a particular case – whether an allegedly applicable causal generalisation and its underlying causal laws were fully instantiated – it must be proved, to the required degree of belief, that all the abstract elements in that causal generalisation and its underlying causal laws were instantiated. As all or almost all of the judges in Sienkiewicz apparently realised,63 only particularistic evidence can establish such instantiation. Thus, the demand for particularistic evidence is not based on the notion that such evidence is ‘uniquely highly probabilifying,’64 but rather on the fact that it is uniquely instantiating. 61   See text to n 39 above. See also, eg, Johnston v United States (1984) 597 F Supp 374, 412 (United States District Court, Kansas) (holding that a greater than 50 per cent statistical probability that the plaintiff ’s cancers were caused by radiation exposure was insufficient proof of specific causation); United States v Shonubi 103 F 3d 1085 (1997) (United States Court of Appeals, Second Circuit) (rejecting the attempt by maverick federal district judge Jack Weinstein, the leading judicial proponent of the statistical probability interpretation of standards of per­ suasion, to use statistics on typical amounts of illegally smuggled heroin to establish, for sentencing purposes, the amounts actually smuggled by the defendant on prior occasions, and insisting instead on proof by evidence specific to each instance of smuggling). The proponents of statistical evidence have been especially apoplectic regarding the Second Circuit’s reining in of Judge Weinstein in the Shonubi case. See eg, United States v Shonubi 962 F Supp 370 (1997) (Federal District Court, Eastern District of New York) (Judge Weinstein); P Tillers, ‘If wishes were horses: discursive comments on attempts to prevent individuals from being unfairly burdened by their reference classes’ (2005) 4 Law, Probability and Risk 33. 62   See Wright, Pruning (n 1) 1059–60. 63   See text to nn 41–48 above. 64   See J Thomson, ‘Liability and Individualized Evidence’ (Summer 1986) 49 Law and Contemporary Problems 199, 206 and fn 14 (criticising the language from the Day case quoted in the text to nn 39 and 61 above).

Proving Causation: Probability versus Belief  209 Of course, we rarely, if ever, will have knowledge of the multitude of necessary abstract elements in the causal laws underlying a possibly relevant causal generalisation, much less direct particularistic evidence of instantiation of each of those elements. Instead, instantiation of the unknown elements, and even some or many of the known elements, will have to be inferred circumstantially from particularistic evidence of instantiation of the network of causal relationships that encompasses the particular occasion, which fit coherently into a story of what happened on the particular occasion. The degree of coherence of the particularistic evidence with the relevant causal story (coherent network of possible causal generalisations and underlying causal laws in the particular situation) is what I have previously inaptly and misleadingly referred to as an ex post causal probability. As I have previously emphasised, rather than being some quantitative statistical or mathematical probability, the inaptly referenced ‘ex post causal probability’, which is better described as a degree of coherent fit, is an unquantified judgment of the degree of coherence of the particularistic evidence with a possibly applicable causal story of what happened on the particular occasion.65 When analysing what actually happened on the particular occasion, we compare the coherence of the particularistic evidence with the various possibly applicable causal stories. The more particularistic evidence there is that fits with a possibly applicable causal story, the greater the fit, although the assessment of the level of fit also depends on the relative significance of the abstract element being instantiated in the overall causal story and particularly with respect to the central causal issue. Direct particularistic evidence of the presence or lack of some necessary condition in the causal generalisation at issue is especially significant; indeed, proof of the lack of instantiation of some necessary element in a causal generalisation or one of its underlying causal laws renders inapplicable the causal generalisation at issue. So-called proof by exclusion proves the applicability of a particular causal generalisation not by direct or even circumstantial proof of its instantiation, but rather by proof of lack of instantiation of all the competing causal generalisations. When the coherence of the particularistic evidence with one of the possibly applicable causal generalisations and the overall causal story in which it is embedded is sufficiently greater than its coherence with competing causal generalisations and the causal stories in which they are embedded, according to the subjective judgment of each individual rather than a uniform criterion, that individual forms a minimal belief in the truth of the first story. The greater the difference in degree of coherence, the stronger the degree of belief. This is the method of argument and proof employed by lawyers and judges in actual litigation.66 In the dice-throwing example, at the time of the throwing of the dice the only particularistic evidence that we have is evidence – eg, a video recording or the testimony of an eye­ witness – of the throwing of the dice, which instantiates the single known abstract antecedent condition – ‘throwing a pair of dice’ – in each of the competing causal generalisations and thus does not distinguish among them in terms of coherent fit. However, the subsequent particularistic evidence, from the same or a different video recording or eye­ witness, that a pair of sixes actually fell uppermost, fits only the ‘throwing the dice causes a pair of sixes to fall uppermost’ causal generalisation; it negates, to the extent it is reliable,   Wright, Pruning (n 1) 1050–52.   ibid 1046, 1049–52. See also, eg, RJ Allen, ‘The Nature of Juridical Proof ’ (1991) 13 Cardozo Law Review 373; CJ Miller and RS Goldberg, Product Liability, 2nd edn (Oxford, Oxford University Press, 2004) paras 17.10 and 17.11 (discussing the Privy Council’s employment of case-specific circumstantial evidence to compare competing causal stories in Grant v Australian Knitting Mills [1936] AC 85); N Pennington and R Hastie, ‘A Cognitive Theory of Juror Decision Making: The Story Model’ (1991) 13 Cardozo Law Review 519; RJ Rhee, ‘Probability, policy and the problem of reference class’ (2007) 11 International Journal of Evidence and Proof 286. 65 66

210  Richard W Wright any fit of the available particularistic evidence with the competing causal generalisation(s). The reliability of the video recordings or eyewitnesses will affect the assessment of the degree of fit, but it will do so only by affecting our judgment of the overall coherence of the particularistic evidence with the competing causal stories rather than through some (eg, Bayesian) modification of the ex ante causal probabilities.67 The mathematical probabilists reject the distinctions that the courts and lay persons draw between particularistic evidence and its coherence with the elements of competing causal stories, on the one hand, and ex ante causal probabilities and even naked statistics (which are mere ad hoc distributions not related to any causal generalisation), on the other hand. Indeed, some assert that such distinctions violate principles of rational decisionmaking by ignoring the lessons of mathematical probability theory.68 Much of the debate has focused on Jonathan Cohen’s paradox of the gatecrasher. Cohen hypothesises a situation in which 1000 people attended a rodeo, but only 499 paid for admission, so that 501 were gatecrashers. He further assumes that no tickets were issued and that there is no other way to establish who actually paid and who was a gatecrasher. Given the happenstance that over half of the spectators were gatecrashers, there is a 50.1 per cent naked statistical probability that any particular spectator was a gatecrasher. If the standard of persuasion is interpreted as merely requiring a 50+ per cent statistical probability, then the naked statistics are probative and, indeed, determinative: it supposedly has been adequately proven that every one of the 1000 spectators was a gatecrasher.69 However, no court would hold any of the spectators liable, even if only 50 paid for admission, so that there was a 95 per cent statistical probability that any particular spectator was a gatecrasher. The courts properly refuse to treat the naked statistics as probative.70

67   Michael Green claims that my ‘formalist concern’ about distinguishing between statistical probabilities and particularistic evidence ‘collapses when all we have after a plaintiff develops a disease is an epidemiological study and no other evidence about the plaintiff ’s exposure to competing causes to differentiate the plaintiff from the study subjects’. MD Green, ‘The Future of Proportional Liability: The Lessons of Toxic Substances Causation’ in MS Madden (ed), Exploring Tort Law (New York, Cambridge University Press, 2005) 352, 363. This is a non sequitur. He also offers an alleged counterexample: I wonder how Wright would respond to a case in which a high quality statistical study showed a very high risk – in excess of 90 percent – that someone exposed to an agent would contract the disease from which the plaintiff suffers and in which an eyewitness claims to have observed the plaintiff contract the disease from an alien bite. The point is that most of us would find the ex ante probabilistic evidence far more probative than the ex post particularistic evidence Wright advocates. ibid 362 fn 38. Green’s hypothetical ignores the requirement that particularistic evidence must instantiate an element in a proven or accepted causal generalisation, which he seems to assume is not true here, since that assumption is what gives force to his argument. If there were a proven or accepted alien-bite-causes-this-disease causal generalisation, then we would have two competing causal generalisations, with a non-specified ex ante causal probability for the alien-bite causal generalisation. Even if the ex ante causal probability for the alien-bite causal generalisation were much lower than the 90+ per cent ex ante causal probability for the toxic-agent causal generalisation, the difference in ex ante causal probabilities, as in the dice-throwing example, would only enable us to place a bet on which causal generalisation and its underlying causal laws was actually instantiated; it would not enable us to resolve the bet. The degree of fit of the available particularistic evidence with each causal generalisation is equal, consisting solely in exposure to the relevant causal agent. If, on the other hand, as Green seems to assume, there is no proven or accepted alien-bite causal generalisation and no other possibly applicable causal generalisation, then the only possibly applicable causal generalisation with at least some particularistic instantiation in the particular situation is the toxic-agent causal generalisation, which fact could support the formation of a belief that it was the causal process actually at work in the particular situation. 68   Wright, Pruning (n 1) 1054–62. 69   LJ Cohen, The Probable and the Provable (Oxford, Oxford University Press, 1977) 75. 70   See Wright, Pruning (n 1) 1050 and fn 271.

Proving Causation: Probability versus Belief  211 For the most part, the mathematical probabilists agree that there should not be liability in this type of situation. Many of them heroically insist that there is nothing manifestly unjust about holding defendants liable when the statistical probability is indeed greater than 50 per cent, but they argue that the probability in each instance of this type of situation is actually less than 50 per cent. Allegedly, the fact-finder discounts the objective 50.1 per cent probability because of concern that the plaintiff, in relying on the naked statistics, is concealing, or has insufficient incentive to discover and produce, probative evidence other than the naked statistics.71 This response clearly is inadequate. First, it does not explain why the jury is not even allowed to consider the naked statistics, rather than being allowed to consider and possibly discount the statistics. Second, it does not explain why the plaintiff, rather than the defendant, is being charged with failure to supply other types of evidence. Third, it at least implicitly admits that other types of evidence are more probative than naked statistics; otherwise, why insist on more than the naked statistics? Fourth, it fails to explain why there is no liability even when the objective probability is much higher than 50 per cent – for example, when only 50 of the 1000 spectators paid for their tickets. In this situation, even when the objective 95 per cent probability is discounted, the subjective probability almost certainly is greater than 50 per cent. Fifth, it fails to address the hypothetical as Cohen presented it, which assumes that no other evidence is available.72 The mathematical probabilists assert that naked statistics and ex ante causal probabilities should be taken into account as initial base rates in Bayes’ Theorem to adjust the judgment regarding causation derived from the reliability of the particularistic evidence. For example, in the gatecrasher hypothetical, assume that only one of a thousand spectators paid, the ticket seller identifies A as the one who paid, and there is a 2 per cent probability that the ticket seller is mistaken. Using the naked statistic that 99.9 per cent of the spectators were gatecrashers as the initial base rate in Bayes’ Theorem, the revised probability that A was a gatecrasher, even after taking into account the ticket seller’s particularistic identification testimony, would still be over 95 per cent.73 The mathematical probabilists assert that A almost surely was a gatecrasher, and indeed should be held criminally liable under the ‘beyond a reasonable doubt’ standard as well as civilly liable under the preponderance standard, despite the ticket seller’s 98 per cent reliable testimony that A was not a gatecrasher.74 This conclusion not only is counterintuitive, but also is clearly incorrect. The problem is not Bayes’ Theorem itself, which is a valid mathematical probability theorem. Rather, the problem is the use of abstract base rates, whether naked statistics or ex ante causal probabil­ ities, to impugn the judgment about what actually happened in the particular case based on the particularistic evidence in that case. The mathematical probabilists are mixing apples and oranges. The base rates merely describe the overall distribution of occurrences in a class. They provide no information about any particular occurrence. Only particularistic evidence can 71   ibid 1055 and fn 288, 1057. Judge Richard Posner endorses this ‘missing evidence’ argument in Howard v Wal-Mart Stores Inc (1998) 160 F 3d 358, 359–60 (United States Court of Appeals, Seventh Circuit). Nevertheless, without recognising the reason for its relevance – its status as particularistic evidence that has differing fit with the two possibly applicable causal stories (causation of a spill by an employee versus a customer of the defendant) – Judge Posner relies on an item of particularistic evidence (the absence of the item from which the liquid spilled) to uphold a verdict against the store and in favour of the plaintiff, who was injured when she slipped on the spilled liquid. 72   Wright, Pruning (n 1) 1055–56 and fns 289–294, 1058. 73   ibid 1061 and fn 321. 74   ibid 1061, fn 322.

212  Richard W Wright give us information about the particular occurrence. When the proposition to be proved is what actually happened, using base rates – which provide no information about what actually happened – as initial probabilities in Bayes’ Theorem is an example of the computer programmers’ maxim: ‘garbage in, garbage out.’ The fact that judges, jurors, and lay persons ignore the base rates and instead focus on the particularistic evidence is, contrary to the assertions of the mathematical probabilists, highly rational. The mathematical probabilists have confused the betting odds that a person is willing to accept on the existence of a certain fact with the degree of belief that the person has in the actual existence of that fact. As Jonathan Cohen has persuasively demonstrated with the gatecrasher example and other examples in which paradoxical results are reached using mathematical probability theory, beliefs regarding what actually happened in a particular situation are not determined by statistical or mathematical probability, but rather by abductive consideration of the fit of the relevant particularistic evidence with the com­ peting possible causal stories.75 Additional problems created by attempting to apply mathematical probability theory to causal explanation, rather than causal prediction or ‘postdiction’, are discussed in section IV below.

IV.  Problems and Paradoxes

A.  Indistinguishable Alternative Tortious Causes In both common law and civil law jurisdictions, an alternative causation doctrine is commonly applied in situations like the typical hunting accident situation, in which two or more defendants each fire in the direction of the plaintiff, who was hit by only one pellet, and it is impossible for the plaintiff to prove which defendant fired the pellet that injured him. In order to achieve a second-best just result in situations like this, when each defendant behaved tortiously and may have thereby caused the plaintiff ’s injury but it is impossible for the plaintiff to prove which defendant actually caused her injury, the courts shift the burden to each defendant to prove that she did not injure the plaintiff and hold each defendant who is unable to do so jointly and severally liable for the plaintiff ’s injury.76 However, if there are more than two defendants and the standard of persuasion is satisfied by a mere 50+ per cent mathematical probability, each defendant ordinarily would easily be able to prove that she was not the cause of the injury, even though it is certain that one of them caused the injury. For example, if there were three defendants, each equally likely to have been the cause of the plaintiff ’s injury, each defendant can ‘prove’ that she was not the cause, since there is a 67 per cent probability that she was not the cause, which leads to the paradoxical result that it can be ‘proven’ that none of the defendants was the cause, even though we know that one of them was the cause.   Cohen (n 69) 49–120.   Restatement of the Law Third, Torts: Liability for Physical and Emotional Harm [Restatement Third] (St Paul, American Law Institute, 2010) §28(b) and comments d(1) and e and related reporters’ notes; Restatement (Second) of Torts [Restatement Second] (St Paul, American Law Institute, 1965) ss 433B(2)–(3); C van Dam, European Tort Law (Oxford, Oxford University Press, 2006) 287–88; W van Gerven, J Lever and P Larouche, Cases, Materials and Text on National, Supranational and International Tort Law (Oxford, Hart Publishing, 2000) s 4.4.3; Wright, Possible Wrongs (n 1) 1299–1301. 75 76

Proving Causation: Probability versus Belief  213 Conversely, by employing Mark Geistfeld’s ‘evidential grouping’ argument,77 the mathematical probability interpretation of the standard of persuasion can be employed iteratively to achieve the opposite paradoxical result: ‘proof ’ that all but one of the initial multiple defendants, each approximately equally likely (or unlikely) to have been the actual cause, was the actual cause. At each step in the iteration, a smaller group consisting of most of the defendants in the prior group can be carved out of the prior group, and the mathematical probability interpretation of the standard of persuasion can be used to ‘prove’ that this smaller group contains the defendant who actually caused the plaintiff ’s injury, until we are down to only two defendants, one of whom can be ‘proved’ to have been the actual cause if there is even a ‘scintilla’ of evidence to tip the probability one way or the other – for example, if the shotgun cartridge fired by one of the hunters contained one more pellet than the cartridge fired by the other hunter.78 The courts avoid each of these contrary paradoxical results by refusing to allow such naked statistics as proof or disproof of actual causation. As the reporters for the American Law Institute’s Restatement Third of Torts state (despite their acceptance elsewhere of the mathematical probability interpretation of the preponderance standard79): Defendants would be able to satisfy their burden of production [under the alternative causation doctrine] when three or more defendants are subject to alternative liability [sic] in one of two ways: a defendant might show why it was not the cause of plaintiff ’s injury or it might show which one of the other defendants was the cause.80

To show why she was not the cause or which one of the other defendants was the cause, the defendant must produce evidence of the actual causal effect of another defendant’s shot or the lack of causal effect of her own shot. To do this, she must provide particularistic evidence specific to the particular occasion, rather than mere ex ante causal probabilities or non-causal naked statistics, neither of which provide any information about what actually happened on the particular occasion. When the preponderance standard is properly understood as requiring the formation of a minimal belief in the truth of a disputed fact, based on particularistic evidence specific to the particular occasion, the logical inconsistency that results from using the statistical probability interpretation of the preponderance standard in the alternative causation cases disappears. As Geistfeld states: [T]he plaintiff has provided particularistic evidence showing that each defendant belongs to the group of [possible] tortfeasors that caused the harm, whereas each defendant [using the mathematical probability argument] only relies upon ‘quantitative probability’ or ‘the greater chance’ that the other defendants caused the injury. That evidence, however, is not probative of what actually happened on this particular occasion. . . . To avoid liability, a defendant must instead provide [particularistic] evidence rebutting the plaintiff ’s particularized proof.81

A further paradox would be produced by use of the statistical probability interpretation of the standard of persuasion when the same defendants are repetitively implicated as having possibly caused a particular type of injury. The Supreme Court of Oregon confronted 77   MA Geistfeld, ‘The Doctrinal Unity of Alternative Liability and Market Share Liability’ (2006) 155 University of Pennsylvania Law Review 447, 464–65, 466, 469. 78   See Wright, Possible Wrongs (n 1) 1312 fn 64, 1330, 1332–33. 79   Restatement Third (n 76) §§26 comment l and illus 5, comment n, 28 comment a reporters’ note. 80   ibid §28 comment j reporters’ note at 565 (emphasis added). While causation is alternative in these situations, liability is joint and several rather than alternative. 81   Geistfeld (n 77) 468.

214  Richard W Wright such a situation in a case involving an injurious DPT vaccine that was supplied by one of two defendants, one of which had a 73 per cent share of the market for the DPT vaccine. Literally applying the statistical probability interpretation of the preponderance standard would paradoxically result in its being ‘proven’ that the defendant with the 73 per cent market share, who presumably was responsible for only approximately 73 per cent of the DPTrelated vaccine injuries, caused 100 per cent of those injuries. Although apparently accepting the statistical probability interpretation of the preponderance standard, the Court, referring to articles discussing the ‘naked statistics’ issue, did not allow either of the two defendants to be held liable, even under the alternative causation doctrine.82 The courts in the United States may have faced a similar situation in the cases involving women who developed clear cell adenocarcinoma of the vagina as a result of their mothers’ taking the drug diethylstilbestrol (DES) when pregnant with them. It has been stated that one company, Eli Lilly, may well have supplied, directly or indirectly, more than half of the marketed DES.83 It thus is worth noting how carefully the Supreme Court of California, in the leading American case, phrased its statements on proof of causation of the plaintiff ’s injury. The Court observed that an inference of causation (based on statistical probability) would fail ‘if we measure the chance that any one of the defendants supplied the injurycausing drug by the number of possible tortfeasors’84 (rather than by relative market share). Like the Oregon Court, the California Court also was unwilling to apply the alternative causation doctrine in this type of situation, since doing so would result in each defendant – even those with a small share of the market for the drug – being held fully liable for all of the many DES-related injuries, even though the portion of the injuries that each defendant actually caused presumably approximated its share of the DES market.85 However, unlike the Oregon court, the California Court devised a new second-best liability doctrine in an attempt to have each defendant be liable, approximately, for the share of the total DES-related damages that it presumably actually caused, by holding each defendant proportionately liable in each case for a share of the damages in that case equal to its share of the DES market.86 This seems to me to be a defensible, just result.87 A few other courts have imposed similar liability in the DES cases and some, in both common law and civil law jurisdictions, have imposed more extensive (and thus, in my view, normatively problematic) liability, but all of the courts, including the California court, have done so as a matter of normative policy while recognising that it is impossible to prove who actually caused the plaintiff ’s injury in each case.88 A similar proportionate liability scheme was adopted on similar normative grounds (albeit on a flawed risk contribution rather than probability of causation theory89) by the British House of Lords in the asbestos cases,90 but   Senn v Merrell-Dow Pharmaceuticals Inc (1988) 751 P 2d 215, 216 fn 1, 222 (Supreme Court of Oregon).   See AM Levine, ‘ “Gilding the Lilly”: A DES Update’ (December 1984) 20 Trial 18, 19–20. 84   Sindell v Abbott Laboratories (1980) 607 P 2d 924, 931 (Supreme Court of California) (emphasis added); see ibid at 936–37. 85   ibid 930–31. 86   ibid 936–37. 87   See Wright, Possible Wrongs (n 1) 1326–30. 88   B v Bayer Nederland BV, Hoge Raad 9 October 1992, [1994] Nederlandse Jurispruzentie (NJ) 535 (CJHB); Collins v Eli Lilly Co (1984) 34 NW 2d 37 (Supreme Court of Wisconsin); Martin v Abbott Labs (1984) 689 P 2d 368 (Supreme Court of Washington); see Ewoud Hondius, ‘A Dutch DES Case–Pharmaceutical Producers Jointly and Severally Liable’ (1994) 2 European Review of Private Law 409; R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing, 1999) 75–80. 89   See Wright, Possible Wrongs (n 1) 1295–96; text to n 103 below. 90   Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572. 82 83

Proving Causation: Probability versus Belief  215 was converted into a ‘draconian’ full (joint and several) liability scheme by the British Parliament.91

B.  Toxic Exposures: the Doubling of the Risk Doctrine Although phrases such as ‘more likely than not’ or ‘balance of probabilities’ have long been used when describing the standard of persuasion in civil cases in common law juris­dictions, it is only in fairly recent years that these phrases have frequently come to be understood as mere statistical probability statements. A major locus of this shift in understanding is the toxic tort cases, in which proof often depends on, and often consists solely of, statistical epidemiological evidence. As was discussed in section III above, such evidence is very useful, although neither necessary nor sufficient, in establishing that exposure to a substance is capable of causing a particular kind of injury – the causal capacity or ‘general causation’ issue – and, if such causal capacity has been sufficiently established, in predicting possible results ex ante or comparing possible causes ex post for purposes of remedial treatment. However, such evidence has also incorrectly come to be viewed by many courts, especially in the United States, as being sufficient to prove ‘specific causation’ – the actual occurrence of the relevant causal process on a particular occasion – if exposure to the substance more than doubles, on average or in the aggregate, the frequency of occurrence of that kind of injury, so that it can be said, whenever that kind of injury occurs following exposure to the substance, that the injury was (statistically) ‘more likely than not’ caused by the exposure to the substance.92 Some courts even erroneously believe that such ‘doubling+’ is necessary to establish specific causation.93 As in the indeterminate defendant cases, the statistical probability interpretation of the standard of persuasion produces odd results in the toxic exposure cases. When exposure to a substance more than doubles the risk, the ‘doubling+’ doctrine will result in defendants being held liable for every instance of the injury that occurs following exposure to the substance, even if there is no evidence that the substance actually caused the injury on any particular occasion, and even though exposure to the substance could only have caused a portion of the injuries. For example, if exposure to the substance barely doubles the frequency of occurrence of the injury, so that just over half of the injuries that occur following exposure to the substance are caused by that exposure, defendants nevertheless will be held liable in every case, for all of the injuries. Conversely, when, as is usually the case, exposure to the substance does not more than double the frequency of occurrence of the injury, no defendant will be liable for any of the injuries that occur following exposure to the substance, no matter how many may actually have been caused by such exposure, even though as many as half of the injuries may be due to exposure to the substance. It is remarkable that such a miniscule difference in statistical probability should be thought to result in such a dramatic difference in the supposed proof of specific causation and resulting liability.94  See Sienkiewicz [2011] UKHL 10 (n 24) [57] and [58] (Lord Phillips).  eg Daubert v Merrell Dow Pharmaceuticals Inc (1995) 43 F 3d 1311, 1313–14, 1318–22 (United States Court of Appeals, Ninth Circuit); Marder v GD Searle & Co (1986) 630 F Supp 1087, 1092 (United States District Court, Maryland), affirmed, (1987) 814 F 2d 655 (United States Court of Appeals, Fourth Circuit); Restatement Third (n 76) s 28 comment c(4). 93   See Green (n 67) 365–70; text to n 69 above. 94   Clermont and Sherwin dismiss this objection as an ‘appealing but unsound lay intuition’ that conflicts with the supposed basic goal of minimising erroneous judgments. Clermont and Sherwin (n 2) 252; see ibid 258. As 91 92

216  Richard W Wright Some courts recognise this. As was discussed in section II.B above, a substantial majority of the United Kingdom Supreme Court in Sienkiewicz rejected, albeit in dicta, treating a mere statistical probability, under the ‘doubling of the risk’ doctrine or any other doctrine, as sufficient to establish causation.95 In the United States, ‘numerous jurisdictions have rejected medical experts’ conclusions based upon a “probability,” a “likelihood,” and an opinion that something is “more likely than not” as insufficient medical proof,’ and instead have required that the expert express a ‘reasonable medical certainty’ about the fact at issue.96 Unfortunately, ‘reasonable certainty’ standards are not employed and have no meaning in the medical and scientific communities, so the plaintiff ’s attorney can and often does fill the semantic void, and the plaintiff ’s expert then employs the required terminology.97 Doctors and scientists understand that a mere statistical probability, while useful for prediction and diagnosis, is insufficient to establish what actually happened in a particular case. Thus, if an expert’s opinion regarding actual causation, whether couched in terms of ‘reasonable certainty’, ‘more likely than not’, or ‘preponderance of the evidence’, is based only on a statistical probability (as is usually true in the toxic exposure cases), a good defence attorney will ask the expert, ‘Can you say whether the plaintiff ’s exposure to the [relevant substance] actually caused the [relevant specific harm] in this case?’ The expert – if honest – will reply, ‘No’, and be chagrined for having been made to appear to have contradicted her earlier testimony.

C.  Professional Malpractice: Lost Chances In many common law and some civil law jurisdictions, the courts, applying the statistical probability interpretation of the standard of persuasion, assume that if a doctor’s negligence in diagnosing or treating an ill patient deprived the patient of a 50+ per cent statistical probability of avoiding the injury that subsequently occurred (generally, death), then the doctor’s negligent causation of the injury has been proven; however, if the patient was deprived of a less than 50+ per cent statistical probability of avoiding the injury, the Lord Salmon observed in McGhee v National Coal Board [1973] 1 WLR 1, even odder results occur when the statistical probability interpretation of the standard of persuasion is combined with the ‘but for’ test of actual causation: Suppose . . . it could be proved that men engaged in a particular industrial process would be exposed to a 52 per cent risk of contracting dermatitis even when proper washing facilities were provided . . . [and] that that risk would be increased to, say, 90 per cent when such facilities were not provided. . . . [W]ithout the negligence, it would still have been more likely than not that the employee would have contracted the disease – the risk of injury then being 52 per cent. If, however, you substitute 48 per cent for 52 per cent the employer could not escape liability, not even if he had increased the risk to, say, only 60 per cent. Clearly such results would not make sense; nor would they, in my view, accord with the common law. ibid 12. In Sienkiewicz [2011] UKHL 10 (n 24), Lord Phillips stated that, although he could ‘understand why Lord Salmon considered that to base a finding of causation on such evidence would be capricious’, he could not understand why Lord Salmon ‘considered that to do so would be contrary to common law. The balance of probabilities test is one that is inherently capable of producing capricious results.’ ibid [26]. This lack of understanding was no doubt due to his admitted failure to understand Lord Salmon’s allegedly ‘cryptic’ comment that ‘the approach by the courts below confuses the balance of probability test with the nature of causation’. ibid. 95   See text to nn 41–48 above. 96   Sterling v Velsicol Chemical Corp 855 F 2d 1188, 1200–01 (United States Court of Appeals, Sixth Circuit) (1988); see Restatement Third (n 76) § 28 comment a. 97   Restatement Third (n 76) s 28 comment a and reporters’ note.

Proving Causation: Probability versus Belief  217 defendant’s negligent causation of the injury is not deemed to be proven (indeed, theoretically it is disproven) and the defendant is not liable.98 The same result is reached in German law, on more defensible normative grounds rather than by the invalid assumption of proof of actual causation, through a rule shifting the burden of proof on (lack of) causation to the defendant doctor in cases of gross medical negligence that deprived the plaintiff of a 50+ per cent statistical probability of avoiding the injury.99 Under either approach, all-ornothing liability again arbitrarily turns on a trivial difference in statistical probability. Some courts, in both common law and civil law jurisdictions, while supposedly adhering to the usual requirement that the plaintiff prove the defendant’s tortious causation of the plaintiff ’s injury, but influenced by the statistical probability interpretation of the standard of persuasion and perhaps by the arbitrary distinction under that standard between trivial differences in statistical probability, have been willing to treat any significant increase in risk (or its converse, loss of any significant chance of avoiding the injury) as proof of, or as being equivalent to, actual causation of the injury and thus as supporting holding the defendant fully liable for the injury.100 Other courts, including many courts in the United States, while treating the defendant’s depriving the plaintiff of a 50+ per cent chance of avoiding the injury as proof of actual causation and consequent full liability, acknowledge that causation is not proven when the plaintiff had less than a 50+ per cent chance of survival, but purportedly hold the defendant liable for having caused a newly recognised legal injury, the plaintiff ’s ‘lost chance’ of avoiding the tangible injury that actually occurred, with liability being imposed for a portion of the tangible injury equal to the lost chance.101 The same theory is followed in France whether the probability is greater or less than 50 per cent.102 This approach (except in France) continues to base significant, albeit reduced, substantive differences in liability on mere trivial differences in statistical probability. Moreover, even in France, it erroneously equates statistical probabilities of avoiding the tangible injury with the particular plaintiff ’s probability of avoiding that injury in the particular situation, which will vary depending on his particular genetic makeup and other relevant conditions, and, while claiming to impose liability for having caused the lost chance, it actually imposes liability not for the lost chance – the mere imposition of risk whether or not the risk is realised in a tangible injury – but rather for the plaintiff ’s tangible injury, in the acknowledged absence of proof of causation of that tangible injury. No liability is imposed for the lost chance in the absence of the occurrence of the tangible injury, and the liability that is imposed when there is a tangible injury (which may or may not have been affected or caused by the lost 98   See eg, Kramer v Lewisville Memorial Hospital (1993) 858 SW2d 397, 399–400 (Supreme Court of Texas); Restatement Third (n 76) §26 comment n (assuming proof of causation, and thus full liability, if ‘the probability of a better outcome was in excess of 50 per cent’); van Dam (n 76) 295–97. 99   BGH, 6 October 1998, 1999 NJW 860 (German Federal Court of Justice); see van Dam (n 76) 296. 100   F Stella, ‘Criminal omissions, causality, probability, counterfactuals: Medical-surgical activity’ (http://works. bepress.com/richard_wright/35/) 1–2 (translation by F Stella of ‘Causalità omissiva, probabilità, giudizi controfattuali: l’attività medico-chirurgica’ in F Stella, Il Giudice Corpuscolariano. La Cultura delle Prove (2005) 201–43; Wright, Pruning (n 1) 1067–72. Examples of the widely varying standards that have been employed by the Italian courts with respect to the causation issue in medical malpractice cases are provided in a forthcoming paper by Claudia DiMarzo: Trib Florence 2222 [1999] (‘reasonable certainty’); Cass Civ 11522 [1997] (‘concrete, actual and not hypothetical possibility of a favorable outcome’); Cass Civ 4725 [1993] (‘reasonable certainty about the existence of a not insignificant probability’); Cass Civ 4044 [1994] (‘moral certainty’); Cass Civ 1286 [1998] (‘reliable and significant possibility of a favorable outcome’). 101   Wright, Pruning (n 1) 1067–72. 102   van Dam (n 76) 293–94. The lost chance theory is applied widely in France. ibid. In many other jurisdictions – especially in the United States – it is applied only in medical malpractice cases.

218  Richard W Wright chance) generally is not for the value of the lost chance, however that might be calculated, but rather for the damages resulting from the tangible injury multiplied by the lost chance or the probability of causation.103 Oddly, in the United Kingdom, the lost chance doctrine is applied with respect to financial losses in contractual relationships, including the attorney-client relationship, but not in medical malpractice cases.104 The English position is even odder when one considers that the usual justification for employing the lost chance doctrine or some other second-best liability rule is the inherent impossibility of proving causation.105 In legal malpractice cases, the trial that did not occur or that was botched can be relitigated – ’a trial within a trial’, as occurs in the United States – to determine whether the plaintiff would have won or lost; there is no need for a lost chance doctrine. This is not true in medical malpractice cases. The British House of Lords’ decision in the Hotson medical malpractice case106 is a dramatic example of the perverse results that can be caused by the statistical probability interpretation of the standard of persuasion when combined with a myopic focus on the ‘but for’ test of actual causation.107 The plaintiff fell from a tree and ruptured some of the blood vessels in his left femoral epiphysis. The defendant’s negligent delay in diagnosing and treating his injury caused a swelling of the epiphysis that compressed the remaining intact blood vessels and thus shut off the supply of blood from those blood vessels. As a result of the combined loss of blood from the initial fall and ruptures and the subsequent compression of the remaining blood vessels, the epiphysis became distorted and deformed, resulting in permanent injury to the boy’s left hip and leg. However, the trial court determined that there was a 75 per cent chance that the permanent injury would have happened anyway even if the defendant had not been negligent, owing to the loss of blood from the ruptured blood vessels. Focusing on this statistical finding, the House of Lords held that the defendant was not liable due to a lack of (‘but for’) causation, which as a past fact is determined by the ‘balance of probabilities’.108 Adding insult to injury, the Court further held that there could be no recovery for any lost chance: ‘In determining what did happen in the past a court decides on the balance of probabilities. Anything that is more probable than not it treats as certain’.109 Unfortunately, the Hotson court failed to recognise that it was dealing with a situation involving cumulative contribution to a possibly overdetermined indivisible injury. As both Lord Phillips and Lord Rodger noted in Sienkiewicz, while still not recognising that Hotson was such a situation, in such situations the ‘but for’ test produces the wrong answer; instead in the United Kingdom (and some other countries) a ‘test’ of ‘material contribution’ is used.110 Although the plaintiff most likely would have suffered the permanent injury anyway, the defendant’s negligence, by causing the loss of the blood supply from the intact blood vessels, contributed to the aggregate loss of blood that caused the permanent injury (and may have been a ‘but for’ cause of this happening earlier than it otherwise would  See Kramer (n 98) 402, 405; Wright, Possible Wrongs (n 1) 1295–96.   van Dam (n 76) 294–95; see Gregg v Scott [2005] UKHL 2, [2005] 2 AC 176, in which, however, the plaintiff had only suffered a reduction in the chance of avoiding the relevant injury, not the injury itself. 105   See Wright, Possible Wrongs (n 1) 1295–97. 106   Hotson v East Berkshire Area Health Authority [1987] 1 AC 750 (HL). 107   See also n 94 above. 108   See Wright, Possible Wrongs (n 1) 1322–23. 109   Hotson (n 106) 785 (Lord MacKay) (quoting Mallett v McMonagle [1970] AC 166, 176 (Lord Diplock)); accord, Hotson (n 106) 792 (Lord Ackner). 110   Sienkiewicz [2011] UKSC 10 (n 24) [17], [27], [75], [90], [93] (Lord Phillips); ibid [138], [144]–[146] (Lord Rodger). 103 104

Proving Causation: Probability versus Belief  219 have), just as stabbing a person who more likely than not already has been stabbed sufficient times to bleed to death, but who still has a significant amount of blood left and several hours to live, contributes to that person’s bleeding to death (and may have been a ‘but for’ cause of the death happening earlier than it otherwise would have).111 Contrary to the statements of Lord Brown in Sienkiwicz,112 the problematic issues with which the British and other courts have had to deal are not the result of a failure to adhere strictly to the ‘but for’ test in every situation, but rather the result of a failure to recognise the inadequacy of the ‘but for’ test as a comprehensive test of actual causation and an understandable but erroneous literal interpretation of the misleadingly named ‘more likely than not’ and ‘balance of probability’ standards of persuasion.

D.  Res Ipsa Loquitur and ‘Prima Facie’ Presumptions The last doctrine to be discussed is the res ipsa loquitur doctrine, which seems to exist in every jurisdiction even though it sometimes is known by a different name (eg as the ‘prima facie’ evidence doctrine in Germany).113 The res ipsa loquitur doctrine, as usually stated, allows an inference of negligent causation of the plaintiff ’s injury by the defendant if (1) in the type of situation that existed the injury ordinarily does not occur unless there is negligence and (2) other possible responsible causes are sufficiently eliminated by the evidence.114 Although condition (2) may put some limitations on the scope of the doctrine, depending on how it is interpreted,115 the doctrine, as stated, constitutes a departure from the ordinary substantive liability and proof rules of a magnitude that is not commonly appreciated. The doctrine allows an inference of negligent conduct by the defendant, and a further inference that the inferred negligence caused the plaintiff ’s injury, based on a mere statistical probability. If, in the aggregate, 50+ per cent of the occurrences of this type of event are caused by negligence, then negligent causation by the defendant can be inferred without any specific evidence of negligence or causation by the defendant or anyone else on the particular occasion. Thus, contrary to what is commonly stated,116 it is not true that res ipsa loquitur is merely a fancy Latin name, employed in the particular context of proving the defendant’s causal negligence, for the ordinary use of circumstantial evidence to make a straightforward factual inference. Circumstantial evidence is concrete evidence specific to the particular 111   See RW Wright, ‘Acts and Omissions as Positive and Negative Causes’ in JW Neyers, E Chamberlain and SGA Pitel (eds), Emerging Issues in Tort Law (Oxford, Hart Publishing, 2007) 287, 295–97, 299. 112   [2011] UKSC 10 [176], [178], [181], [183], [186] (Lord Brown): see ibid [167] (Baroness Hale). Lord Brown attempts to explain cases in which liability was based on ‘material contribution’ rather than ‘but for’ causation as not actually being exceptions to the ‘but for’ test, on the ground that they involved, or should have involved, apportionment of damages based on relative contribution. ibid [176]. He has confused apportionment of damages with the prior issue of causation, for which the ‘but for’ test fails to give the right answer in situations actually or possibly involving overdetermined injury. See Wright (n 49) II, III.D, III.E. 113   Murray and Stürner (n 15) 311–12; van Dam (n 76) 1107; van Gerven, Lever and Larouche (n 76) s 4.2.3 at 428/16–17 (http://www.casebooks.eu/documents/tortLaw/heading4.2.3.pdf). 114   eg Restatement Second (n 76) § 328D(1). 115   See Wright, Possible Wrongs (n 1) 1338–40. 116   eg WL Prosser et al, Prosser and Keeton on Torts, 5th edn (St Paul, West Publishing Co, 1984) s 39 at 243–44 and fn 20; Restatement Second (n 62) s 328D comment a. But see DB Dobbs, The Law of Torts (St Paul, West Publishing Co, 2000) s 154 at 372 (noting, correctly, that res ipsa loquitur cases differ ‘overwhelmingly’ from ordinary circumstantial evidence cases by allowing an inference of negligence (and causation) without any particularistic evidence of negligence (or causation) on the particular occasion).

220  Richard W Wright occasion about the network of instantiated causal relationships leading to and flowing from the particular factual issue being litigated. For example, a person’s running away from the scene of a murder immediately after it happened with blood on her that matches the victim’s blood and holding a knife, the blade of which matches the victim’s stab wound, is strong circumstantial evidence that she stabbed the victim. The inference of negligence allowed by the res ipsa loquitur doctrine as stated by many courts, interpreted literally, does not require any such case-specific evidence of what actually happened on the particular occasion, but rather only abstract statistical data (or assumptions) on what usually (50+ per cent of the time) has happened in such situations. The difference in the validity of the inference depending on whether or not the word ‘ordinarily’ is included parallels the distinction between the admissibility of habit evidence (allowed) and character evidence (generally not allowed) to prove what a person did on a particular occasion.117 The common failure to appreciate the extraordinary nature of the res ipsa loquitur doctrine is probably attributable to an assumption that the word ‘ordinarily’ in the first condition is simply an incorporation of the ‘preponderance of the evidence’, ‘more likely than not’, or ‘balance of probability’ standard of persuasion, which is interpreted as merely requiring a 50+ per cent statistical probability. Once again, however, taking this interpretation seriously immediately raises a logical contradiction. Why, if the first condition is satisfied, is the inference that someone was negligent only a permissive one, rather than being required? Why, in the absence of any contrary evidence by the defendant, allow the trier of fact not to draw the inference once the conditions for drawing the inference have been established, especially since this permits inconsistent verdicts by different juries in similar situations, which is a denial of formal justice? The reason, I suspect, is a discomfort with the broad formulation of the doctrine, especially when there is a conscious realisation that it permits an inference of negligence by the defendant based merely on a statistical probability. Allowing the trier of fact not to draw the inference may be an implicit concession that she should be able to draw the inference or not depending on whether she actually believes the defendant was causally negligent in the particular situation. But if the existence of an actual belief is the concern, the broad formulation should be abandoned in favour of the narrow one (with the word ‘ordinarily’ omitted), or at least the trier of fact should be instructed that an inference of negligent causation should be drawn only if evidence specific to the particular case combines with the ‘ordinarily would not happen’ statistical probability to raise a minimal belief that the defendant actually was negligent in the particular situation and that such negligence contributed to the plaintiff ’s injury. On the other hand, if the broad formulation is meant to provide a second-best (or third-best) resolution of the factual uncertainty regarding negligent causation, it seems that decision should be consistently implemented through a rebuttable presumption.

  eg KS Broun et al (eds), 1 McCormick on Evidence, 6th edn (St Paul, West Publishing Co, 2006) ss 186, 188,

117

195.

11 Liability for Future Harm ARIEL PORAT AND ALEX STEIN

I. Introduction

The House of Lords and the United States Supreme Court have recently addressed the issue of liability for a risk of future illness. This issue arose in connection with employees wrongfully exposed by their employers to asbestos. As a result of this wrongful exposure, the employees exhibited symptoms indicating an increased risk of developing fatal cancer diseases in the future. Both decisions have assumed as a common ground that this risk per se is not actionable in torts.1 This presupposition moved the focus of the Law Lords’ and the Justices’ attention from real harm to the ethereal damage. The issue adjudicated in both instances was whether the plaintiffs are entitled to recover from their employers compensation for the mental anguish caused by the fear of developing cancer. The American case, Norfolk & Western Railway Company v Ayres,2 involved railroad workers who sued their employer under the Federal Employers Liability Act. The suit attributed to the defendant liability for emotional distress suffered by the plaintiffs as a consequence of contracting asbestosis. Each plaintiff contracted this disease following his exposure to an impermissibly hazardous quantity of asbestos, for which the defendant was unquestionably responsible. The plaintiffs’ emotional distress (fear and anxiety) originated from a worrying statistical fact: about ten per cent of the people in a medical condition similar to theirs develop mesothelioma – a fatal cancerous disease – at some point in the future. The Supreme Court’s narrow (5-4) majority decision allowed recovery to plaintiffs whose distress was proven to be ‘genuine and serious’. Note again that recovery was allowed not for the increased risk of contracting cancer, but rather for the plaintiffs’ fear of becoming afflicted in the future. The Court’s decision relied on both policy and doctrine. As a policy matter, the Court underscored the problematics of the two-disease requirement that the defendant asked it to interpose: this requirement would deny compensation to an asbestosis sufferer who never develops cancer. As a result, this sufferer’s fear of contracting cancer at some point in the future – an undeniably harmful consequence of the defendant’s wrongdoing – would never be actionable in torts. As far as doctrine is concerned, courts across the United States have long recognised a tort victim’s right to sue for any serious emotional harm resulting from her present physical injury. Any such harm is part of the   This assumption is more explicit in the British than in the American decision. See below.   Norfolk & Western Railway Company v Ayres 538 US 135 (2003) [Ayres].

1 2

222  Ariel Porat and Alex Stein victim’s ‘pain and suffering’. This parasitic actionability is a deeply rooted common law principle. Allowing the plaintiffs to recover compensation for their asbestosis-induced fear therefore does not break away from the common law tradition. The dissenters’ objection to this decision alluded (inter alia) to its socially deleterious consequences. According to the dissent, the Court’s compensatory generosity might drain the funds available for compensating real asbestos victims for their physical injuries. These injuries – held the dissent – are more immediate and more serious than the disruption of the plaintiffs’ peace of mind. The British case, Johnston v NEI International Combustion Ltd,3 involved plaintiffs who developed pleural plaques (fibrous thickening of pleural membranes surrounding the lungs). Pleural plaques are generally benign. Nor do they cause any asbestos-related diseases. Those plaques, however, indicate the presence of asbestos fibres in the person’s lungs and pleura. Asbestos, in turn, may independently cause a cancerous disease – a grim prospect upon which the plaintiffs based their suit. According to the plaintiffs, anxiety and emotional distress associated with this prospect constitute compensable harm. The House of Lords disagreed. It held that a risk of future illness and the attendant anxiety are not actionable in torts as a stand-alone harm. The Law Lords clarified, however, that the parasitic actionability doctrine will still allow recovery in appropriate cases. That is, a person who sustains compensable injury can recover damages for the fear that the injury will develop into a more serious harm in the future. This factor makes Johnston the doctrinal equivalent of Ayres. The two decisions, however, are not completely identical to each other. Ayres contains no express pronouncement on whether a risk of future harm can ever become actionable as a stand-alone damage.4 Johnston, in contrast, holds unequivocally that such risks are not actionable in torts. The Law Lords’ adherence to the parasitic actionability doctrine was unambiguous and unqualified.5 In the pages ahead, we conduct a normative exploration of this issue. We examine the desirability of a rule that imposes negligence-based liability for a risk of future illness even when there is not any physical harm in the present. Our desirability criteria are efficiency and justice. Based on these criteria, we articulate the reasons for making a stand-alone risk of future illness actionable in torts. We argue that a person wrongfully exposed to such risk should be entitled to probabilistic compensation that matches his prospect of becoming ill. This compensation should equal the harm resulting from the illness multiplied by the illness’s probability. After demonstrating the efficiency and justice of this proposal, we discuss counterarguments. We first consider the objections to the probabilistic recovery rule as applying in cases of indeterminate causation. Those cases involve physical injury that may or may not have been caused by the defendant’s negligence (‘past-injury cases’). We show that arguments against probabilistic recovery in past-injury cases do not hold with respect to risks of future injury. We then discuss two additional objections to our proposal. One of them holds that small risks of future illness should not be actionable: allowing people to sue for such risks would be more costly than beneficial. We agree with this point and limit our proposal to suits involving substantial risk of illness.   Johnston v NEI International Combustion Ltd [2007] UKHL 39, [2008] 1 AC 281 [Johnston].  See Ayres (n 2) 153 (Justice Ginsburg opinion): ‘But the asbestosis claimants did not seek, and the trial court did not allow, discrete damages for their increased risk of future cancer’. 5   Johnston (n 3) 10, 17, 32, 38, 40–41. 3 4

Liability for Future Harm  223 A much stronger objection to our proposal identifies a serious problem of collective action. Our proposal would allow tort victims to choose between immediate compensation for the risk of future illness and the ‘wait and see’ strategy. Victims who choose to wait and see would either become ill or avoid the illness. Victims who ultimately avoid the illness would not be able to sue the defendant. The remaining ill victims would become eligible to full compensation for the harm suffered. Alas, those victims’ ability to recover compensation would depend on the defendant’s solvency. If the defendant becomes insolvent, it would not compensate those victims. The defendant’s prospect of remaining solvent would crucially depend on the victims’ initial choice. If many victims choose to recover the immediate probability-based compensation, the defendant’s funds may shrink to a degree that will deny compensation to wait-and-seers. Every victim would anticipate this contingency. The victims, however, would not be able to coordinate their suits because the required coordination is too costly to establish and enforce. Absence of coordination and the dilutedfund prospect would prompt all victims to opt for the instant probability-based recovery. This problem is real. We believe, however, that it can be resolved by courts or through governmental intervention and offer two such solutions. Structurally, our argument unfolds in three sections. In section II, we carry out a comparative analysis of Ayres and Johnston. In section III, we analyse the virtues and vices of probabilistic recovery in past-injury cases (also identifiable as cases of indeterminate causation) and relate this analysis to the future illness problem. In section IV, we make out the case for making risks of future illness actionable in torts and develop a number of suggestions on how to operationalise this proposal. A short conclusion follows.

II.  Ayres v Johnston

A.  Ayres In Ayres,6 the United States Supreme Court addressed the question of whether an employee wrongfully exposed to asbestos at work is entitled to recover from his employer compensation for the fear of developing cancer. The plaintiffs suffered from asbestosis caused by a work-related excessive exposure to asbestos. About 10 per cent of the people suffering from this disease develop fatal cancer. Five Justices allowed recovery for the plaintiffs’ fear. Specifically, they affirmed the plaintiffs’ recoveries for pain and suffering – mostly fearrelated7 – ranging from $500,000 to $1,200,000.8 The dissent opined that no such recovery should be allowed. Justice Ginsburg, writing for the Court, held that mental anguish resulting from a person’s fear of developing cancer in the future is compensable under the Federal Employers’ Liability Act9 as part of a successful plaintiff ’s entitlement to compensation for pain and suffering.10 She set two cumulative conditions for a plaintiff ’s eligibility for this compensation.   Ayres (n 2).   Justice Kennedy emphasised that the plaintiffs have sustained no significant harm other than this fear. See Ayres (n 2) 179–80. 8   The lower recovery amounts were awarded to smokers owing to their comparative negligence. Ayres (n 2) 179. 9   Federal Employers’ Liability Act 45 USCS § 51 and following. 10   Ayres (n 2). 6 7

224  Ariel Porat and Alex Stein First, the plaintiff must prove by a preponderance of the evidence that his fear is genuine and serious.11 Second, the plaintiff must preponderantly establish that this fear results from an illness for which the defendant is responsible.12 Justice Kennedy spoke for the dissent. According to him, the plaintiffs’ suits were causatively weak. First, the alleged fears could not be considered a direct consequence of the plaintiffs’ disease.13 Second, there was no dependable scientific proof of the causal nexus between the plaintiffs’ disease and cancer.14 Third, the alleged harms were speculative and quantitatively insignificant, given the presence of a significant asbestos-unrelated risk of contracting cancer (for smokers, in particular).15 Justice Kennedy also alluded to social policy. According to him, making bare fears of cancer compensable would exhaust the financial resources that defendants could use for compensating asbestos victims who actually develop cancer.16 Justice Breyer took a middle-ground approach. He agreed with the Court that a tort victim’s fear of a future illness should be actionable when it is genuine, serious, and originates from a proven disease. Yet, according to him, compensation for such fear should be generally unavailable when the following conditions are present: (1) actual development of the disease can neither be expected nor ruled out for many years; (2) fear of the disease is separately compensable if the disease occurs; and (3) fear of the disease is based upon risks not significantly different in kind from the background risks that all individuals face.17 Justice Breyer clarified that these compensation-denying rules are not meant to be rigid: ‘This is not to say that fear of cancer is never reimbursable. The conditions above may not hold. Even when they do, I would, consistent with the sense of the common law, permit recovery where the fear of cancer is unusually severe – where it significantly and detrimentally affects the plaintiff ’s ability to carry on with everyday life and work.’18

B.  Johnston In Johnston, as in Ayres, the claimants were exposed to excessive quantities of asbestos by their employers. The claimants, however, contracted no recognisable diseases (with the exception of one plaintiff who developed clinical depression). Instead, they developed pleural plaques – a condition not amounting to a physical impairment or disablement. Not being a disease, pleural plaques indicate the presence of asbestos fibers in the person’s lungs. This contamination puts the person at risk of developing cancer in the future.19 11   Ayres (n 2) 157. Another issue was whether damages should be apportioned among multiple tortfeasors in a way that would allow the defendant to reduce its compensation duty to the plaintiff. The Justices have reached a unanimous opinion that the Federal Employers’ Liability Act allows no such apportionment, ibid 159–66, but the defendants could still bring ‘indemnification and contribution actions against third parties under otherwise applicable state or federal law.’ ibid 162. 12   Ayres (n 2) 157. 13   ibid 171–72. 14   ibid 173. 15   ibid 179. 16   ibid 168–69. 17   ibid 187. 18  ibid. 19   As attested in Johnston (n 3)[80], ‘The Claimant is at risk of future development of asbestosis (1%), diffuse pleural thickening (1%) and mesothelioma (5%) and as a result suffers anxiety for his future health and welfare’.

Liability for Future Harm  225 This risk may instill in the person fear of death and related anxieties. Alternatively, it may exacerbate a person’s preexisting fear and anxiety. The claimants in Johnston had developed such anxieties. The House of Lords consequently had to decide whether this mental and emotional anguish constitutes compensable damage under the negligence doctrine.20 The Law Lords decided that it does not. This holding was both categorical and unanimous (it also extended to the claimant with a fully-blown depression). The Law Lords dismissed the claimants’ argument that a combination of non-actionable plaques, a nonactionable risk of illness and a non-actionable anxiety constitutes actionable harm.21 As Lord Scott explained, this argument fails ‘because [n]aught plus naught plus naught equals naught’.22 The Law Lords clarified, however, that presence of an actionable injury – asbestosis being an example – would make the attendant anxiety actionable as well.23

C.  Ayres, Johnston, and Liability for Future Harms We now attempt to extrapolate the two courts’ attitude toward liability for a wrongfully imposed risk of illness. We leave behind the anxiety-related causes of action and focus instead on the risk of illness as compensable harm. We distinguish, as the courts did, between two scenarios. In the first scenario, the risk of illness originates from the plaintiff ’s actionable injury, for which the defendant is responsible. In the second, the plaintiff sustains no actionable injury and sues the defendant for the risk of harm as a stand-alone damage. Under the first scenario, the Law Lords in Johnston would impose liability for the risk. As Lord Scott explained: [I]t is common ground that if some physical injury has been caused by the negligence, so that a tortious cause of action has accrued to the victim, the victim can recover damages not simply for his injury in its present state but also for the risk that the injury may worsen in the future.24

Ayres did not address this issue. Under the second scenario, the Law Lords in Johnston would impose no liability whatsoever. As already indicated, they held unanimously that a stand-alone risk of future harm does not constitute actionable harm.25 The Ayres decision did not address this issue expressly,26 but it seems that also American law would preclude liability for a stand-alone risk of future illness.

20   The Law Lords parenthetically mentioned the possibility of a contract-based action, but made no decision on that matter since it was not argued by the plaintiffs. Johnston (n 3) [59], [74]. 21   Johnston (n 3) [89]. 22   ibid [73]. 23   ibid [67]. 24  ibid. 25   See section I.A. 26   See n 4 and accompanying text.

226  Ariel Porat and Alex Stein

III.  Past Harm

A.  Paradigmatic Cases The problem of indeterminate causation arises when the defendant may or may not be the cause of the harm wrongfully inflicted on the plaintiff.27 In the next few paragraphs, we present the paradigmatic cases in which this problem arises. Consider a case in which several wrongdoers, acting independently of each other, negligently expose the plaintiff to a risk of sustaining harm. The plaintiff consequently faces several independent risks of sustaining harm. One of those risks materialises and the plaintiff suffers harm. This harm was brought about by one of the wrongdoers, whose identity is unknown. In this type of cases – labelled ‘unidentifiable wrongdoer’ – both English and American courts allowed recovery. They held that the wrongdoers are jointly and severally liable for the plaintiff ’s damage.28 As an alternative to this remedy, courts could have awarded prorated recovery: an award that equals the plaintiff ’s harm multiplied by each defendant’s probability of being the cause of the harm (the ‘probabilistic recovery principle’, or PRP, for short). Courts, however, almost uniformly declined to adopt PRP in ‘unidentifiable wrongdoer’ cases.29 This policy cannot be explained by the judges’ mistrust of mathematical probabilities, because they do something very similar to applying PRP when they apportion damages among jointly liable tortfeasors (and also between tortfeasors and their victims). The decision to reject PRP and hold wrongdoers jointly and severally liable for the victim’s entire damage is best explained by the courts’ desire to afford the victim maximal protection against wrongdoers’ insolvency. Another paradigmatic case features a wrongful risk-creator and a natural cause as two mutually exclusive explanations for the victim’s injury. Neither of these causal explanations can be confirmed or ruled out. Consider a doctor who negligently misdiagnosed a patient, thereby reducing the patient’s chances to recover. The patient ultimately did not recover, but this outcome is attributable to his pre-existing medical condition for which the doctor is not responsible. Arguably, because of this condition, the patient would not have recovered even if the doctor had diagnosed him properly. Assume that the probability of this 27   In other cases that we discussed elsewhere, the defendant caused part of the plaintiff ’s harm, but the exact amount of that part could not be established. See Ariel Porat and Alex Stein, Tort Liability Under Uncertainty (Oxford, Oxford University Press, 2001) 76–83, 201–06; Ariel Porat and Alex Stein, ‘Indeterminate Causation and Apportionment of Damages: An Essay on Holtby, Allen and Fairchild’ (2003) 23 OJLS 667. 28   Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32 (in a subsequent decision of the House of Lords it was assumed that Fairchild left undecided the question of whether the defendants should be either jointly and severally liable to the claimant for the entire indivisible disease or whether they should be severally liable according to their relative contribution to the risk. See Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572. The Barker decision adopted the second approach. See n 29); Summers v Tice (1948) 199 P 2d 1 (Cal 1948). Some American courts awarded this remedy even against defendants whose negligence was unproven. See eg, Ybarra v Spangard 154 P2d 687 (Cal 1944). This decision, however, is an outlier. 29   The main American exception is the ‘Market Share Liability’ doctrine: see Sindell v Abbott 607 P 2d 924 (Cal 1980); Hymovitz v Eli Lilly Co 539 NE 2d 1069 (NY 1989). The House of Lords applied PRP in a case where the harm to the plaintiff was the result of either a wrongful exposure to asbestos caused by his employer or of a nonwrongful exposure that occurred when the plaintiff was self-employed. See Barker v Corus (UK) plc (n 28). The legislature, however, overturned this ruling by introducing s 3 into the Compensation Act 2006 (c 29). This section only applies to mesothelioma victims.

Liability for Future Harm  227 argument, raised by the doctor, is 50 per cent or higher, which means that the patient is unable to establish his causal allegation against the doctor by a preponderance of the evidence. We identify cases exhibiting these characteristics as ‘lost chance’ cases. Under the traditional ‘winner takes all’ approach, to which English courts adhere in such cases,30 the patient would recover no compensation whatsoever. Under PRP, in contrast, the patient would recover compensation that equals his harm multiplied by the probability of the causal claim that attributes this harm to the doctor’s negligence. In the United States, several courts have taken this route, while others have refused to do so.31 In a recent decision, the House of Lords considered the adoption of PRP for ‘lost chance’ cases, but decided against it by a 3-2 majority.32 Another illustrative case involves a group of employees suffering from a disease that can be equally attributed to the employees’ predisposition to develop the disease and to their exposure to asbestos wrongfully caused by their employer. While it can be established by a preponderance of the evidence that a certain percentage of the employees – typically less than 50 per cent of them – suffering from the disease contacted it because of the wrongful exposure to asbestos, it cannot be established who those employees are. We define cases falling into this category as ‘unidentifiable victim’ cases. Under the traditional ‘winner takes all’ approach that applies in England and in the United States, the court must dismiss all employees’ suits against their employer. Under PRP, the employer must compensate all employees for the harm they suffered multiplied by the probability of the claim that the harm was caused by the employer’s wrongdoing.33

B.  The Vices of Probabilistic Recovery PRP offers an attractive solution to the problem of indeterminate causation that arises in past-injury cases. Under this principle, victims recover partial compensation for their possibly wrongful injuries and wrongdoers pay for their misdeeds rather than go scot-free. On many occasions, PRP is preferable from an efficiency perspective because it promotes optimal deterrence of wrongdoers. The principle is also appealing as a retributive device because wrongdoers are required to pay for their wrongs even in the absence of a clear-cut causal nexus between those wrongs and the victims’ harms (some theorists do not consider harm a sine qua non prerequisite of liability34). These virtues, however, come accompanied with a number of vices that some consider fatal. One of those vices, which we discuss first, is epistemological. All others are (interchangeably) operational and moral. As a factual matter, in each of the aforementioned cases, the defendant had either caused the plaintiff ’s entire harm or no harm at all. The defendant certainly did not cause part of the harm. Consequently, all decisions that PRP prompts courts to deliver would be  See Hotson v East Berkshire Area Health Authority [1987] AC 750 (HL); Gregg v Scott [2005] 2 AC 176 (HL).   For applications of PRP see Herskovits v Group Health Cooperative of Puget Sound 664 P 2d 474 (Wash 1983). For other approaches taken by American courts, see Porat and Stein, Tort Liability Under Uncertainty (n 27) 73–76. 32   Gregg v Scott (n 30). 33   Porat and Stein, Tort Liability Under Uncertainty (n 27) 70–73, 193–95. See Barker v Corus (n 28). Even though this case deals with a bit different scenario, it may signal the emergence of a different approach to PRP in general, and, specifically, to the principle’s application to ‘unidentifiable victim’ cases. 34   See CH Schroeder ‘Corrective Justice & Liability for Increasing Risks’ (1990) 37 UCLA Law Review 439. For a retributivist account of harm as a pre-requisite for tort liability, see Jeremy Waldron, ‘Moments of Carelessness and Massive Loss’ in David G Owen (ed), Philosophical Foundations of Tort Law (Oxford, Clarendon Press, 1995) 387. 30 31

228  Ariel Porat and Alex Stein factually incorrect. This comprehensive incorrectness is a serious epistemological vice that also creates tensions with corrective justice.35 Application of PRP would also present moral and operational hurdles. Courts would have to make complicated determinations of probabilities, instead of deciding cases on a ‘more probable than not’ basis. That would increase litigation costs. In addition, PRP would allow courts to hold defendants liable on the basis of naked statistics. This liability format is contentious from a moral point of view.36 Relatedly, PRP authorises courts to impose liability for bare risks. This authorisation is morally problematic as well: it clashes with the deeply entrenched moral intuition that considers the actual infliction of harm to be a prerequisite for imposing liability in torts.37 According to this intuition, pure endangerment cannot be a reason for holding its creator liable. From the deterrence perspective, PRP might be operationally redundant, as the ‘winner takes all’ rule works fine in deterring wrongdoers across cases. Under this rule, suits that are not preponderantly probable fail completely. This failure erodes deterrence because some (and possibly many) of the failed suits are meritorious. The ‘winner takes all’ rule, however, also provides that suits with a probability greater than 0.5 are unqualified winners. Plaintiffs recover full compensation even when their claims are only slightly more probable than not. This provision increases the expected amount of compensation for prospective wrong­ doers and compensates for the erosion of deterrence on the other side. Furthermore, PRP breaks away from the minimal proof requirement – preponderance of the evidence – set by the law as a condition for authorising courts to extract payments from defendants. This requirement is far from technical: it derives from a more general principle of evenhandedness that limits the courts’ power to force transfers of money and property from one individual to another. Under this principle, a court must not order such a transfer unless the reasons for forcing it out are better than the reasons for preserving the status quo. In the domain of fact-finding, the required ‘better reasons’ must be present in the probability of the plaintiff ’s case. When the plaintiff ’s claims are not more probable than the defendant’s, the court has no reasons for changing the status quo.38

C. Refinements In this section, we outline a number of refinements in the conditions for applying PRP. Those refinements help identify cases in which the principle’s vices fade away, partially or even completely. These cases call for PRP’s application because the principle’s attractiveness is no longer offset by its vices. One refinement is a temporal distinction between two different categories of cases. Both categories involve causal indeterminacy. One of those categories accommodates cases in which the indeterminacy related to a past occurrence and was present all the time. Cases falling into another category are different. In those cases, causal indeterminacy attaches to 35   See SR Perry, ‘Risk, Harm, and Responsibility’ in David G Owen (ed), Philosophical Foundations of Tort Law (Oxford, Clarendon Press, 1995) 321. But see M Stauch, ‘Causation, Risk, and Loss of Chance in Medical Negligence’ (1997) 17 OJLS 205, 217–18, for the claim that ruling according to chances in past-injury cases is not contrary to the deterministic causation approach, since such a ruling does not make any claim about the state of the world, but, rather, about the state of our knowledge about the world. For a different view, see H Reece, ‘Losses of Chances in the Law’ (1996) 59 MLR 188. 36   Charles R Nesson, ‘Reasonable Doubt and Permissive Inferences: The Value of Complexity’ (1979) 92 Harvard Law Review 1187, 1225. 37   Perry (n 35). 38   Alex Stein, Foundations of Evidence Law (Oxford, Oxford University Press, 2005) 219–25.

Liability for Future Harm  229 a hypothetical or future event that can be characterised as a past occurrence only at a later point in time.39 This distinction eliminates the epistemological vice. Consider a wrong­ doing that has just been perpetrated and assume that the prospective victim did not yet suffer the harm associated with that wrongdoing. At this point in time, the prospective victim can only complain about his wrongful exposure to a risk of sustaining the harm. His future prospects became blurred (or more blurred than previously) owing to the unwelcome imposition of the new risk of harm. The wrongdoing thus made the victim’s future prospects less desirable and, consequently, less valuable than before.40 Note that this assessment of the victim’s ex ante situation will remain unmodified even if at a later stage he sustains the anticipated harm without being able to associate it with the wrongdoing. The worsened-prospect description of the prospective victim’s situation is therefore empirically correct for the point in time at which the wrongdoing took place. Compare this scenario with a different setup, in which the plaintiff suffers harm immediately after the defendant’s wrongdoing. Assume that the probability of the allegation that the defendant’s wrongdoing caused the harm is 0.5 or less, and that the plaintiff consequently cannot establish that the defendant caused his harm. This setup features no exposure-to-risk point along its timeline. There is no point in time (save for points in time that are completely artificial) at which the value of the plaintiff ’s future prospects decreases. The worsened-prospect description therefore does not fit this setup. As a result, any verdict other than no-liability or full-liability would be empirically unfounded. This ‘all or nothing’ setup is present in a case in which an obstetrician negligently induces labour and the baby he delivers dies. Assume that the baby’s death could result from the obstetrician’s negligence, or, alternatively, from the acute respiratory problem which could not have been overcome even if the obstetrician had taken due care. In this scenario, there was no point of time at which the baby’s prospects of survival decreased by less than 100 per cent. Those prospects were either unaffected or completely eliminated by the obstetrician’s negligence. Another important refinement of PRP relies on a differentiation between shortages of evidence. In some cases, courts need scientific information about causation, but this information is not available. In other cases, courts do not have enough case-specific evidence for deciding the case. In the first category of cases, causal indeterminacy is uniformly present in the fact-pattern of every case. In the second category, it presents itself uniquely in each individual case.41 Applying PRP in the first category of cases is more straightforward than in the second. When the uncertainty problem is general, some wrongdoers systematically escape liability. This happens when the probability of causation is 50 per cent or less and plaintiffs are systematically denied remedies under the preponderance standard. This effect is present in unidentifiable-wrongdoer cases, in unidentifiable-victim cases, and in many of the lost-chance cases. In those cases, the redundancy objection to PRP does not hold. Under the ‘winner takes all’ rule, all defendants – many of whom wrongfully damaged their plaintiffs – are held not liable. This outcome is both unjust and inefficient. To avoid this outcome, the legal system needs to implement PRP. This need makes PRP crucial rather than redundant. For similar reasons, one cannot easily object to PRP by alluding to the problematics of naked statistical evidence. When uncertainty is present in every case and case-specific 39   The House of Lords analysed this distinction in Gregg v Scott (n 30) at 181-83, without determining its implications. 40   See section III for further discussion of cases where the victim’s future is blurred with uncertainty at the time of the trial. 41  See Gregg v Scott (n 30) 184.

230  Ariel Porat and Alex Stein evidence is systematically scarce, adherence to the rule against naked statistics is a logical equivalent of a rule that absolves all defendants – many of whom are wrongdoers – from liability in torts. Adoption of PRP is the only plausible way to avoid this unjust and inefficient result. When causal indeterminacy is recurrent in a well-defined category of cases, litigation costs triggered by the PRP are far from being crucial. Indeed, under PRP, courts would have to deal with mathematical probabilities and rely on experts more often than under the extant regime. Courts, however, would also develop expertise and amass information about causes, effects and probabilities. These knowledge and expertise would be applicable in many cases. This economy of scale would partially offset the increase in litigation costs. More crucially, in cases where uncertainty systematically allows wrongdoers in a specific field to escape liability, the inefficiency that PRP’s adoption would avert outweighs any foreseeable increase in litigation cost. Last, the torts system needs to pay special attention to cases in which the same wrongdoer recurrently causes harm. These cases are qualitatively different from those featuring wrongdoing as a unique or sporadic event. This distinction is important because in cases falling into the recurrent wrongdoer category, virtually none of the objections to PRP is valid. Consider the well-known case of DES, a drug designed to prevent miscarriages that was manufactured by hundreds of companies, mainly in the ’50s and turned out to be latently carcinogenic to female foetuses. Twenty-five years later, numerous young women whose mothers had taken the drug were diagnosed with uterine cancer. It was found by the courts that the drug had not been tested adequately prior to its marketing and that the manufacturers had failed to take into account certain findings that had pointed to a risk of carcinogenic effect. Furthermore, the plaintiffs’ mothers had never been cautioned against this risk. Finally, the drug had been marketed under a generic rather than brand name, which foiled attempts to trace each pill back to its actual manufacturer.42 For the purpose of providing a remedy to the victims, courts developed the Market Share Liability doctrine (MSL). Under this doctrine, first adopted by the California Supreme Court in Sindell,43 every defendant manufacturer was to assume liability for the plaintiff ’s harm unless it could prove (by a preponderance of the evidence) that it did not manufacture the drug taken by the plaintiff ’s mother. As the Sindell court clarified, this liability had to be imposed only on those manufacturers who produced a substantial proportion of DES in the relevant market. The court ultimately decided that the burden of compensating each plaintiff for her damage would be allocated amongst the manufacturers in accordance with their respective shares in the DES market.44 In the DES cases, recovery was probabilistic: the market share of each defendant substituted for the probability that the litigated harm was caused by that defendant. But since for each defendant the wrongdoing was a recurring event, in the long run, MSL made each defendant liable for the harm it actually caused. Indeed, each defendant was obligated to compensate other defendants’ victims, but at the end of the day, after all claims have been satisfied, the final outcome was the same as in a case featuring no causal indeterminacy whatsoever.45  See Sindell v Abbott, 607 P2d 924 (Cal 1980).  ibid.   This decision can be understood as imposing liability on each defendant for all of the plaintiff’s damage (in which case, equitable apportionment of the compensation burden could be achieved through wrongdoers’ indemnification claims against each other) or, alternatively, as imposing liability on each defendant for a prorated part of that damage. See Brown v Superior Court, 751 P2d 470, 485-87 (Cal 1988) (adopting the second interpretation of Sindell). 45   We assume that none of the defendants becomes insolvent. 42 43 44

Liability for Future Harm  231 The DES scenario is far from being the only case in which PRP would make the wrongdoers internalise the harm they actually caused. Unidentifiable-victim cases often exhibit the same characteristic. Take the asbestos case discussed above, in which many employees were negligently exposed to asbestos by the same employer. In this case, PRP would make the employer pay money damages that correspond to the harm it caused. This socially desirable effect will be achieved by allowing a compensatory distortion. Because each of the injuries was either caused in its entirety by the employer’s wrongdoing or by a natural cause, none of the employees will receive compensation for the harm wrongfully inflicted upon him or her. But this distortion is not a serious problem. Under any plausible criterion of justice and efficiency, allowing each employee to recover partial compensation while holding the employer liable for the harm it wrongfully caused is better than leaving all employees uncompensated and letting the employer go scot-free. All the objections to PRP fade away when the principle is applied to recurrent-­ wrongdoer cases. The epistemological objection vanishes because wrongdoers eventually pay for the harms they actually cause. Moreover, in many (but not all) cases, victims of torts are compensated for their actual harms (DES cases are the prime example of this category of cases). The objection alluding to the minimum proof threshold weakens as well because there would be no significant departures from the proof requirements that ordinarily apply in civil cases. Under PRP, the plaintiff must prove by a preponderance of the evidence that the defendant committed a recurrent wrongdoing. The plaintiff also must preponderantly establish the approximate amount of the aggregate harm resulting from the defendant’s wrongdoing. Arguments criticising PRP for establishing liability for bare risks become inapplicable, too. In cases involving recurrent wrongdoers, defendants would not be paying for bare risks. Rather, they would pay for the harms they actually caused. Finally, when recurring-wrongdoer cases systematically feature probability of causation of less than 50 per cent, and case-specific evidence is scarce, also the naked statistics, the redundancy and the litigation costs objections lose most of their power. The table below summarises the applicability of the different objections to PRP:

Type of Event

Type of Uncertainty

Frequency of Wrongs

Past

Future/ Hypothetical

Case Specific

Inherent

Discrete

Recurrent

Epistemologial Problem

A

N/A

A

A

A

N/A

Litigation Costs

A

A

A

N/A

A

N/A

Naked Statistics

A

A

A

N/A

A

N/A

Liability for Bare Risk

A

A

A

A

A

N/A

Redundancy

A

A

A

N/A

A

N/A

Minimum Threshold

A

A

A

A

A

N/A

A = applicable. N/A = non-applicable.

232  Ariel Porat and Alex Stein

D.  Quantum of Damages In the paragraphs below, we distinguish between two types of probability-based compensation: 1. forward-looking compensation for the risk of future injury; and 2. backward-looking compensation, based on the probability of causation.46 Take a person who sustains injury after being wrongfully exposed to a risk of sustaining that injury. Before the wrongdoing, this victim’s probability of sustaining the injury equaled 1-p (eg, 0.25), which is parallel to her probability of remaining uninjured (p) (ie, 0.75). After the wrongdoing, the victim’s probability of sustaining the injury became 1-q (eg, 0.75), which parallels her probability of escaping the injury (q) (ie, 0.25). Because the victim actually sustained the injury, her case falls into the 1-q category (ie, the 0.75 category). This statistical category comprises two jointly exhaustive and mutually exclusive scenarios that reflect the victim’s initial position. In the first scenario, the victim sustains the injury irrespective of the wrongdoing. Under this scenario, the victim was doomed to sustain the injury, so that the wrongdoing made no impact on her well-being. As already indicated, the probability of that scenario equals 1-p (ie, 0.25). In the second scenario, it is the wrongdoing that causes the victim’s injury. Under this scenario, the victim would have remained uninjured had she not been exposed to the wrongdoing. The probability of this scenario equals (1-q)-(1-p), that is, p-q (ie, 0.50). This ex ante probability represents the reduction in the victim’s chances of remaining uninjured, as effected by the wrongdoing. Now consider the probability of the scenario that the wrongdoing was the actual cause of the victim’s injury. This probability is represented by the fraction of scenarios featuring a victim who could not sustain her injury without being subjected to a wrongdoing in the more general cluster of cases that feature an injured victim, a wrongdoing, and the exhaustive variety of causal factors that could inflict the same injury on the victim. The above fraction of scenarios equals p-q (ie, 0.50). The cluster of cases covering all possible scenarios equals 1-q (ie, 0.75). The ex post probability of the scenario in which the wrongdoing actually inflicts the victim’s injury therefore equals (p-q)/(1-q) (ie, 0.50 / 0.75 = 2/3). As we already mentioned, the victim’s risk of sustaining injury as a result of the wrongdoing equals p-q (ie, 0.50). So long as the victim lives under the risk of becoming injured, his expected damage therefore equals (p-q)D (ie, D/2), when D denotes the average amount of harm suffered by similarly situated victims. This expected damage is what the victim facing the risk of injury or illness should recover from the wrongdoer. Victims who actually become ill should be compensated differently. Using the same notation and the same numerical example as before, let p (ie, 0.75) and q (ie, 0.25) denote, respectively, the victim’s chances of remaining uninjured before and after the wrongdoing. Allow D to denote the average amount of damage that the wrongdoing inflicts in the long run of cases, and let T denote the total number of cases in which the tortious activity takes place. The ideal compensation that the legal system should exact from the wrongdoer would thus equal (p-q)DT (ie, DT/2). In reality, however, only injured victims can successfully sue the wrongdoer. Therefore, the number of cases in which the wrongdoer would have to pay compensation would equal   This discussion is based on Porat and Stein, ‘Indeterminate Causation’ (n 27) 685–88.

46

Liability for Future Harm  233 (1-q)T (ie, 0.75T). If the wrongdoer compensates each injured victim at the amount of (p-q)D (ie, D/2), then the wrongdoer’s compensation duty would fall below the optimal. Using the probability of causation as an award-multiplier would eliminate this shortfall. As already established, the probability of causation equals (p-q)/(1-q) (ie, 2/3). Each injured victim’s compensation would consequently be set at [(p-q)/(1-q)]D (ie, 2/3·D). The total amount of the wrongdoer’s compensation duty would then be [(p-q)/(1-q)]DT(1-q), ie: (p-q)DT (ie, DT/2). This compensation duty equals the losses inflicted by the wrongdoer.47 American courts tend to conflate the two types of compensation by awarding injured victims the risk-based, rather than probability-based, amounts.48 The Restatement of the Law Third, Torts: Liability for Physical and Emotional Harm proposes to fix this error by differentiating, as we do, between the two types of awards.49

IV.  Future Harms

In numerous cases in which one person acts wrongfully towards another, the prospective victim faces a continuous risk of illness that may or may not materialise in the future. Employees are exposed to the risk of contracting an occupational disease due to unsafe working conditions; residents of a polluted neighbourhood face the risk of becoming ill as a result of their exposure to pollution; consumers of defective products have a prospect of contracting diseases from those products; victims of medical malpractice face the risk of developing an affliction or handicap; and there are many other examples. Should faulty creators of such risks assume liability in torts for the ensuing prospect of future illness?50 This question is puzzling. Why not wait and see what happens in the end? If 47   Note that when the probability of causation is greater than 0.5, the court may decide to apply the ‘winner takes all’ rule and award the plaintiff full compensation. 48   For examples of this mistaken approach in the United States, see Wendland v Sparks 574 NW2d 327, 333 (Iowa, 1998) (an oft-cited decision analogising the value of lost chances to that of a lottery ticket); Mays v United States 608 F Supp 1476 (DC Colo 1985), revd on other grounds 806 F2d 976 (10th Cir 1986), cert den 482 US 913 (upon finding that malpractice reduced the patient’s chances of recovery from 40 to 15 per cent, the court reasoned that the damage related to net pecuniary loss caused by the medical centre was 25 per cent of the $173,200 total net pecuniary loss, or $43,300); Herskovits v Group Health Cooperative of Puget Sound 664 P2d 474 (Wash 1983) (holding a 14 per cent reduction, from 39 per cent to 25 per cent, in the decedent’s chance for survival as sufficient evidence to allow the case to go to the jury); Alberts v Schultz, 975 P2d 1279, 1287 (NM 1999) (if medical malpractice reduced the patient’s chance of recovery from 50 to 20 per cent, that patient’s compensation would be equal to 30 per cent of the value of his or her life); Jorgenson v Vener, 616 NW2d 366, 372 (SD 2000) (if instead of completely eliminating the chance of recovery, the physician’s negligence merely reduced the chance of recovery from 40 per cent to 20 per cent, then the value of the lost chance would be 20 per cent of the value of a complete recovery); Smith v Washington 734 NE2d 548 (Ind 2000) (affirming an award of 50 per cent of the patient’s damage upon finding that the defendant’s malpractice increased the patient’s risk of incurring an already likely injury from 50 per cent to 100 per cent). For reasons provided above, the claimant should have recovered 29 per cent of the damage in Mays; 19 per cent of the damage in Herskovits; 37.5 per cent of the damage in the Alberts example; and 25 per cent of the damage in the Jorgenson example. In Smith, the outcome was correct because the defendant’s malpractice totally eliminated the claimant’s chances of recovery. Otherwise, the court’s adherence to the lottery analogy would have generated an error (as it did in our previous examples). 49   See Restatement of the Law Third, Torts: Liability for Physical and Emotional Harm § 26, cmt n. 50   For a theory favouring imposition of liability for a future disease when the plaintiff proves that the wrongdoing has more than doubled her risk of contracting the disease, see Andrew R Klein, ‘A Model for Enhanced Risk Recovery in Tort’ (1999) 56 Washington & Lee Law Review 1173. This theory justifies its ‘doubling the risk’ requirement by claiming that ‘there should be no pre-manifestation recovery for possible illness that a plaintiff cannot eventually connect to a defendant’s conduct using normal measures of actual causation’. ibid 1195.

234  Ariel Porat and Alex Stein the risk materialises into harm, its creator should then be obligated to compensate the victim; and if the victim remains unharmed, he should receive no compensation at all. Note that statutes of limitations would not deny compensation to victims who ultimately sustain harm because those victims’ causes of action accrue only after the occurrence of the harm.51 The repose provisions, in contrast, would yield a different result. These provisions render all suits non-actionable after a specified post-transgression period. The fact that the victim’s harm had developed only after the statutory deadline is immaterial. This fact can toll a limitations period, but not a statute of repose because repose provisions have a special goal: to reduce the volume of litigation by denying actionability to suits in which the defendant’s wrongdoing and the plaintiff ’s harm are separated by a long period of time.52 This policy purposefully extinguishes all suits alleging deferred illness (or other latent injuries) that are filed after the expiration of the repose period. Whether prospective victims should be allowed to bypass this policy by suing wrongdoers for their prospects of illness consequently becomes a big question. This question concerns the limits that the law should impose upon suits for future harm, while our goal here is to determine whether such suits should ever be allowed to proceed. We therefore leave this question open. The Law Lords in Johnston have upheld the rule that allows compensation for a future illness that might develop from the plaintiff ’s present injury, for which the defendant is responsible.53 The prospect of future illness can thus be perceived as attaching to the present injury and becoming part of the plaintiff ’s present physical condition. Alternatively, this prospect instills in the plaintiff fear and anxiety that attach to his present pain and suffering. In both cases, the prospect’s attachment is a legal move that creates a fusion between the plaintiff ’s present condition and future possibilities. This move is artificial because in the empirical world no attachment actually occurs. Theoretically (and more than just theoretically), even with a present injury, the court could tell the plaintiff: ‘Wait and see what, if anything, happens with your risk. If it materialises, come and see us; and if not then not’. But why not abandon the artificial devices that operate in this area of the law? Why not allow a tort victim to recover compensation not only for the risks attaching to her existing injury but also for a free-standing risk of future illness? We now turn to discussing this question.

A.  The Case for Liability The case for making risks of future illness actionable in torts is straightforward. A person’s prospect of becoming seriously ill erodes his well-being. Consider two people who happen to be equal in all respects except one: one of those people has a prospect of developing a serious illness in the future, while the other has no such prospect. The second person’s wellbeing outscores the well-being of the first person (if forced to live one of those people’s lives, a rational individual would prefer to be the second person rather than the first). Assume now, that the first person did not come upon this misfortune by himself. His prospect of becoming seriously ill resulted from exposure to a toxic substance by a negligent wrongdoer. We argue that under such circumstances, the law should allow the prospective   See eg, Baird v American Medical Optics, 713 A2d 1019 (NJ 1998).   See eg, Rudenauer v Zafiropoulos, 445 Mass 353 (2005).   Johnston (n 3) [67].

51 52 53

Liability for Future Harm  235 victim to choose between immediate recovery of compensation for his expected harm and a postponed entitlement to recover full compensation in the event of illness (for the sake of simplicity, our ensuing discussion assumes that once the illness develops, the court can verify with sufficient certainty that it was caused by the wrongdoing). The victim should be able to recover from the wrongdoer compensation for the wrongfully imposed risk. The amount of this compensation equals the harm associated with the illness multiplied by the victim’s probability of becoming ill owing to the wrongdoing. That amount typically represents the victim’s increased cost in purchasing health or life insurance due to his or her wrongful exposure to the risk. After paying this compensation to the victim, the wrongdoer will become immune from further liability. If the victim ultimately becomes ill, he would not be allowed to collect from the wrongdoer the difference between the expected and the actual harm. Alternatively, the victim should be able to wait and see whether he actually becomes ill. If he becomes ill, the wrongdoer would have to fully compensate him for the harm suffered. If the victim develops no illness, however, the wrongdoer would pay nothing. The victim should be allowed to choose between these remedies for a number of reasons. From the wrongdoer’s perspective, the two remedies are economically identical. Therefore, he has no legitimate reasons to oppose their substitution by one another. Assume that the wrongdoer exposed 10 victims to a 10 per cent risk of sustaining physical harm in the amount of $1,000,000. Regardless of whether victims recover immediate compensation for their expected harm or are compensated in the future, if and when their harm materialises, the wrongdoer’s expected liability would be the same: $1,000,000. Equally important, this liability also reflects the wrongdoing’s social cost. Indeed, under our proposal, each victim would choose between immediate compensation in the amount of $100,000 and future compensation in the amount of $1,000,000, if harm materialises. This choice would not change the wrongdoer’s expected liability. Yet, it might change the wrongdoer’s liability de facto. To illustrate, assume that nine victims out of ten recover $100,000 each, while the tenth victim chooses to wait and see. Subsequently, this victim sustains the harm and recovers $1,000,000. In this scenario, the wrongdoer ends up paying 1.9 million dollars rather than one million. But that possibility should not bother us. If our proposal is adopted, the insurance market would allow the risk-averse wrongdoer to insure against the risk of paying the ninth victim $1,000,000. The cost of this insurance would be around $100,000.54 As we already explained, the wrongdoer has no legitimate reasons upon which to base a claim that those two remedies are not identical to each other. The wrongdoer, for example, cannot invoke its hypothetical prospect of becoming insolvent in the future as a reason for not compensating the victim for his presently expected harm. Nor can the wrongdoer benefit from the victim’s prospective inability to furnish evidence (typically, a long time after the wrongdoing’s occurrence) that would causally relate his illness to the toxic exposure (given the presence of competing causes to which the victim would be exposed before suing). The wrongdoer also should not be allowed to take advantage of the victim’s prospect of becoming fatally ill – a condition that may extinguish the victim’s ability and motivation to wage a legal battle.

54   We thank Lord Hoffmann for raising this issue when this chapter was presented at the Aberdeen University symposium on causation.

236  Ariel Porat and Alex Stein All those prospects entail a risk of injustice for the victim: they indicate that denial of the victims’ right to be compensated in the present for the risk of future harm may erode their entitlement to compensation if and when their physical harm materialises. Those prospects also dilute deterrence: wrongdoers expecting to undercompensate their victims in the future will be inefficiently under-deterred. The erosion-of-compensation prospect thus does not merely offset the wrongdoer’s complaint about being held liable for bare risk. This prospect is an affirmative reason for imposing such liability. The proposed liability system would bring about additional social benefits. First, when the future harm is a fatal disease, expedited compensation would allow victims to use the money they get during their lifetime.55 Second, some victims might be able to use their expedited compensation towards mitigation of the risk of harm. For example, a victim might be able to undergo extensive medical tests and obtain preventive treatments that reduce her prospect of illness. Victims would also be able to change places of residence, work and lifestyles. Finally, some of the victims would also be able to purchase life and medical insurance. The cost of this insurance would increase as a result of the wrongdoing – yet another independent reason to have the wrongdoer pay for it.56

B. Objections We now return to the objections raised against PRP as applied to past-injury cases. We consider the validity of those objections in the present context and find them unpersuasive. Subsequently, we consider some additional – more powerful – objections to our proposal. The epistemic objection to probabilistic recovery falls apart once it is recognised – as an empirical matter – that a person’s risk of becoming ill in the future erodes her well-being. This erosion is self-evident: risk of illness is an unquestionably unwanted condition. Indeed, it is epistemically rational for a person to buy insurance against such risks or to avoid them by paying a steep price (for example, by exercising and eating healthy, but untasty, food). The relationship between risk of future harm and people’s well-being in the present is straightforward when the risk is for future property harm. In this case, the present market value of the property will be discounted by the risk of harm even when the risk is below 50 per cent. For example, if a wrongdoer created a 30 per cent risk that the victim’s house will collapse in the future, this risk would certainly reduce the house’s market value. Admittedly, this analogy is far from perfect because, unlike real property and commodities, life and limb are not tradable. Yet, as we have demonstrated, risks to a person’s life and health constitute detraction from her well-being. Those risks bring into the person’s life deleterious economic and non-economic consequences.

55   For an argument in the same vein, in another context, see David Friedman, ‘What is “fair compensation” for death or injury?’ (1982) 2 International Review of Law and Economics 81. 56   Even under a regime that imposes no liability for future harms, courts should allow recovery for money and efforts expended by victims on reducing such risks. For principles governing such awards, see Mark Geistfeld, ‘The Analytics of Duty: Medical Monitoring and Related Forms of Economic Loss’ (2002) 88 Virginia Law Review 1921 1940–49; Kenneth S Abraham, ‘Liability for Medical Monitoring and the Problem of Limits’ (2002) 88 Virginia Law Review­1975; John CP Goldberg and Benjamin Zipursky, ‘Unrealized Torts’ (2002) 88 Virginia Law Review 1626. Furthermore, even when a tort victim sustains no physical injury, courts should allow her to recover compensation for fear of future illness and for the increased cost of her health and life insurance. See Porat and Stein, Tort Liability Under Uncertainty (n 27) 121.

Liability for Future Harm  237 Litigation costs do not present a serious problem either. These costs should be considered money well spent, given the benefits that the proposed system would yield. Courts would also be able to reduce those costs by relying on information generated by the market for health and life insurance. Based on this information and expert testimony, courts should be able to evaluate risks and damages both adequately and expediently. Courts also should not be reluctant to base wrongdoers’ liability for future illnesses upon naked statistical evidence. This evidence is often the only proof that the plaintiff can furnish. Her inability to adduce case-specific evidence is hardly a good reason for allowing the defendant to go scot-free, especially when the defendant’s wrongdoing is recurrent. The objection against liability for bare risks is equally unpersuasive. We do not discuss here the pros and cons of a legal regime that imposes liability for risks, as opposed to harms. We have done it elsewhere.57 As a general matter, arguments against liability for bare risks lose much of their validity when there is uncertainty as to whether the plaintiff was or will be harmed by the defendant’s wrongdoing.58 Furthermore, as we already explained, in cases involving risk of future harm, the plaintiff would normally be able to establish that this risk already constitutes harm for him or her. The redundancy objection to our proposal would insist that a prospective victim should wait and see whether she actually becomes ill. As we already explained, however, there are good reasons for allowing the victim to choose between immediate recovery of compensation for her expected harm and the ‘wait and see’ option. Allowing the wrongdoer to dictate this choice to the victim would defeat justice and efficiency at once. Actionability of future illnesses thus cannot be considered redundant. Finally, the minimum-threshold objection should not bother us at all. Under our system, risk of future illness would be actionable as a matter of substantive law. To succeed in an action for such risk, the plaintiff would have to prove its nature and extent by a preponderance of the evidence. The plaintiff would also have to preponderantly establish that the risk originated from the defendant’s wrongful action. The conventional principles of proof would thus stay intact. Typically (albeit not always), the uncertainty as to whether the plaintiff will develop illness in the future is inherent in the entire category of cases, as opposed to being present due to the absence of case-specific evidence. As we explained above,59 in such cases, the objections to PRP fade away. Furthermore, as we have shown in relation to past-injury cases,60 the recurrent character of the defendant’s wrongdoing should be taken into account as well. When a repeat wrongdoer compensates each victim for her probabilistic harm, he ultimately pays the victims the right amount of compensation. The same mechanism will work well in future-harm cases when the defendant is a repeat wrongdoer. Two additional objections against our proposal have more merit. These objections help set the limits to our proposed liability regime. One of those objections focuses on the magnitude of the risk of illness. Sometimes, many people are exposed to a risk of illness, but only a few of them actually become ill. Consider a polluter who exposes 1,000,000 residents to a small risk of a serious illness. Damage associated with that illness equals $1,000,000 and its probability is 1:100,000 for each resident. That means that 10 out of 1,000,000 residents would suffer significant injury   ibid 103–15.   ibid 116–29. 59   Above text accompanying notes. 60   Above text accompanying notes. 57 58

238  Ariel Porat and Alex Stein at some point in the future. For obvious reasons, allowing each of 1,000,000 residents to sue the polluter for his or her expected harm in the amount of $10 makes no sense. Small risks of future illness should not be actionable. We agree with this limitation to our proposal. Subject to this limitation, however, whenever there is enough information for evaluating victims’ expected harm, prospective victims should be allowed to sue wrongdoers for risks of future illness.

C.  The Problem of Collective Action The remaining objection to our proposal identifies a serious problem of collective action. Suppose that victims can choose between immediate recovery for risks and future recovery for harms. For reasons discussed earlier, some victims would prefer immediate recovery. Their consumption of this remedy, however, will diminish the wrongdoers’ resources and increase the risk that the wrongdoers would be unable to meet their obligations to victims who sue for materialised harms. A similar problem troubled Justice Kennedy in Ayres, who disagreed with a rule that allows recovery for fears of contracting cancer in the future. Justice Kennedy’s reasons for disagreeing with this rule included the prediction that fear-based recovery of compensatory awards would exhaust the financial resources that defendants could use for compensating asbestos victims who actually develop cancer.61 But the problem at hand may be even more severe. If some potential victims (Group A) sue the wrongdoer for their risks of future illness – and they may have perfectly good reasons for filing those suits – other potential victims (Group B), who originally preferred to sue only in the event of illness, would find themselves in a different position. The increased risk of the wrongdoer’s insolvency might prompt some of those victims to sue immediately. This choice will snowball: any additional suit – that is, any victim migration from Group B to Group A – will further increase the risk of the wrongdoer’s insolvency. As a result, none of the potential victims would wait. All of them would migrate from Group B to Group A and sue the wrongdoer immediately. As in the ‘Tragedy of the Commons’, the ensuing flood of suits might exacerbate the wrongdoer’s insolvency to the detriment of all victims. This problem calls for a regulatory solution. One such solution is to authorise courts to deal with the prospect of the wrongdoer’s insolvency on a case-by-case basis. If this prospect is insignificant, the court should allow the plaintiff to sue the defendant for the risk of future illness. If the prospect is real, the court should stay the proceeding and require the plaintiff to substitute her suit by a limited-fund class action.62 Another solution is to set up a statutory fund to which wrongdoers would have to contribute sums that equal the amount of their victims’ expected harm (a Pigouvian tax). This solution is particularly suitable for cases involving recurrent wrongdoers.

  Ayres (n 2) 168–69.   The goal of this proceeding is to make sure that all plaintiffs get compensated and that early filers of suits do nor exhaust the defendant’s funds. See eg, Ortiz v Fibreboard Corp 527 US 815 (1999). 61 62

Liability for Future Harm  239

V. Conclusion

Probability-based compensation is a controversial remedy that departs from the established legal tradition. This remedy is fairly common, but still controversial, in the United States. Courts in England and other jurisdictions use it on rare occasions. Most cases in which this remedy was used – or could be used, if courts did not feel reluctant about it – involved causally indeterminate harm that the plaintiffs had already sustained. Under such circumstances, even a partial success of the plaintiff ’s suit clashed with conventional wisdom. This wisdom requires courts to dismiss as unproven any suit that fails to satisfy the preponderance-of-the-evidence requirement. An attempt to bypass this requirement by awarding plaintiffs probability-based compensation consequently becomes suspicious. This suspicion accounts for the tort system’s widespread refusal to treat risk of future illness as actionable harm. Another reason that explains this refusal is a general belief that future harm, without more, is identical to bare risk – an actuarial, rather than concrete, endangerment for which courts should impose no liability. This essay questioned the validity of these two analogies. We hope to have shown that these analogies are untidy, if not altogether invalid. We also hope to have demonstrated that there are good reasons for treating serious risks of future illness or injury as actionable harm. If we are right, courts that persistently refused to allow probabilistic recoveries in past-injury cases should reconsider their position with regard to future harm.

12 Scientific and Medical Evidence in Causation Decisions: The Australian Experience IAN FRECKELTON SC

I.  A Taxonomy of Difficulty

Causation determinations by the courts usually rely heavily upon expert evidence. They can be a major challenge, and a source of error, for fact-finders (whether judge or jury) in both civil and criminal cases. One of the principal reasons for this is the difficulties encountered generally, and in many countries, in relation to the evaluation of complex and conflicting expert opinions. Some of the problems experienced are particular to adversarial systems of justice; others are not. However, the challenges of fact-finding in cases involving expert evidence are often amplified by the complexities of fact,1 law2 and social policy3 which lie at the heart of causation determinations which may be based upon subtly differing expert perspectives on abstruse and recondite issues of science, medicine and engineering, to name but some subjects4, or on the provision of limited guidance of key issues of law in respect of causation evaluation. These problems are the more potentially productive of appealable error or miscarriage of justice when expert evidence about causation is in one way or another defective, imprecise or ambiguous. Surveys of judges and magistrates published in 1999 and 20015 showed that Australian judicial officers harboured significant levels of concern about the potential for error in cases in which expert evidence figured prominently. They identified particular consternation about psychiatric and accounting evidence and articulated worries about unacceptable levels of partisanship amongst too many experts who appeared in front of them.6 They specified a variety of generic problems in relation to expert opinion evidence (not confined 1   See AC Becht and FW Miller, The Test of Factual Causation in Negligence and Strict Liability Cases (St Louis, Washington University, 1961); J Manning, ‘Factual Causation in Medical Negligence’ (2007) 15 Journal of Law and Medicine 337. 2   See L Green, The Rationale of Proximate Cause (Kansas City, Vernon Law Book Co, 1927). 3   See eg, Sullivan v Moody (2001) 207 CLR 562, [49]. 4   See eg, I Freckelton and H Selby, Expert Evidence: Law, Practice, Procedure and Advocacy 4th edn (Melbourne, Thomson, 2009). 5   I Freckelton, P Reddy and H Selby, Judicial Perspectives on Expert Evidence: An Empirical Study (Melbourne, The Australasian Institute of Judicial Administration, 1999); I Freckelton, P Reddy and H Selby, Magistrates’ Perspectives on Expert Evidence: A Comparative Study (Melbourne, The Australasian Institute of Judicial Administration, 2001). 6  See Dasreef Pty Ltd v Hawchar [2011] HCA 11 at [58].

242  Ian Freckelton SC to evidence about causation), including being troubled about: a culture of partiality amongst some expert witnesses; a failure on the part of some witnesses to depart in their professional opinions from their ordinary areas of discourse for forensic exigencies; deficient methodologies and reasoning processes adopted by some experts; and limitations on the part of too many experts in their understanding of their forensic role and on the part of some litigation lawyers in relation to commissioning experts, eliciting their views orally in court and in making them accountable for their expressed opinions. Cases in which proof of causation on the basis of expert evidence have proved problematic for plaintiffs in civil litigation and the prosecution in criminal matters, have highlighted the following features: •  bizarre or unscientific expert evidence; •  expert evidence lacking in objectivity; •  expert evidence not easily amenable to evaluation and where cross-examination has not succeeded in its objective of rendering the evidence accountable; •  expert evidence where the facts are misunderstood; and •  expert evidence in the context of the law being in a state of flux. These scenarios exist across a variety of litigation and in many countries.7 For the purpose of this chapter, however, examples from Australian case law are drawn upon particularly to illustrate the issues posed by the proof of causation on the basis of expert evidence. An attempt is made to explore some of the dynamics and dilemmas which have the potential to lead to appealable error and to miscarriages of justice in the context of causation determinations.

II.  Australian Law Concerning Causation

Australian law relating to causation has made its way through a number of iterations. It is arguable that causation is being increasingly recognised by appellate judgments as a threshold criterion for success in tort litigation, without which there is little point in evaluating the content of duties of care and whether they were breached.8 In the High Court’s decision in March v Stramare (E & MH) Pty Ltd  9 the reality of multiple causative factors was acknowledged.10 Mason CJ held that ‘Generally speaking, . . . causal connection [in the sense of establishing that wrongful conduct caused or materially contributed to injury] is established if it appears that the plaintiff would not have sustained his or her injuries had the defendant not been negligent’.11 The Court embraced a test of

7   See H Beebee, C Hitchcock and P Menzies (ed), The Oxford Handbook of Causation (Oxford, Oxford University Press, 2009). 8   See eg, CAL No 14 Pty Ltd v Scott (2009) 239 CLR 390; D Mendelson, The New Law of Torts, 2nd edn (Oxford University Press, Melbourne, 2010) 430. 9   [1991] HCA 12, (191) 171 CLR 506. See S Yannoulidis, ‘Causation in the Law of Negligence’ (2001) 27 Monash University Law Review 319. 10   ibid [13] per Mason CJ. 11   ibid [16].

Expert Evidence as to Causation  243 application of ‘common sense’12 to the facts of the particular case,13 rejecting the proposition that the question to be asked is a straightforward ‘but for’ inquiry, in part because of its unhelpfulness when there is more than one event that would be sufficient for an injury, as well as its inadequacy in the context of a superseding event. However, the Court accepted that the ‘but for’ test had a continuing role to play in the resolution of factual causation. It was acknowledged that causation had both factual and policy components. Justice Kiefel later, in 2010,14 summarised the common law: The ‘but for’ test is regarded as having an important role in the resolution of the issue of causation, although more as a negative criterion than as a comprehensive test. The resolution of the question of causation has been said to involve a common sense idea of one matter being the cause of another. But it is also necessary to understand the purpose for making an inquiry about causation and that may require value judgments and policy choices.

In the aftermath of March the ‘but for’ test has functioned largely as a means of eliminating factors determined not to be a cause of a litigant’s injuries,15 and a distinct retreat can be identified in Australia from the latitude inherent in the ‘common sense’ approach to causation.16 In turn, this places increased pressure upon the expert evidence that frequently forms the basis for judicial findings of causation as a matter of fact. With the statutory tort reforms throughout Australia early in the twenty-first century, the High Court has been required to revisit the issue of causation in the context of the New South Wales version of the legislation governing liability for damages for harm resulting from negligence. Section 5D of the Civil Liability Act 2002 (NSW) provides that: (1)  A determination that negligence caused particular harm comprises the following elements: (a) That the negligence was a necessary condition of the occurrence of the harm (factual causation), and (b) That it is appropriate for the scope of the negligent person’s liability to extend the harm so caused (scope of liability). (2) In determining in an exceptional case, in accordance with established principles, whether negligence that cannot be established as a necessary condition of the occurrence of harm should be accepted as establishing factual causation, the court is to consider (amongst other relevant things) whether or not and why responsibility for the harm should be imposed on the negligent party. (3) If it is relevant to the determination of factual causation to determine what the person who suffered harm would have done if the negligent person had not been negligent:

(a) the matter is to be determined subjectively in the light of all relevant circumstances, subject to paragraph (b) and (b) any statement made by the person after suffering the harm about what he or she would have done is inadmissible except to the extent (if any) that the statement is against his or her interest.

12   The risks of the ‘common sense’ approach were later to be pointed out by Lord Hoffmann in Fairchild v Glenhaven Funeral Services Ltd [[2002] UKHL 22, [2003] 1 AC 32, [53] where he expressed concern that an appeal to common sense by a trial judge can suggest ‘that causal requirements are a matter of incommunicable judicial instinct’. He argued that causal requirements should not be regarded as ‘autonomous expressions of some form of logic or judicial instinct’, but as rules of law which need to be given content: at [54]. 13   [1991] HCA 12, [17], applying Stapley v Gypsum Mines Ltd [1953] AC 663, 681. 14   Tabet v Gett [2010] HCA 12, [112]. 15   See M Bagaric and S Erbacher, ‘Causation in Negligence: From Anti-jurisprudence to Principle-Individual Responsibility as the Conerstone for the Attribution of Liability’ (2011) 18 Journal of Law and Medicine 759. 16   See eg, Allianz Australia Insurance Ltd v GSF Australia Pty Ltd (2005) 221 CLR 568, [96]–[99] per Hummow, Hayne and Heydon JJ; Travel Compensation Fund v Tambree (2005) 225 CLR 627, [28] – [29], [45]–[46], [49].

244  Ian Freckelton SC (4) For the purpose of determining the scope of liability, the court is to consider (amongst other relevant things) whether or not and why responsibility for the harm should be imposed on the negligent party.

In Adeels Palace Pty Ltd v Moubarak17 a unanimous joint judgment of the High Court held that the twofold division of causation in section 5D ‘expresses the relevant questions in a way that may differ from what was said by Mason CJ in March v Stramare (E &MH) Pty Ltd to be the common law’s approach to causation’.18 The Court held that the references in March v Stramare to common sense were: evidently intended to disapprove of the proposition that value judgment should have no role to play in resolving causation as a matter of fact. However, by contrast, s5D treats factual causation and scope of liability as separate and distinct issues, the factual causation issue being determined by the ‘but for’ test: but for the negligent act or omission, would the harm have occurred?19

The Court observed that the ‘exceptional circumstances’ provision may have been enacted to deal with cases exemplified by the House of Lords decision in Fairchild v Glenhaven Funeral Services20 where the plaintiffs suffering from mesothelioma had been exposed to asbestos in successive places of employment.21 In Tabet v Gett22 the High Court grappled with a complex scenario of medical causation where the plaintiff was unable to prove on the balance of probabilities that her ultimate condition was caused or materially contributed to by a failure to undertake earlier medical testing (by CT scan) which would have revealed a condition which ultimately contributed to her deterioration. The problem for the Court was as to whether the evidence was capable of showing that the undertaking of a CT scan and the intervention that would have followed (administration of steroidal treatment) would have made any difference to the outcome for the patient. It found that the expert evidence did not establish this, the case also providing the platform for a rejection generally of loss of chance bases for recovery in tort.23

III.  Bizarre and Unscientific Evidence

A problem often identified in relation to expert evidence is that it can be ‘cloaked . . . in scientific language’ and thereby acquire: a mystical air of infallibility. . . . The fear is, therefore, rather than assist laypersons in forming judgments, that the use of expert evidence will usurp the function of the trier of fact and distort the proceedings by bolstering the evidence of parties and overwhelm a jury.24

  Adeels Palace Pty Ltd v Moubarak [2009] HCA 48.   ibid [43]. 19   ibid [45]; see also [55]. Note too, Travel Compensation Fund v Tambree (2005) 225 CLR 627, [28]–[29] per Gleeson CJ; [45]–[46] per Gummow and Hayne JJ. 20   Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32. Compare the issue arising in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [2011] 2 WLR 523. 21   [2009] HCA 48, [57]. 22   [2010] HCA 12. 23   See though the ‘optimism’ expressed in this regard by G Walsh and A Walsh, ‘Tabet v Gett: The End of Loss of Chance Actions in Australia’ (2010) 18 Journal of Law and Medicine 50; D Dubrow, ‘Loss of Chance in Medical Litigation’ (September 2010) Law Institute Journal 32. 24   R v C (G) (1996) 110 CCC (3d) 233, [65] (Newfoundland Court of Appeal). See too R v DD [2000] SCC 43. 17 18

Expert Evidence as to Causation  245 This is particularly so in respect of expert evidence about causation which often relates to complex and technical areas of specialist discourse and can have a significant statistical component25. An archetypal example of this phenomenon is to be seen in Commissioner for Government Transport v Adamcik26 in which the Australian High Court was called upon to consider how extraordinary evidence from a medical practitioner, who had expressed the view that leukaemia could be caused by emotional disturbance, should be evaluated. The question on appeal was not the admissibility of the evidence which did not conform to any orthodox theories of medical aetiology27 but whether it could be said that there had been evidence at first instance upon which the jury could have found a causal nexus between negligence, for which each of the defendants had been responsible, and the death of the respondent’s husband. The deceased had received injuries in a tram accident – a fracture (without displacement) of his acetabulum and abrasions and lacerations to his scalp. His stay in hospital was for less than a fortnight and he was described on discharge as being in happy spirits. However, within a few days swellings were observed in his throat, armpit and groin. He was diagnosed with lymphatic leukaemia from which he died a few months later. The physician called for the plaintiff/respondent (the wife of the deceased) conceded that he had only ever treated one case of lymphatic leukaemia. However, some years previously he had published a work entitled Autonomic Dyspraxia28 in England and there too propounded the theory that emotional stress could give rise to a wide variety of diseases (not referring explicitly to lymphatic leukaemia) on the ground that a major emotional disturbance may halt or disorganise the activity of the hypothalamus. He accepted in evidence that he was the first person to argue in favour of such a theory in relation to the aetiology of lymphatic leukaemia. The difficulty posed for the High Court on appeal was that because of sympathy for the deceased or simple confusion the jury had accepted the curious and highly speculative arguments about causation advanced by the physician. The majority of the High Court (in 1961) was loathe to interfere with the jurors’ fact-finding primacy, commenting that it was up to them to determine whether the witness was a ‘charlatan’ – ‘they might think so or they might regard him as a discoverer and prophet or in some other way’.29 The risk of unreliable evidence in the causation context, then, is that apparently plausible evidence will be misunderstood or misevaluated with an appellate court hesitant to overturn first instance decision-making autonomy. Amongst other things, this highlights the importance of counsel effectively making experts accountable in cross-examination for unorthodox or unscientific opinions which they proffer in evidence-in-chief so that the potential for uncritical acceptance of such opinions by triers of fact is minimised.30 A more modern case in which bizarre and unscientific evidence was given in Australia was in R v Parenzee31 where the appellant was found guilty at trial of having engaged in   See the difficult issues arising in this regard in R v T [2010] EWCA Crim 2439.   Commissioner for Government Transport v Adamcik (1961) 106 CLR 292. Compare the argument advanced in Dingley v The Chief Constable, Strathclyde Police 1998 SC 548 in terms of the potential nexus between multiple sclerosis and a motor vehicle accident. 27   See I Freckelton, ‘Unscientific Health Care: Legal Ramifications’ (2011) 118 Journal of Law and Medicine 645. 28   BG Haynes, Autonomic Dyspraxia: An Hypothesis for the Operation of Psychosis, Neurosis and Psychosomatic Disease (London, Lewis, 1958). 29   Commissioner for Government Transport v Adamcik (1961) 106 CLR 292, 308. 30   See the discussion by the Law Commission, Expert Evidence in Criminal Proceedings in England and Wales (Law Commission, London, Law Com 325, 2011): http://www.lawcom.gov.uk/docs/lc325_web.pdf. 31   R v Parenzee [2007] SASC 143, and then on appeal, [2007] SASC 316. 25 26

246  Ian Freckelton SC unprotected sexual intercourse with three women at a time when he knew that he was infected with the HIV/AIDS virus and that his conduct was likely to endanger the lives of the women in question. Mr Parenzee appealed and argued that there was reputable scientific opinion, of which he had been unaware before the trial, that AIDS is caused by a unique infectious agent, that HIV testing procedures were manifestly unreliable, that there was no proof that HIV, if it exists, is sexually transmitted, and that if HIV does exist, the risk of its being sexually transmitted is extremely low. To say that such views were unrepresentative of the community of scientific and medical opinion on the subject was an understatement. Justice Sulan, on appeal, considered the relevance of the High Court decision in Adamcik but distinguished it, finding it not to be ‘perfectly analogous’ to the case before him on the basis that the evidence with which he was dealing was that HIV was ‘an extremely well understood virus and that the issues raised by the applicant’s witnesses were not the subject of legitimate scientific controversy’.32 By contrast, he said, the understanding of leukaemia at the time of Adamcik was limited. He observed that in Adamcik the physician was proposing a theory which, while it was not accepted, was not contrary to a well-developed body of knowledge. Further, whereas the medical practitioner in Adamcik was propounding a new theory, the witnesses for Mr Parenzee before Sulan J were seeking to discredit a well-­ established theory. Justice Sulan concluded that the reasoning by the High Court in Adamcik meant that the level of acceptance of a witness’ evidence should not be determinative of the question whether the witness is qualified to give the evidence – even, as in Adamcik, where the evid­ ence is far-fetched or implausible, although such considerations are highly relevant to assessment of the probative value of the evidence, and also to the court’s assessment of the likely effect of such evidence, had it been led before a jury. Two witnesses were called in support of Mr Parenzee’s arguments. The first held a degree in nuclear physics from the University of Bucharest in Romania and had worked as a labora­ tory assistant and then as a physicist at a department of medical physics where persons with cancer and other diseases were treated. She had never treated nor been directly involved with clinical trials in relation to any disease. Sulan J made a variety of criticisms of the way in which she gave her evidence, finding her to be often non-responsive, failing to address the questions posed of her,33 lacking balance,34 misusing and misrepresenting others’ work,35 and misunderstanding matters which she should have understood.36 He accepted that ‘there may be circumstances in which a person can become expert in a particular area of expertise simply by reading and self-teaching’.37 However, he concluded that the areas in which she gave evidence were not of such a kind: ‘it is virtually impossible to develop an expertise in medical science, sufficient upon which others can rely, simply by reading textbooks and research papers’.38 He emphasised the importance of ‘practical experience’: If a person has work experience and has developed their knowledge from learning from others and being taught, that may be sufficient to qualify the person as an expert. In many disciplines, practical experience is essential. For example, an expert in winemaking may gain their expertise by   [2007] SASC 316, [73].   ibid [92].   ibid [102], [120], [137]. 35   ibid [121]. 36   ibid [122]. 37   ibid [130], 38   ibid [130]. 32 33 34

Expert Evidence as to Causation  247 working and being taught by an experienced winemaker. Simply reading about the subject may not be sufficient.39

He commented too that even had he accepted the expertise of the witness, he considered her opinions, amongst other things, about the causation of HIV were so ‘out of line with the prevailing opinions and prevailing evidence’ that no jury could rely upon her evidence.40 He found her not to be independent, to be an advocate for a cause, to lack objectivity, and to refuse to acknowledge or to be inclined to dismiss any evidence unsupportive of her views. Accordingly, he ruled that her opinions ‘lacked any credibility’ and thus no miscarriage of justice had been demonstrated by reason of the absence of her evidence at trial.41 Another witness called on behalf of the Applicant on appeal was Dr T who held the orthodox medical qualifications and was an experienced emergency physician and Fellow of the Royal Australasian College of Surgery and of the Royal Australasian College for Emergency Medicine. Thus, on the face of it, like the physician in Adamcik, he had apposite expertise by reference to his training and his expertise. However, Sulan J concluded that his knowledge of the relevant subject matter too was ‘limited to reading.’42 He found him to lack relevant practical experience in the treatment of viral diseases and in the disciplines of virology, immunology or epidemiology: ‘His opinions are based on reading scientific literature, studying of scientific literature, and spending a considerable amount of time thinking’.43 Justice Sulan took into account that Dr T had not published any papers or conducted any research of significance in the relevant areas. He concluded that Dr T too was not qualified to provide expert opinions about matters such as virus isolation, antibody tests, viral load tests or sexual transmission of the HIV virus. The result of Sulan J’s assessment of the two witnesses’ evidence was to hold that they were not adequately qualified to express their opinions and so their evidence (as adduced on appeal) was inadmissible. He concluded, though, that even if their evidence were admissible, no jury would conclude that there was any doubt that HIV was the cause of AIDS or that it was sexually transmissible. Thus their evidence lacked cogency. The South Australian Court of Appeal44 upheld the decision of Sulan J, concluding that the ‘new evidence’ lacked cogency and probative force and was not based on ‘sound principle. It is not just that they hold a different point of view on a matter that is accepted to be contentious or open to debate. Their views . . . are untenable’. This leaves a contrast in Australia between the two approaches in relation to unorthodox expert evidence of causation. In the Parenzee decisions, the more modern approach to admissibility was adopted with a robust approach to the expertise of the persons held out spuriously as having the specialist knowledge to assist the court. However, much work is done by the discretion to exclude evidence as more prejudicial than probative which, inevitably, takes into account the quality or reliability of the evidence. To this point, however, in Australia the Daubert indicia45 of reliability (falsifiability, peer reviews, error rates and general acceptance within the expert community) have neither been adopted as a standalone test nor as yardsticks for determining probative value in relation to expert opinions.   ibid [131].   ibid [136].   ibid [140]. 42   ibid [142]. 43   ibid [143]. 44   ibid [49]. 45   Daubert v Merrell Dow Pharmaceuticals 509 US 579 (1993). 39 40 41

248  Ian Freckelton SC

IV. Evidence Lacking in Objectivity

One of the serial concerns expressed in relation to expert evidence generally, including in respect of causation, is the failure by some witnesses to adopt a neutral, objective stance in relation to their provision of opinions.46 An attempt has been made to address this difficulty by the promulgation of court rules, largely confirming the common law,47 that experts’ duties are owed directly to the courts and that they should eschew bias, identifying the bases of their opinions, distinguishing facts from opinions, providing reasons for their views, and identifying any reservations or limitations to their opinions. In addition, experts have been mandated to declare, where possible, that they have been provided with all material that they have needed for their analysis of the issues that they have been asked to address. This has been part of an attempt in Australia, and other common law jurisdictions, to unleash experts from some of the constraining and distorting factors of the unmodified adversary system.48 Concerns about bias have been to the fore in a sequence of Australian cases dealing with the capacity of DNA profiling evidence to establish a nexus between a crime scene stain and a person accused of a serious crime. The most prominent decision in this regard is that of the South Australian Supreme Court in R v Karger49 where a very lengthy voire dire took place in relation to the admissibility of expert evidence concerning the Profiler Plus system for DNA analysis. It remains Australia’s most contested decision in relation to DNA profiling evidence. Criticism was made by a United States expert and also a Victorian forensic scientist of procedures and controls relevant to the technology. Ultimately, Mullighan J rejected the criticisms and in doing so repudiated the expertise and credibility of both experts. In language similar to that later used by Sulan J in R v Parenzee, he found that the Victorian scientist (Dr A) had never used comparable technology and was prepared to suggest significant error by the South Australian forensic laboratory without having made any inquiry of it: It reflects adversely upon Dr A as a scientist and an expert witness that he was prepared to reach an opinion adverse to the laboratory and the scientists without having made a fundamental and obvious inquiry or considering the most obvious explanation. . . . He was prepared to express adverse opinions about serious matters adverse to fellow scientists and, inferentially, their competency without making any inquiry.’50

He lamented the preparedness of some expert witnesses to give their opinions to the courts without ‘appropriate investigation and impartiality.’51 Another example of the phenomenon of partisan experts was repeatedly identified by judges in an extraordinary sequence of in excess of 200 cases arising out of the collision on 10 February 1964 between the HMAS Melbourne and the HMAS Voyager in which 82 sailors from the HMAS Voyager were killed off the New South Wales coast during peacetime manoeuvres. During the 1990s and until 2008 sailors from the larger ship, the HMAS 46   See I Freckelton, P Reddy and H Selby, Judicial Perspectives on Expert Evidence (Melbourne, The Australasian Institute of Judicial Administration, 1999); I Freckelton, P Reddy and H Selby, Magistrates’ Perspectives on Expert Evidence (Melbourne, The Australasian Institute of Judicial Administration, 2001). 47   See eg, The ‘Ikarian Reefer’ [1993] 20 FSR 563, 565–66. 48   See generally, I Freckelton and H Selby, Expert Evidence: Law, Practice, Procedure and Advocacy, 4th edn (Thomson, Sydney, 2009). 49   R v Karger [2001] SASC 64. 50   ibid [435], [442]. 51   ibid [612].

Expert Evidence as to Causation  249 Melbourne, instituted actions against the Commonwealth of Australia alleging that they had suffered psychiatric injuries arising from their experience in the rescue efforts for, or observation of attempts to rescue, sailors from the HMAS Voyager. Because the actions were instituted so long after the allegedly tortious event, applications for leave to institute proceedings out of time were required before litigation which concentrated upon whether the sailors had received psychiatric injuries as a result of the collision between the two ships.52 The major injury asserted by the plaintiffs was post-traumatic stress disorder (PTSD). The formulation of the disorder under the taxonomies of both the World Health Organization’s International Classification of Diseases (the ICD) and the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Diseases (the DSM) provided the plaintiffs with some assistance, as the symptomatology of the disorder diagnosed by a variety of treating and assessing psychiatrists and psychologists itself addressed the ‘nexus problem’. The re-experiencing aspects of the disorder, as asserted by the former sailors, related to the collision between the ships and its aftermath, thereby suggesting that the present-day pathology was causally related to the tortious event of the collision. However, the Commonwealth of Australia, the defendant, formulated a complex litigation strategy. It argued that there was no such thing as PTSD; that if there was, the plaintiffs did not have it; that if they did have it, it was not caused by the collision but by other traumatic experiences; and, finally, that if the causation was proved, the plaintiffs had not done what was reasonably open to them to mitigate their harm. This required trial courts to scrutinise (in what were mostly judge-alone cases) with great care whether the diagnostic criteria for PTSD under the ICD or the DSM were made out. The high point of judges evaluating whether diagnostic criteria were satisfied by reference to each indicium of each criterion for PTSD under the DSM occurred in Burk v Commonwealth of Australia53 and Cavenett v Commonwealth of Australia.54 In both of these decisions, the trial judges of the Victorian Supreme Court reviewed every element of the diagnostic criteria for PTSD by reference to the evidence of the plaintiffs and of the forensic psychiatrists and psychologists called by each side. There was a close examination of both whether the plaintiffs met the criteria for PTSD and then whether the PTSD symptomatology that they at times had experienced had been caused by the collision between the ships. The extent to which the judges undertook the exercise was a subject of appeal but in both cases the punctilious examination of the issues undertaken by the judges was not found to have descended into impermissible ‘judicial diagnosis’.55 However, as a result of these decisions and the Melbourne/ Voyager litigation generally, it can be said for Australia that the processes for determination of psychiatric disorders in ‘mental harm litigation’ have become extraordinarily demanding, with each of the diagnostic criteria for disorders and their aetiology coming in many cases under intense forensic scrutiny. In spite of the protestations in the various versions of the DSM that it should not be used as a ‘cook book’, the reality has become in Australia that the DSM has come to function in a way similar to a statutory instrument. The Melbourne/Voyager litigation was characterised by a remarkable polarisation of experts with the same report writers and witnesses figuring in a great many of the cases – 52   See generally I Freckelton and H Selby, Expert Evidence: Law, Practice, Procedure and Advocacy, 4th edn (Melbourne, Thomson, 2009). See too, I Freckelton, ‘PTSD’ in I Karpin, T Carney and B Bennett (ed), The Brave New World of Health (Sydney, The Federation Press, 2008). 53   Burk v Commonwealth of Australia [2006] VSC 25. 54   Cavenett v Commonwealth of Australia [2005] VSC 333. 55   Cavenett v Commonwealth of Australia [2007] VSCA 88; Burk v Commonwealth of Australia [2008] VSCA 29.

250  Ian Freckelton SC with highly predictable and entrenched positions in relation to the aetiology of the plaintiffs’ symptomatology. While inevitably there was heavy judicial reliance upon experts’ opinions, many of the cases were resolved on the basis of what was adjudged to be the plaintiffs’ credibility – both in relation to their involvement in the events of 10 February 1964 and in what occurred subsequently during their naval careers (including during the Vietnam War, where a number of plaintiffs experienced further traumata) and later in civilian life. To the extent that inconsistencies emerged between official records and their own accounts or between what was determined to have happened in the plaintiffs’ lives and what they had reported to treaters or forensic assessors, their credibility was often found to be diminished and the value of supportive opinions by expert assessors and treaters similarly reduced. The effect of this level of scrutiny was to reduce the extent to which the diagnosis of PTSD was advantageous to the plaintiffs, including in relation to the determin­ ation of causation. The Karger and Melbourne/Voyager decisions are typical of the frustrations experienced by many trial judges (in Australia and elsewhere) about the preparedness of some forensic practitioners to assume an advocacy role in causation matters and to provide opinions to the courts which have been problematically influenced by factors other than a focus upon provision of objective, even-handed assistance to the courts.

V. Evidence Not Easily Amenable to Evaluation

Certain kinds of expert evidence have been regarded as particularly difficult for juries (and judges) to evaluate.56 This can be because of the complexity of the area of expertise, statistical evidence having proved an example in this regard57 or because of an inability of experts to communicate accessibly and in such a way as to facilitate reasoned evaluation of their opinions. It has been held that evidence should not be rejected just because the technique, instrument or methodology has not been used before.58 Likewise, even where expert evidence is rejected at one time, this does not mean that it may not be accepted at a later time.59 A particular issue in this regard in Australia has been evidence which is substantially statistical in character and where there is the possibility for misestimation or misevaluation as to aetiologies, as against influence or association. Cases dealing with the admissibility of DNA evidence, which is dependent upon statistical interpretation of databases, which may or may not contain subpopulations, have explored this issue extensively, including in the context of the presentation of such evidence and avoidance of the ‘prosecutor’s fallacy’, whereby DNA evidence is presented in such a way as to risk overestimation or erroneous estimation of key issues in the course of a criminal trial.60 However, in Australia the most thorough analysis of the role of statistical evidence occurred in Seltsam v McGuiness,61 an appellate decision of the New South Wales Court of   See I Freckelton, The Trial of the Expert (Melbourne, Oxford University Press, 1987).   See eg M Redmayne, P Roberts, C Aitken, and G Jackson, ‘Forensic Science Evidence in Question’ (2011) Criminal Law Review 347–56, commentating on the difficulties raised by the Court of Appeal decision in R v T [2010] EWCA 2439. 58  See R v Danielson and McHardie [1983] 2 NSWLR 733. 59   R v Mallard [2003] WASCA 296. 60   See eg, R v Doheny & Adams [1997] 1 Cr App R 369 at 372–73; R v Latcha (1999) 104 A Crim R 390; R v GK [2001] 53 NSWLR 317. 61   Seltsam v McGuiness (2000) 49 NSWLR 262, [2000] NSWCA 29. 56 57

Expert Evidence as to Causation  251 Appeal dealing with the reception of expert epidemiological evidence, a category of evid­ ence about which Lord Kerr62 was later to observe: There is a danger that so-called ‘epidemiological evidence’ will carry a false air of authority. It is necessary to guard against treating a theory based on assumptions as a workable benchmark against which an estimate of the increase in trisk could be measured. Whether and in what circumstances epidemiological evidence can assist in the determination of whether a particular case of mesothelioma is likely to have been caused by a particular exposure will have to be decided according to the particular circumstances of an individual case.

Seltsam v McGuiness arose from an appeal from a decision of the Dust Diseases Tribunal which awarded a person with renal cell carcinoma compensation because of his exposure to a form of asbestos, crocidolite, one of the potential causes of his cancer. The issues on appeal included whether epidemiological evidence that only amounted to a low level of potential association – a relative risk ratio in the order of 1.4 – should have been admitted, and also whether evidence of statistical possibility could ever amount to evidence at the level of probability, which the plaintiff needed to establish his case. Spigelman CJ and Davies JA held that the test for causation is not satisfied where no more is shown than that the risk of injury to a plaintiff was increased by the defendant’s act or omission when it is not established that the risk has eventuated. They classified epidemiological evidence, which is necessarily general, rather than specific, as simply a form of circumstantial evidence with the potential to establish probability of causation. They declined to impose an arbitrary criterion for admissibility of such evidence in the form of a relative risk ratio of 2.0, as is done in some parts of the United States, and utilised the Bradford Hill criteria to evaluate the strength of epidemiological evidence:63 1 Strength of the Association. In general the higher the risk estimate, the less likely the finding is a result of confounding or bias . . . 2  Dose Response Effect. If the risk of the disease rises with increasing exposure64, a causal interpretation of the association is more plausible . . . 3  Time Sequence. The exposure or risk factor must precede the disease . . . 4  Consistency. Results from other epidemiological studies of the exposure-disease association should be similar. If similar results are found in different populations using various study designs, the plausibility of a causal interpretation is increased. An alternative explanation of bias or confounding would have to apply to each of the different studies, a highly implausible explanation. 5  Biological Coherence. Does the exposure-disease association make biological sense given what is known of the natural history of the disease? Do animal experiments support the association? Do other types of collateral evidence support the association, such as secular trends of the exposure factor in the disease? Unfortunately, for many diseases little is known about their aetiologies, so the informational background by which to judge biological coherence is often limited. Thus, failure of this broad principle does not necessarily weaken the plausibility of a causal interpretation. The first three principles can be applied to an individual study and used to assist the findings. The last two principles referred to results outside their particular study and relate more to external issues of coherence or consistency. All of the criteria or principles should be viewed as guidelines. Except, perhaps, for time sequence, none is required for a causal interpretation.   Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10, [206].   ibid [139]. 64   See the contrast made by Lord Phillips in Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10 at [12]–[13] between malaria which can be caused by a single mosquito bite and lung cancer caused by smoking, noting (at [15]) uncertainty as to whether contraction of mesothelioma is related to the amount of asbestos fibres ingested. 62 63

252  Ian Freckelton SC In taking this approach, Spigelman CJ and Davies JA adopted the unusual step of urging a particular methodology for evaluating the strength of a form of expert evidence about causation. Spigelman CJ and Davies JA concluded in circumstances where it was known that three factors in principle increased the potential for contracting renal cell carcinoma – smoking, obesity and exposure to crocidolite – that the evidence before the Tribunal had not established that the exposure to crocidolite had caused or materially increased the risk of the plaintiff contracting cancer. Stein JA disagreed, finding the evidence sufficient for the Tribunal to have reached its decision which was favourable to the plaintiff. The Seltsam decision is an Australian example of courts wrestling with mechanisms for evaluating complex and esoteric evidence about causation. Necessarily, the court made compromises, taking a moderately generous approach in relation to the reception of epidemiological evidence, but analysing rigorously the probative value of the evidence in relation to the particular case.65 Its most innovative contribution to the jurisprudence on causation was its adoption of a mechanism for evaluating the probative value of a key area of relevant forensic endeavour. The Seltsam decision constituted an important background to the 2010 decision of the High Court in Amaca Pty Ltd v Ellis,66 a case in which once again conflicting expert evidence, including in relation to epidemiology figured prominently. Mr Cotton, for whom the executor sued, died of lung cancer. The issue in dispute was whether the plaintiff had established at trial that it was exposure to asbestos fibres, as against smoking, that had caused his death, both options constituting potential aetiologies. As the plaintiff ’s case was brought on the basis that she could only succeed if she showed that Mr Cotton’s exposure had caused or contributed to (in the sense of being a necessary condition for) his developing lung cancer, issues of the kind considered by the House of Lords in McGhee v National Coal Board,67 Fairchild v Glenhaven Funeral Services Ltd,68 Barker v Corus (UK) plc69 and Sienkiewicz v Greif (UK) Ltd  70 did not arise. The High Court considered the epidemiological evidence of four experts: Expert

Relative Risk from Smoking

Relative Risk from Exposure to Asbestos

Probability cancer due to smoking

Probability cancer due to asbestos

Probability due to Smoking-Asbestos Combination

67%

3%

20%

92%

0.1%

0.9%

A

7.7

1.3

B

20

1.1–1.2

C

15

1.1

D

8

1.6

 Compare Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10.   Amaca Pty Ltd v Ellis [2010] HCA 5. 67   [1973] 1 WLR 1. 68   [2003] 1 AC 32. 69   [2006] UKHL 20, [2006] 2 AC 572. 70   [2011] UKSC 10, [2011] 2 WLR 523. 65 66

2–12%/5–20% (if higher exposure)

Expert Evidence as to Causation  253 The Court71 noted that: the epidemiological evidence showed that many sufferers of lung cancer had smoked tobacco, a few had been exposed to asbestos; some of those who had been exposed to asbestos had also smoked; some had been neither smokers nor exposed to asbestos. The observations made about the relationship between exposure to the known carcinogens of smoking and asbestos and the development of lung cancer were used and interpreted by epidemiologists in a number of different ways.

It observed:72 The question to which the plaintiff sought to have an affirmative answer inferred was, in each case: is it more probable than not that the negligence of the defendant was a cause of Mr Cotton’s cancer? Observing only that exposure to asbestos may have been a cause does not answer that question affirmatively.

The Court accepted that asbestos on its own or in combination with smoking may have been a cause of Mr Cotton’s cancer. It found that the risks and probabilities associated with his exposure to asbestos, whether alone or in conjunction with smoking, were low and not enough to found the inference that it was more probable than not that exposure to respirable asbestos fibres was a cause of Mr Cotton’s cancer. The Court opined:73 no scientific or medical examination can now say, with certainty, what caused Mr Cotton’s cancer or lung cancer in any other particular case. . . . despite this uncertainty, the courts must, and do, ‘reduce to legal certainty [a question] to which no other conclusive answer can be given’. The courts do that by asking whether it is more probable than not that X was a cause of Y. Saying only that exposure to asbestos may have been a cause of Mr Cotton’s cancer is not a sufficient basis for attributing legal responsibility. Observing that a small percentage of cases of cancer were probably caused by exposure to asbestos does not identify whether an individual is one of that group. And given the small size of the percentage, the observation does not, without more, support the drawing of an inference in a particular case. The paradox, if there be one, arises from the limits of knowledge about what causes cancer.

Ultimately, the case does not establish significant new legal precedent but it does emphasise the challenges and limitations to fact-finding which in turn is dependent upon the limitations of expert evidence. Another complex area of conflicting expert evidence has been spawned by the controversial decision of the High Court in Australia in Malec v JC Hutton Pty Ltd,74 a case which ruled upon whether a worker from a meat works who became unemployable, having suffered nervous problems potentially attributable to brucellosis contracted by his employer’s negligence, should have had his damages for future loss of earning capacity discounted by the fact that he was likely, quite independently, to suffer back problems which would have been likely to have created a ‘neurotic condition’ that would have rendered him unemployable. The Court held that it was proper for his damages to be reduced by the potential that, separate and distinct from the employer’s negligence, he was going to become unemployable. This meant that his damages for future loss of earning capacity needed to be reduced, not eliminated, and similarly his damages for the care and attention given to him by his wife and for his neurotic condition.   [2010] HCA 5, [20].   [2010] HCA 5, [51]. 73   ibid [70]. 74   [1990] HCA 20, (1990) 169 CLR 638. See D Hamer, ‘ “Chance Would be a Fine Thing”: Proof of Causation and Quantum in an Unpredictable World’ (1999) Melbourne University Law Review 24. 71 72

254  Ian Freckelton SC The High Court’s decision has repercussions for commercial disputes75 and personal injury cases.76 In Tabet v Gett77 the High Court revisited its reasoning and held that Malec analysis does not apply to loss of chance litigation which it has now precluded in the medical negligence context. However, the controversial decision of Malec continues to enable arguments that but for a tortfeasor’s negligence the same or similar consequences would have ensued, enabling defendants’ liability thereby to be reduced. Fundamental to the Malec approach is the difficulty in translating the hypothetical or speculative into percentile chances.78 Those who carry this burden are first the expert witnesses and then decisionmakers who are required, on the basis, inevitably, of limited facts to look to the future and ‘guesstimate’ either what will happen or what would have happened. It is correct that the calculation of pain and suffering and loss of earning capacity involve something of such a process but the calculation of what might/would have been has few signposts to guide the expert who attempts to provide evidence-based opinions. It is a recipe for a problematic incidence of disagreement and for reasoning that takes into account imponderables and immeasurables. To enumerate some of the cases where Malec has been applied makes the point: the likelihood of a plaintiff having become a world-ranked squash player;79 the likelihood of a plaintiff having challenged for a world title and defended it in kick-boxing;80 the likelihood of a plaintiff having exploited a mechanical dinosaur in the film industry;81 and, in a more pedestrian but very complex factual scenario, the likelihood of a plaintiff ’s obstructive airways’ disease and obesity having affected his life, separate and apart from the impact of the effects of his exposure to asbestos.82

VI. Expert Evidence Where the Facts Are Misunderstood

In a percentage of cases in which expert evidence about causation figures prominently, a number of the difficulties are attributable to confusion about the facts upon which expert opinions are based.

  Sellars v Adelaide Petroleum NL [1994] HCA 4, (1994) 179 CLR 332.   See eg, Wilson v State of Tasmania [1999] TASSC 145. 77   (2010) HCA 12, [39] (Gummow ACJ): ‘in that case [Malec] the claim giving rise to the assessment had been for physical injury, the contraction of a disease as a result of the negligence of the defendant. The imprecision allowed in the assessment of damages in such cases does not necessarily or logically apply where a claim for physical injury fails but is said to be saved by transmutation of the damage alleged into the loss of a chance of a better outcome’. See also Kiefel J, [136]. See also Lindsay-Field v Three Chimneys Farm Pty Ltd [2010] VSC 436. 78  See Malec v JC Hutton Pty Ltd [1990] HCA 20, (1990) 169 CLR 638 per Brennan and Dawson; Wright v Commonwealth of Australia [2005] VSCA 309, [57]–[58]. 79   Mansini v Martin [1998] QCA 222 (per Chesterman J): ‘Both inquiries involve an evaluative process. In most cases, including the present, a great deal of uncertainty will surround each of them. The assessment of damages will continue to be an art rather than a science. It will often not be possible in the uncertain state of affairs in which damages for personal injuries fall to be awarded to make an assessment with the degree of precision mentioned in Malec. Although the High Court apparently regarded it as possible to assess the degree of likelihood of an event occurring as being calculable to an accuracy of 1 per cent I think this is, with great respect, unrealistic. The best one can do is to make general assessments’. See too the discussion of an actor’s prospects of success in Norris v Blake (by his Tutor Porter) (No 2) (1997) 41 NSWLR 49, (1997) 25 MVR 101. 80   Doo v Murphy, unreported Supreme Court of Queensland, 8 October 1998 per Williams J (BC 9805163). See too Tszyu v Fightvision Pty Ltd; Fightvision Pty Ltd v Onisforou [1999] NSWCA 323. 81   Glenmont Investments Pty Ltd v O’Loughlin (2000) 79 SASR 185. 82   Seltsam Pty Limited v Ghaleb [2005] NSWCA 208 esp per Ipp JA. 75 76

Expert Evidence as to Causation  255 An example of this phenomenon is to be found in the case of Chappel v Hart.83 The plaintiff consulted an ear, nose and throat surgeon who informed her that she required a Dohlman’s Pouch operation in order to avoid a potentially fatal blockage to her throat. The plaintiff asked him about the potential for her vocal cords to be adversely affected by such a procedure. He reassured her but, as it turned out, in the course of the operation he punctured the plaintiff ’s oesophagus, she suffered mediastinitis and her vocal cords were adversely affected. In addition, the surgeon failed to inform the plaintiff that he had undertaken such an operation on only a modest number of occasions previously when completing his surgical training and thereafter as a consultant surgeon. More importantly, he did not inform her that surgeons more experienced than he, because of having done the procedure on many occasions and the dexterity they had acquired, ran a lower risk of puncturing the oesophagus and thus setting in train the sequence of events which in the plaintiff ’s case led to damage to her vocal cords. This led the majority in the High Court to conclude that the surgeon had been negligent, not in his execution of the procedure, but in denying the plaintiff the option of securing a more experienced surgeon at a later time by failing to inform her of the consequences of a differential in surgical expertise amongst surgeons in respect of Dohlman’s Pouch operations. The plurality accepted that had she been properly advised, the plaintiff would have delayed the operation and sought a highly experienced surgeon so as to minimise the risks to her vocal cords. The High Court, however, was at two stages removed from the trial judge who heard the actual expert evidence at trial about the incidence of oesophageal perforation. It appears that the plurality may have misunderstood the evidence by the key expert witness for the plaintiff, a Professor Benjamin, in relation to his never having punctured an oesophagus despite having conducted the operation on between 100 and 150 occasions. Justice McHugh in dissent observed that: I do not think that it is possible to read Professor Benjamin’s evidence as asserting that either he or other unidentified surgeons could perform the procedure with greater care or more skill than the defendant ordinarily performed it. In the Court of Appeal, Handley JA read one of Professor Benjamin’s answers as meaning that he had performed between one hundred and one hundred and fifty operations without a perforation of the oesophagus. If that interpretation had been correct, it would have provided an evidentiary foundation for the argument, if not the conclusion, that the defendant’s failure to warn had denied her the alternative of having the procedure performed with a reduced risk of perforation of her oesophagus. However, the plaintiff did not really dispute that his Honour misunderstood Professor Benjamin’s answer and that the effect of the Professor’s evidence was merely that he had carried out that number of operations without the onset of mediastinitis. Indeed, it is possible to read one of Professor’s Benjamin’s answers as indicating that perforations have occurred on a number of occasions when he or a team of surgeons of which he was a member has carried out the procedure.84

This is illustrative, in the causation context, of the difficulty experienced in many appellate contexts by judges, who do not see or hear from expert witnesses and have an opportunity themselves to clarify matters of ambiguity or uncertainty, in coming to an   Chappel v Hart (1998) 195 CLR 232.   ibid [40]. Another difficulty in relation to the evaluation of expert evidence arises if a wrong standard is applied to, for instance, clinical practice by a court. It is not uncommon for this to occur when a specialist doctor is called to give evidence about the conduct of a general practitioner. See eg, Lowns v Woods [1986] Aust Torts Rep 81–376. 83 84

256  Ian Freckelton SC effective understanding of the import of complex opinion evidence given at first instance. The more complex the evidence, the greater the potential there is for fundamental error – which, in turn, can lead to unclear appellate authority.

VII. Expert Evidence in the Loss of Chance Context

Until 2010 Australian law had been unclear for some time in relation to the propriety of loss of chance litigation – where the principal loss for a plaintiff as a result of negligence, for instance by a health practitioner, is a chance of recovery. A number of Australian decisions had upheld the proposition that a plaintiff could sue successfully if the outcome of a defendant’s negligence was shown by expert evidence to have lessened the plaintiff ’s opportunity of a better outcome and the plaintiff suffered an adverse result.85 The difficulty in such litigation lies in the fact that ordinarily it is incumbent on plaintiffs to prove that they have suffered damage in the sense of physical or psychiatric harm or economic loss. Under loss of chance theory, the loss is not harm or economic loss, per se; it is a reduced chance of a better outcome – such as losing the opportunity to secure advantageous early operative intervention in relation to a developing cancer or an aneurysm at risk of rupturing.86 It is common for experts to disagree about such scenarios, and to be unable to allocate in an evidence-based way percentile chances of a different outcome had earlier intervention taken place, the difficulty referred to above in the context of Malec. This leaves decisionmakers in a difficult situation in discerning between diametrically different views about potentially contrasting outcomes, had intervention occurred. The issue especially tends to arise in cases of missed breast lumps and subsequent life-threatening or terminal breast cancer. It is the more problematic when the basis for the bringing of the action and the provision of the experts’ opinions is unclear from a legal perspective. Australia’s most significant case on loss of chance litigation was litigated at appellate level in 2009/2010. In Tabet v Gett,87 the plaintiff, who was aged six at the time, suffered headaches, nausea and chickenpox symptoms. She was admitted to hospital and had tests to ascertain whether she had chicken pox meningitis on 13 January 1991 after nursing staff observed that her pupils were unequal and the right pupil was not reactive. The tests did not include a CT scan. She deteriorated further and suffered a seizure at 11.45am on 14 January. At that stage a CT scan was done. It showed a medulloblastoma (a brain tumour which had probably been growing for between 745 and 900 days, and being 3.5cm in diameter, namely ‘extensive’ and ‘well advanced’) which resulted in increasing intracranial pressure from a build-up of cerebrospinal fluid (hydrocephalus). It was this which caused the deterioration on 14 January and the seizure suffered by the plaintiff which was a contributor to her ultimate brain injuries. In due course (16 January), the tumour was removed and chemotherapy attempted but it was only partially successful. Now badly impaired, she sued for the loss of chance arising from the failure to undertake a CT scan. The evidence called on behalf of the plaintiff and the defendant about whether there had been negligence in failing to order a CT scan, and,   See eg, Rufo v Hosking [2004] NSWCA 391.   See eg, Naxakis v Western General Hospital (1999) 197 CLR 269. 87   Tabet v Gett [2010] HCA 76. See D Dubrow, ‘Loss of Chance in Medical Litigation’ (September 2010) Law Institute Journal 32. 85 86

Expert Evidence as to Causation  257 more importantly, whether a CT scan at a slightly earlier juncture (some hours) would have made any difference was complex. The brain injuries of the plaintiff were caused by a combination of: •  the tumour; •  the hydrocephalus; •  the surgery on 16 January to remove the tumour; •  subsequent medical treatment; and •  the failure to relieve the hydrocephalus on 13/14 January. The trial judge was not satisfied on the balance of probabilities that had the tumour been found on 13 January the plaintiff could have been treated in such a way as would have avoided the seizure and deterioration in her condition on 14 January. Instead, he found that the failure to do the CT scan deprived the plaintiff of less than 50 per cent of a chance of earlier treatments which might have lessened the damage done by the medulloblastoma. Normally this would mean that she would have been found not to have suffered a compensable injury. However, the trial found that her decline ultimately contributed 25 per cent to her disabilities and of that 25 per cent, 40 per cent was referable to the loss of chance of a better outcome. Thus she succeeded to a modest degree and although in principle damages referable to her entire brain injury exceeded $A6 million, because of his taking into account her ‘loss of chance’ of a better outcome she was entitled to damages of ‘only’ $A610,000.88 However, the New South Wales Court of Appeal89 overruled earlier authority90 and held that those decisions, including that at trial level in the case before it, which had permitted recovery on the basis of loss of chance were ‘plainly wrong’. It concluded that if it were incorrect in this regard, at most the plaintiff had lost a 15 per cent chance of a better outcome. The High Court agreed on the threshold issue91 and followed United Kingdom92 and Canadian93 authority on the subject, distinguishing a decision of the Supreme Judicial Court of Massachusetts.94 However, much of its decision revolved around a close analysis of the expert evidence as to factual causation. At trial a neurosurgeon expert witness, Mr Klug, when asked whether usage of steroids on 13 January (had a CT scan been taken showing the tumour) would have avoided the acute decline on 14 January, responded that that would be ‘entirely speculative . . . you could certainly make an argument that they would have improved the situation and may have prevented the episode, but, as I say, nobody could answer that with any certainty’.95 He said that in such a difficult situation he may have used a combination of steroids and a ventricular drain if he concluded there was a risk of serious deterioration. However, oddly, he was not asked for his opinion as to the likely efficacy of this combination. This led Gummow ACJ in the High Court to conclude that the evidence provided a basis for no more than speculation as to the loss of any chance, whether it be 40 per cent as found by the trial judge or 15 per cent as mooted by the Court of Appeal.96   Tabet v Mansour [2007] NSWSC 36.   Gett v Tabet (2009) 254 ALR 504. 90   Gavalas v Singh (2001) 3 VR 404; Rufo v Hosking (004) 61 NSWLR 678. 91   Tabet v Gett [2010] HCA 76. 92   Gregg v Scott [2005] 2 AC 176; see further J Stapleton, ‘Loss of the Chance of Cure from Cancer’ (2005) 68 MLR 996. 93   Laferriere v Lawson [1991] 1 SCR 541. 94   Matsuyama v Birnbaum, 890 NE 2d 819 (2008). 95   [2010] HCA 12, [43]. 96   ibid [45]. 88 89

258  Ian Freckelton SC Justice Heydon97 reached a similar result, noting that if a decision to perform a CT scan had been made on 13 January, it would have required the summoning from within the hospital or from outside an anaesthetist, a radiologist and a radiographer: It could have taken five to six hours to arrange and carry out the CT scan. But the CT scan would have detected the tumour and the hydrocephalus. The plaintiff would then have been treated with steroids rather than by the insertion of an intracranial drain to remove cerebral spinal fluid, because although Dr Maxiner, a neurosurgery registrar on duty on 13 January . . . would have preferred the latter course, her superior, Mr Johnston, a neurosurgeon, would have adopted the former.

That led him to conclude that the steroid treatment would probably not have avoided the 25 per cent share of the brain damage attributed to the negligence. He noted that the expert opinion from Mr Klug at trial was that the steroid treatment would work within 24 hours. Yet it could not have commenced before 6.00pm on 13 January, giving it only 17 or 18 hours to achieve the desired result by the time of the plaintiff ’s deterioration at 11.45am on 14 January. He observed that Mr Klug did not say that steroid treatment could work within 17 or 18 hours: Hence a decision that it had not brought about an improvement could not have been made until after the episode on 14 January. And there was no deterioration until that episode. Accordingly there was no possibility, if Mr Johnson’s approach was being followed, that a drainage episode would have been employed.98

Mr Johnson at trial said that he could not be sure that administration of steroids would have had a significant effect by the relevant time. This evidence was consistent with the opinion of Mr Klug. Justice Heydon thus concluded that the evidence did not permit any conclusion that the effect of earlier administration of steroids would have been to reduce the occurrence of the brain damage which took place at about 11.45am on 14 January. This meant the finding that there was a chance of avoiding the brain damage on 14 January had the defendant arranged a CT scan soon after 11.00am on 13 January was wrong99 and therefore that the question of law in relation to loss of chance litigation was hypothetical. He went further, finding that the discrepancy of findings as to loss of chance by the trial judge as against the Court of Appeal (40 per cent as against 15 per cent) was illustrative of the risks of resort to speculation: The very fact of this large difference between conclusions reached after such careful consideration in each of the courts below of a very difficult case supports the conclusion that it was in truth not possible to arrive at any conclusion on the question of whether the negligence caused the plaintiff to lose a chance of avoiding or reducing the damage. The difference suggests that the condition of the evidence left them no alternative but to grope towards speculative outcomes which it was impossible for them firmly to grasp.100

Justice Kiefel, with whom Hayne, Bell and Crennan JJ agreed, arrived at the same finding, noting that Mr Klug stated that ‘it is entirely possible that actually the same course of events would have happened’,101 had steroids been given. This meant that for her the plaintiff was unable to prove that it was probable that if corticosteroids had been given it   ibid [76].   ibid [81]. 99   ibid [94]. 100   ibid [96]. 101   ibid [116]. 97 98

Expert Evidence as to Causation  259 was probable that the brain damage of 14 January would have been avoided – ‘the evidence was insufficient to be persuasive’.102 The Tabet v Gett decision is authoritative for Australia103 in relation to loss of chance litigation, in spite of the strong position taken by Heydon J that because a wrong finding of fact had been made the legal question in relation to loss of chance did not arise. At another level, though, the decision is particularly illustrative of a phenomenon frequently found in causation (and especially loss of chance) cases – significant factual complexity which can lead to fact-finding error if the elements of factual causation are not rigorously identified and determined on the evidence. If assumptions wrongly become viewed as facts, and if extra-factual speculation intrudes, incorrect decisions can ensue. In Tabet v Gett, the expert evidence suggested that a possibility existed that administration of steroids might have made a difference but the possibility was speculative and hypothetical. It was unquantifiable and not such as to enable any proper finding of fact in terms of causation, whether on the traditional approach of the law or even on a loss of chance basis. For advocates and decision-makers alike, the case stands as a stark reminder that any imprecision in evidentiary analysis, especially analysis of expert evidence about causation, can result in fact-finding error and all the unfairness that that entails.

VIII. Conclusions

The many difficulties encountered by Australian courts in cases relating to expert evidence as to causation are exemplary of issues encountered in all countries. In a number of instances the expert evidence (eg Adamcik, Parenzee and Karger) has been characterised by questionable and methodologically dubious expert evidence. On occasions too the existence of multiple potential aetiologies has created both theoretical and practical problems for fact-finding, calling into question what the criteria should be for reception of evidence about causation which was non-specific and amounted only to ‘possibility evidence’ (eg Seltsam, Amaca). Courts traditionally have struggled with ‘novel’ scientific and medical evidence104, lacking tools by which to determine whether theories or techniques underlying such evidence have reached a point of ‘crossing from the experimental to the demonstrable’ by yardsticks such as the United States mechanisms of ‘general acceptance within the relevant scientific community’105 or multifactorial ‘reliability’.106 An aspect of this is the dilemma as to how to deal with iconoclastic or heterodox theories, such as those peddled in the Adamcik and Parenzee cases. It can be said that, where there are juries, there is a risk both of ‘hindsight error’,107 which results in assumptions of causation being made on the basis of little more   ibid [152].   See, though, G Walsh and A Walsh, ‘Tabet v Gett: The End of Loss of Chance Actions in Australia’ (2010) 18 Journal of Law and Medicine 50; D Dubrow, ‘Loss of Chance in Medical Litigation’ (September 2010) Law Institute Journal 32. 104  See Dasreef Pty Ltd v Hawchar [2011] HCA 11. 105   Frye v United States 293 F 1013 (1923). In Australia, an early decision in relation to the admissibility of fingerprinting evidence took a similar approach: R v Parker [1912] VLR 152. 106   Daubert v Merrell Dow Pharmaceuticals 509 US 579 (1993). 107   See TB Hugh and SWA Dekker, ‘Hindsight Bias and Outcome Bias in the Social Construction of Medical Negligence: A Review’ (2009) 16 Journal of Law and Medicine 846. 102 103

260  Ian Freckelton SC than an adverse outcome for a plaintiff, and also sympathy for a victim leading to an assumption of causative nexus in face of expert evidence which might suggest that the correct decision is otherwise. Both factors can distort rigorous reasoning and result in outcomes which are unfairly disadvantageous for defendants – in both criminal and civil litigation. When the law is unclear on crucial issues of causation law (eg admissibility of possibility evidence: Seltsam, Amaca), what might occur in the future (Malec, Sellars), or loss of chance reasoning (Tabet), the risks are amplified that trial courts will be distracted by expert evidence that is other than evidence-based and reach results which seek to compensate plaintiffs whose circumstances are adverse.

13 The Challenge of Developing Science for the Law of Torts CARL F CRANOR

I.  Introduction: Disquiet about Torts

In the US there is some disquiet about the inventiveness of courts in adapting the tort law to the complexity of the biological world. Indeed courts may be less inventive than they once were to address such issues. This concern has at least three sources. First, as federal and many state judges have been given greater responsibility to screen expert testimony and its foundations in each case, too many courts err in their decisions. They have difficulty recognising the nature of scientific inferences, or what I have elsewhere called ‘inferences to the best explanation’.1 Perhaps because of this, they have also struggled with recognising the scientific and legal relevance of different kinds of scientific studies. When they make such mistakes this does not permit scientists to utilise the full body of scientific evidence that they would normally consider in coming to conclusions about general and specific causation. A second source of concern is that the complexity of scientific data about causation, or ‘models of causation’ as I will use that term, poses several possible problems. Models of causation that have been legitimised by courts may or may not allow for causal relationships that scientists recognise and use. Scientists’ understanding of causation tends to be more complex than the law legitimates. I will discuss some different models arising from recent research that might well pose such problems. I will not address whether the law has adequate conceptual categories to address them. This I leave to others. A third source of concern is that some of the causal relationships scientists are identifying may well pose problems for procedural rules – in particular, statutes of limitations. Some diseases can have long periods before all causal antecedents trigger the disease process, while other may have long latency periods before they can be identified clinically. The main topic of this chapter is the developmental origins of disease, dysfunction and premature death. While this research is in its infancy, such early sources of disease are likely to stress and strain the law. At present scientists almost certainly do not yet know all the

1  CF Cranor, Toxic Torts: Science, Law and the Possibility of Justice (New York, Cambridge, 2006) 205–82; CF Cranor, ‘A Framework for Assessing Scientific Arguments: Gaps, Relevance, and Integrated Evidence’ (2007) XV Journal of Law and Policy 7.

262  Carl F Cranor disease processes that early exposure to toxicants can initiate, promote, accelerate, catalyse, potentiate, facilitate or exacerbate. They may not know the lowest exposures during development that can contribute to risks, either immediately or later in life. As research tools are refined to detect more subtle endpoints and more studies are conducted, more adverse outcomes may well be identified. The outcome of this research is likely to transform how scientists and the rest of us think about, analyse and assign causal responsibility for various diseases. In what follows, I review several causal models proposed or found by scientific research in order to pose the major question of whether the law can appropriately accommodate discoveries in science.

II.  Procedure and the Development of Disease

Some basic features of disease processes that diverge from more common examples could well pose problems for civil procedure in torts. A disease might have a long induction period or a long latency period or both; either can stress the law’s procedural rules. The induction period of a disease is the time it takes for several critical biological events to occur; these are components that contribute to a full-fledged disease process. The latency period of a disease is the time between completion of all events needed to produce the disease (completion of the induction period) and when physicians can clinically detect it.2 Some disease processes may have a short induction and short latency period. Others might have a short induction period and a long latency period. Still others, such as multistage diseases that take the completion of several stages to initiate disease, might have both a longer induction period and a long latency period. Heroin addicts, who took the synthetic heroin, MPTP (l-methyl-4-phenyl-1,2,5,6-­ tetrahydropyridine), suffered such an immediate effect that the period between their taking of MPTP and the destruction of brain cells that led to a Parkinson-like condition was extremely short – within days. In these cases the induction and latency periods were extremely quick: short induction period, short latency period.3 Similarly, various bacterial or viral infections might have short induction and short latency periods, eg measles. Diseases like cancers can have a fairly long induction period before all the causal antecedents needed for the disease have been completed, and another longish latency period before the disease manifests itself in a clinically detectable manner. DES-caused vaginal cancer illustrates these points. Diethylstilbestrol (DES) is a synthetically created oestrogen that was used for a brief period of time to try to prevent miscarriages (it did not work), but which caused a very rare form of vaginal cancer in young women in their late teens or early twenties. The developing children experienced exposure to DES in utero but the diseases did not appear for about 20 years. On the one hand, if in utero exposure to the synthetic oestrogen DES were sufficient to cause vaginal and cervical cancer about 20 years later, then one would say that the induction period of this disease was quite short (only in utero exposures were required to complete the disease process). However, on this account the latency period of the disease would   K Rothman, Modern Epidemiology (Boston, Little, Brown and Company, 1986) 72.   JW Langston, P Ballard, JW Tetrud and I Irwin, ‘Chronic Parkinsonism in Humans Due to a Product of Meperidine-Analog Synthesis’ (1983) 219 Science 979. 2 3

The Challenge of Developing Science for the Law of Torts  263 be 20 years. This was one hypothesis proposed by researchers that discovered DES-caused cancer: 4 very short induction period, long latency period. On the other hand, if a natural oestrogen surge during puberty were needed to supplement in utero DES exposures before all the antecedents for vaginal or cervical cancer were complete (a second critical step as it were), then one would say that the disease had a long induction period (from in utero exposures to DES through to a second oestrogen surge from natural hormones during puberty). (This was a second hypothesis proposed by the research team.)5 However, since the disease process on this possibility would only have been completed during puberty, the cancers would have had a relatively short latency period (from puberty to the diseases manifesting themselves about the age of 19 or 20, on average).6 This could represent an example of a longer induction period but shorter latency period. The possibility of long induction or long latency periods or both by themselves may pose problems for statutes of limitations. The law must allow for these in order to recognise legitimate causal relationships and keep open the possibility of compensating people who had been injured as a result of exposures to toxic products.

III. Background: The Developmental Origins of Disease7

A.  The Developmental Sensitivity of Children to Toxicants During development, children are less well protected by the womb than once might have been thought. The placental wall, once thought to be something like a protective insulating barrier from toxicants and other invaders, is vastly more permeable than once believed. These research results challenge a common older assumption about the protective or impermeable nature of the placenta and the womb. James Schardein, a leading researcher on developmental toxicants, puts the point as follows: ‘It is clearly evident that there really is no placental barrier per se: The vast majority of chemicals given the pregnant animal (or woman) reach the fetus in significant concentrations soon after administration’.8 Children also go through developmental stages that make them more susceptible to toxic and other insults than adults. During organ development children go through ‘rapid and dramatic’ changes.9 This is an especially vulnerable period because cells multiply rapidly, migrate to specific locations, begin to communicate with one another, and tissue structures undergo changes. In this period, different organ systems are subject to different windows of sensitivity to toxicants and there are numerous opportunities for exogenous substances to disrupt the developmental process and to damage organ systems. There are different periods of vulnerability for different parts of the developing body coinciding with key events in 4   AL Herbst, H Ulfelder, and DC Poskanzer, ‘Adenocarcinoma of the Vagina: Association of Maternal Stilbestrol Therapy with Tumor Appearance in Young Women’ (1971) 284 New England Journal of Medicine 878, 880. 5  ibid. 6   Rothman (n 2) 14. 7   P Grandjean et al, ‘The Faroes Statement: Human Health Effects of Developmental Exposure to Chemicals in Our Environment’ (2008) 102 Basic & Clinical Pharmacology & Toxicology 73, 74, available at www.blackwellsynergy.com. 8  JL Schardein, Chemically Induced Birth Defects: Third Edition Revised and Expanded (New York, Marcel Dekker, 2000) 5. 9   JM Rogers and RJ Kavlock, ‘Developmental Toxicology’ in C Klaassen, Casarett and Doull’s Toxicology, 6th edn (New York, Pergamon Press, 2001) 351, 358.

264  Carl F Cranor the changing organ structure: eyes can be affected quite early, long bones of the body somewhat later, while the palate has two different periods of sensitivity. Moreover, a particular toxicant may affect more than one developing system.10 In addition, the window of susceptibility may be very narrow. For example, researchers have identified at least one thalidomide victim whose mother took only one pill when she was pregnant. Nonetheless, the child was a thalidomide baby.11 Developing children have lesser defences against toxic insults. The upshot is the children are much more sensitive to toxicants than adults. These circumstances by themselves should alert scientists, public health officers and the public to potentially greater threats to children. In addition, a developing foetus or newborn is also typically subject to greater exposures than most adults. A ‘mother’s chemical body burden will be shared with her foetus or neonate, and the child may, in some instances, be exposed to larger doses relative to the body weight’.12 Toxicants can concentrate in umbilical cord blood, increasing toxic concentrations in a developing foetus to higher levels than in the mother’s body tissues. For example, methylmercury (MeHg) concentrations in a foetal brain can be up to ten times higher than concentrations in the mother’s blood.13 If newborns are breast-fed, since breast milk contains considerable fat, any toxicants that are lipophilic (soluble in body fats) can have greater concentrations in breast milk than in the nursing mother’s body. Some toxicants in breast milk can be as much as one hundred times more concentrated than in the mother’s body.14 Not all substances will concentrate in this manner in developing children, but some can. Moreover, lipid soluble toxicants will not necessarily decrease with longer breast-feeding. Once children are born and begin life as more or less independent organisms, they have a number of features that typically increase their exposures to toxicants. They have higher metabolisms and more rapid breathing rates. Over the first 12 years of life a child on average has a breathing rate that is double that of an adult. They drink more than twice as many fluids and eat more food on a per body weight basis than adults.15 They eat five or ten times as much fruit.16 In sum, they have ‘augmented absorption rates, and diminished ability to detoxify many exogenous compounds, relative to that of adults’.17 Also, as any parent knows, young children also tend to play close to the floor and ‘mouth’ everything in sight within their grasp. They test objects in their environment with their mouths: toys, cooking utensils, electric cords, pet food, almost any object that might be on the floor, in the grass, in the dirt, or on occasion even in mud. They are more active than adults.18  ibid.   JL Schardein and OT Macina, Human Developmental Toxicants: Aspects of Toxicology and Chemistry (Boca Raton, FL, Taylor & Francis, 2007) 131–32. 12   Grandjean et al, ‘The Faroes Statement’ (n 7) 74 (emphasis added). 13   S Honda, L Hylander and M Sakamoto, ‘Recent Advances in Evaluation of Health Effects on Mercury’ (2006) 11 Environmental Health and Preventive Medicine 171, 173. 14   P Grandjean and P Landrigan, ‘Developmental Neurotoxicity of Industrial Chemicals’ (2006) 368 The Lancet 2167, 2168. 15  ibid. 16   SG Selevan, CA Kimmel, and P Mendola, ‘Identifying Critical Windows of Exposure for Children’s Health’ (2000) 108 Environmental Health Perspectives 451, 454. 17  See Grandjean and Landrigan, ‘Developmental Neurotoxicity of Industrial Chemicals’ (n 14) 2168; G Ginsburg et al, ‘Incorporating Pharmacokinetic Differences Between Children and Adults in Assessing Children’s Risks to Environmental Toxicants’ (2004) 198 Toxicology & Applied Pharmacology 164; see generally National Research Council, Pesticides in the Diets of Infants and Children (Washington, DC, National Academy Press, 1993). 18   MD Miller et al, ‘Differences Between Children and Adults: Implications for Risk Assessment at California EPA’ (2002) 21 International Journal of Toxicology 403, 405. 10 11

The Challenge of Developing Science for the Law of Torts  265

B.  Developing Children Tend to Be More Susceptible Than Adults There are very good general biological reasons for developing children to be more susceptible than adults at the same concentration of toxicants. There is some quantitative evidence on this point, but it is somewhat limited because of the difficulty of conducting human studies. Nonetheless, a variety of human and experimental animal studies have substantiated the increased susceptibility of mammals to external toxicants. Consider a few of these. The poorly tested pharmaceutical DES caused genital abnormalities as well as cancer of the vagina and cervix in women who were exposed in utero. Moreover, as the DES daughters reach middle age, they have a breast cancer rate that is double that of the general population of women of the same age.19 DES sons also experienced reproductive abnormalities, but almost no instances of cancer from their in utero exposures. In addition, DES offspring appear to have ‘a higher incidence of autoimmune diseases and asthma’ likely resulting from ‘inappropriate immune system responses’ or other immune system dysfunction.20 Although DES mothers who took the drug appear to be at a 20–30 per cent elevated risk of breast cancer, they do not suffer near the range of maladies that their daughters (and to a lesser extent their sons) did.21 Cancers were not the only effects. The exposed daughters also had cervical or vaginal abnormalities, more spontaneous abortions and preterm deliveries, and perhaps a higher incidence of pregnancy losses. Thalidomide was a sedative that presumably had beneficial effect in adults (along with some peripheral neuropathy), but it caused phocomelia and numerous other dysfunctions in newborns because of maternal prenatal use.22 Two anti-convulsive drugs – diphenlhydantoin and valproic acid – again appear to have had beneficial effects for the mothers (reducing tendencies to convulsions). The offspring of mothers who took diphenhydantoin were at risk for a ‘broad spectrum of abnormalities,’ while the children of mothers taking valproic acid had ‘neural tube defects and heart, craniofacial and limb anomalies’.23 A pharmaceutical similar to vitamin A, isotretinoin, ‘used to treat cystic acne, is associated with a characteristic pattern of malformations’.24 It also caused profound mental retardation in the offspring.25 Coumadin, a blood thinner for patients with heart problems or those with artificial heart valves, has been associated with underdevelopment of the cartilage of the nose, under­ development of bodily cartilage, as well as atrophy of the optical nerves in children exposed 19   A Kortenkamp, ‘Breast Cancer and Exposure to Hormonally Active Chemicals: An Appraisal of the Scientific Evidence’ Background paper published by the Health and Environment Alliance and CHEM Trust, www. chemtrust.org.uk. 20   RW Luebke, DH Chen, R Dietert, Y Yang, M King and MI Luster, ‘The Comparative Immunotoxicity of Five Selected Compounds Following Developmental or Adult Exposure’ (2006) 9 Journal of Toxicology and Environmental Health 1, 5. 21   Centers for Disease Control and Prevention, DES Update: Consumers, ‘Recent DES Research,’ located at www.cdc.gov/DES/consumers/research/recent.html, citing L Titus-Ernstoff, EE Hatch, RN Hoover, JR Palmer, ER Greenberg et al, ‘Long-term Cancer Risk in Women Given Diethylstilbestrol (DES) During Pregnancy’ (2001) 84 British Journal of Cancer, 125. 22   PJ Landrigan, CA Kimmel, A Correa and B Eskenazi, ‘Children’s Health and the Environment: Public Health Issues and Challenges for Risk Assessment’ (2004) 112 Environmental Health Perspectives 257, 259. 23   ibid 259. 24   ibid 259. 25   ibid 259.

266  Carl F Cranor in utero.26 The mothers appear to have been unaffected or largely unaffected by the pharmaceutical (although there are risks from blood that becomes too thin). A recent review of five chemicals in experimental animals studies – DES, Valium (diazepam), lead, dioxin and tributyltin oxide (an industrial preservative, disinfectant and antifouling agent) – known to have adverse effects on the immune system adds to the science that shows that developing mammals are susceptible to toxic effects at lower levels than adults. The researchers report states: ‘For all five chemicals, the developing immune system was found to be at greater risk than that of the adult, either because lower doses produced immunotoxicity, adverse effects were more persistent, or both’.27 Until there are data to the contrary, scientists should assume that ‘development and maturation of the immune system constitutes a period of greater sensitivity to xenobiotic exposure’.28 Similarly, a 2009 study found that in animal models prenatal or nursing exposures to very low levels of dioxin caused the animals to be more susceptible to influenza virus than their mothers when they reached adulthood. Critical cells that fight such diseases were significantly reduced in the offspring compared with their mothers. The mothers were not harmed by the low levels of dioxin, either initially or later in life, but developing animals were at greater risk later in life as a result of early life exposures.29 Quite recently, researchers found that young women exposed to the pesticide DDT during puberty have higher rates of breast cancer than those exposed to DDT as adults.30 More specifically, women who had higher DDT exposure when they were under 14 years of age had rates of breast cancer that were five times higher than women who were older when they were exposed (about 27 years of age).31 Landrigan et al, citing a variety of studies and a National Academy of Sciences report, note that ‘in utero and early life exposures to lead, polychlorinated biphenyls (PCBs), certain pesticides, and other environmental neurotoxicants are known or thought to contribute to the burden of [neurodevelopmental] disorders’ including learning disabilities, dyslexia, mental retardation, attention deficit disorder and autism.32 These dysfunctions are of particular concern because, according to the National Research Council, ‘3% of developmental disabilities are the direct consequence of neurotoxic environmental exposures, and . . . another 25% arise out of the interplay of environmental factors and individual genetic susceptibility’.33 Since humans have wide variation in genetic susceptibility, the second estimate is particularly worrisome. Tort law is not just based on protecting people of average susceptibilities, but must account for genetic variation or other features that give them ‘egg shell skulls’. Consequently, if up to one-fourth of developmental disabilities result from environmental-genetic interactions, citizens have a strong interest in receiving compensatory justice if they are harmed in part by exposures to neurotoxicants.

 ibid.   Luebke et al, ‘The Comparative Immunotoxicity of Five Selected Compounds’ (n 20) 1.   ibid 22. 29  J Hogaboam, A Moore and BP Lawrence, ‘The Aryl Hydrocarbon Receptor Affects Distinct Tissue Compartments During Ontogeny of the Immune System’ (2008) 102 Toxicological Sciences 160. 30   BA Cohn, MS Wolff, PM Cirillo and RI Sholtz, ‘DDT and Breast Cancer in Young Women: New Data on the Significance of Age at Exposure’ (2007) 115 Environmental Health Perspectives 1406. 31  ibid. 32   Landrigan et al, ‘Children’s Health and the Environment’ (n 22) 261. 33   ibid 261–62 (citing the National Research Council, Toxicological Effects of Methylmercury (Washington, DC, National Academy Press, 2000). 26 27 28

The Challenge of Developing Science for the Law of Torts  267

C.  Some Biological Presumptions Much of the science of the developmental origins of disease is in its infancy. As a consequence the scientific picture is far from complete, resembling in many respects a pointillist painting, partially filled in but with many gaps and incomplete sections. Despite this, several biological plausibility or proof of principle arguments should be recognised for science, the law and the discussion that follows. Scientists have established certain facts of toxicology about humans, their tissues or adverse effects that can occur. These are similar to certain kinds of proofs or showings in mathematics. A mathematician often seeks to show that there is at least one mathematical entity of the kind he or she is considering. After the existence of one example is demonstrated, the question then becomes how many similar entities are there? This is called an ‘existence proof in mathematics’.34 Analogous to such existence arguments, scientists have shown certain results for particular substances. Once the existence or plausibility of certain biological facts has been shown for one substance, the next issue is how many other substances will show similar or analogous effects? This awaits the progress of science on the issues. Consider three ‘biological plausibility’ claims that the current research has shown. 1. It is biologically plausible that virtually all industrial chemicals enter citizens’ bodies and penetrate their tissues and fluids. There is quite good evidence that this occurs. The Centers for Disease Control’s (CDC’s) biomonitoring program reveals such outcomes for persistent and bioaccumulating substances as well as for those with more transient half-lives, but a number of those reveal more continual exposures. At the moment some 212 substances, most of them toxic, have been found in human bodies at some detectable concentration. More will be discovered as reliable protocols are developed to identify chemical substances in our bodies. All but the very largest molecules appear to invade in this manner. 2. It is biologically plausible (and substantially likely) that many industrial chemicals that are in women’s bodies can cross the placenta and enter the tissues of developing foetuses. It is also biologically plausible that industrial chemicals can enter the breast milk and expose nursing newborns. These facts challenge a common older assumption about the protective or impermeable nature of the placenta and the womb. Researchers now point out that ‘there really is no placental barrier per se: The vast majority of chemicals given the pregnant animal (or woman) reach the fetus in significant concentrations soon after administration’.35 3. It is also biologically plausible, in the sense that it is a live possibility that developing foetuses and newborns can be threatened by toxicants contaminating their bodies. Several reasons support this plausibility claim. Most industrial chemicals, pesticides and drugs will be able to cross the placenta and contaminate the womb. Developing children are in one of the most vulnerable life stages as their various organ systems develop from single cells into million- or billion-celled organs. One organ, the brain, has few opportunities to repair any cellular damage. If aspects of it are damaged during development, there are increased chances that damage will be more nearly permanent compared with   JF Lucas, Introduction to Abstract Mathematics (Lanham Md, Roman & Littlefield, 1990) 75–76.  Schardein, Chemically Induced Birth Defects (n 8) 5.

34 35

268  Carl F Cranor other organ systems that have greater opportunities for repair.36 The immune system may also have fewer opportunities to repair itself after development. Consequently, toxic contamination of the womb plus high degrees of vulnerability during development creates the potential for live threats to developing children. Some of these threats materialise and can manifest themselves in substantial diseases later in life. Moreover, some substances are known to cause diseases or dysfunction to developing children as a result of in utero exposures to drugs, industrial chemicals or pollutants. These include, inter alia, thalidomide, mercury, DES, PCBs, lead, anti-convulsant drugs, alcohol, tobacco smoke, DDT and ionising radiation. Several of these have no known safe level. 37 Experimental animal studies have revealed a wider range of chemicals that can cause diseases or dysfunctions as a consequence of in utero exposures. For example, there is considerable concern about exposures to organophosphate pesticides, synthetic oestrogens (eg, bisphenol A (BPA) and phthalates), dioxin-like substances, air- and water-borne mercury, and classes of substances that cause harm through certain cellular receptors. Whether or not developing children will be at risk from other chemicals is not known, but their general susceptibility during this period of their lives – together with existing animal evidence – suggests more substances will likely pose threats. These findings may stress and strain causal models in the law. To the extent that tort law cannot easily accommodate them, this will preclude justice to wrongfully injured parties and undermine the role of torts with respect to toxic exposures during development.

IV. Some Causal Models Showing Different Biological Effects

Some toxicants can initiate, accelerate, catalyse, potentiate, facilitate or exacerbate disease, dysfunction and premature death. These causal models have been identified in human studies, experimental animals research or both. Sooner or later the law will need to address the biological reality that is emerging from scientific research. It is likely to be sooner rather than later. Moreover, recent trends in the law suggest that the law may be less than fully flexible in coping with these different possibilities.38

A.  Promotion Effects Some substances can speed up, accelerate or promote a disease that is likely to occur in any case. This was one toxicological issue in the US Supreme Court case, General Electric v Joiner.39 The plaintiffs had argued that Mr Joiner’s earlier smoking and perhaps his childhood exposure to smoke had initiated his lung cancer, but that exposure to PCBs had promoted the disease so that he contracted it very early in life. 36   Grandjean et al, ‘The Faroes Statement’ (n 7) 74; PM Rodier, ‘Developing Brain as a Target of Toxicity’ (1995) 103 Environmental Health Perspectives 73; D Rice and S Barone, Jr, ‘Critical Periods of Vulnerability for the Developing Nervous System: Evidence from Humans and Animal Models’ (2000) 108 Environmental Health Perspectives 511. 37   DT Wigle and BP Lanphear, ‘Human Health Risks from Low-Level Environmental Exposures: No Apparent Safety thresholds’ (2005) 2 PLoS Medicine 1, located at www.plosmedicine.org. 38  Cranor, Toxic Torts (n 1). 39   General Electric v Joiner 522 US 136 (1997).

The Challenge of Developing Science for the Law of Torts  269

100%   Percent  of  Animals  with  Tumours  

 

Cancers in most cases appear to go through three different stages before the full-fledged disease occurs: initiation, promotion, and progression. Initiation results from a ‘mutation in one or more cellular genes controlling key regulatory pathways of the cell’. Promotion is the ‘[s]elective functional enhancement of signal transduction pathways induced in the initiated cell and its progeny by the continuous exposure to the promoting agent’. Finally, progression is the ‘[c]ontinuing evolution of a basically unstable karyotype’.40 While the Supreme Court ruled that the trial judge had not abused her discretion in excluding plaintiffs’ expert’s testimony because in her view it rested on an insufficient scientific foundation, it did appear to recognise legally the idea of cancer promotion. Thus, a promotion causal model may not be particularly unusual. The law should certainly recognise promotion contributions to disease for torts because it clearly is one model with which scientists work. The graph that follows (Figure 1) is constructed from the data generated by Lucy Anderson et al at the National Cancer Institute.41 They show that PCBs can promote lung cancer. From the graph one can see that the percentage of animals with lung tumors tended to be not significantly different towards the end of their lives whether they had been exposed to the cancer initiator, dimethylnitrosamine, only, or whether they had been exposed to the cancer initiator plus a single dose of commercial PCBs. The scientists interpret this as showing that the exposure to both substances evidences a classic case of disease promotion. Consequently, even though someone exposed to a cancer-initiating substance might well have contracted a cancer in any case, he or she would have contracted it much sooner if exposed to a cancer promoter such as PCBs.

Percentage  of  Animals  with  Tumours  at  Weeks   of  Exposure  

90%   80%   70%   NDMA+PCBs   NDMA  Only   PCBs  Only   Controls  

60%   50%   40%   30%   20%   10%   0%  

           0                                16                                  28                                      52                                66         Weeks  of  Exposure  

40   HC Pitot III and VP Drgan, ‘Chemical Carcinogenesis’ in C Klaassen, ed, Casarett and Doull’s Toxicology, 6th edn (New York, Pergamon Press, 2001) 241–319, 277–79. 41   L Anderson et al, ‘Promotion by PCBs of Lung and Liver Tumors in Mice’ (1994) 15 Carcinogenesis 2245, 2246.

270  Carl F Cranor Consequently, if someone such as Mr Joiner had his lung cancer accelerated or promoted because of PCB exposures, the law should recognise this in order to compensate him for a shortened lifespan. In fact, the US Supreme Court seems to have given judicial recognition to this causal model in its Joiner decision, even though it overturned it on other grounds.

B.  Variable-hit Models Other in utero or perinatal exposures to toxicant can cause illnesses with varying degrees of ‘hits’ and varying latency periods. Consider Parkinson’s disease, or the related parkinsonism condition, as a well-studied example. Parkinson’s disease is typically seen in the elderly, rarely before the age of 50, with a peak incidence of the disease when people are in their 60s. It is among ‘the 15 leading causes of death in the U.S’. 42 In the early 1980s, scientists were alerted to sudden onset forms of Parkinson’s disease as a result of people recreationally taking a synthetic form of heroin that contained MPTP (l-methyl-4-phenyl-1,2,5,6-tetrahydropyridine). Physicians in Northern California received several persons in the clinics or hospitals with Parkinson’s disease. These individuals were in their late 20s to early 40s. One they called ‘The frozen addict’. This individual had what appeared to be Parkinson’s disease and could not move because his muscles were ‘frozen’. He otherwise had normal brain functions.43 Several other people who took the same synthetic heroin were partially affected, but seemingly not fully ‘frozen’. Research several years later revealed that similar parkinsonism effects were also seen in a chemist who synthesised MPTP (but did not ingest it) and who only worked with it for two to three weeks. Thus, MPTP may cause this condition by injection, ingestion, inhalation, or absorption through the skin. The addicts exposed to MPTP suffered from degeneration of cells in an area of the brain called the ‘substantia nigra’. MPTP appears to cause these cells sufficiently to deteriorate that the affected person manifests almost instant Parkinson’s-like conditions. This provided clues to the development of more common cases of Parkinson’s disease and led to the development of animals models of Parkinson’s disease that made it much easier to study and created the possibility of finding treatments.44 At the cellular level, Parkinson’s disease is characterised by a loss of cells that produce dopamine (dopaminergic cells) in the substantia nigra area of the brain.45 Moreover, Parkinson’s disease appears to act by means of a threshold mechanism. That is, ‘clinical symptoms of PD [Parkinson’s Disease] will only appear after [dopaminergic] function has been diminished by a set percentage of “maximal function”, and . . . this threshold usually does not occur until late-in-life’.46 However, the dopamine cell loss on rare occasions can occur as a result of an acute event, as the MPTP poisoning showed.

42   BK Barlow, DA Cory-Slechta, EK Richfield, M Thiruchelvam, ‘The Gestational Environment and Parkinson’s Disease: Evidence for Neurodevelopmental Origins of a Neurogenerative Disorder’ (2007) 23 Reproductive Toxicology 457, 458. 43   See also, AE Lang and RD Gordon Blair, ‘Parkinson’s Disease in 1984: An Update’ (1984) 131 Canadian Medical Association Journal 1031. 44  DB Calne et al, ‘Positron Emission Tomography After MPTP: Observations Relating to the Cause of Parkinson’s Disease’ (1985) 317 Nature 246. 45   Barlow et al, ‘The Gestational Environment and Parkinson’s Disease’ (n 42) 458. 46   ibid 459.

The Challenge of Developing Science for the Law of Torts  271 Researchers now have evidence in experimental studies that following early life exposure to a toxicant there can be a long period of ‘silent toxicity’ usually lasting decades during which this portion of the brain is operating at a suboptimal level but there are no overt Parkinson’s symptoms.47 Researchers describe this as follows: silent toxicity [is] a persistent morphological and/or biochemical injury that remains clinically unapparent unless unmasked. Through compensatory mechanisms, the affected system initially appears normal when considered against traditional biomarkers for toxicity (such as gross morphological and histological examination, behavioral assays or mortality studies), though the system is functionally compromised. It is hypothesized, therefore, that silent toxicity induces a state of ‘altered potential,’ or a ‘mutant steady state,’ where abnormal homeostasis (associated with increased vulnerability and higher risk of cell death) is established in a time-, tissue- and toxicantspecific manner, is maintained by altered gene expression, and is persistent and irreversible.48

Consequently, damage to cells in this particular area of the brain or some compromising of their function can set the stage for later disease or dysfunction long-removed from the event that damaged the cells. MPTP poisoning does not particularly illustrate a developmental problem. However, it shows that adverse effects can manifest themselves immediately, if the damage is severe enough. It also suggests that similar acting substances at lower doses from early life environmental exposures could so sufficiently compromise the neurological system that adverse effects can manifest themselves years later, long after exposure has ceased (and probably has been forgotten). If it turns out that certain classes of pesticides have this effect, as a number of scientists have proposed, it will be important to reduce or eliminate such exposures. For example, the children of farmers or farm-workers who are exposed to pesticides would be an important group to protect by reducing such exposures. There are good rodent models for the hypothesis that exposures to the herbicide paraquat and the fungicide maneb, as well as primate models for MPTP exposures.49 The herbicide paraquat is chemically similar to a metabolite of MPTP. This led scientists to believe that paraquat could cause similar problems. Mice treated in utero with either maneb or paraquat showed no evidence of Parkinson’s disease in adulthood. However, when male mice were treated with maneb in utero then dosed with paraquat as adults, within one week they began to show signs of Parkinson’s disease.50 Thus, two hits, one early, one much later, can trigger parkinsonism. The causal models from this research seem to be the following. One hit of a substance that damages the dopamine neurons might be enough if the dose is large enough. It appears that the MPTP in synthetic heroin was sufficient at least for a few individuals to suffer this effect. One hit combined with the normal aging process may also be sufficient to accelerate the disease. One pesticide hit in utero may not be sufficient to trigger the disease, but, as the mouse studies show, a second hit by a different pesticide later in life jointly may cause the onset of disease. Jerry Heindel summarises some of the possibilities for developmental origins of disease by generalising on the above points.   ibid 459.   ibid 458. 49   JJ Heindel, ‘Animal Models for Probing the Developmental Basis of Disease and Dysfunction Paradigm’, (2008) 102 Basic & Clinical Pharmacology & Toxicology, 76, 79. 50   BK Barlow, EK Richfield, DA Cory-Slechta, M Thiruchelvam, ‘A Fetal Risk Factor for Parkinson’s Disease’ (2004) 26 Developmental Neuroscience 11, and M Thiruchelvam, EK Richfield, BM Goodman, RB Baggs, and DA Cory-Slechta, ‘Developmental Exposure to the Pesticides Paraquat and Maneb and the Parkinson’s Disease Phenotype’ (2002) 23 NeuroToxicology 621. 47 48

272  Carl F Cranor The environmental insult could act via a one-hit or two/ three-hit scenario. That is, there could be an in utero exposure or neonatal exposure that would lead by itself to pathophysiology later in life or there could be in utero exposure combined with a neonatal exposure [same or different compound(s)] or adult exposure that would trigger or exacerbate the pathophysiology. The pathophysiology or functional change that results from the exposures/insult could lead to: (i) the occurrence of a disease that otherwise would not have happened, (ii) an increase in risk for a disease that would normally be of lower prevalence, or (iii) either an earlier onset of a disease that would normally have occurred or (iv) an exacerbation of the disease. The pathophysiology could have a variable latent period ranging from onset in the neonatal period, to early childhood, to puberty, to early adulthood or to late adulthood depending on the toxicant, time of exposure and tissue/organ affected. Importantly, the effects may be transmitted to future generations through the germ line. Finally, the effects of in utero exposure to toxic environmental chemicals can occur in the absence of birth defects or reduced birth weight that, of course, makes it more difficult to assess than effects due to severe nutritional deficits during development.51

The fact that various substances can cause neurotoxic effects in people is not merely conjecture. Other neurotoxicants appear to cause Parkinson’s-like conditions, although the number of ‘hits’ cannot be easily specified. Manganese fumes can cause early Parkinson’slike effects.52 In addition, emerging evidence suggests that exposure to pesticides is associated with Parkinson’s disease. A family-based epidemiological study adds support ‘to the hypothesis that pesticide exposure is positively associated with risk of PD [Parkinson’s Disease]’. In particular, researchers found that exposure to ‘organochlorines, organophosphorus compounds, chlorophenoxy acids/esters, and botanicals as potential risk factors for PD’.53 A most recent epidemiological study provides additional supporting evidence, echoing the animal studies. As Costello et al have noted: ‘[E]xposure to a combination of maneb and paraquat increases PD risk, particularly in younger subjects and/or when exposure occurs at younger ages’.54 Exposure to these two pesticides within 500 metres of homes increased Parkinson’s disease about 75 per cent.55 Scientists believe that pesticides can have such effects because many of them are neurotoxicants. Quite recently, trichloroethylene (TCE), the degreaser, anesthetic, and contaminant of drinking water (found in many Superfund priority sites), has also been proposed as a contributor to Parkinson’s disease, especially as a consequence of workplace exposure.56 TCE is associated with parkinsonism among workers exposed in a Kentucky plant; they show evidence of neuronal damage to the substantia nigra. In short: TCE is implicated as a principal risk factor for parkinsonism based on its dopaminergic neurotoxicity in animal models, the high levels of chronic dermal and inhalation exposure to TCE by the three workers with Parkinson’s disease, the motor slowing and clinical manifestations of parkinsonism in coworkers clustered around the TCE source, and the mounting evidence of neurotoxic effects in other reports of chronic TCE exposure.57   JJ Heindel, ‘Animal Models’ (n 49) 77.   Grandjean and Landrigan, ‘Developmental Neurotoxicity of Industrial Chemicals’ (n 14) 2171.   DB Hancock et al, ‘Pesticide Exposure And Risk of Parkinson’s Disease: A Family-Based Case-Control Study’ (2008) BMC Neurology 2008, located at 8:6 doi:10.1186/1471-2377-8-6. 54   S Costello, M Cockburn, J Bronstein, X Zhang and B Ritz, ‘Parkinson’s Disease and Residential Exposure to Maneb and Paraquat From Agricultural Applications in the Central Valley of California’ (2009) 169 American Journal of Epidemiology 919. 55   ibid 919. 56   DM Gash, K Rutland, NI Hudson et al, ‘Trichloroethylene: Parkinsonism and Complex 1 Mitochondrial Neurotoxicity’ (2008) 63 Annals of Neurology 184. 57   ibid 191. 51 52 53

The Challenge of Developing Science for the Law of Torts  273 The significance of this for the law seems to be that different defendants could have products that contribute to one or two ‘hits’ followed by natural aging, and after a possibly long latency period accelerate the onset of Parkinson’s disease. Thus, the law would need to recognise a one- or two-hit model, combined with natural processes and a long latency period before the adverse effects actually appear. Of course, plaintiffs would need to provide sufficient proof of these facts, which in all likelihood would not be easy. However, if these points could be established, the law must be prepared to recognise them within the causation requirements of torts.

C.  Immulogical Weakening In utero or perinatal exposures to toxicants can permanently damage the immune system, creating time bombs that leave a person at risk of various diseases over an entire lifetime. Exposure to a toxicant that does not cause the death or obvious dysfunction of a developing foetus ‘may produce an unrecognisable immunotoxic alteration until the postnatal immune system is placed under subsequent stress’.58 For example, in utero exposures to lead cause postnatal effects when the immune system is subjected to a viral infection.59 Experimental studies ‘place the developing immune system on par with the developing neurological system for lead sensitivity’.60 That is, the assessment of risks in adults does not predict how sensitive developing children will be to concentrations to particular toxicants or to a range of long-term consequences.61 These effects have been seen in five chemicals that have received attention – DES, Valium (diazepam), lead, dioxin, and tributyltin oxide (an industrial preservative, disinfectant, and antifouling agent). They also add to the science that shows that developing children are susceptible to toxic effects at lower levels than adults. The researchers established that ‘[f]or all five chemicals, the developing immune systems was found to be at greater risk than that of the adult, either because lower doses produced immunotoxicity, adverse effects were more persistent, or both’.62 Some of these effects appear to persist into old age. Lead, the insecticide, methoxychlor, and ethanol, the active ingredient in alcohol, cause effects that would be unpredictable from studies on adult animals or humans.63 More broadly, in a review of the five immunotoxic chemicals above, scientists concluded that until there are data to the contrary, one should assume that ‘development and maturation of the immune system constitutes a period of greater sensitivity to xenobiotic exposure’.64 Quite recently, researchers have found that activation of a particular cellular receptor, the aromatic hydrocarbon receptor (Ah receptor or AhR): alters different elements of the immune system at different times during development by affecting different tissue targets. In particular, diminished T-cell responses arise due to deregulated events within bone marrow–derived cells. In contrast, increased interferon gamma levels in the infected lung result from AhR-regulated events extrinsic to bone marrow–derived cells, and require AhR 58  RR Dietert and MS Piepenbrink, ‘Perinatal Immunotoxicity: Why Adult Exposure Assessment Fails to Predict Risk’ (2006) 114 Environmental Health Perspectives 480. 59   ibid 480. 60   Luebke et al, ‘The Comparative Immunotoxicity of Five Selected Compounds’ (n 20) 11. 61   Dietert and Piepenbrink, ‘Perinatal Immuniotoxicity’ (n 58) 480. 62   Luebke et al, ‘The Comparative Immunotoxicity of Five Selected Compounds’ (n 20) 5. 63   Dietert and Piepenbrink, ‘Perinatal Immunotoxicity’ (n 58) 480. 64   Luebke et al, ‘The Comparative Immunotoxicity of Five Selected Compounds’ (n 20) 5.

274  Carl F Cranor agonist exposure during early gestation. . . . AhR activation causes long-lasting functional alterations in the developing immune system, whereas the impact on the mature immune system is transient.65

They add: There is growing evidence that prenatal and early postnatal environmental factors influence the development and programming of the immune system, causing long-lasting negative health consequences . . . [Moreover,] diminished T-cell responses arise due to deregulated events within bone marrow . . . [and cause] long-lasting functional alterations in the developing immune system, whereas the impact on the mature immune system is transient. 66

Consequently, a company could make a product that is known to cause suppression of the immune system, a pregnant woman and her developing foetus could be exposed to it, and when the child matures to adulthood, he or she could suffer a much more serious series of illnesses than ordinary people, all because of early exposure to the toxicant. If such claims could be established with sufficient evidence, tort law should allow for legal cases based on the model to be brought.

D.  Additive Effects Toxic substances may add to the adverse effects resulting from exposure to other toxicants or to naturally occurring substances that can have the same effect. Thus, two or more toxicants may add to produce a disease (much like the two-hit model above) or a toxicant plus a naturally occurring chemical may together contribute to disease, accelerate the onset of an illness or worsen it. Such effects are clearly seen concerning the effects of natural and synthetic oestrogen on women’s breast cancer. In the US and many western European countries, breast cancer rates among women have been rising. In the US a woman has about a one in eight or a one in nine chance of developing breast cancer during her lifetime.67 In the United Kingdom, her chances are one in nine or about eleven out of one hundred women. Although breast cancer rates have levelled off or actually begun to decline in the US, they remain at very high levels. The observed increase in these rates is not due to improved diagnostic screening. There is often an increase in disease rates after better diagnostic techniques are developed, but then the diagnostic effect tends to disappear even if disease rates are increasing. Better diagnoses have not hidden long-term increases in incidence breast cancer; it is a real increase. 68 A percentage of breast cancers is related to genetic inheritance; some are traceable to ‘environmental factors’.69 Kortenkamp reports that about five per cent or one in twenty breast cancer cases are traceable to inherited genetic predispositions.70 Other contributors to the disease are from ‘environmental influences’.71 Environmental factors could be any non-genetic factors that contribute to breast cancer, but this is too broad. A better 65   Hogaboam, Moore and Lawrence, ‘The Aryl Hydrocarbon Receptor Affects The Immune System’ (n 29) 160–70. 66   ibid 160. 67   American Cancer Society, Breast Cancer Facts & Figures 2007–2008, located at www.cancer.org, p 11. 68   Kortenkamp, ‘Breast Cancer and Exposure to Hormonally Active Chemicals’ (n 19) 4. 69  ibid. 70  ibid. 71  ibid.

The Challenge of Developing Science for the Law of Torts  275 conception of environmental factors would be ‘avoidable contributions’ to breast cancer and would include ‘work place exposures, food contaminants, pharmaceuticals, chemicals in consumer products, air, water and soil, and physical factors such as radiation’.72 What percentage of breast cancer cases result from these avoidable factors? Studies of identical twins and women with a genetic predisposition to cancer indicate that about thirty-three per cent of breast cancers in these pairs are attributable to genetic factors, whereas environmental factors not common to each twin in a pair accounted for sixtyseven percent of the difference in breast cancer rates between identical twins. Research on women who have a predisposition to breast cancer because of mutated tumour suppressor genes also supports this conclusion.73 A large body of evidence supports the claim that a woman’s increased exposure to oestrogens, natural or synthetic, can increase her risks of breast cancer. In utero and early exposures seem even more important as some ‘natural’ experiments, the DES cases and twin studies illustrate. Bernardino Ramazzini, a seventeenth century Italian occupational physician, observed that childless women or those in nunneries had greater rates of breast cancer than women who had children. He conjectured that their celibacy and childlessness contributed to their breast cancer.74 His observations were correct, but he did not understand the role of hormones at that time. The earlier women bear children, the lower their rates of breast cancer, while the later they give birth, the greater their odds of contracting the disease. Thus, there is a protective effect from early childbirth. It contributes to more complete growth of the breast tissues earlier in life, thus reducing the ultimate number of cells in the breast that are susceptible to cancer.75 Early puberty or late menopause each increases the risks of breast cancer. The longer women are exposed to their own oestrogens, the greater their risks of breast cancer.76 If puberty is delayed or menopause occurs earlier, both tend to lower breast cancer rates, because either of these occurrences somewhat lessens total exposure to oestrogens. In addition, when developing females are exposed to higher levels of their mothers’ oestrogens in utero, their breast cancer rates go up. Women who are twins, genetically identical or not, are exposed to higher concentrations of oestrogens in utero and have sixty to one hundred per cent higher rates of breast cancer later in life.77 Female offspring who are nonidentical twins of a brother when in utero have a somewhat higher rate of breast cancer than those that are in utero with an identical female twin. The greater exposure to natural oestrogens increases the chances of breast cancer. In utero exposures to the synthetic oestrogen diethylstilbestrol (DES) have added to the evidence that a woman’s increased exposure to natural or synthetic oestrogens can increase her risk of breast cancer. When DES daughters reach middle age, they have a breast cancer rate that is double (one hundred per cent greater than) that of the general population.78  ibid.   ibid 5.   National Institutes of Health, National Cancer Institute, ‘Cell Biology and Cancer’, located at science-education.nih.gov. 75   Kortenkamp, ‘Breast Cancer and Exposure to Hormonally Active Chemicals’ (n 19) 5 and The US National Institutes of Health, the National Cancer Institute, Fact Sheet, ‘Pregnancy and Breast Cancer Risk’, located at www. cancer.gov. 76   Kortenkamp, ‘Breast Cancer and Exposure to Hormonally Active Chemicals’ (n 19) 5. 77  HA Weiss et al, ‘Prenatal and Perinatal Risk Factors for Breast Cancer in Young Women’ (1992) 89 Epidemiology 181. 78   Kortenkamp, ‘Breast Cancer and Exposure to Hormonally Active Chemicals’ (n 19) 6. 72 73 74

276  Carl F Cranor The oestrogen-breast cancer explanation receives additional confirmation from hormone replacement therapy to reduce the effects of menopause. In a large British study, women who received some form of hormone replacement – whether oestrogen only or a combined oestrogen-progestin treatment – had ‘increased breast cancer risks’.79 Use of this therapy in the United Kingdom may have resulted in an extra 20,000 breast cancer cases (versus no use of hormone replacements).80 US researchers identified this risk from hormone replacement therapy and recommended against using it, resulting in a downturn in breast cancer rates. This total body of evidence strongly supports the claim that a woman’s increased exposure to oestrogens, natural or synthetic, can increase her risks of breast cancer. In utero and early exposures seem even more important as the DES and twin cases illustrate. Because of this evidence, scientists are concerned about substances that mimic oestrogens. Bisphenol A (BPA) is another synthetic oestrogen synthesised about the same time as DES. It is produced and used at the rate of about 2.3 billion pounds81 annually in the US and most citizens are contaminated with it.82 Moreover, although BPA has a very short halflife in a human body, the detection of BPA in most citizens suggests that they are continually exposed from various sources. Thus, ‘there is chronic, low level exposure of virtually everyone in developed countries to BPA’.83 Current body burdens of BPA in adults are ‘within the range that is predicted to be biologically active in over 95% of people sampled’.84 That is, current human body burdens could cause adverse effects, based on experimental studies. A consensus committee of researchers is concerned that BPA may contribute to breast cancer in women, among other diseases.85 In addition, because of exposures in utero or during other periods of early development adverse effects may not manifest themselves until long after exposures occurred. This substantial latency period will make tracing the causes of disease or dysfunction quite difficult. Breast cancer and other diseases that may be triggered by BPA are irreversible. Another scientific committee, while less concerned about adult exposures, is also worried about BPA increasing the risks for breast cancer and other diseases.86 However, the implications of BPA and related substances for contributing to breast cancer are much wider than the above conclusions suggest. A British researcher, Andreas Kortenkamp, points out that substances that attach to similar receptors or sites in mammalian bodies can make additive contributions to disease or dysfunction. For oestrogens ‘human risk assessment could work on the basis of the rebuttable hypothesis that dose addition is applicable, but should rapidly modify this practice as soon as evidence to the contrary becomes available’.87 Of even greater concern is that:   ibid 6.  ibid. 81   ie production by weight. 82  National Institute of Environmental Health Sciences, National Institutes of Health, US Department of Health and Human Services, ‘Draft NTP Brief on Bisphenol A’ 14 April 2008, www.cerhr.niehs.nih.gov. 83   FS vom Saal et al, ‘Chapel Hill Bisphenol A Expert Panel Consensus Statement: Integration of Mechanisms, Effects in Animals and Potential Impact to Human Health at Current Exposure Levels’ (2007) 24 Reproductive Toxicology 131, 136. 84   ibid. 85  ibid. 86  National Institute of Environmental Health Sciences, National Institutes of Health, US Department of Health and Human Services, ‘Draft NTP Brief on Bisphenol A’ 14 April 2008, www.cerhr.niehs.nih.gov. 87   A Kortenkamp, ‘Low-Level Exposure to Multiple Chemicals: Reason for Human Health Concerns?’ (2007) 115 Environmental Health Perspectives, Supplement 1, 106, 113. 79 80

The Challenge of Developing Science for the Law of Torts  277 [g]ood evidence is available to show that joint effects occur even when all mixture components are present at levels below doses that cause observable effects. In view of this evidence, the traditional chemical-by-chemical approach to risk assessment is hard to justify, and the ground is prepared to seriously consider group-wise regulation of EDs [endocrine disruptors].88

There is profound significance to his last point. Too often some scientists or some companies affected by a legal action will argue that there is no evidence that at very low exposure levels substances in fact cause adverse effects. There are many problems with such assertions. Epidemiological studies can easily be too small or too short to discover risks of harm and as a result they may be too insensitive to detect some effects from low concentration exposures.89 Moreover, it is even more difficult for human studies to identify subtle effects. Diseases may have such long latency periods that they have not yet had time to materialise. They may have multiple causes, some of which mask the subtle effects caused by low-dose exposures. And it is difficult for studies to identify new causes of adverse effects from variation of disease patterns in existing diseases, especially if they are common. An additional and most serious problem is the additive effect of low concentrations of synthetic substances with background levels of similar substances that together can cause adverse outcomes. This appears to be the case with BPA. Even at tiny doses in experimental studies, this synthetic oestrogen seems to add to natural levels of oestrogens in animals to produce breast cancer as well as some other adverse reproductive outcomes. Thus, con­ centrations that by themselves are not known to cause adverse effects appear to add to natural levels of oestrogens or other sources of oestrogenic compounds and cause diseases or dysfunction. Moreover, if there is more than one kind of synthetic oestrogen that attaches to similar receptors, as scientists have found, the problems increase. Each synthetic oestrogen taken by itself may not pose problems even when added to natural background rates. However, when they are added together, the several exposures may be more than sufficient to produce adverse outcomes or accelerate disease processes. Consequently, in the technical language of science, just because a substance is present in persons’ bodies at levels below which adverse effects are seen (the No Observed Adverse Effect Level, or NOAEL), it does not follow that this is the same as a No Effect Level in humans. When the effects of substances are additive, there is not necessarily a no effect level unless the concentration of the substance is at zero. Other classes of substances also appear to have additive effects. Researchers have found such effects for dioxin-like substances, such as dioxins, certain (co-planar) PCBs and dibenzofurans, among others.90 Dioxin-like substances attach to the Ah receptor and activate a metabolic pathway that tends to make chemicals more toxic (although it does detoxify some). However, understanding the mechanism of action by which adverse effects result is a rare exception, not the rule. Another class of substances where both biological sites and mechanisms of action are understood are organophosphorus and carbamate pesticides that tend to cause similar adverse neurological outcomes.91 88   A Kortenkamp, ‘Ten Years of Mixing Cocktails: A Review of Combination Effects of Endocrine-Disrupting Chemicals’ (2007) 115 Environmental Health Perspectives, Supplement 1, 98, 104. 89  Cranor, Toxic Torts (n 1) 225–27. 90   SA Kafafi, HY Afeefy, AH Ali, HK Said and AG Kafafi, ‘Binding of Polychlorinated Biphenyls to the Aryl Hydrocarbon Receptor’ (1993) 101 Environmental Health Perspectives 422. 91   DJ Ecobichon, ‘Toxic Effects of Pesticides’ in C Klaassen, Casarett and Doull’s Toxicology, 6th edn (New York, Pergamon Press, 2001), 763, 774–82.

278  Carl F Cranor Taken together, the human studies involving complex mixtures of dioxin-like compounds, including both dioxin and non-dioxin-like PCBs, suggest that levels present in the general population may be associated with subtle signs of neurological dysfunction, delays in psychomotor development, alterations in thyroid hormone status, and changes in immunological functions.92 Dioxin-like compounds tend to have half-lives of three to nineteen years.93 The specific half-life for PCBs appears to be about eight years for workers, but can be longer. And estimates of half-lives likely underestimate the persistence of a substance.94 Substances that are toxic as a result of attaching to the Ah receptor tend to be associated ‘with reproductive, immunologic, teratogenic and carcinogenic effects for PCBs’.95 Finally, there are substances that, although they do not attach to identical cellular receptors, as do oestrogens and dioxin-like substances, act via a common mechanistic pathway likely producing additive effects. Researchers at a workshop synthesised several case studies of different toxicological pathways that suggested the possibility of additive effects. For instance, many environmental substances can disrupt thyroid hormone levels. In turn, this can have neurodevelopmental adverse effects, eg, reducing IQ in developing children. Adverse effects on thyroid uptake by pregnant women can adversely effect the neurological development of developing foetuses in utero. Moreover, women often do not have sufficient thyroid reserves to counteract these effects during pregnancy and developing children do not have thyroid hormones stored in their bodies as do adults.96 Studies have found similar effects among antiandrogen substances that can adversely affect male reproduction, as well as immunotoxicity effects similar to those indicated above.97

E.  Genetic Susceptibility In addition to general biological disease tendencies that result from additive, multi-hit or immunologic suppression resulting from in utero exposures to toxicants, such tendencies also result for specific subpopulations as a result of genetic susceptibility. Human populations have much greater biological variation than animals on which controlled experiments are conducted. As a result there can be more or less susceptible members of the species as a result of genetic variability and diversity. They might be either more or less susceptible to toxic effects or have greater or lesser defences against toxicants. For some substances, which in turn raise a general concern, because of genetic variability, humans may be more sus­ ceptible than animals on which experiments with toxicants are conducted. For example, Perera et al, showed that environmental exposures to polycyclic aromatic hydrocarbons and

92  L Birnbaum, ‘Developmental Effects of Dioxins’ in KS Korach (ed), Reproductive and Developmental Toxicology (New York, Marcel Dekker Inc, 1998)106. 93  CDC, Report on Human Exposure, 139. 94   US Environmental Protection Agency, ‘PCBs: Cancer Dose-Response Assessment’ 23–24. (The half-life of a chemical is the time it takes for one-half of it to be eliminated from a person’s body.) 95   PRS Kodavanti, ‘Neurotoxicity of Persistent Organic Pollutants: Possible Mode(s) of Action and Further Considerations’ (2005) 3 Dose-Response 273, 274. 96   TJ Woodruff, L Zeise, DA Axelrad, KZ Guyton et al, ‘Meeting Report: Moving Upstream – Evaluating Adverse Upstream End Points for Improved Risk Assessment and Decision-Making’ (2008) 116 Environmental Health Perspectives 1568, 1570. 97   Woodruff et al (n 96) 1571–73.

The Challenge of Developing Science for the Law of Torts  279 second hand tobacco smoke cross the placenta and create adducts on DNA.98 Polycyclic aromatic hydrocarbons are a large group of substances formed during incomplete combustion of organic compounds such as coal, gas and oil, garbage, tobacco or charbroiled meat. Some are manufactured for use in dyes, pesticides, medicine and various tars, oils, creosote and roofing products.99 In urban areas or households with smokers a common source would be second hand smoke. When substances form ‘adducts’ on DNA they are bound to the DNA, altering its function and often causing mutations or incorrect repair leading to cancers or other diseases.100 Some subpopulations of foetuses had greater numbers of the PAH-DNA adducts and increased sensitivity to genetic damage compared with the mother.101 Other researchers have ‘found that an individual’s susceptibility to organophosphate pesticides may vary by age and genotype’.102 In particular children and adults with a variant of a particular gene had lower levels of an enzyme that assists in metabolising organo­ phosphate pesticides. As a result they ‘may be at higher risk of health effects from organophosphate exposure’. 103 In addition, given the variation of enzymes that detoxify organophosphate pesticides, researchers estimate that ‘most, if not all, newborns, as well as a subpopulation of adults, will exhibit significantly increased sensitivity to organophosphate exposure’.104 Children are more susceptible than adults because they have lower levels of this enzyme at birth and because of genetic diversity, but adults can also have low levels, and thus can be more susceptible to organophosphates, simply because of genetic variation.105 When genetic susceptibility is added to the fact that many of us are exposed to multiple substances that can act by similar mechanisms, this raises a more worrisome issue. Woodruff et al compiled several case studies illustrating the reasons policy-makers would need to consider both sets of factors. Recall from above that numerous substances can disrupt thyroid hormones, and produce adverse effects on neurologic development. These include perchlorate, (an ingredient in rocket fuel that was poorly disposed of), nitrates, used for fertilisers, some PCBs, some chlorinated pesticides, dioxin and bisphenol A.106 Exposures to low levels of some of these compounds that might cause perturbations in thyroid hormones in developing children, ‘would not present an adverse challenge to a healthy adult’.107 Similarly, several substances can adversely affect the immune system, especially in developing children. Even though these may not act by identical mechanisms, one might add to the effects of others producing a cumulative effect.108 Pre-existing exposures 98  Miller et al, ‘Differences Between Children and Adults’ (n 18) 409–10, and F Perera, W Jedrychowski, V Rauh and RM Whyatt, ‘Molecular Epidemiologic Research on The Effects of Environmental Pollutants on The Fetus’ (1999) 107 Environmental Health Perspectives (Suppl 3) 451. 99  Centers for Disease Control, Agency for Toxic Substances and Disease Registry, ‘Polycyclic Aromatic Hydrocarbons: Fact Sheet’, located at www.atsdr.cdc.gov. 100   RJ Preston and GR Hoffman, ‘Genetic Toxicology’ in Curtis D Klaasen (ed), Casarett and Doull’s Toxicology 6th edn (New York, McGraw-Hill, 2001) 321–50, 327. 101   Miller et al, ‘Differences Between Children and Adults’ (n 18) 409–10 and F Perera, W Jedrychowski, V Rauh and RM Whyatt, ‘Molecular Epidemiologic Research on The Effects of Environmental Pollutants on The Fetus’, (1999) 107 Environmental Health Perspectives (Suppl 3) 451. 102   B Eskenazi, LG Rosas, AR Marks, A Bradman, K Harley, N Holland, C Johnson, L Fenster and DB Bar, ‘Pesticide Toxicity and the Developing Brain’ (2008) 102 Basic & Clinical Pharmacology & Toxicology 228, 235. 103  ibid. 104  ibid. 105   CE Furlong, N Holland, RJ Richter, A Bradman, A Ho and B Eskenazi, ‘PON1 Status of Farmworker Mothers and Children as a Predictor of Organophosphate Sensitivity’ (2006) 16 Pharmacogenetics and Genomics 183, 189. 106   Woodruff et al, 1569. 107   ibid 1573. 108   ibid 1572–73.

280  Carl F Cranor that can add to effects of other substances increase the baseline of risks from those classes of substances. If genetic vulnerability is added to the mix, meaning that some subpopulations are more vulnerable to the particular disease effects, then ‘slight perturbations in upstream biological indicators are more likely to increase risk of subsequent downstream events given an already more activated state among segments of the population’.109 They summarised the current research in these areas by arguing that each case study shows ‘the importance of considering preexisting or continuous exposure to environmental chemicals as well as preexisting biological or disease susceptibilities that contribute independently to risk of overt disease’.110 Co-exposures and genetic or other susceptibility ‘can increase the baseline risk of the population or enhance already initiated disease processes’.111

F.  Multi-Generational Effects In utero exposure to methoxychlor or vinclozolin in experimental animals studies not only resulted in effects in the adult due to in utero exposure (eg increased testicular apoptosis of developing germ cells leading to infertility starting at 100 days of age in a small percentage of the animals, up to 20 per cent), but also showed that the effects could be transmitted through the male germ line for at least four generations.112 Vinclozolin is a fungicide utilised on grapes grown for wine as well as some other agricultural crops. It is an antiandrogenic compound, ie it inhibits the effects of the male hormone androgen. It has been observed that ‘[e]mbryonic exposure to vinclozolin can influence sexual differentiation, gondal formation, and reproductive functions in the F1 generation’. An F0 gestating female animal was exposed just once (the only animal exposed) to vinclozolin. Then researchers bred her offspring and their offspring to wild types through several generations. In order to determine transgenerational effects, researchers must breed animals through three additional generations. An F0 exposure results in exposure to the F1 offspring and the F2 germline. Consequently, in order to see transgenerational effects researchers must examine the F4 generation. The exposure to an F0 mother caused tumors in 12–33 per cent of the F1-F4 generation offspring compared with controls (none). Males had prostatic lesions at a rate of 45–55 per cent. Vinclozolin also induced kidney lesions. Fifteen to thirty-eight per cent of the F1-F4 generation animals had spermatogenic cell death, gross defects in spermatogenesis or complete lack of spermatogenesis. That is, some animals had damaged sperm cells but some were infertile.113 Exposure also appeared to increase the odds of inner ear infections and bacterial infections (12–33 per cent), whereas controls had no inflammation. Most of these adverse effects were seen in the F4 generation, which indicates that the effects were preserved genetically. Most likely this occurred by means of an epigenetic mechanism, not a change in the genetic sequence in the animals.   ibid 1572.   ibid 1573. 111  ibid. 112  Heindel, ‘Animal Models’ (n 49) 81, citing MD Anway, AS Cupp, M Uzumcu, M Skinner, ‘Epigenetic Transgenerational Actions of Endocrine Disruptors on Male Fertility’ (2005) 308 Science, 1466; MD Anway, C Leathers, M Skinner, ‘Endocrine Disruptor Vinclozolin Induced Epigenetic Transgenerational Adult-onset Disease’ (2006) 55 Endocrinology 5515. 113   Anway, Leathers, Skinner (n 112) 5518. 109 110

The Challenge of Developing Science for the Law of Torts  281

V. Conclusion

These causal models found by scientific evidence are real. They pose the question as to whether tort law can adapt to such scientific discoveries. Tort law will have to face them sooner or later – probably sooner. The major challenge for law and science will be whether tort law can accommodate these models to ensure justice to people who might be injured as a result of one of these patterns of exposure.

14 The NESS Account of Natural Causation: A Response to Criticisms RICHARD W WRIGHT*

I. Introduction

The NESS (necessary element of a sufficient set) account of natural (scientific, ‘actual’, ‘factual’) causation is usually acknowledged to be a more satisfactory and comprehensive account than the traditional sine qua non (‘but for’) account.1 However, objections have been raised to the claim that the NESS account fully captures the concept of natural causation (hereafter ‘causation’) and properly handles all types of situations. Various types of counter-examples have been proposed. More fundamentally, it is argued that the NESS account is viciously circular, since causal terminology often is used in its elaboration and it relies upon the concept of causal laws.2 Many of the objections raised against the NESS account assume that it is essentially the same as Herbert Hart’s and Tony Honoré’s ‘causally relevant factor’ account and John Mackie’s INUS account. In section II of this chapter I distinguish these three accounts, which differ in important ways that make the latter two accounts vulnerable to objections to which the NESS account is immune, and I offer an account of causal laws that I believe *  I am grateful to my critics for raising issues that forced me to develop further my elaborations of the NESS account, to Horacio Spector for helpful discussions during the early stages of preparation of this chapter, and to David Cheifetz and Tony Dillof for helpful comments during the final stage. 1   eg American Law Institute, Restatement of the Law, Third, Torts: Liability for Physical and Emotional Harm [Restatement Third] (St Paul, Minnesota, American Law Institute, 2010) s 26 comment c, s 27 comments a, f–i and reporters’ notes; A Beever, Rediscovering the Law of Negligence (Oxford, Hart Publishing, 2009) 417–21, 426; DA Fischer, ‘Insufficient Causes’ (2006) 94 University of Kentucky Law Review 277, 281–84, 317; R Fumerton and K Kress, ‘Causation and the Law: Preemption, Lawful Sufficiency, and Causal Sufficiency’ (2001) 64:4 Law and Contemporary Problems 83, 83–84, 95–97; T Honoré, ‘Necessary and Sufficient Conditions in Tort Law’ in DG Owen (ed), Philosophical Foundations of Tort Law (Oxford, Oxford University Press, 1990) 363, 363–64, 366, 367, 374–77, 381–85; M Kelman, ‘The Necessary Myth of Objective Causation Judgments in Liberal Political Theory’ (1987) 63 Chicago-Kent Law Review 579, 601–03; W Lucy, Philosophy of Private Law (Oxford, Oxford University Press, 2007) 170–76; MS Moore, Causation and Responsibility (Oxford, Oxford University Press, 2009) 473–74, 486–90, 493; SR Perry, ‘The Impossibility of General Strict Liability’ (1988) 1 Canadian Journal of Law and Jurisprudence 147, 157; J Stapleton, ‘Choosing What We Mean by “Causation” in the Law’ (2008) 73 Missouri Law Review 433, 443–44, 471–72, 474; JJ Thomson, ‘Some Reflections on Hart and Honoré, Causation in the Law’ in M Kramer et al (eds), The Legacy of HLA Hart: Legal, Political and Moral Philosophy (Oxford, Oxford University Press, 2008) 143, 144, 148–50, 157, 163–64. 2   Beever (n 1) 422–25; Fischer (n 1); Fumerton and Kress (n 1) 84, 97–104; Kelman (n 1) 603–08; Moore (n 1) 477, 484, 494–95; Stapleton (n 1) 472–79; Thomson (n 1) 147–54.

286  Richard W Wright rebuts the claim that the NESS account is viciously circular. In section III I argue that the NESS account handles properly the various types of situations that have been raised as alleged counter-examples to its comprehensive validity. In both parts, the most significant criticisms are addressed in the text, while other criticisms are discussed in the footnotes.

II. Distinguishing the NESS Account from Hart and Honoré’s and Mackie’s Accounts

The NESS account often is erroneously equated with Hart and Honoré’s account of a ‘causally relevant factor’ and John Mackie’s account of an INUS condition. This is understandable, since, although I have often noted the significant differences between the NESS account and the other two accounts, I drew on both of the other accounts in developing and elaborating the NESS account, credited Hart and Honoré with the initial elaboration of the weak sense of necessity that underlies each of the three accounts and distinguishes them from the traditional sine qua non account, and loosely used the NESS acronym to refer to Hart and Honoré’s account as well as my own account.3 However, the criticisms directed indiscriminately against all three accounts do not apply to the NESS account, owing to the significant differences between it and the other two accounts. The weak sense of necessity that underlies all three accounts was initially elaborated in 1959 in Hart and Honoré’s seminal treatise, Causation in the Law.4 Contrary to the traditional insistence in law, and by many philosophers, that a cause be a condition that ‘made a difference’ by being strongly necessary for the occurrence of the consequence, in the sense that without it the consequence would not have occurred on the particular occasion, Hart and Honoré employed John Stuart Mill’s scientific-method based account of causal laws and singular instances of causation5 to explain that a ‘causally relevant factor’ need merely be ‘necessary just in the sense that it is one of the set of conditions jointly sufficient for the production of the consequence: it is necessary because it is required to complete this set’.6 Under this weak sense of necessity, which is also referred to as strong sufficiency, necessity is subordinated to sufficiency: a causally relevant factor need merely be necessary for the sufficiency of a set of conditions sufficient for the occurrence of the consequence, rather than being necessary for the consequence itself as in the sine qua non account.7 Hart and Honoré demonstrated the superiority of their weak-necessity account of a causally relevant factor to the sine qua non account in a variety of situations.8 They also correctly emphasised that causation is an empirical rather than a merely analytical, logical or identity relation.9 However, their account had some significant deficiencies, all of which are 3   RW Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (1988) 73 Iowa Law Review 1001, 1018–34. 4   HLA Hart and AM Honoré, Causation in the Law (Oxford, Oxford University Press, 1959). 5   HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Oxford University Press, 1985) 13–22, 111–13; see JS Mill, A System of Logic bk III ch IV s 1, ch V ss 2–3, ch VIII ss 1–4, ch X ss 1–3; RW Wright, ‘Causation in Tort Law’ (1985) 73 California Law Review 1735, 1788–91. 6   Hart and Honoré (n 5) 112. 7   See Wright (n 3) 1020–21 for discussion of the different senses of necessity and sufficiency. 8   Hart and Honoré (n 5) 122–28, 206–07, 235–53. 9  ibid 114–15; see Fumerton and Kress (n 1) 90–93; Thomson (n 1) 148 fn 7. Also to be ruled out are mere mereological relations – the relation between some entity and its constituent parts. See J Collins, N Hall and LA Paul, ‘Counterfactuals and Causation: History, Problems, and Prospects’ [Collins, Hall and Paul, Introduction] in

The NESS Account: Response to Criticisms  287 remedied in the NESS account. They apparently required that a causally relevant factor be either strongly necessary or independently strongly sufficient (sufficient for the occurrence of the consequence disregarding competing or duplicative conditions, but combined with the other weakly necessary conditions).10 This requirement results in erroneous denials of causal contribution in the frequently occurring situations in which numerous conditions combine to cause the consequence, but none of them individually were (or could be proven to be) strongly necessary or independently strongly sufficient.11 Hart and Honoré also submerged and sometimes confused the critical distinction between duplicative and pre-­ emptive causation by constructing an overlapping typology of overdetermined causation cases,12 often emphasised ‘making a difference’ rather than sufficiency and thus sometimes erroneously employed strong-necessity rather than weak-necessity analysis in overdetermined causation situations,13 interpreted the analysis of necessity as a hypothetical counterfactual analysis rather than a real world factual analysis,14 interpreted sufficiency in terms J Collins, N Hall and LA Paul (eds), Causation and Counterfactuals (Cambridge, MIT Press, 1994) 1, 21–22, 44. Causation is an empirical relationship between concrete properties of distinct events or states of affairs. TL Beauchamp and A Rosenberg, Hume and the Problem of Causation (New York, Oxford University Press, 1981) 251–52, 255–56, 281–82; JL Mackie, The Cement of the Universe: A Study of Causation (Oxford, Clarendon Press, 1974) 32–33, 256–58, 260–63, 266–67; R Fumerton, ‘Moore, Causation, Counterfactuals, and Responsibility’ (2003) 40 San Diego Law Review 1273, 1278; LA Paul, ‘Aspect Causation’ in Collins, Hall and Paul (n 9) 205; Wright (n 3) 1033–34 and fn 171. After long arguing otherwise, Moore now acknowledges this. Moore (n 1) 361–65, 368 fn 61. Yet he persists in wanting to refer to whole events, rather than their causally relevant instantiated properties, as causes. ibid 366–68, 395–96. He apparently is driven to this position by his rejection of the ‘harm within the risk’ (HWR) limitation on attributable responsibility and his erroneous belief that the law’s tortious-aspect-causation requirement is functionally equivalent to that limitation, which erroneous belief further leads him to claim that I support the HWR limitation. See ibid 104, 166–67 and fn 50. Contrary to Moore’s claims, ibid 329–31, the law is not ambivalent between event-causation and aspect-causation, but rather clearly requires the latter, as I have shown while distinguishing, criticising and rejecting the HWR limitation. See Wright (n 5) 1759–74; RW Wright, ‘The Grounds and Extent of Legal Responsibility’ [Legal Responsibility] (2003) 40 San Diego Law Review 1425, 1479–1528. 10   Hart and Honoré (n 5) 123–25, 206–07, 235–39, 245, 249. Hart and Honoré did not discuss situations involving conditions that were neither strongly necessary nor independently strongly sufficient. Writing separately, Honoré noted, without elaboration, that such conditions could be causes. AM Honoré, ‘Causation and Remoteness of Damage’ in A Tunc (ed), International Encyclopedia of Comparative Law, vol 11: Torts pt 1 (Tübingen, JCB Mohr, 1983) 7-107, 7-108, 7-115, 7-121. He apparently accepts the NESS account’s extension of the weak-necessity analysis to encompass such situations. See text to n 147 below. 11   See section III.E below. 12   Hart and Honoré (n 5) xl, 122–25, 206–07, 235–53; see Honoré (n 10) 7-126 to 7-140; Wright (n 5) 1796–97. 13   Hart and Honoré (n 5) 29, 34–37, 239–41, 246–48, 250–51; Honoré (n 1) 368–69, 371–73, 379–80; Honoré (n 10) 7-109 to 7-111(1), 7-126; see Thomson (n 1) 156–57; Wright (n 5) 1797–1801. As Hart and Honoré generally understood, ‘making a difference’ should be understood in the sense of weak necessity rather than strong necessity. See M Strevens, ‘Mackie Remixed’ in J Keim, M O’Rourke and D Shier (eds), Causation and Explanation (Cambridge, MIT Press, 2007) 93, 97–98, 111–12; RW Wright, ‘Acts and Omissions as Positive and Negative Causes’ in JW Neyers et al (eds), Emerging Issues in Tort Law (Oxford, Hart Publishing, 2007) 287, 294 fn 20. 14   Hart and Honoré (n 5) lviii–lxi; Honoré (n 1) 370–72. But see Honoré (n 1) 376–79 (engaging in causal sufficiency analysis by matching actual concrete conditions against the abstract conditions in the antecedent of a causal law). For discussion of the many defects of the counterfactual-dependency analysis of causation, see Moore (n 1) 382–90, 392–425; Collins, Hall and Paul, Introduction (n 9) 2–12, 15–29; N Hall, ‘Two Concepts of Causation’ in Collins, Hall and Paul (n 9) 225, 225–26, 232–48; J Schaffer, ‘Trumping Preemption’ in Collins, Hall and Paul (n 9) 59, 67–71. Although I initially referred to the analysis of necessity in both the NESS account and the sine qua non account as a counterfactual analysis, see Wright (n 5) 1803–07, I have always insisted that the analysis is (or should be) a real-world ‘covering law’ matching of actual conditions against the required elements of the relevant causal generalisations rather than a counterfactual ‘possible worlds’ exploration of what might have occurred in the absence of the condition at issue. ibid; Wright (n 13) 296–97; Wright, Legal Responsibility (n 9) 1444–45 and fn 67; RW Wright, ‘Once More Into the Bramble Bush: Duty, Causal Contribution, and the Extent of Legal Responsibility’ [Once More] (2001) 54 Vanderbilt Law Review 1071, 1106–07; Wright (n 3) 1035–37, 1039–42; see n 119 and text to nn 118–19 below. The existence of these two very different methods of analysing necessity is noted by Moore, who however incorrectly describes both methods as ‘counterfactual’ and focuses on the second approach. Moore (n 1) 372–82, 390.

288  Richard W Wright of abstract, incomplete causal generalisations rather than complete instantiation of the causal laws underlying the causal generalisations,15 and denied that human decisions and actions are governed by causal laws.16 Nevertheless, Hart and Honoré’s account of a causally relevant factor was a major advance in the analysis of causation in both law and philosophy. Unfortunately, however, their account was overshadowed and distorted by their primary emphasis on elaborating supposedly factual ‘common sense’ principles for treating only some causally relevant factors as causes,17 so that, initially, it received minimal attention in the legal literature.18 Hart and Honoré’s account had a significant impact on non-legal philosophers, a number of whom subsequently published analyses of causation similar to Hart and Honoré’s account, although generally without citing Hart and Honoré. The best known analysis was published in 1965 by a fellow professor at Oxford, John Mackie, who employed an acronym, INUS (for ‘insufficient but necessary part of an unnecessary but sufficient condition’), to facilitate reference to his analysis.19 However, these philosophers only used the weak-necessity analysis to describe causal laws. For singular instances of causation they required that the condition at issue be necessary for the occurrence of the consequence in the particular circumstances, thereby converting their accounts into the sine qua non account.20 Mackie went further: he denied that identification of singular instances of causation requires any (even implicit) reference to causal laws or generalisations.21 Mackie’s arguments on these two issues do not withstand analysis and are contradicted by other arguments that he himself makes.22 Yet, due perhaps to the handy INUS acronym, philosophers have generally referred to Mackie’s INUS account of weak necessity rather than Hart and Honoré’s earlier account.23 In 1985, I substantially revised and extended Hart and Honoré’s account of a causally relevant factor and created the NESS acronym to refer to this revised and extended account.24 According to the NESS account as initially elaborated, a condition c was a cause   Hart and Honoré (n 5) 15, 22, 32, 44–45, 111–12, 124–25, 207, 237; Honoré (n 10) 7-53.   See section III.F below. 17   See RW Wright, ‘The Nightmare and the Noble Dream: Hart and Honoré on Causation and Responsibility’ in M Kramer et al (eds), The Legacy of H.L.A. Hart: Legal, Political and Moral Philosophy (Oxford, Oxford University Press, 2008) 165. The introduction to Causation in the Law focuses entirely on this issue; it does not mention the distinction between natural causation and attributions of responsibility or their account of causally relevant factors. See Hart and Honoré (n 5) 1–7; cf ibid 8–12. 18   See Wright (n 5) 1788 fn 227. 19   JL Mackie, ‘Causes and Conditions’ in E Sosa and M Tooley (eds), Causation (Oxford, Oxford University Press, 1993) 33, 34–37, originally published in (1965) 2 American Philosophical Quarterly 245 and revised and reprinted as chapter three of Mackie (n 9). Honoré states that Mackie ‘applied our idea’ in developing his INUS account. Honoré (n 1) 365. However, unlike Hart and Honoré’s account or the NESS account, Mackie literally requires that an INUS condition be insufficient by itself for the consequence and that the sufficient condition of which it is a part be unnecessary. Both restrictions are too strict. 20   eg Mackie (n 9) 38–48, 62, 76–77, 126–27; see Wright (n 3) 1023 fn 113. 21   Mackie (n 9) 40–58, 76–78, 120–22. Fumerton, Kress and Moore seem to agree with Mackie. See Fumerton and Kress (n 1) 97. Despite Moore’s insistence that singular instances of causation must always be ‘accompanied by’ causal laws, Moore (n 1) 361–64, 472, 496–97, 506, and his declaration that the notion that one can directly observe singular causal relations is a ‘hogchoker of a premise’, ibid 363, he asserts that singular causal relations are basic and prior to causal laws, which are mere inductively derived generalisations from already identified (in some unexplained manner) true singular instances of causation. ibid 361, 472–73, 497–99, 506. In the end Moore seems to abandon even this apparently superfluous role for causal laws and instead to opt for being a ‘reluctant primitivist’. See ibid xii, 505–12. 22   See Hart and Honoré (n 5) xxxix–xlii; Wright (n 3) 1023–34; text to nn 42–64, 156–59 below. 23   See Wright (n 3) 1023 fn 113. 24   Wright (n 5) 1788–1803. 15 16

The NESS Account: Response to Criticisms  289 of a consequence e if and only if it was necessary for the sufficiency of a set of existing antecedent conditions that was sufficient for the occurrence of e.25 The required sense of sufficiency, which (following a suggestion by Fumerton and Kress)26 I call ‘causal sufficiency’ to distinguish it from mere lawful strong sufficiency, is the instantiation of all the conditions in the antecedent (‘if ’ part) of a causal law, the consequent (‘then’ part) of which is instantiated by the consequence at issue.27 A causal law is an empirically derived statement that describes a successional relation between a set of abstract conditions (properties or features of possible events and states of affairs in our real world) that constitute the antecedent and one or more specified conditions of a distinct28 abstract event or state of affairs that constitute the consequent such that, regardless of the state of any other conditions, the instantiation of all the conditions in the antecedent entails the immediate instantiation of the consequent, which would not be entailed if less than all of the conditions in the antecedent were instantiated. It is critically important when analysing singular instances of causation to distinguish causal relations from mere lawful relations. To do so one must include in the causal analysis the entire causal process up to the time of the occurrence of the consequence.29 The definition of a causal law in the immediately preceding paragraph assures this by requiring that the instantiation of the consequent of the causal law occur immediately when all of the conditions in the antecedent of the causal law have been instantiated. A causal process consists of the instantiation of one or more simultaneously or successively operative causal laws. Another critical feature of causal laws – and the related concept of causal sufficiency as distinct from mere lawful sufficiency – is their successional or directional nature, according to which the instantiation of the conditions in the antecedent of the causal law causes the instantiation of the consequent, but not vice versa.30 Our knowledge of the required conditions in the antecedent of a causal law – and thus of the direction of causation – is based on experience and empirical investigation, by ourselves or others. Scientists employ Mill’s Difference Method in carefully designed experiments to see if the non-instantiation of a supposed antecedent condition makes a difference in the occurrence of the consequence.31 For example, we determine by observation or experimentation that eliminating a flagpole or changing its height eliminates or changes the length of the flagpole’s shadow, but not vice versa.32 In the causal law that is thereby derived, the successionally antecedent NESS conditions are grouped together in the ‘if ’ part of the causal law and the consequent is stated separately in the ‘then’ part. The successional nature of causation is incorporated 25   See text to n 40 below for my current definition of a concrete NESS condition. The definition in the text here is itself a clarification of my initial definition, which literally merely required that a NESS condition be a necessary member of a sufficient set and thus failed to expressly incorporate the requirement, which has always been stated in my elaborations of the NESS account, that a NESS condition be necessary for the sufficiency of a sufficient set. eg Wright (n 5) 1790; Wright (n 3) 1019, 1021, 1041. Richard Fumerton and Ken Kress drew my attention to the ambiguity in my initial definition. See Fumerton and Kress (n 1) 94; Wright, Once More (n 14) 1103 fn 112. 26   Fumerton and Kress (n 1) 93, 101–02; see Wright, Once More (n 14) 1103 fn 113. Michael Strevens employs the term ‘causal sufficiency’ for a similar, if not identical, concept of sufficiency. See text to n 73 below. 27   I have always insisted on this. eg Wright (n 5) 1789, 1795–98, 1803–04, 1808–09, 1823; Wright (n 3) 1031, 1033, 1041–42, 1045–46, 1049–53. 28   See n 9 above. 29   See Fumerton and Kress (n 1) 103–04; Hall (n 14) 238–40; Honoré (n 10) 7-53, 7-107, 7-110; Schaffer (n 14) 67–71; Strevens (n 13) 113–16; Thomson (n 1) 153–54, 164; text to n 73 below. 30   Fumerton and Kress (n 1) 93; see sections III.B and III.C below. 31   See Mill (n 5) bk III ch VII ss 2–4, ch VIII s 2. 32   However, as Hart and Honoré note, one should be careful not to treat human manipulability as a requirement for being a cause. Hart and Honoré (n 5) 36.

290  Richard W Wright in the concept of causal sufficiency, which is defined as the complete instantiation of all the conditions in the antecedent of the relevant causal law.33 While the NESS account’s definition of causal sufficiency relies upon the concept of a causal law, the concept of a causal law is defined above without any use of causal language. There is no conceptual circularity here, vicious or otherwise. Even if some circularity should be found to exist in these definitions, it would not be ‘vicious’. The NESS account provides a description of the structure of causal laws and their instantiation that is acknow­ ledged, even by its critics, to be illuminating and very useful for proper identification of singular instances of causation.34 Our knowledge of causal laws generally is incomplete, and even when it is complete we rarely refer to completely specified causal laws, since such complete specification would be extremely burdensome and unnecessarily detailed and lengthy. We rather employ causal generalisations, which refer to only some of the antecedent conditions in the relevant causal laws and have only as much specificity as is possible and needed in the particular situation. For example, we usually refer to the causal generalisation that specifies that bringing a flame into contact with combustible material causes that material to burn, without referring to other necessary antecedent conditions such as the presence of oxygen or the absence of a soaking rain – unless the latter expected conditions did not exist in the particular situation or, conversely, existed but were not expected. Moreover, the generalisations that we employ usually refer elliptically to a large number of simultaneously or successively operative causal laws.35 However, when we make an assertion regarding a singular instance of causation, we are implicitly asserting that all the unstated as well as the stated conditions in the relevant causal generalisations and all the unknown as well as known conditions in the causal laws underlying the causal generalisations were instantiated on the particular occasion. Contrary to Hart and Honoré’s account,36 the NESS account insists that singular instances of causation always consist of the complete instantiation on a particular occasion of one or more causal laws, and that identification of a singular instance of causation always implies that such complete instantiation has occurred.37 The implication may be based on direct particularistic evidence of the existence of one or more of the required conditions, or, as is always true for the unknown conditions in the causal laws and generally true for many (sometimes all) of the known conditions, is inferred from particularistic evidence of the network of causal relationships that encompasses the particular occasion.38 33   See text to n 27 above. Interpreted in the usual manner, causal succession precludes temporally backward causation, through which events today change events in the past. However, the definition of causal succession in the text does not preclude such backward causation, which would occur if the present instantiation of the antecedent results in the immediately following instantiation of the consequent (paradoxically) in the past. 34   eg Beever (n 1) 426; Fischer (n 1) 277, 281–84, 302, 317; Fumerton and Kress (n 1) 83–84, 95–97, 102–04; Moore (n 1) 361–62, 368 and fns 59 and 61, 473–74, 486–90, 493; Stapleton (n 1) 443–44, 471–72, 474; Thomson (n 1) 144, 148–50, 163–64. 35   Hart and Honoré (n 5) xxxvii, 11–13, 31–32, 45–49; Honoré (n 10) 7-53; Mackie (n 9) 35–38, 66–76; Mill (n 5) bk III ch IV s 1, ch V ss 2–3; Moore (n 1) 477–78, 485; Strevens (n 13) 111–12; Wright (n 5) 1823–24; Wright (n 3) 1031–34, 1045–46. Hart and Honoré erroneously claim that Mill insisted that all of the conditions in the antecedent of the relevant causal law must be known before selecting only one or a few for explicit reference in a singular causal statement. Hart and Honoré (n 5) 15, 21, 22, 31, 44–47; Honoré (n 10) 7-53. 36   See text to n 15 above. But see Honoré (n 1) 367, 376, 385 (apparently accepting the NESS account’s conception of causal generalisations and singular instances of causation). 37   See text to nn 25–27 above. 38   Wright (n 3) 1045–46, 1049–52.

The NESS Account: Response to Criticisms  291 Even when the inference of an applicable causal generalisation (and the underlying causal laws) is based on a single observation of a singular instance of causation, the inference of causation in the singular instance comes after and depends upon the inference of the applicable causal generalisation. In such instances the order is (1) observation of the occurrence, (2) inference of the causal generalisation based on the observation, and (3) assertion of causation through enunciation of the singular causal statement, which implicitly invokes the causal generalisation.39 My initial elaborations of the NESS account were overly demanding. I incorporated the weak-necessity requirement in the definition of singular instances of causation. As I have previously stated, this is too restrictive. The weak-necessity requirement is sufficiently incorporated in a properly formulated causal law, which contains in its antecedent only those abstract conditions the instantiation of which is necessary for the sufficiency of the set of conditions that is sufficient for the immediate instantiation of its consequent. When analysing singular instances of causation, an actual condition c was a cause of an actual condition e if and only if c was a part of (rather then being necessary for) the instantiation of one of the abstract conditions in the completely instantiated antecedent of a causal law, the consequent of which was instantiated by e immediately after the complete instantiation of its antecedent, or (as is more often the case) if c is connected to e through a sequence of such instantiations of causal laws.40 This formulation of the requirement for a NESS condition is more straightforward and simpler to apply than my initial formulation, which requires ‘at least so much’ descriptions of actual conditions in some situations in order to (validly) treat other conditions as NESS conditions.41 However, care must be taken to make sure that the antecedents of the relevant causal laws include only abstract NESS conditions in the strict sense – that is, those abstract conditions the instantiation of which is necessary for the sufficiency of the set of conditions that is sufficient for the instantiation of the consequent. The sine qua non account’s strong-necessity analysis, properly applied,42 is a corollary of the NESS analysis that gives the correct answer when there was only one set of conditions that was actually or potentially sufficient for the consequence on the particular occasion.43 Contrary to what many assume,44 the sine qua non analysis relies on an embedded analysis of (lawful rather than causal) sufficiency. To determine if some condition was strongly necessary for the occurrence of some consequence that actually occurred, one must ‘rope off ’ the condition at issue and then, using the relevant causal generalisations, determine whether the remaining existing conditions were lawfully sufficient for the occurrence of the consequence – that is, whether the relevant causal laws would have been fully instantiated   ibid 1031–34; see Mill (n 5) bk III ch III.   Wright (n 13) 297–98; Wright (n 3) 1045; see text to nn 27–29 above. Allan Beever fails to note my recent confinement of the weak-necessity requirement to the abstract conditions in causal laws. He rejects any necessity restriction in causal analysis and instead merely requires that a condition be a member of a sufficient set. This ‘weak sufficiency’ analysis opens the door to treating every condition as a cause. See Beever (n 1) 425–26; Wright (n 3) 1020–21. 41   Wright (n 5) 1793–94; Wright, Legal Responsibility (n 9) 1444–45; Wright, Once More (n 14) 1106–07; Wright (n 3) 1035–37; see text to nn 117–20 below. 42   See n 14 above. 43   Hart and Honoré (n 5) 112–13; Wright (n 3) 1021. 44   eg DW Robertson, ‘Causation in the Restatement Third of Torts: Three Arguable Mistakes’ (2009) 44 Wake Forest Law Review 1007, 1010 (‘[W]hen we make causation-attribution decisions, necessity vel non is the heart of the inquiry. . . . In torts cases, the cause-in-fact inquiry is always an attribution question, never a predictive one, so sufficiency issues are not in play’.). 39 40

292  Richard W Wright in the absence of the condition at issue. If they would have been, the condition at issue was not strongly necessary for the occurrence of the consequence.45 In overdetermined causation situations,46 in which there were two or more (usually overlapping) sets of conditions that were each actually or potentially sufficient – for example, two fires each sufficient to destroy a house if the fire reaches the house while it is still standing, which merge and destroy the house (duplicative causation), or one of which reaches the house and destroys it before the other arrives (pre-emptive causation) – the NESS analysis reaches the proper conclusions regarding causation, while the sine qua non (‘but for’, ‘made a difference’) analysis as usually described and applied does not.47 The sine qua non analysis is able to reach the proper conclusion in many overdetermined causation situations if the consequence is qualified by the time at which it occurred,48 and in a few more such situations if it is further qualified by the location at which it occurred.49 However, employing such qualifications will greatly expand the proliferation of causes, especially if trivial differences in the time or location of a specific event – for example, due to gravitational forces exerted by nearby objects – are taken into account, rather than differences in the time or location of distinct events.50 More significantly, in many situations it will be impossible to determine whether the condition at issue had any effect on the timing or location of the consequence, and a condition can be a cause even if it had no effect on the timing or location, or not be a cause even though (if not pre-empted) it would have caused the consequence at the same time and location.51 The NESS analysis will properly resolve the causal issue in such situations; the sine qua non analysis will not.52 Some, including Hart and Honoré, Mackie, Michael Moore, Jane Stapleton and Robert Stevens, have stated that the sine qua non analysis can be further sharpened to reach the proper conclusion in overdetermined causation situations by specifying minute details of the consequence or qualifying it by the manner of its occurrence, ‘as and how it came 45   Wright (n 13) 296–97; see D Lewis, ‘Void and Object’ in Collins, Hall and Paul (n 9) 277, 279, 288–89; n 119 and text to nn 118–19 below. 46   Contrary to the usual practice, Stapleton and Michael Strevens apply the term ‘overdetermination’ only to duplicative causation situations. Stapleton (n 1) 442 fn 19; Strevens (n 13) 104–05. 47   Wright (n 5) 1775–76, 1791–98; Wright (n 3) 1018–28. 48   Wright, Once More (n 14) 1112–14; Wright (n 13) 292 fn 14. Stapleton ignores these sources and my unsuccessful attempts to have the American Law Institute add a time-of-occurrence qualifier to the sine qua non test in s. 26 of the Restatement Third of Torts when she implies that I do not recognise that the sine qua non account can be modified in this manner to resolve the causal issue in many cases. See Stapleton (n 1) 452 and fn 42; American Law Institute, 79th Annual Meeting, Proceedings 2002 (Philadelphia, American Law Institute, 2003) 273–76; American Law Institute, 82nd Annual Meeting, Proceedings 2005 (Philadelphia, American Law Institute, 2006) 81–84. The major defects in the Restatement Third’s treatment of ‘factual causation’ are discussed in n 125 below. 49   F Stella, ‘The Vitality of the Covering Law Model: Considerations on Wright and Mackie’ http://works. bepress.com/richard_wright/34/ at 34–37. 50   Collins, Hall and Paul, Introduction (n 9) 46; Hall (n 14) 237, 239; Moore (n 1) 477–78; see text to nn 53–56 below. 51   Collins, Hall and Paul, Introduction (n 9) 23–24; Hall (n 14) 235–38; Schaffer (n 14) 70–71; Wright, Once More (n 14) 1112–14. 52   Wright, Once More (n 14) 1112–14; see text to nn 83–88 and 96 below; cf Stapleton (n 1) 452–53 and fn 45 (employing the NESS account’s causal sufficiency analysis rather than her ‘duplicate necessity’ analysis of ‘involvement’, which is described in n 125 below, to avoid treating a pre-empted condition as a cause). Federico Stella acknowledged the need to turn to the NESS account to handle these types of situations. Stella (n 49) 38. Yet he otherwise preferred the time-and-location-qualified sine qua non account, in part because he erroneously assumed that the NESS analysis as applied in the law is not framed by a focus on the legally relevant conditions, ibid 23, 39, but primarily (I believe) because he was concerned that ‘in Italy to abandon [the sine qua non test] would lead to real disasters. Our judges . . . are very far from having a sound culture of proofs’. Federico Stella, email to Richard Wright, 25 January 2005.

The NESS Account: Response to Criticisms  293 about’.53 Both approaches are defective and extremely promiscuous.54 The first approach will merely establish causation of some irrelevant detail of the consequence, rather than the consequence itself.55 This objection applies as well to minute differences in the time or location of a specific event.56 The second approach is viciously circular,57 as Honoré and Moore now acknowledge.58 Although Honoré notes that the law requires that a plaintiff identify the specific event for which she seeks redress, including the time, place and persons involved, he now rejects, as part of the causal analysis, qualifying the consequence by time, location or manner of occurrence.59 Instead, he relies upon the NESS account’s causal sufficiency analysis.60 It therefore is puzzling that he states, without elaboration, that the sine qua non account will ‘point to the same causal conclusion’ as the NESS account when ‘the [competing] causal processes are different’ or ‘one culminates in harm before the other’.61 Even initially, although stating that ‘[s]uch solutions are perfectly sound logically and legally’, Hart and Honoré correctly noted that the second approach requires ‘first deploying the more fundamental notion of sufficient conditions and so identifying the causal process which culminated in the harm’.62 Similarly, Mackie notes that this approach relies upon determining which of the competing causal stories was completed in the particular situation.63 Although Mackie does not acknowledge it, this is a causal sufficiency analysis rather than a strong necessity analysis. When even detailing the manner of occurrence will not distinguish the competing conditions using the sine qua non analysis, Mackie paradoxically asserts that none of the conditions individually were causes but all of them in the aggregate were.64 53  A Becht and F Miller, The Test of Factual Causation in Negligence and Strict Liability Cases (St. Louis, Washington University Press, 1961) 15–19; D Coady, ‘Preemptive Preemption’ in Collins, Hall and Paul (n 9) 325, 327–33; Hart and Honoré (n 4) 118; Hart and Honoré (n 5) xli–xlii; Honoré (n 10) 7-111(2), 7-126; Mackie (n 9) 45–46; NJ McBride and R Bagshaw, Tort Law, 3rd edn (Harlow, Pearson Education Ltd, 2008) 535 fn 16; M Moore, ‘Thomson’s Preliminaries About Causation and Rights’ (1987) 63 Chicago-Kent Law Review 497, 510–11; R Stevens, Torts and Rights (Oxford, Oxford University Press, 2007) 131, 133–34 and n 12, 135; Stapleton (n 1) 442 fn 19 (‘death by two bullets’), 452 (‘death by electrocution’, ‘death by explosion’). For hair-splitting semantic variations on this approach, see D Lewis, ‘Causation as Influence’ in Collins, Hall and Paul (n 9) 75, 85–90. 54   Collins, Hall and Paul, Introduction (n 9) 46; Hall (n 14) 237–38, 239; Strevens (n 13) 96–97, 98–99, 111–12. 55   Fumerton and Kress (n 1) 96-97; Wright (n 5) 1778–80. 56   See text to n 50 above. 57   Wright (n 5) 1777–78; Wright (n 3) 1025. 58   Honoré (n 1) 378–79; Moore (n 1) 87 fn 16. 59   Honoré (n 1) 378–79; see ibid 368. 60   See ibid 377–79. 61   ibid 374. 62   Hart and Honoré (n 4) 118; see ibid 119; cf Hart and Honoré (n 5) xli–xlii, 124–25 (no longer asserting that this approach is ‘sound logically and legally’). Although apparently not recognising what he is doing, Robert Stevens, in his attempt to defend the ‘but for’ test as the sole test of causation, employs causal sufficiency analysis rather than ‘but for’ analysis to reach the proper conclusion in several much-discussed overdetermined causation situations. He also resurrects the discredited privity of contract limitation on tort liability in order to explain the mechanic’s lack of liability in the unused defective brakes case. See Stevens (n 53) 135–37; text to nn 83–94, 116– 23 and 179–95 below. 63   Mackie (n 9) 45–46. 64   ibid 47, criticised in Moore (n 1) 354–56; Wright (n 3) 1026–27. Stevens claims to reject Mackie’s aggregative move, not because it is illogical but rather because the law focuses on individual rather than aggregative responsibility. Stevens (n 53) 131. However, he implicitly employs it when assessing ‘substitutive’ and consequential damages for infringement of a right. For example, if a plaintiff ’s dog was killed by two stab wounds inflicted by different defendants, each of which would have been sufficient by itself to cause the death of the dog, Stevens asserts that the two defendants are each liable for ‘the value of the dog’ as ‘substitutive’ damages for their respective infringements of the plaintiff ’s right to the dog, and also for any consequential damages that are a ‘but for’ result of the dog’s death, but that double recovery is not permitted since ‘he only had one right to the dog’. ibid 134. Yet the ‘but for’ test can only establish each defendant’s causation of a distinct stab wound, not the death of the dog,

294  Richard W Wright The courts and the secondary literature generally do not qualify the consequence by specifying its non-salient details or the time, location or manner of its occurrence when describing or applying the sine qua non analysis.65 Instead, in overdetermined causation situations they either reach incorrect conclusions on natural causation by using an unqualified sine qua non analysis66 or employ question-begging, conclusory phrases – for example, ‘substantial factor’, ‘material contribution’ or ‘common sense causation’67 – that confusingly merge the scientific issue of causal contribution with the normative issue of the appropriate extent of legal responsibility for the consequences of one’s (legally relevant) conduct while failing to provide any test or guidance for resolving either issue.68 Philosophers have attempted to defend the sine qua non account and the usual employment of counterfactual analysis in that account by shifting the focus of the analysis from the occurrence of the consequence at issue to the occurrence of each step in the competing causal sequences. This attempt abandons (without acknowledgement) strong necessity in favour of a deficient version of weak necessity. Moreover, it works only in ‘early pre-­ emption’ situations, in which the pre-emption of the pre-empted causal sequence occurs before the completion of the completed causal sequence. It does not work in ‘late preemption’ situations, in which the pre-emption of the pre-empted causal sequence occurs at the same time as (or after) the completion of the completed causal sequence, nor in any duplicative causation situation.69 In the duplicative causation situations, some sine qua non advocates hold fast. They assert that neither of the duplicative conditions was a cause and that the alleged consequence is either an uncaused miracle or was caused by the conjunction of the duplicative conditions although neither by itself was a cause.70

and thus it cannot support holding either defendant liable for ‘substitutive’ damages for the full ‘value of the dog’ or for consequential damages caused by the dog’s death. Rather, under the individually applied ‘but for’ test that Stevens claims to be applying, each defendant can only be held liable for the ‘substitutive’ value of the rights infringement that is constituted by his or her stabbing of the dog, however that might be valued, and any consequential damages (of which there likely will be none) that are a ‘but for’ result of his or her distinct rights infringement. Moreover, although there is ‘only one right to the dog’, it is a right that each defendant has separately and distinctly infringed, and thus, under Stevens’ odd damages theory, each defendant should be fully but separately, rather than concurrently, liable for ‘substitutive’ damages for his or her distinct rights infringement, regardless of any resulting ‘double recovery’. 65   See eg, Restatement Third (n 1) § 26; Restatement (Second) of Torts § 432(1) (1965); Hart and Honoré (n 5) xlii; Honoré (n 1) 378–79; n 48 above. Stapleton, rather than I, ‘falls into manifest error’ when she asserts otherwise. See Stapleton (n 1) 452 and fn 42. 66  See Wright, Legal Responsibility (n 9) 1438–40; RW Wright, ‘Liability for Possible Wrongs: Causation, Statistical Probability, and the Burden of Proof ’ (2008) 31 Loyola Los Angeles Law Review 1295, 1321–24. 67  eg March v E & MH Stramare Pty Ltd (1991) 99 ALR 423 (High Court of Australia) (‘material contribution’ and ‘common sense’); Athey v Leonati (1996) 3 SCR 458 (Supreme Court of Canada) (‘materially contributed’); Sew Hoy & Sons Ltd v Coopers & Lybrand (1996) 1 NZLR 392 (New Zealand Court of Appeal) (‘material contribution’); Fairchild v Glenhaven Funeral Services Ltd [2002] UKHL 22, [2003] 1 AC 32 (‘material contribution’); Kingston v Chicago & Northwestern Ry Co (1927) 211 NW 913 (Supreme Court of Wisconsin) (‘substantial factor’); Mitchell v Gonzales (1991) 819 P 2d 872 (Supreme Court of California) (‘substantial factor’). 68   Wright (n 5) 1742–50, 1781–88; Wright, Once More (n 14) 1073–80; Wright (n 3) 1012–14. 69  See J Collins, ‘Preemptive Preemption’ in Collins, Hall and Paul (n 9) 107, 109–10; Collins, Hall and Paul, Introduction (n 9) 18, 22–24, 32–33; Schaffer (n 14) 67–70. Philosophers tend to focus on overdetermined causation situations involving pre-emptive rather than duplicative causation. eg Collins, Hall and Paul, Introduction (n 9); Hall (n 14); Schaffer (n 14); Lewis (n 53) 80–82; Thomson (n 1) 144, 150–51, 152–59. 70   See Collins, Hall and Paul, Introduction (n 9) 32–33; Lewis (n 53) 80; Mackie (n 9) 47. David Coady adopts the latter position for pre-empted conditions as well as duplicative causes. Coady (n 53) 326–28.

The NESS Account: Response to Criticisms  295

III.  Alleged Deficiencies of the NESS Account

A.  Spurious Weak Necessity A common counter-example to the NESS account is the alleged conversion of a causally irrelevant condition c into a NESS condition by replacing a causally relevant condition b with two conditions, one of which is c and the other of which is a disjunction of b with the negation of c.71 For example, if D’s cutting off C’s head was a necessary condition for C’s death (or for the sufficiency of a set S of conditions sufficient for C’s death), D’s beheading C is replaced with the following two conditions: 1.  Napoleon married Josephine. 2.  Either Napoleon did not marry Josephine or D beheaded C. Given the disjunction in condition 2, the set S is not sufficient for C’s death unless we also include condition 1, so, apparently, Napoleon’s marrying of Josephine is (erroneously) treated as a cause of C’s death under the NESS account. To counter this logical manoeuvre, Fumerton and Kress propose excluding from the set S of sufficient conditions any condition (such as condition 2 above) that contains a truthfunctional constituent part (such as ‘D beheaded C’) that by itself is lawfully sufficient for the consequence.72 Without condition 2, S is no longer a sufficient set, so condition 1 is not a NESS condition. To be generally effective, this analytic counter needs to be broadened to exclude as a constituent part of any disjunctive condition in S a condition that is sufficient when conjoined with all the other conditions in S other than the condition with which it is disjoined. However, such appended analytic restrictions not only seem ad hoc, they also are unnecessary given restrictions that are already embedded in the NESS account. The requirement that a NESS condition be an actually existing condition rules out as a constituent part of any NESS condition the negation of any actually existing condition. Napoleon’s marrying Josephine can be tentatively included in S as condition 1 only if it was an actually existing condition, but if it was then Napoleon’s not marrying Josephine in condition 2 cannot be true, so condition 2 collapses down to the actual condition of D’s beheading C, thereby making condition 1 not necessary for the sufficiency of S. The logical manoeuvre that creates the spurious weak necessity is similarly precluded by the NESS account’s requirement that each condition in S be part of the instantiation of a condition in a fully instantiated causal law. As Michael Strevens states in his revision of Mackie’s INUS account in a way that makes it similar to the NESS account, it is not enough that a set of conditions be sufficient for the occurrence of e; it must be causally sufficient for e. . . . Causal sufficiency ought to be defined, then, so that a set of conditions is causally sufficient for an event e only if the conditions represent a causal process that produces e. A set of conditions entailing e represents a causal process producing e, I propose, just in case each step in the entailment corresponds to a strand in the relevant causal web. . . . [T]he step from the [statement of the causally irrelevant condition] and the disjunction [of the causally relevant condition and the negation of the causally irrelevant condition] to the [consequent] . . . is not a causal entailment, as it does not correspond to a causal process recognized by   eg Thomson (n 1) 151; cf Lewis (n 53) 77.   Fumerton and Kress (n 1) 95; Moore (n 1) 487–88.

71 72

296  Richard W Wright the laws of physics. Indeed, it is hard to imagine a physics in which something in the world captured by the description ¬r [or] c could be part of any story about causal influence. . . . [W]hen what we regard as an intuitively irrelevant factor r is made essential to the entailment of an event e, it is always by way of disjunction or other logical construction that links r and e truth-functionally but not causally. The approach to causal claims taken by my revision of Mackie’s account puts a considerable burden on the physical laws; they must determine what primitive causal connections there are in the world, hence determine the structure of the causal web. I think that they are quite capable of bearing that load, and that we do indeed look to the laws as the final arbiters on any question of causal connection.73

This third counter to the logical manoeuvre that creates spurious weak necessity has been criticised as being circular, since it assumes the prior specification of a ‘relevant causal web’, the structure of which is determined by reference to the ‘physical laws’.74 However, the NESS account does not refer circularly to causally relevant conditions or causal laws. It specifies the method for identifying causally relevant conditions – they must be instantiations of abstract conditions listed in the antecedent of an empirically derived causal law – and its definition of a ‘causal law’ does not employ causal terminology.75

B.  Causal Directionality (No ‘Backtracking’) When the sun is at a 45 degree angle to an upright flagpole and the shadow of the flagpole on the ground is 5 feet long, the laws of geometry and nature entail that the flagpole is 5 feet tall. However, as Fumerton and Kress state, ‘it would surely be mistaken to claim that the shadow causes the flagpole to be [5] feet tall. If anything, the direction of causation is from the flagpole’s height to the length of the shadow’.76 Although Fumerton and Kress and others assume otherwise, this conclusion is correctly reached under the NESS account, which insists that singular instances of causation always involve the complete instantiation of an empirically derived causal law and that causal laws state a direction of causation, whereby the complete instantiation of the conditions in the antecedent of the causal law cause the instantiation of the consequent of the causal law, not vice versa.77 Fumerton and Kress and others claim that the reference to causal laws and their inherent directionality is viciously circular.78 However, as I have discussed above, this is not true.79

73   Strevens (n 13) 113–14; see Collins, Hall and Paul, Introduction (n 9) 22 (‘we should say that the disjunctive event is not an event at all, hence not apt to cause (or be caused)’); Moore (n 1) 355–56 (same). 74   Thomson (n 1) 152–53. 75   See text to nn 25–29 above. 76   Fumerton and Kress (n 1) 102. Fumerton, Kress and Moore have the geometry wrong; they state that the flagpole must be 10 feet tall. ibid 93, 101–02; Moore (n 1) 476–77, 483. 77   See text to nn 25–33 above. 78   See n 2 above; cf Lewis (n 53) 77. Moore acknowledges that the directionality of causal laws is inherent in our concept of causation and can be used to reject the criticisms of the NESS account discussed in this section and the next section. However, he claims that referring to this directionality or to causal laws per se to elucidate the meaning of causation is viciously circular. He thus commends Mackie for treating this directionality, which Mackie calls ‘causal priority’, as an addition to, rather than as an essential constitutive element of, the sense of sufficiency employed in causal analysis. Moore (n 1) 482–85. Mackie acknowledged the use of the more discriminating sense of causal sufficiency in common discourse, but he thought it better for clarity’s sake to reserve ‘sufficiency’ for mere lawful sufficiency. Mackie (n 9) 51–53. See ibid 190–92 for Mackie’s analysis of causal priority. 79   See text to nn 25–34 above.

The NESS Account: Response to Criticisms  297

C.  Epiphenomena (Collateral Effects of a Common Cause) When a condition is a common cause of two different causal sequences, employing mere lawful strong sufficiency rather than causal sufficiency can result in erroneously treating conditions in one sequence as causes of the conditions in the other sequence, and vice versa.80 A frequently mentioned example is Mackie’s ‘hooters’ hypothetical, in which the horns (‘hooters’) at a Manchester factory and a London factory each sound at five o’clock every workday afternoon, signalling the stoppage of work at their respective locations. A few moments later, workers begin to exit from each factory. The sounding of the horn at the Manchester factory is lawfully strongly sufficient for its being five o’clock, and its being five o’clock is lawfully strongly sufficient for the sounding of the horn and the exodus moments later of workers at the London factory. Thus, the sounding of the horn at the Manchester factory is lawfully strongly sufficient for the sounding of the horn and exodus of workers moments later at the London factory.81 But it is not a cause of these events at the London factory, as the NESS account correctly concludes, because it is not causally sufficient. Neither horn’s sounding, nor even its existence, is part of the instantiation of the antecedent of any causal generalisation for which the consequent is the state of its being five o’clock. As we have empirically determined, the direction of causation runs the other way.82

D.  Pre-emptive Causation The NESS account has been criticised for allegedly being unable to handle properly some types of pre-emptive causation situations. Once again, the criticisms erroneously assume that the NESS account merely requires lawful sufficiency, rather than causal sufficiency.83 Consider some common examples: 1. C is a traveller in the desert, whose only source of water is a keg full of water. A adds a fatal dose of undetectable poison to the water in the keg, for which there is no antidote. C remains unaware of the poison in the water. Subsequently, before C drinks any of the poisoned water, B dumps the poisoned water out of the keg. When C attempts to drink water from the keg, she discovers that it is empty. C dies due to dehydration. 2. C drinks a fatal dose of poison for which there is no antidote but which takes several hours to produce death. While C is still alive, D shoots C in the head. C dies a few minutes later from the bullet wound, well before the time at which the death by poisoning would otherwise have occurred. 3. A ship is traveling down a river to deliver goods to Metropolis by a specific date. The ship is unable to arrive by that date, since its crew must and does stop when it reaches bridge A, which had collapsed into and blocked the river. The ship would not have been able to reach Metropolis on time even if bridge A had not collapsed, due to another collapsed bridge, bridge B, of which the ship’s crew was unaware, located on the river between bridge A and Metropolis. 80   Collins, Hall and Paul, Introduction (n 9) 17–18; Fumerton and Kress (n 1) 93, 101; Lewis (n 53) 77; Mackie (n 9) 33–34. 81   Mackie (n 9) 83–86; Moore (n 1) 481–82. 82   Strevens (n 13) 108–09; Wright (n 5) 1808–09. 83   Fumerton and Kress (n 1) 100–02; Moore (n 1) 474; Thomson (n 1) 148–51.

298  Richard W Wright If we merely require lawful sufficiency and do not qualify the consequence by the time of its occurrence – or, in examples 1 and 3, perhaps even if we do84 – we would incorrectly treat the poison as well as the emptying of the keg in example 1, the poison as well as the shooting in example 2, and the collapse of bridge B as well as the collapse of bridge A in example 3 as duplicative causes of the relevant consequence (respectively, death, death, and failure to reach Metropolis on time). In each example, either condition guarantees the occurrence of the (abstractly described) consequence and thus is lawfully sufficient for its occurrence.85 On the other hand, if we require strong necessity – that the condition ‘made a difference’ as a ‘but for’ cause – we would be forced to conclude that neither condition was a cause. In the NESS account, we require causal sufficiency rather than strong necessity or mere lawful strong sufficiency. For causal sufficiency, the condition at issue must be part of the instantiation of a fully instantiated causal law that is part of a sequence of such fully instantiated causal laws that link the condition at issue with the consequence.86 The instantiated causal laws that constitute the sequence are usually referred to incompletely and elliptically in the form of a causal generalisation.87 In example 1, the causal generalisation for death by poisoning includes the victim’s drinking the poison, a condition that was not instantiated since the keg was emptied before C was able to drink from it. On the other hand, the causal generalisation and the underlying causal laws for death by dehydration, which include as a necessary condition lack of water (and the physical bodily processes that occur as a result of the lack of water), were fully instantiated. Thus, B’s emptying of the keg, but not A’s poisoning of its contents, caused C’s death, even though C may have lived longer due to the emptying of the keg.88 Jane Stapleton states that there is substantial disagreement about this conclusion.89 Hart and Honoré initially rejected it, but only because they erroneously equated causing death with shortening life.90 Honoré eventually focused on the details of the competing causal processes and applied a causal sufficiency analysis to reach the correct conclusion: [I]f we know enough about the stages by which the events came about we can generally tell, even in cases of over-determination, which causal process ran its course and which was frustrated. So in the case of the desert traveler . . . it now seems to me that B causes C’s death. . . . B’s conduct introduces a condition, lack of water, that in the circumstances . . . is sufficient to bring about and does bring about C’s death from dehydration.91

Stapleton initially claimed that A, instead of or in addition to B, can be treated as a cause of C’s death, or that one can plausibly argue that neither was a cause of C’s death – the   See text to nn 48–56 above.   See Fumerton and Kress (n 1) 100–02; Collins, Hall and Paul, Introduction (n 9) 18; Lewis (n 53) 80; J Stapleton, ‘Perspectives on Causation’ in Jeremy Horder (ed), Oxford Essays on Jurisprudence: Fourth Series (Oxford, Oxford University Press, 2000) 61, 82–84; Thomson (n 1) 150–51; Wright (n 13) 298–300, 305. 86   See text to nn 27–29, 40 and 73 above. 87   See text to n 35 above. 88   Becht and Miller (n 53) 205–10; Mackie (n 9) 44–46; Wright (n 5) 1802. I now believe that my opposite conclusion, in Wright (n 5) 1802, for McLaughlin’s original version of this hypothetical, in which A emptied the keg and refilled it with salt and B subsequently stole the keg before C attempted to drink from it, is wrong for reasons set out in section III.I below. 89   Stapleton (n 1) 439 fn 15. Since Moore refuses to treat negative conditions such as lack of water as causes or effects, he denies that either A or B caused C’s death, thereby treating the death as an uncaused miracle. Moore (n 1) 144, 466–67; see section III.H below. 90   See Hart and Honoré (n 5) 239–41 and fn 74. 91   Honoré (n 1) 378; see Honoré (n 10) 7-133. 84 85

The NESS Account: Response to Criticisms  299 occurrence of which thus is an unexplained miracle.92 She later claimed (and apparently still does) that both A and B were causes. Her arguments for treating A as a cause fail to distinguish guaranteeing an outcome (here, death) from actually causing the outcome and erroneously assert that a lack of ‘potable’ or ‘fresh’ (unpoisoned) water is the same as a lack of water, while failing to focus on what actually caused C not to drink the water.93 Even if we follow Stapleton by speculating about counterfactual scenarios rather than focusing on what actually happened, the poisoning of the water would not have caused C not to drink the water if the keg had not been emptied, since, under the usual understanding of the example, the poison was not detectable.94 A similar analysis applies in example 2, in which C actually drank the fatal dose of poison but was shot and killed before the poison had time to have fatal effect. The empirically derived causal generalisation for death by poisoning includes as a necessary condition, in addition to the person’s drinking the poison, a minimum amount of time of the person’s thereafter being alive: the time, determined empirically, that is required for the poison to have fatal effect, that is, for certain internal bodily processes to occur. (Strictly speaking, the occurrence of these bodily processes, rather than the passage of a certain amount of time, are the NESS conditions in the underlying causal laws; however, in causal generalisations we can and often do substitute for the bodily processes the time required for them to occur, especially when we have insufficient knowledge of the required bodily processes.) In example 2, the would-be poisoning victim is shot and dies after drinking the poison but before the specified required time of remaining alive in order for the poison to have fatal effect has elapsed. On the other hand, the causal generalisation and the underlying causal laws for death by shooting were fully instantiated.95 In example 3, the relevant causal generalisation for a ship’s being delayed by a bridge’s blocking the river includes the bridge’s collapsing into and blocking the river, the ship’s reaching the bridge while the bridge is blocking the river, and the ship’s crew’s seeing the bridge’s blocking of the river and stopping the ship before it runs into the bridge. This causal generalisation was fully instantiated for bridge A, but not for bridge B. Although bridge B’s collapsing into and blocking the river guarantees and thus is lawfully sufficient for the ship’s being delayed and not reaching a point beyond the bridge by a certain time, it is not causally sufficient since all the antecedent conditions in the ‘if ’ part of the relevant causal generalisation involving bridge B were not instantiated: the ship did not reach bridge B, and the ship’s crew did not see bridge B blocking the river.96 Michael Moore acknowledges that the NESS account reaches the proper conclusion in a non-circular, non-question-begging way in ‘early pre-emption’ situations. His examples of   See Stapleton (n 85) 82–84.   ibid; J Stapleton, ‘Unpacking “Causation” ’ in P Cane and J Gardner (eds), Relating to Responsibility (Oxford, Hart Publishing, 2001) 145, 178–83. For further elaboration of the flaws in Stapleton’s arguments, see Wright (n 13) 298–300, 305; Wright, Once More (n 14) 1115–20. 94   See Stapleton (n 93) 178 (poison was ‘odourless’). If C did not know that the keg had been emptied but did not attempt to drink from it because she knew the water in it had been poisoned, A’s poisoning of the water rather than B’s emptying of the keg was the pre-emptive cause of C’s death. See section III.I below. 95   Fumerton and Kress (n 1) 100; Hart and Honoré (n 5) 124; Honoré (n 10) 7-130, 7-134; Thomson (n 1) 151; Wright (n 5) 1795. 96   Fumerton and Kress (n 1) 100–01; Hart and Honoré (n 5) 250–51; Wright (n 5) 1796–97. Hart and Honoré agree that bridge A but not bridge B caused the boat’s delay; however, they claim that bridge A did not cause any financial loss resulting from the delay if bridge B’s collapse was not wrongfully caused. Hart and Honoré (n 5) 251. They have confused the causation issue with the normative issue of legal responsibility. Wright (n 5) 1797–98; Wright, Legal Responsibility (n 9) 1434–67. 92 93

300  Richard W Wright early pre-emption are a two-fires case, in which ‘the first fire burns all the fuel around the house, so the second fire cannot get to the house to burn it’, and a case that is similar to example 1, in which ‘the shooter shoots the victim before he drinks any of the poisoned tea’.97 He states: One of the conditions needed to complete the set of which the second fire/poisoning was a part, was the fuel leading up to the house/the drinking of the tea by the victim. Therefore, there is no set (of which setting the fire/poisoning the tea are members) that is truly sufficient for death. Missing are some positive conditions, conditions not described in question-begging causal terms. Sufficiency theorists thus can get the right answers in cases of early pre-emption.98

However, Moore claims that attempts to use the NESS account to resolve the causal issue in ‘late pre-emption’ situations are circular and question-begging, because they allegedly simply amount to requiring that the effect was not already caused by something else.99 His distinction between early pre-emption and late pre-emption is significantly different from the usual one. For Moore, early pre-emption situations are those in which ‘the pre-emptive cause removes something needed by the pre-empted factor to do any causal work’, while late pre-emption cases are those in which ‘there is no last event needed by the pre-empted factor and prevented by the pre-empting cause, save the ultimate effect itself ’.100 He treats the poison–shooting situation in example 2 above and a two-fires case in which the pre-empted fire approaches the already destroyed house from an opposite direction than the pre-emptive fire (and thus shares no fuel on the way to the doomed house) as late pre-emption cases,101 and he claims that attempts to apply the NESS account in such cases by specifying, as alleged necessary conditions in the antecedent of the relevant causal generalisation, conditions such as ‘the house exists at the time the second fire arrived at the site’ or ‘the poison remaining in [the would-be victim P’s] body a certain amount of time while she is still alive’ are simply disguised circularity, since the house existing, or [P’s] still being alive, is just there not being a house destruction or a death – and this, in a deterministic universe, is to say nothing else (other than the putative cause we are testing) caused such destruction or death.102

Moore’s argument ignores the distinction between the necessary conditions in the antecedent of a causal generalisation and the distinct condition that constitutes the consequent of the causal generalisation. Although to say that an entity exists or is alive at a certain time obviously entails (in our world) that nothing has caused the entity’s destruction or death prior to that time, it should also be obvious that the entity’s existence or being alive at a certain time is a state of affairs that is distinct from and not simply the converse of the destruction, death or non-existence of the entity at a later (remote or immediate) time;103 rather, it is one of the necessary but by itself insufficient conditions for the occurrence of the later event or state of affairs. As Collins, Hall and Paul explain,   Moore (n 1) 493.  ibid. 99   ibid 494–95. 100   ibid 493. See text to n 69 above for the usual distinction between early pre-emption and late pre-emption. 101   Moore (n 1) 493–94. 102   ibid 494. 103   It is unfortunate that on this issue Moore fails to recognise or employ the rigorous distinction between distinct states of affairs that he correctly employs in distinguishing between acts and consequences to rebut various arguments by criminal theorists earlier in his book. See ibid 14–19, 282. 97 98

The NESS Account: Response to Criticisms  301 When Suzy throws a rock at the window, breaking it, we naturally tend to think that there is just one sequence of events – the one initiated by Suzy’s throw – converging on the effect. But it is far better, at least for the purpose of systematic metaphysics, to see this effect standing at the intersection of two sequences of events: There is the interesting sequence just mentioned, and then there is the quite boring sequence consisting in the continued presence of the window, up to the moment it shatters. More generally, a proper theory of events almost certainly must count as such things that we ordinarily would classify as states, or standing or background conditions.104

In the two-fires example, the house’s destruction is an instantiation of the consequent of the causal generalisation for destruction by fire, which will occur only if all of the necessary conditions in the antecedent of that causal generalisation and its underlying causal laws are instantiated. The existence of the house as fuel when a fire reaches it is one of those necessary antecedent conditions, as much as the existence of the other fuel along the fire’s path to the house. All of the antecedent conditions were instantiated for the first fire to reach the house. At least one – the existence of the house when the fire reached it – was not instantiated with respect to the second fire, which arrived too late from the opposite direction. The same analysis applies in the poison–shooting hypothetical. The NESS account requires that we empirically determine, as best we can, what set of conditions is minimally sufficient for a certain causal process to occur. By empirical observation and experimentation, we determine that a specific minimal amount of time is required for a specific poison to have fatal effect after the would-be victim drinks the poison. If we knew enough about the causal process involved, we could (but, as a practical matter, need not) replace this required elapsed time by the NESS conditions for which it is a placeholder: the steps of the physical process that must occur inside the victim’s body during this elapsed time. The victim’s remaining alive for the specified amount of time is included as one of the necessary conditions when specifying the causal generalisation. When we later investigate whether the poison caused the death in a particular situation, we check to see whether every one of the conditions specified in the antecedent of the causal generalisation was instantiated on the particular occasion. The elapsed time condition was not, so the complete instantiation required for causal sufficiency is not satisfied. Although we rely on our prior empirically derived causal generalisations, there is no circularity in our definition, identification or application of those generalisations. Moore’s examples of late pre-emption actually are instances of early pre-emption given his definitions of those terms. Indeed, there are no actual pre-emption situations that fit Moore’s definition of late pre-emption, which rather is swallowed up by his definition of early pre-emption. Recall that he distinguishes late pre-emption situations from early pre-emption situations by the fact that, in the former, ‘there is no last event needed by the pre-empted factor and prevented by the pre-empting cause, save the ultimate effect itself ’.105 In both his two-fires hypothetical and his poison-shooting hypothetical, there is an ‘event’ 104   Collins, Hall and Paul, Introduction (n 9) 44; see Fumerton and Kress (n 1) 90 (‘John’s stabbing Mary in the heart caused her death because it was a necessary element in a set of actual conditions – including her being alive at the time that she was stabbed’); Hall (n 14) 231 (‘among the causes of the June [forest] fire is not just the lightning but also the very presence of the forest, filled with flammable material’); ibid 244 (noting that the causal history of a bombing by a plane includes the process of the flying of the plane and ‘less conspicuously, the process consisting in the persistence of the target’); Strevens (n 13) 103–04 (a jar must be in existence when a ball reaches it in order for the throw of the ball to cause the shattering of the jar); Thomson (n 1) 149 (including in a causal generalisation for a thrown rock’s breaking of a window the condition of the window’s remaining in the place where it was when the rock was thrown for at least the amount of time it would take the rock, thrown at a specified velocity, to reach that place). 105   See text to n 100 above.

302  Richard W Wright (a state of affairs) needed by the pre-empted factor and prevented by the pre-empting cause – the existence of the house when the second fire reaches it and the would-be victim’s remaining alive for a specific amount of time after drinking the poison, respectively – that is not the ‘the ultimate effect itself ’ (the destruction of the house or the death of the victim, respectively) and may not even have been the last event or state of affairs needed by the pre-empted factor in order for it to be a cause. Consider, for example, the necessity of the house’s not only still being in existence when the second fire arrives but also being sufficiently dry and remaining so for some period thereafter and with sufficient oxygen (necessary for combustion) being present and remaining present for some period thereafter. Moore discusses a modified version of the two-bridges hypothetical as an example of a third, supposedly distinctive, ‘trumping’ type of pre-emption that he claims creates the worst problems for the NESS account, since pre-emption allegedly exists even though the pre-empted factor ‘runs its whole course’.106 Moore changes the hypothetical by having the boat’s captain be aware of both collapsed bridges prior to the stopping of the boat, although he continues to assume that the captain stopped the boat solely because of the collapse of bridge A.107 This is a plausible assumption, especially if, after becoming aware of the collapse of both bridges, the captain proceeded down the river until he reached bridge A. However, it could also be the case that the captain’s knowledge of the collapse of bridge B was a duplicative, reinforcing positive reason for, and hence a cause of, his stopping the boat, rather than a pre-empted factor.108 The second, reinforcement rather than pre-emption, description of the causal situation seems more likely in the other example of ‘trumping’ pre-emption that Moore discusses, which also has been discussed by Jonathan Schaffer. In this example, a major and a sergeant each simultaneously order the same soldiers to advance. Schaffer and Moore assume that the soldiers advance solely due to the order of the major, although they normally would have obeyed the sergeant’s order.109 While the major’s order will (or should) pre-empt the sergeant’s order when their orders conflict, when their orders are consistent it seems to me that the sergeant’s order will continue to be a positive reason for advancing that reinforces the reason provided by the major’s order, rather than its being completely without effect.110 Nevertheless, if we assume, along with Schaffer and Moore, that the soldiers advanced solely due to the major’s order, no new type of pre-emption has been discovered and no new difficulties are created for the NESS account. The applicable causal generalisation is assumed to contain as a necessary antecedent condition for the effectiveness of an order that it be the highest-ranking relevant order. This antecedent condition was instantiated for the generalisation as applied to the major’s order, but not for the same generalisation as applied to the sergeant’s order, which thus, contrary to Moore’s assumption, did not ‘run its whole course’.111 If the sergeant’s order indeed had ‘run its whole course’ – that is, if the relevant causal laws were fully instantiated by a set of conditions including the sergeant’s   Moore (n 1) 494–95.   ibid 495. Moore treats my previously published analysis of the original version of this hypothetical as if it were my analysis of his modified version of the hypothetical. ibid. 108   See section III.F below. 109   Moore (n 1) 494–95; Schaffer (n 14) 67; see Coady (n 53) 333–35; Lewis (n 53) 81. 110   See section III.F below. 111   The same analysis applies to Schaffer’s primary illustration of trumping pre-emption, which assumes that a necessary condition in the antecedent of the relevant causal law for the effectiveness of a magical spell is that it be the first spell cast during the relevant time period. Schaffer (n 14) 59; see Lewis (n 53) 81. It also applies to Collins, Hall and Paul’s neuron example, which assumes a causal law according to which a neuron fires with the same polarity as the most intense incoming signal. Collins, Hall and Paul, Introduction (n 9) 27–28. 106 107

The NESS Account: Response to Criticisms  303 order but not the major’s order – then the sergeant’s order (as well as the major’s order) was a cause of the soldiers’ advance.112 As John Collins states, examples like this offered to illustrate a supposed distinct type of pre-emption ‘owe what plausibility they have to a confusion of causes with norms – the only sense in which a major’s order clearly trumps a sergeant’s order rather than pre-empting it in some other way is a normative sense’.113 For proper resolution of the two-bridges hypothetical, as modified by Moore, we need to know more about the thought processes in the captain’s brain. If he was determined to proceed down the river until forced to stop by a physical obstruction, then the relevant causal generalisation includes as a necessary antecedent condition the boat’s reaching the physical obstruction. This condition was instantiated with respect to bridge A but not with respect to bridge B, so bridge A’s collapse, but not bridge B’s collapse, was a cause of the boat’s stopping and thus of the delay in reaching Metropolis. If, instead, the captain’s learning of a blockage of the river was causally sufficient for his stopping the boat as soon as he learned of the blockage (or at the nearest convenient stopping point short of the first blockage), then bridge A’s collapse and bridge B’s collapse (more precisely, his knowledge of or belief regarding each collapse) were duplicative causes of the boat’s stopping and the delay in reaching Metropolis. If, on the other hand, as implicitly assumed by Moore, the relevant causal generalisation includes as a necessary antecedent condition not only knowledge of an obstruction but also that the knowledge be knowledge of the nearest known obstruction, then (knowledge of) bridge A’s collapse, but not bridge B’s collapse, was a cause of the stopping of the boat and the delay in reaching Metropolis.

E.  Duplicative Causation It is generally acknowledged that the NESS account, by requiring causal strong sufficiency rather than strong necessity, properly handles situations involving duplicative causation by two or more independently strongly sufficient conditions, such as two independently sufficient fires or floods that merge and destroy a building, which the sine qua non account mishandles since none of the conditions ‘made a difference’ by being a ‘but for’ cause.114 However, some have questioned the NESS account’s handling of a condition that was not, or could not be proven to be, either strongly necessary or independently strongly sufficient – eg, each of 10 drops of poison or discharges of pollution when only three such drops or discharges were necessary for the occurrence of the relevant harm. Such situations are quite common – eg, in the frequent cases of multiple fires, flood waters, noises, weights, pollutants, toxic substances and other conditions that combine to cause a particular harm115 or a more-than-sufficient number of individual affirmative votes under non-unanimity voting rules.116 112   cf Strevens (n 13) 104 (‘If there were a set of veridical conditions entailing that the mother’s [supposedly pre-empted] ball hits the jar, it would indeed have hit the jar. But then it would, intuitively, be a cause of the breaking, in which case Mackie’s [weak-necessity] account would be correct in deeming it so.’) 113   Collins (n 69) 114. Schaffer acknowledges that this example could be a case of ‘standard pre-emption’, in which causal processes in the soldiers’ brains filter out the sergeant’s order or only respond to highest-ranked orders. Schaffer (n 14) 67. 114   eg Fumerton and Kress (n 1) 96–97; Moore (n 1) 474. There is considerable disagreement over its ability to handle overdetermined failure of a causal process. See section III.I below. 115   Wright, Legal Responsibility (n 9) 1444–45; text to nn 10–11 and 40–41 above. 116   See Stapleton (n 1) 443.

304  Richard W Wright My initial elaborations of the NESS account handled these situations by using subsets of existing conditions or ‘at least so much’ descriptions of an existing condition to construct minimally sufficient sets of existing conditions, for the sufficiency of which the inclusion of the non-independently-sufficient condition was necessary.117 Some critics have objected that doing so involves hypothetical, counterfactual reasoning – in particular, an assumption that the conditions not included in the described subset or the at-least-somuch nature of an included existing condition did not actually exist.118 This is not true. The non-included existing conditions or greater-than-at-least-so-much nature of an included condition are simply ‘roped off ’ while determining whether the described existing conditions constituted a minimally sufficient set. Indeed, I have always emphasised that one must double-check to make sure that the roped-off existing conditions or their roped-off nature did not prevent the complete instantiation of the antecedent of the relevant causal generalisation and its underlying causal laws.119 In any event, the objection to these methods of disaggregating existing conditions is mooted when the necessity aspect of the NESS account is confined, as it is in my later elaborations of the NESS account, to the relevant causal generalisations and their under­lying causal laws. Under this simpler, more straightforward definition of a NESS condition, no disaggregation of existing conditions is necessary. A NESS condition need merely be part of the instantiation of one of the abstract conditions in the antecedent of an applicable causal law, all the conditions of which were at least minimally instantiated. It does not matter if some or all of them were more than minimally instantiated, although – I again emphasise – care must be taken to ensure that the more-than-minimal instantiation of some antecedent condition did not prevent the instantiation of some other antecedent condition.120 Moore acknowledges the validity and usefulness of the NESS account’s ability to identify as causes conditions that were neither strongly necessary nor independently strongly sufficient, while also noting and criticising the consequent increased proliferation of causes.121 David Fischer also objects to the proliferation of causes and questions the validity of recognising trivial contributions as causes – eg, a teaspoon of water added to a flooding river or a match added to a raging forest fire.122 Yet the teaspoon of water and the match contributed to and are part of the flood and forest fire, respectively. What if the same flood or fire 117   Wright (n 5) 1792–94; Wright, Legal Responsibility (n 9) 1444–45; Wright, Once More (n 14) 1106–07; Wright (n 3) 1035–39. Fischer asserts that Warren v Parkhurst 92 NYS 725 (NY Sup Ct. 1904), affirmed 93 NYS 1009 (NY App Div 1905), affirmed 78 NE 579 (NY 1906), which I have used to illustrate judicial recognition of causation by conditions that were neither necessary nor independently strongly sufficient, is not a ‘pure’ example. Relying on the distinction between ‘[legal] injury’ and ‘[actual] harm’ – a distinction the Warren court did not make – Fischer argues that, although the pollution by each of the 26 defendants was neither necessary nor independently sufficient for the ‘fixed threshold of stench’ that constituted the required legal injury, each defendant’s pollution was necessary for some separable part of the recoverable harm. Fischer (n 1) 285. He ignores the court’s statements that each defendant’s pollution by itself ‘would not cause any material change to the plaintiff by way of noxious smell’ but rather would be ‘merely nominal’ and would not support any liability. 118   Kelman (n 1) 603–04; DA Fischer, ‘Causation in Fact in Omission Cases’ (1992) Utah Law Review 1335, 1359, 1362; Fischer (n 1) 303–04, 307–08; Stapleton (n 85) 83–84. 119   eg Wright (n 13) 295–97; Wright, Legal Responsibility (n 9) 1444–45; Wright (n 3) 1035–37; see n 13 above. Strevens usefully discusses the difference between assuming the negation of some condition and simply setting it aside with no assumption as to whether or not it exists. Strevens (n 13) 96–97. Moore and Thomson recognise this distinction. Moore (n 1) 489–90; Thomson (n 1) 148, 156–59. Fischer and Stapleton do not. See Fischer (n 1) 303–04, 307–08; Stapleton (n 1) 472. 120   See text to nn 40–41 above. 121   Moore (n 1) 487–91. 122   Fischer (n 1) 290–91; see Restatement Third (n 1) s 27 comment i (initially suggesting the correct causal analysis but then analogising these duplicative causation situations to quite different pre-emptive causation situations).

The NESS Account: Response to Criticisms  305 were caused by a million (or many more) different people all contributing a teaspoonful of water or a single match? Denying that any of the teaspoonfuls or matches contributed to the destruction of the property that was destroyed by the flood or fire would leave its destruction as an unexplained, non-caused miracle. As a pure matter of causation, it cannot possibly matter whose hands supplied the different bits of water, flame or fuel. What is driving the intuition of no causation is the judgment regarding attributable responsibility, which is especially brought to mind if the question is posed as ‘Did the teaspoon of water or match destroy the property?’ rather than ‘Did the teaspoon of water or match contribute, even if only extremely minimally, to the flood or fire that destroyed the property?’ What is generally agreed upon is that the trivial contributor should not be held liable when her contribution was trivial in comparison to the other contributing conditions and was neither strongly necessary nor independently strongly sufficient for the injury at issue, but this is a normative issue of attributable responsibility rather than causal contribution.123 Stapleton also acknowledges the validity and usefulness of this application of the NESS account, but she claims that I have failed to provide any reasons or justifications for it – or at least for the law’s use of it – and have thereby (how?) made the NESS account ‘incoherent’.124 To the contrary, the reasons that Stapleton gives for endorsing it, and the NESS account as a whole as the proper ‘algorithm’ for identifying causal ‘involvement’ or ‘contribution’,125 are the same ones that I have repeatedly emphasised: the obvious need, in what is intended to be a comprehensive account of causation, to be able to identify all instances of causation, and the related need – especially in legal analysis – to distinguish the normative and context-specific purposive considerations that are often confused with the causation issue, which instead should be explicitly recognised, labelled and discussed as non-causal reasons for focusing on and attributing responsibility to only some of the many conditions that contributed to the occurrence of some consequence.126 It is Stapleton, rather than I, who has a history of failing to make this distinction.127 Her failures have been encouraged by Fischer, who refuses to distinguish between causation as a   Restatement Third (n 1) s 36; Fischer (n 1) 289–90; Wright, Legal Responsibility (n 9) 1448–50 and fn 84.   Stapleton (n 1) 474–76. 125   ibid 444, 474; Stapleton (n 93) 174. However, Stapleton’s supposed elaboration of the NESS account differs greatly from that account by employing a doubly counterfactual ‘duplicate necessity’ analysis of strong necessity rather than a real-world analysis of causal sufficiency. She treats a condition as a cause if it would have been strongly necessary in the counterfactual world in which the condition(s) that prevented it from being strongly necessary in the real world did not exist. Stapleton (n 1) 433, 436, 441–44, 472–74, 479; J Stapleton, ‘Legal Cause: Cause-in-Fact and the Scope of Liability for Consequences’ (2001) 54 Vanderbilt Law Review 941, 959–60; Stapleton (n 93) 175–85. This method of analysis, which unfortunately is adopted in s 27 of the Restatement Third, fails to distinguish duplicative causation from pre-emptive causation and is substantially over-inclusive and under-inclusive in both types of situation. See Restatement Third (n 1) s 27; Wright, Once More (n 14) 1110–15 and fn 149; Wright, Legal Responsibility (n 9) 1450–51 fn 86. The authors of the Restatement Third try to paper over these defects in comments that limit s 27 to duplicative causation situations, without providing any (nonviciously-circular) criteria for distinguishing the two types of situations, and that refer pre-emptive causation situations to supposed resolution by s 26’s sine qua non test, which, however, cannot properly resolve pre-emptive causation situations. See Restatement Third (n 1) s 26 comment k, s 27 comments e and h. A recent case thus wisely moves past the blackletter sections of the Restatement Third to focus instead on the NESS-based ‘causal sets’ analysis that is employed in the comments to sections 26 and 27. June v Union Carbide Corp 577 F3d 1234, 1242–44 (10th Cir 2009); see Restatement Third (n 1) s 26 comments c, d, i and k, s 27 comments a, b, e, f, g, h and i, and related reporters’ notes. 126   eg Wright (n 5) 1737–58, 1764 fn 121, 1781–83, 1791–92; Wright, Legal Responsibility (n 9) 1437–40; Wright, Once More (n 14) 1072–82, 1111, 1119–23; Wright (n 3) 1004–18; see Stapleton (n 1) 433, 441–42, 445– 46, 455–58, 463–64, 473–77, 479–80; section III.J below. 127   See Stapleton, Legal Cause (n 125) 957 fn 38, 966–68 and fns 60 and 61; Stapleton (n 85) 61, 62–66 and fn 13, 77 fn 40, 79–80, 81–84; Wright, Once More (n 14) 1101 fn 108, 1111–13, 1119–23. 123 124

306  Richard W Wright purely scientific matter and attributable legal responsibility and instead seeks to have ‘causal’ judgments depend on intuitive judgments of legal responsibility, but who nevertheless ultimately agrees for practical reasons with the American Law Institute’s finally making this distinction in the Restatement Third (after having confused students, lawyers and courts for decades by its failure to do so in the first and second Restatements).128 Stapleton, while now claiming to support the distinction, instead abandons it and undercuts her primary argument – that the law should choose ‘involvement’ (as identified by the NESS ‘algorithm’) as the sole ‘interrogation’ regarding ‘causation’ in the law129 – when she claims that there is no core empirical or metaphysical concept of causation, so that (as bluntly asserted by Lord Hoffmann at the conference at which the initial version of this chapter was presented), judges and others can and should define causation in any manner that suits their particular purposes: [Schaffer’s attempt to provide an account of causation] is the same doomed concept of many other philosophers: a search for ‘a broad and nondiscriminatory concept’ of causation. In contrast to Schaffer’s passing flirtation with the idea that such a concept might be ‘a philosopher’s myth,’ the central argument of this article is that it is indeed a myth.130 It is futile for philosophers to search for a coherent freestanding metaphysical account of ‘causation’ unless a choice of underlying interrogation (blame, explanation, physical role, any sort of involvement etc) is specified at the outset.131

Stapleton’s seeming radical scepticism about there being any core concept of causation runs counter to her discussions of her concept of ‘involvement’, which she states is meant to encompass and be limited to the natural scientific relation that we call ‘causation’ and is determined objectively by reference to the physical laws of nature, excluding any normative considerations.132 As she seems to recognise, all the other ‘interrogations’ (regarding blame, responsibility, contextual salience and so forth) that loosely employ causal language use this core concept of causation but also use normative or other purposive or pragmatic considerations to focus on only some of the contributing conditions and to limit attribution of responsibility even for those contributing conditions.133 While she often seems to argue otherwise,134 the purposive considerations that determine the particular focus of the causal enquiry do not result in different concepts of causation being employed once the focus has been set.135

  See Fischer (n 1) 284–92, 312–13, 317; Wright, Once More (n 14) 1074–80, 1121 fn 172.   Stapleton (n 1) 433, 441–44, 455, 473–74; see nn 125–26 above. 130   Stapleton (n 1) 439 fn 15 (quoting J Schaffer, ‘Contrastive Causation’ (2005) 114 Philosophical Review 327, 350). 131   ibid 439; see ibid 447 (same), 456 (‘causal language can denote whatever we choose it to’), 459 (same). 132   ibid 433–37, 444, 446, 452–53 and fn 45, 455, 474, 479–80. However, if by ‘physical laws of nature’ Stapleton means the mathematical formulations of such laws, which do not incorporate the critical interrelated concepts of causal sufficiency and directionality, she is relying on the inadequate notion of mere lawful relation. See Collins, Hall and Paul, Introduction (n 9) 16, 34; text to nn 80-87 above. In a note, she refers to theoretical physicist Richard Feynman’s description of physical laws in terms of ‘the principle of least action’ and restraints on and increasing entropy within a physical environment, but this is not a description that provides a basis for identification of singular instances of causation in law or ordinary life. See Stapleton (n 1) 440 fn 16 (citing R Feynman, The Character of Physical Law (New York, Random House Modern Library, 1994) 46). 133   See Stapleton (n 1) 440–41, 444–46, 448–51, 455–64; Collins, Hall and Paul, Introduction (n 9) 36–37; Hall (n 14) 228–31; Fumerton and Kress (n 1) 87–88; Honoré (n 10) 7-49 to 7-53; Thomson (n 1) 160–63; Wright (n 5) 1741–50; Wright (n 3) 1011–14. 134   Stapleton (n 1) 437, 439–41 and fn 15, 444, 445–46, 447, 451, 476–77 and fn 149. 135   Wright (n 5) 1788–1803. 128 129

The NESS Account: Response to Criticisms  307 Stapleton’s basic point seems to be that it is impossible to provide a comprehensive, noncircular account of this core concept of causation – even though she seems to assume that her concept of ‘involvement’ has succeeded in doing so. More specifically, she asserts that, although the NESS account (which she claims to employ and in tight spots does employ) is a very useful ‘algorithm’ for identifying all instances of ‘involvement’, it does not capture the essence or meaning of causation.136 Yet, if, as she concedes,137 the NESS account is able to identify all instances of causation (and its lack) given sufficient information about existing conditions and the relevant causal laws, hasn’t it then (as she once stated138) captured the essence of causation and given it a comprehensive specification and meaning? Her claim that it has not seems to be based primarily on the charge that the NESS account, by referring to the interrelated concepts of ‘causal sufficiency’ and ‘causal laws’, engages in ‘vicious conceptual circularity’139 – a claim that I have rejected above,140 but which applies to her own ‘involvement’ account.141

F.  Reasons as Causes The NESS account’s ability to identify conditions that were neither strongly necessary nor independently strongly sufficient as causes applies to and is especially useful for accounting for human decisions and actions, which often are based on multiple reasons, none of which may have been – or can be proven to have been – strongly necessary or independently strongly sufficient for the particular decision or action.142 Hart and Honoré acknowledge that loose generalisations apply to human decisions and actions, but they deny that these decisions and actions are governed by causal laws. They point out that it is not practically possible to specify all the conditions sufficient to produce a certain decision, and they deny that the same decision or action necessarily would be taken in the same situation by the same person at different times or by different persons at the same or different times.143 The first point is true, but it does not differentiate human decisions and actions from other causal processes. We can and do make plausible singular causal statements based on incomplete knowledge of the causal laws that underlie the causal generalisations that we employ.144

136   Stapleton (n 1) 472–74 and fn 145, 477; contra Honoré (n 1) 367 (treating the NESS account as capturing the meaning of natural causation as well as being a useful test for identifying natural causes). 137   Stapleton (n 1) 444, 474. 138   Stapleton (n 93) 174, 179. 139   Stapleton (n 1) 472–73 and fn 145. 140   See nn 25–34 above. 141   Stapleton asserts that her use of the term ‘involvement’ ‘avoid[s] circular causal terminology’. Stapleton (n 1) 436. Yet not only does ‘involvement’ imply causal contribution, she explicitly defines ‘involvement’ as ‘contribution’ in her delineation of her three forms of ‘involvement’, states that ‘involvement’ is determined through our knowledge of ‘the physical laws of nature’, for which she provides no usable definition or elaboration, and endorses the NESS account, with its allegedly circular reliance on causal laws, as the proper ‘algorithm’ for identifying all three forms of involvement. ibid 433–37, 441–44, 446, 452–53 and fn 45, 472–74, 479–80; text to n 125 above. 142   Wright (n 13) 297–98; Wright (n 3) 1035–37. Stapleton ignores these sources when she asserts that ‘Wright does not adequately address . . . whether it is coherent, acceptable, convenient or wise notionally to disaggregate an individual person’s decision’. Stapleton (n 1) 476. 143   Hart and Honoré (n 5) xxxvii, 2, 22–23, 55–61; Honoré (n 1) 363, 382, 384; Honoré (n 10) 7-53, 7-168. 144   See text to nn 35–39 above.

308  Richard W Wright The second point also is true, but it is true only because all the relevant conditions are not the same. As Honoré notes, even for the same person, there is the additional information provided by the similar past experience.145 There also is the additional information provided by other intermediate experiences, other sources of new knowledge, and possibly different decision criteria due to, eg, changed goals and attitudes towards risk. Two different persons have different experiences, genetic makeup and goals. Honoré agrees that attributions of reasons for decisions or actions involve a belief that the (known and unknown) reasons in the aggregate were sufficient (along with other relevant conditions) for the decision or action, although none of them may have been strongly necessary or independently strongly sufficient. He treats such reasons as ‘necessary elements in a set of reasons together sufficient’ for the relevant decision,146 and he cites with apparent approval my application of the NESS account to such reasons.147 However, he claims that the sense of sufficiency is different than for physical sequences.148 The difference he seems to have in mind is the alleged lack of repeatability in the same circumstances, but this is contrary to the assumed sufficiency of the set of conditions, including reasons, on the particular occasion. As I have previously noted, if all the relevant conditions were jointly sufficient in the first instance, they will also be jointly sufficient in subsequent instances involving the same set of relevant conditions:149 [I]f all the relevant conditions (accumulated experience and knowledge, beliefs, goals, mood, and so forth) were the same, surely the decision or action would also be the same. To assert otherwise is to assert that human action is random or arbitrary. Human action is less regular and predictable than physical events because humans learn from prior experiences and new information, the range of relevant conditions is much broader, and the applicable causal generalizations are much more complex and less well understood.150

The fact that the applicable causal generalisations and the underlying causal laws for human decisions are much more complex and less well understood than those for physical events often makes it especially difficult to determine what decision a person would have made if she had been provided with some information or opportunity that was not provided. In a recent essay, I erroneously stated that we need to have this question answered, and thus must use (highly restricted) counterfactual analysis, to resolve the causal issue in situations in which the defendant tortiously failed to provide some safeguard (information or device), which if provided and used would have prevented the injury that occurred.151 The causal issue in such situations involving the failure of a preventive causal process instead should be resolved, without any counterfactual analysis, by using the method of analysis that is discussed in sections III.H and III.I below. As is discussed in section III.I, the lack of an attempt to access and use a safeguard preempts the potential failure of the preventive causal process (involving the successful use of   Honoré (n 1) 382.   ibid 383.   ibid 377 fn 40 (citing Wright (n 3) 1035–37). 148   ibid 383–85; see Honoré (n 10) 7-121. 149  Although Mackie argues for strong necessity rather than strong sufficiency when identifying singular instances of causation, see text to nn 19–22 above and nn 156–58 below, he agrees that there is a single basic concept of natural causation that applies equally to physical events and human decisions and actions. Mackie (n 9) 122–26. 150   Wright (n 13) 298. 151   ibid 301–02; see V Black, ‘Decision Causation: Pandora’s Tool-Box’ in J Neyers, E Chamberlain and SGA Pitel (eds), Emerging Issues in Tort Law (Oxford, Hart Publishing, 2007) 309. 145 146 147

The NESS Account: Response to Criticisms  309 the safeguard) at the subsequent stage in that process, never reached, in which the attempt would have failed due to the non-provision of the safeguard. However, the non-provision of the safeguard nevertheless was a cause of the non-prevented injury if (unlike the braking situation discussed in section III.I) the lack of an attempt to use the safeguard was caused by its non-provision – for example, by knowledge of its non-provision or, conversely, by lack of knowledge of its possible availability due to its non-provision. As in other overdetermined causation situations, the causal issue in these non-­providedsafeguard situations unfortunately is often confused with the ‘no worse off ’ limitation on attributable responsibility, which generally treats the defendant as not being legally responsible for an injury, despite having tortiously contributed to its occurrence, if the injury would have happened anyway due to non-liability-generating conditions – for example, if a plaintiff (who cannot sue himself) would not have used the information or safeguard even if it had been provided.152 To resolve this attributable responsibility issue in these and other pre-emptive causation situations, (highly restricted) counterfactual analysis will be required. Treating human decisions, like other biologically based mental processes, as subject to causal laws is not incompatible with a plausible conception of free will, given the very complex, goal-directed nature of human decision-making. Indeed, to treat human decisions as not being subject to causal laws would imply that they are arbitrary and irrational, rather than free in the sense of being based on one’s interests and goals. On the other hand, as I will explain in the next section, the presence of a random or probabilistic element in human decision-making would not undermine the concept of causation or make it impossible to provide causal explanations of human decisions and actions.

G.  Indeterminism and ‘Probabilistic Causation’153 In a radically indeterministic world, in which nothing was (weakly or strongly) necessary or sufficient for anything else, the concepts of causation and probability likely would not exist due to lack of instantiation. Our world is at most only partially indeterministic – that is, probabilistic. A great many apparently probabilistic processes appear to be so only because we have insufficient knowledge of the underlying causal laws and turn out on closer examination to be deterministic.154 As far as we know all apparently indeterministic processes may be deterministic. As Mackie stated, the difficulty one encounters in trying to describe truly indeterministic statistical ‘laws of working’ reinforces doubt about their existence. Such a law would seem to require an objective chance or propensity inherent in each individual occurrence of the antecedent conditions, but one can deny the existence of such objective entities with arguments similar to those used to deny the existence of objective causal qualities or powers in deterministic processes. Mackie concluded that the only workable formulation would be a limiting frequency on actual outcomes of a series of instances of a certain set of antecedent conditions. Yet, he noted, this formulation deprives such statistical laws of explanatory power. There is no apparent way to explain, without invoking deterministic

  See Wright, Legal Responsibility (n 9) 1434–78.   Portions of this section are taken, with minor modifications, from Wright (n 3) 1028–31.   See Mackie (n 9) 49–50, 76, 237–38, 242–46.

152 153 154

310  Richard W Wright laws, why an actual series does or should approach a limiting frequency. Truly indeterministic or probabilistic laws, therefore, are unintelligible and mysterious.155 However, Mackie and others have used the theoretical possibility of partially indeterministic processes to argue against the universal applicability of a strong-sufficiency account of causation and, indeed, for the preferability of the sine qua non account. Mackie postulates a candy machine L that never produces candy unless a shilling is inserted, but sometimes, due to indeterminism, does not produce candy even when a shilling is inserted. He assumes that this means that the insertion of a shilling in machine L is strongly necessary but not strongly sufficient for the machine’s production of candy, and that, since whenever candy is produced we will treat the strongly necessary condition of the insertion of the shilling as a cause, strong sufficiency is not necessary for causation, but strong necessity is.156 Fumerton and Kress make the same argument using a similar example: a bomb activated by the indeterministic decay of a radioactive element, which has a very low probability of going off during the intended victim’s lifetime but which nevertheless does go off a few minutes after being planted and kills the intended victim. However, countering Mackie, they note that even the sine qua non account will not properly identify the decay-activated bomb as a cause if there was a backup deterministically activated bomb.157 Mackie, Fumerton and Kress incorrectly assume that there is a lack of strong sufficiency in these examples. They seem to assume that the planting of the bomb or the insertion of the shilling must be sufficient all by itself – a situation that is rarely if ever true. The planting of the bomb or the insertion of the shilling need only be one of the conditions that are each necessary for the sufficiency of the set of conditions that constitutes the antecedent of the relevant causal law. Another of those conditions in each example is the occurrence of the specified indeterministic state or event: the decay of the radioactive element in the bomb or the unknown indeterministic state or event in the candy machine, respectively. When the bomb explodes or the machine produces candy, the planting of the bomb or the insertion of the shilling, respectively, was both strongly necessary and strongly sufficient for the occurrence of that result, as was the occurrence of the relevant indeterministic state or event. Mackie’s other candy machine hypothetical demonstrates, contrary to what he intended, that causation requires causal (rather than merely lawful) strong sufficiency rather than strong necessity. This machine, M, always produces candy when a shilling is inserted, but occasionally, as a result of some mysterious indeterministic process, produces candy even though no shilling was inserted. Mackie assumes that the insertion of a shilling in machine M is strongly sufficient but not strongly necessary for the production of candy. He asserts 155   Mackie (n 9) 239–47; see Wright (n 3) 1029, 1042–49. Some believe that the notion of deterministic causation collapsed with the announcement in 1927 of Heisenberg’s uncertainty principle. That principle, even if valid, merely states that it is impossible to measure simultaneously the position and velocity of an object so that the product of the uncertainties in measurement is less than an infinitesimally small constant – about 10 to the minus 34th power joule-seconds. 10 Encyclopaedia Britannica, Micropaedia,15th edn (1979) 253. While this might cause problems in hypothetical lawsuits between subatomic particles, it creates no problems at the macroscopic level of events in everyday life. Others believe that the notion of causation collapsed with the introduction of functional equations in science. Yet functional relationships in science are merely mathematically quantified statements of causal laws or generalisations, which usually are expressed through time-based derivatives of the regularities of succession that constitute ordinary causal generalisations. Mackie (n 9) 143–48, 153–54, 216–18. The concept of causation is alive and well in the natural and social sciences. 156   Mackie (n 9) 41–43. 157   Fumerton and Kress (n 1) 97–98.

The NESS Account: Response to Criticisms  311 that when candy is produced by machine M following the insertion of a shilling, we will be unable to say whether the production of the candy by machine M was caused by the insertion of the shilling or by the indeterministic process. He concludes that strong sufficiency is not sufficient to establish causation; rather, strong necessity is required.158 Yet, when candy is produced after a shilling is inserted in machine M, we know by definition that the insertion of the shilling produced the candy through the deterministic process, if the insertion of a shilling is causally strongly sufficient for the production of candy. What we do not know is whether the probabilistic event and its associated indeterministic process also occurred, since we have no knowledge of and cannot directly observe the probabilistic event or its associated indeterministic process. If the probabilistic event did not occur, the candy was produced solely by the deterministic process. If the probabilistic event did occur, then either the candy was duplicatively produced by both the deterministic and the indeterministic processes (if the probabilistic event also was causally strongly sufficient) or else the candy was produced solely by the deterministic process, which pre-empted the indeterministic process. Thus, no matter what happened with the indeterministic process, the insertion of the shilling was a (duplicative or pre-emptive) cause of the production of the candy. This conclusion is logically compelled by the definition of causal strong sufficiency, if the insertion of a shilling is causally strongly sufficient for the production of candy in machine M. If that assumption is removed, so that we only have the empirical observation that machine M produces candy whenever a shilling is inserted, but sometimes produces candy although nothing was inserted, then Mackie is right when he insists that we cannot say on any particular occasion that the insertion of the shilling was a cause of the production of the candy. In the absence of the causal strong sufficiency assumption, it is possible that the deterministic process initiated by the insertion of the shilling was pre-empted by the indeterministic process. But this reinforces the argument that causation requires causal strong sufficiency. When we assumed that causal strong sufficiency existed, we were compelled to conclude that the insertion of the shilling caused the production of the candy. When we dropped the causal strong sufficiency assumption, we no longer could identify the cause(s) of the candy’s production. In sum, Mackie’s indeterministic machine hypotheticals, and others like them,159 demonstrate the reverse of what he claims. They demonstrate that attributions of causation, even in a partially indeterministic world, depend on the causal strong sufficiency requirement embodied in the NESS account rather than the strong necessity requirement embodied in the sine qua non account.

H.  Omissions and Negative Causation We routinely identify omissions, such as the failure to water a plant or to pay attention while driving, as causes of some result, such as the death of the plant or a traffic accident, respectively. However, there is reason to question whether omissions or other negative conditions can actually be causes or effects. As John Stuart Mill, who provided the philosophical foundation of the NESS account, observed, ‘From nothing, from a mere negation, no   Mackie (n 9) 41–43.   See eg, Collins, Hall and Paul, Introduction (n 9) 26–27. Their examples nicely illustrate the defectiveness of attempts to define causation in terms of increased probability. See also Wright (n 3) 1042–49. 158 159

312  Richard W Wright consequences can proceed’.160 Mill nevertheless saw the need to include negative conditions as causal (NESS) conditions: We say for example, The army was surprised because the sentinel was off his post. But since the sentinel’s absence was not what created the enemy or put the soldiers asleep, how did it cause them to be surprised? All that is really meant is, that event would not have happened if he had been at his duty. His being off his post was no producing cause, but the mere absence of a preventing cause: it was simply equivalent to his non-existence. From nothing, from a mere negation, no consequences can proceed. All effects are connected, by the law of causation, with some set of positive conditions; negative ones, it is true, being almost always required in addition. In other words, every fact or phenomenon which has a beginning invariably arises when some certain combination of positive facts exists, provided certain other positive facts do not exist. . . . The cause, then, philosophically speaking, is the sum total of the conditions positive and negative taken together; the whole of the contingencies of every description, which being realised, the consequent invariably follows. The negative conditions, however, of any phenomenon, a special enumeration of which would generally be very prolix, may be all summed up under one head, namely, the absence of preventing or countervailing causes.161

In this passage, Mill draws two important distinctions, without clearly distinguishing them or clarifying the precise nature of the second distinction. The first distinction is between ‘positive’ conditions and ‘negative’ conditions; the second distinction is between a ‘producing’ or ‘positive’ cause and a ‘non-producing’ or ‘negative’ cause. There seems to be general agreement on the first distinction: a ‘positive’ condition is the presence of some act, event or entity in a particular situation; a ‘negative’ condition is the absence of some act, event or entity in a particular situation. There is less agreement on the existence and nature of the second distinction.162 I consider the distinction to be as follows: a ‘producing’ or ‘positive’ cause is a NESS condition that is connected to the consequence at issue by a causal process in which each step in the process has a positive condition as its consequence; a ‘negative’ cause is a NESS condition that is connected to the consequence at issue by a causal process in which at least one of the steps in the process has a negative condition as its consequence. Many philosophers believe that a negative condition cannot be a positive cause, but rather, at best, can only be a negative cause.163 Some argue that negative conditions cannot be part of any plausible account of reality and thus cannot be any sort of cause. Moore is most insistent on this point. He notes, correctly, that there are no negative conditions in the sense of the negative opposite of a positive condition, for example non-tramplings by nonelephants. Rather, a negative condition must be understood to be the absence of some positive condition, rather than the ‘ghostly’ reality of its negative opposite.164 Moore and others   Mill (n 5) bk III ch V s 3.  ibid. 162   Jonathan Schaffer describes ‘positive’ causation as a ‘physical connection’ through transmission of some physical entity, structure or ‘mark’ between a cause and the consequence, without the involvement of any negative conditions, and ‘negative causation’ as counterfactual dependency of the consequence on the cause in the absence of any such physical connection. J Schaffer, ‘Causes need not be Physically Connected to their Effects: The Case for Negative Causation’ in Christopher Hitchcock (ed), Contemporary Debates in Philosophy of Science (Oxford, Blackwell, 2004) 197, 203–04, 214. Ned Hall makes a similar distinction between ‘production’ and ‘dependence’ as two different concepts or kinds of causation. His concept of ‘production’ focuses on the spatio-temporal contiguity and ‘intrinsicness’ (internal structure) of the causal process rather than physical transmission of some entity. Hall (n 14) 225–27, 253–54. 163   eg Hall (n 14) 253, 260; Schaffer 162) 197–98; Thomson (n 1) 163 fn 12. 164   eg Moore (n 1) 53–55, 399–400, 444–45, 460–61. 160 161

The NESS Account: Response to Criticisms  313 agree that propositions about the absence of some positive condition can be included as NESS conditions in laws. However, Moore notes, again correctly, that such propositions, being mere abstractions rather than aspects of reality, cannot be concrete instantiations of the abstract conditions in causal laws and thus cannot enter into causal relations as causes or effects.165 What Moore refuses to acknowledge is that the absence of some positive condition is an aspect of reality, a fact that makes the relevant proposition true or false, 166 that is as much a part of the real, existing state of affairs on a particular occasion as are the positive conditions in existence on that occasion.167 Indeed, negative conditions often are positive or producing causes rather than negative causes, through the perception of the negative condition by a sentient being. For example, the absence of a salute by a soldier is a positive cause of an officer’s perception of that absence and the consequent disciplining of the soldier, a baseball player’s failure to touch a base is a positive cause of an umpire’s perception of that failure and the consequent ruling that the player is out, the absence of a stop sign or signal is a positive cause of a motorist’s perception of that absence and the motorist’s consequent proceeding through the intersection with his foot pushing down on the accelerator, and the absence of a poison label on a container is a positive cause of a consumer’s perception of that absence and con­sequent drinking, ingestion or handling of the contents of the container.168 Negative conditions also can be and often are effects. Indeed, a refusal to recognise negative conditions as effects makes it impossible to account for what is probably the most paradigmatic instance of an effect, to which Moore repeatedly adverts: death. Although Moore argues otherwise,169 death as an event is a transition between a positive condition,   ibid 303–04, 347–48, 351–53, 445, 461, 479. David Lewis agrees, but nevertheless treats absences as causes: Absences are bogus entities. Yet the proposition that an absence occurs is not bogus. . . . And it is by way of just such propositions, and only by way of such propositions, that absences enter into patterns of counterfactual dependence. . . . Should we conclude, then, that when we say absences are causes, really it is true negative propositions that do the causing? – No; in other cases, we distinguish between the cause itself and the true proposition that describes it. . . . [T]he proposition is a necessary being, ‘abstract’ in one sense of that multifariously ambiguous term, and doesn’t cause anything. . . . I refuse to concede that facts – true propositions – are literally causes. So I have to say that when an absence is a cause or an effect, there is strictly speaking nothing at all that is a cause or effect. Sometimes causation is not a relation, because a relation needs relata and sometimes the causal relata go missing. Lewis (n 53) 100. 166   I use ‘fact’ in the sense of a part of reality, a concrete property of a state of affairs, what Moore refers to as ‘facta’ to avoid confusing this sense with the different sense (that eg Ned Hall and David Lewis have in mind) of a true proposition about reality. See Hall (n 14) 254; Lewis (n 53) 100; Moore (n 1) 333, 341–47. It is not clear to me which sense Judith Thomson has in mind. See Thomson (n 1) 148, 150. 167   Moore (n 1) 352. At one point (only), to preserve his criticism of the NESS account’s recognition of omissions as causes, Moore acknowledges that abstract negative conditions can be instantiated: When [NESS and other strong-sufficiency theorists] are asking after the truth makers for the negative conditions in laws, they are not looking for negative tokens of negative types. Rather, the truth-maker for negative conditions is the absence of any positive instances of some type. The law is “instantiated” in this sense whenever its negative conditions are made true by the world. In that sense, negative conditions can be ‘instantiated’, and so omissions will be causes. ibid 479. 168   Wright (n 13) 291; cf Collins, Hall and Paul, Introduction (n 9) 48 (‘Billy’s failure to show up for their lunch date causes Suzy to become disappointed’); Schaffer (n 162) 201 (‘One can perceive black holes or any other black entity, which is merely an absence of radiated light energy’). The same cognitive processes, which are limited and often biased, are involved in the perception of positive as well as negative conditions. A good example is the well-known optical illusion in which one sees either an old hag or a young, beautiful woman when looking at the drawing or picture of the woman, but never both at the same time. 169   eg Moore (n 1) 442 (‘Dying is also a presence, even if it can be described as “not surviving”, and surviving is an absence, even if it sounds like it is referring to some actual state of affairs’); see ibid 53–55, 303. 165

314  Richard W Wright being alive (with a functioning heart or brain, depending on one’s preferred definition) to a negative condition, the absence of life – which, unfortunately, for each of us is an ultimately certain real state of affairs. The relevant positive causal processes for the main­ tenance of life, at least as we know it, include as necessary positive conditions the presence of sufficient amounts of water, oxygen (for humans and other complex animals) and certain nutrients. The negative condition of an absence of the required amount of water, oxygen or nutrients, or of the blood that transports oxygen and nutrients to the brain and other parts of the body, is a negative cause of the negative condition of death (the absence of life). As Jonathan Schaffer demonstrates, to deny that negative conditions participate in causation (generally but not always as negative causes or effects) is to blind oneself to the layers of the causal web and, more significantly, to deny that humans ever cause anything: The pattern of negative causation features in even the most paradigmatically causal cases. Suppose that the sniper feels murderous, pulls the trigger, fires a bullet through the victim’s heart, and the victim dies. Here is a paradigmatic causal sequence, every step of which is negative causation. Working backwards, surely the firing of the bullet through the victim’s heart causes the victim to die. But heart damage only causes death by negative causation: heart damage (c) causes an absence of oxygenated blood flow to the brain (~d), which causes the cells to starve (e) [and die]. . . . At the next step backwards, surely the pulling of the trigger causes the bullet to fire. But trigger pullings only cause bullet firings by negative causation: pulling the trigger (c) causes the removal of the sear from the spring (~d), which causes the spring to uncoil, thereby compressing the gunpowder and causing an explosion, which causes the bullet to fire (e). . . . At the third and final step backwards, surely the sniper’s feeling murderous causes him to pull the trigger. But nerve signals only cause muscle contractions (such as that of the sniper’s trigger finger) by negative causation: the firing of the nerve (c) causes a calcium cascade through themuscle fiber, which causes calcium-troponin binding, which causes the removal of [the inibitor] tropomyosin from the binding sites on the actin (~d), which causes myocin-actin binding, and thereby causes the actin to be pulled in and the muscle to contract (e). . . . Since all voluntary human actions are due to muscle contractions, it therefore follows that all voluntary human actions (perhaps the most paradigmatic of all causes) involve negative causation.170

As Schaffer’s example illustrates, acts and other positive conditions often are negative causes. Acts of shooting, stabbing, strangling, removal of water or food, etc cause the negative condition of the absence of life by causing the negative condition of the absence of one or more of the necessary conditions (water, oxygen, nutrients, blood etc) for the positive causal process of maintaining life. Acts as well as omissions also cause positive conditions through negative causation, by preventing a positive causal process that would have prevented the occurrence of the positive condition. For example, the act of removing a safety device or damaging it so that it no longer works results in the same negative state of affairs 170   Schaffer (n 162) 199–200; see ibid 202 (‘Biologists routinely invoke negative causation, such as in explaining diseases. What causes scurvy is an absence of vitamin D, what causes diabetes milletus is an absence of insulin, and what causes dwarfism is an absence of growth hormone, and so on. The way in which HIV causes death is by disconnecting the immune system’). Schaffer notes that ‘negative causation is supported by all the central conceptual connotations of causation, including counterfactual, statistical, agential, evidential, explanatory, and moral connotations’. ibid 202; see ibid 198–203; D Fair, ‘Causation and the flow of energy’ (1979) 14 Erkenntnis 219, 248 (noting that a shadow is an absence of light and that ‘ice on the road caused the auto accident because the road failed to transmit its usual frictional force to the tires’).

The NESS Account: Response to Criticisms  315 – the lack of a working safety device, which if present and working would have prevented some injury – that exists if there is no working safety device in the first place. The act of removing the top X feet from a dam results in the same negative state of affairs – the lack of adequate height of the dam, which if present would have prevented a flood – that exists if the dam initially lacked those X feet of height.171 Thus, the critical distinction for purposes of causal analysis is not the distinction between positive conditions (eg, acts) and negative conditions (eg, omissions), but rather between positive causation and negative causation. This distinction is critical because the analysis of each type of causation, although based on the same underlying (NESS) account of causation, differs owing to their differing focus. A negative condition, eg, being dead, is the absence of a positive condition, eg, being alive. Such absences exist owing to the lack of occurrence of the positive causal process(es) that would have produced the relevant positive conditions. Thus, while the analysis of positive causation focuses on the successful completion of the relevant positive causal process(es), the analysis of negative causation focuses on the failure of a positive causal process. To know that a positive causal process succeeded, we need to know that all of the necessary conditions in the antecedents of the relevant causal generalisations and their underlying causal laws were instantiated on the particular occasion. To know how one or more of them was instantiated, we need to go one or more levels deeper in the causal web. To know that a positive causal process failed, we need know that at least one of the necessary conditions in the antecedents of the relevant causal generalisations and their underlying causal laws was not instantiated. To know what caused its failure we need to know at what point it failed and go one or more levels deeper in the causal web to find out why it failed at that point. As Mill noted, for any particular positive causal process, there are a multitude of possible preventing causes that could prevent its successful completion by preventing the instantiation of one of its necessary antecedent conditions. Mill’s omnibus negative condition – the absence of any preventing cause – encompasses all the different possibilities, a listing of which is practically impossible.172 If all the necessary conditions (which may include negative conditions)173 for the positive causal process were fully specified, there perhaps would be no need to engage in any analysis of negative causation, since the instantiation of all of the fully specified necessary conditions would be inconsistent with the existence of any preventing cause. However, such a complete listing of all the necessary conditions is rarely if ever possible in practice. Mill’s omnibus negative condition – the absence of any preventing cause – partially fills the gap, while also serving the useful purpose of emphasising the importance of always considering the possible existence of preventing causes.

171   Wright (n 13) 291–92. Negative causation by acts or other positive conditions is often referred to as ‘prevention’ – simple ‘prevention’ when the consequence is a negative condition and ‘double prevention’ when the consequence is a positive condition that occurred because the act prevented the occurrence of a causal process that, if not prevented, would have prevented the occurrence of the consequence. See Hall (n 14) 241; Moore (n 1) 54, 62, 130–31, 303, 459–61. 172   See text to n 161 above. Moore properly criticises me for some ‘incautious’ references to this omnibus negative condition when engaging in some specific causal analyses, while generously granting that I do not place much reliance on it. Moore (n 1) 492, 494. In each of the cited analyses, I relied instead on non-instantiation of a specific necessary condition in the positive causal process. See Wright (n 5) 1795; Wright, Once More (n 14) 1129–30; text to nn 88, 95–96 above. 173   See text to n 168 above.

316  Richard W Wright

I.  Overdetermined Negative Causation Philosophers, including those who reject counterfactual dependency analyses of positive causation, generally believe that counterfactual dependency analysis is required for situations involving negative causation.174 Yet, as with overdetermined positive causation, counterfactual dependency (sine qua non, ‘but for’, ‘made a difference’) analysis will not properly resolve situations involving overdetermined negative causation. Philosophers have discussed a ‘pre-emptive pre-emption’ example in which a ball is thrown towards a window, is caught by B before it reaches the window and breaks it, but would have been caught by C, who was standing behind B with her arm raised to catch the ball, if B had not caught it. Neither B’s nor C’s conduct is a necessary condition for (‘made a difference’ with respect to) the window’s not being broken, since the conduct of the other person guaranteed the ball’s being stopped before it reached and broke the window. Nevertheless, when people are asked who prevented the ball’s reaching and breaking the window, they almost always say B did. However, if C is replaced by a brick wall that would have stopped the ball if B had not caught it, many people’s first reaction is that B did not prevent the ball’s reaching the window, presumably because the fact that it would never have reached the window anyway is much more obvious although no more certain than in the original hypothetical. Yet, if they are then asked to identify who or what did prevent the ball’s reaching the window, people again usually choose B. The options ‘neither B nor the wall’ and ‘both B and the wall’ are quickly ruled out, and choosing the wall is implausible given that the ball never reached the wall. This leaves B as the only plausible answer.175 B is also the correct answer in both versions of the example. As in the similar examples of pre-emptive overdetermined positive causation, eg, the destruction by fire A of a house that would have been destroyed anyway by the subsequent arrival of fire B, refusing to treat B’s catching the ball as the cause of the ball’s not reaching the window or, worse, choosing C or the wall as a pre-emptive or duplicative cause, confuses strong necessity or lawful strong sufficiency (guaranteeing an outcome), respectively, with causal strong sufficiency (actually causing the outcome). Although C’s backstopping B or the presence of the wall guaranteed that the ball would stop its forward motion before it reached the window, and thus prevented B’s catching the ball from ‘making a difference’ in the ball’s not reaching the window, neither condition was guaranteed to be the actual cause of the ball’s stopping, and neither condition actually stopped the ball. Instead, B’s catching the ball stopped it. C or the wall would have stopped the ball, but only at a later and further point, which the ball did 174   eg Hall (n 14) 248–53; Lewis (n 45) 281–85; Schaffer (n 162) 197, 214. Schaffer assumes this even though, when criticising Phil Dowe, he makes use of the fact that the counterfactual-dependency analysis cannot properly resolve situations involving overdetermined negative causation. Schaffer (n 162) 212–13. Surprisingly, Schaffer ignores overdetermined positive as well as negative causation when he states that even positive causation requires counterfactual dependency. ibid 214. David Armstrong follows Dowe in using counterfactual dependency to provide an account of negative causes and effects as ‘parasitic’ on positive causation, without however accepting counterfactual dependency as an analysis of causation. DM Armstrong, ‘Going through the Open Door Again: Counterfactual versus Singularist Theories of Causation’ in Collins, Hall and Paul (n 9) 445, 447–49. Even Moore, despite all of his criticisms of counterfactual dependency analysis, uses it as a ‘non-causal’ basis for attributing moral and legal responsibility to what I and others call negative causes. Moore (n 1) 144, 302–05, 353, 451–53, 458–60. 175   M McDermott, ‘Redundant Causation’ (1995) 43 British Journal for the Philosophy of Science 523, 525; see Collins (n 69) 107; Lewis (n 53) 102–03; T Maudlin, ‘Causation, Counterfactuals, and the Third Factor’ in Collins, Hall and Paul (n 9) 419, 435–38.

The NESS Account: Response to Criticisms  317 not reach, in the ball’s movement towards the window. All of this is true even if the wall or C was in place, guaranteeing the ball’s failure to reach the window, prior to B’s arrival on the scene.176 This example illustrates the critical difference in the proper method of analysing positive causation and negative causation when there is causal overdetermination. In situations involving positive causation, we need to determine whether each possibly applicable positive causal process was fully instantiated. Only those that were fully instantiated in each step of the process were causes; the rest were pre-empted. Since the focus is on complete instantiation of every step in the process, not much attention needs to be paid to the sequencing of the steps. However, when analysing overdetermined negative causation, it is critically important to focus on the sequencing of the steps in the positive causal process that failed, in order to determine at which step it failed. As courts generally correctly hold, the failure at that step pre-empted any potential failure at subsequent steps, the occurrence of which is dependent on successful completion of all the prior steps.177 In the thrown ball example, the positive causal process that failed requires, among other necessary conditions, the ball’s continuing to move at each point along its line of travel in the direction of the window with sufficient speed to reach the window. The failure of the ball to continue along this line at the point where B caught the ball pre-empted its potential failure at the subsequent point along the line where C or the wall was located, which it never reached.178 One of the most discussed examples of overdetermined negative causation in the legal literature is the Saunders case, in which a vehicular collision occurred when a motorist driving a rental car did not attempt to brake until it was too late to avoid the collision, but the brakes were defective owing to a lack of proper inspection and maintenance by the rental car company and therefore would not have stopped the car in time even if the driver had applied them earlier.179 Assume that there were two defects in the braking system: the lack of a bolt connecting the brake pedal to the lever rod between it and the master cylinder and the lack of sufficient hydraulic fluid in the master cylinder. There is considerable disagreement whether the defects in the brakes (for which the rental company’s lack of proper inspection and maintenance was a negative cause), the driver’s failure to attempt to use the brakes, neither, or both was a cause of the accident. I have argued that the driver’s failure to attempt to use the brakes was a negative cause of the collision, which pre-empted the potential negative causal effect of the defects in the brakes. David Fischer rightly criticised my initial analysis for making conclusory statements without elaborating (or, admittedly, clearly seeing) the distinct methods of analysis required for negative causation (which involves the failure of a positive causal process) and positive

  See Wright (n 13) 298–300, 305; text to nn 84–96 above.   See Wright (n 13) 302–05; Wright, Once More (n 14) 1123–31; text at nn 152 and following n 171 above. 178   John Collins, confusing both guarantees and counterfactual dependency with causation, is ‘very reluctant’ to say that B’s catching the ball prevented the ball from breaking the window in the wall version of the hypothetical: ‘Given that the wall was there, the window was never in any danger of being broken. The presence of the wall really does seem to make [B’s] catch irrelevant.’ Collins (n 69) 108. To persuade others of his position, Collins offers a third version of the hypothetical in which the Earth’s gravitational force replaces the wall and the window is replaced by a point one hundred million miles from the Earth. However, this third version is critically different. Unlike the first two versions, Collins is assuming that the ball lacked sufficient speed when it was released to reach the stated end point. The lack of sufficient speed when it was released caused the failure of the positive causal process of the ball’s reaching that very distant point the instant the ball was released and thus pre-empted the potential negative causal effect on that process of B’s subsequent catching of the ball. 179   Saunders System Birmingham Co v Adams 117 So 72 (Alabama, 1928). 176 177

318  Richard W Wright causation (which involves the successful completion of a positive causal process).180 Fischer’s criticism forced me to focus on and elaborate the distinction between these two types of causation and the distinct methods of analysis (both based on the NESS account) that are required for proper resolution of each, which I have restated above.181 The analysis of the failure of the braking-stops-car causal process is the same as in the thrown ball example. The braking process is a complex combination of a large number of more discrete causal processes, each of which is dependent for its occurrence on the occurrence of prior steps in the causal sequence. Some of the necessary events, in order of occurrence, are: (1) the driver’s applying force to depress the brake pedal; (2) the depression of the brake pedal operating a lever to put pressure on the hydraulic brake fluid in the master cylinder; (3) the pressure in the brake fluid being transmitted through pipes and tubes to the brake cylinders; (4) the pressure in the brake cylinders pushing braking pads against the rotating brake drum or disc in the wheel assembly; and (5) the friction created by such contact slowing and stopping the rotation of the wheels. Each of these steps in the braking process, which occur in sequence, is itself a causal process; each has its own set of necessary antecedent conditions, mostly related to the structure and integrity of the mechanical, hydraulic and electrical components of the various parts of the braking system. The failure of any step in the braking process prevents that process from progressing any further in the sequence of dependent events. It thus pre-empts the potential negative causal effect of any non-instantiated conditions in subsequent steps, which would have caused the braking process to fail if it had progressed that far. In Saunders, when the very first step in the braking process, the driver’s depressing the brake pedal, does not occur, the braking process fails – actually never gets started – at that initial step in the causal sequence. The braking process does not get as far as step (2), although if it had got that far, it would then have failed owing to the missing bolt connecting the brake pedal to the lever, which in turn would have pre-empted the potential negative causal effect of the insufficient brake fluid in the master cylinder, which would have caused the braking process to fail at step (3) if it had progressed that far. But it never even got to step (1). Duplicative as well as pre-emptive negative causation can occur. For example, if one mechanic put insufficient hydraulic brake fluid into the master cylinder for it to work and another failed to seal it properly so that whatever fluid was in it would leak out, their respective omissions, which negate required positive conditions for the occurrence of step (3) of the braking process, are duplicative negative causes of the failure of the braking system to work, owing to the non-occurrence of step (3) in the braking process, when the brake pedal is subsequently depressed and there is no other defect.182 Fischer continues to assert that my detailed analysis of the situation in Saunders can be used to reach a conclusion opposite to the one that I reach: that the lack of proper inspection and maintenance of the brakes pre-empted the subsequent failure of the driver to attempt to use the brakes. He initially argues that this is true ‘because without good brakes an accident caused by a failure to stop became inevitable at the moment the car was given to the Driver’.183 This argument, like John Collins’ arguments regarding the thrown ball   Fischer (n 118) 1357–59, criticising Wright (n 5) 1801.   Wright (n 13) 302–07; Wright, Once More (n 14) 1128–31; text to nn 177–78 above. Allan Beever’s criticism of my analysis of the Saunders situation fails to note my elaboration of the distinct nature of negative causation and the distinct method of analysis that is required when analysing overdetermined negative causation. See Beever (n 1) 423–25. 182   Wright (n 13) 305–07. 183   Fischer (n 1) 310 (emphasis added). 180 181

The NESS Account: Response to Criticisms  319 hypothetical,184 once again confuses guaranteeing the occurrence of some consequence with actually causing that consequence.185 Fischer also argues that the existence of working brakes can be viewed as a step in the braking process that is prior to the driver’s pressing on the brake pedal, in which case, using the method of analysis that I claim is necessary for negative causation (failure of a positive causal process), the lack of working brakes due to the lack of proper inspection and maintenance caused the failure of the braking process and pre-empted the potential negative causal effect of the driver’s failure to timely use the brakes.186 Stapleton reiterates and expands on this argument. Without describing my analysis, she asserts: The only indication of how Wright arrives at [his] conclusion is his assertion that we ‘must’ look at the sequence of the causal process that did not take place. Two obvious objections to Wright’s analysis can be made. First, it is not at all clear why we ‘must’ look at this sequence. After all, in the actual world our two specified factors (the failure-to-repair and the failure-to-attempt-to-brake) did not occur in sequence; omissions simply do not ‘occur in sequence’; here both persisted at the time of the actual phenomenon of interest, namely the collision; so notions of one omission pre-empting another omission are, without more explanation, incoherent. Secondly, Wright merely stipulates that the relevant sequence would have started with ‘the driver’s depressing the brake pedal.’ Yet we could just as easily have stipulated that the first step in the braking-stops-car causal process was the brake repair: after all, ‘delivery’ of the motorist’s foot on a workable brake pedal is not needed in the braking-stops-car causal process until the working brakes are in place. Wright has no coherent rationale for choosing to stipulate the sequence in the way he does, and so, ironically, he stipulates the exact opposite sequence in a different example concerning the failure of a house-building project. Here he asserts that if there was a simultaneous failure to deliver concrete for the foundation and lumber for the framing of the house, the failure to deliver the concrete, which results in the failure of the house-building causal process at the foundation-­ building stage, preempts the potential negative causal effect of the failure to deliver the lumber, which is not needed until the subsequent framing stage, the occurrence of which depends on the prior occurrence of the foundation-building stage.187

Stapleton’s first objection puzzles me. As I have previously explained and have reiterated here, we must look at the physical sequencing of the braking process because we want to know how it failed, and determining that requires determining at what step in that process it failed. The failure at that step pre-empted the potential failure of the process at some later step that the process never reached. Stapleton erroneously states that the two relevant omissions in Saunders did not occur in sequence. She seems to have in mind temporal sequence. If so, she is wrong: the failure properly to inspect and maintain the brakes occurred prior to the driver’s failure to timely attempt to apply the brakes. However, as I noted in my dis­ cussion of the thrown ball hypothetical, the temporal order of occurrence of the negative conditions at issue is not the sequence we need to consider.188 Instead, we need to consider the physical sequence of the steps in the braking process. Fischer’s and Stapleton’s argument that we could and perhaps should view the braking process as starting with working brakes or their proper inspection and maintenance fails to pay attention to the structure of the relevant causal web. The defects in the braking system,   See n 178 above.   See text to nn 84–96 and 176 above and following n 188 below. 186   Fischer (n 1) 310–11. 187   Stapleton (n 1) 478, quoting Wright, Once More (n 14) 306 (emphasis by Stapleton). 188   See text to n 176 above. 184 185

320  Richard W Wright for which the lack of proper inspection and maintenance was a negative cause, are like the wall in the thrown ball example. Although the existence of the defects in the braking system occurred prior to the driver’s failure to attempt to use the brakes and guaranteed the failure of the braking process, they did not actually cause the failure of the braking process. The defects in the brakes do not come into play in the braking process unless and until that process progresses to the steps, steps (2) and (3), at which those defects would actually cause the failure of the braking process. The braking process never reached those steps. Instead, it failed (never got started) owing to the driver’s failure to (timely) use the brakes. The occurrence of steps (2) and (3) in the braking process is dependent on successful completion of each prior step in that process, including step (1), the driver’s depression of the brake pedal. On the other hand, the occurrence of the driver’s depression of the brake pedal (step (1) in the braking process) is not dependent on the prior occurrence of steps (2) and (3) nor on the rental company’s prior proper inspection and maintenance of the brakes. Stapleton’s third claim, that I inconsistently specify the sequencing of steps in different causal processes, apparently confuses my discussion of the dependency of occurrence of some steps in a causal process on the prior occurrence of other steps with claims regarding the strong necessity of various conditions. It is she, rather than I, who (unsuccessfully) attempts to use artificially constructed ‘duplicate necessity’189 rather than causal strong sufficiency to identify causes. Perhaps her misinterpretation of my statements would have been avoided if I had employed ‘used’ rather than ‘needed; when noting the step in the house construction process in which the presence of lumber becomes causally relevant. Fischer also asserts that my analyses of causation in these and other situations are faulty because they fail to conform with many persons’ intuitive ‘causal’ judgments, upon which he assumes I rely in all causally controversial situations.190 However, I have never stated that the analysis of causation should conform to persons’ intuitive ‘causal’ judgments in every case or even most cases. Indeed, I have often argued against taking persons’ intuitive ‘causal’ judgments as reliable judgments regarding true (natural) causation, since such judgments often include normative or purposive factors in addition to the causation issue.191 Instead, it is he and Stapleton who, while erroneously asserting that the NESS account privileges intuitive ‘causal’ judgments, continue to confuse intuitive judgments regarding liability with proper determinations of causation192 and unfortunately have led others into similar confusion.193 I have merely stated that, when attempting to determine the meaning of causation, including the proper senses of necessity and sufficiency, one should take into account our basic carefully considered intuitions regarding causation rather than policy   See n 125 above.   Fischer (n 1) 299, 312–16. 191   See my articles cited in n 126 above. 192   See Wright, Once More (n 14) 1120–23 and fn 172; text to nn 92–94, 122–23, 127–35 above. 193   eg DF Partlett, ‘Foreword: David Fischer, the Fox’ (2008) 73 Missouri Law Review 281, 282–83, 286–87. Partlett and others have been led astray not only by Fischer’s confusion of causation with liability judgments, but also by his careless and misleading use of causal terminology, some of which unfortunately has been employed in the Restatement Third. See Wright, Legal Responsibility (n 9) 1449 and fn 81, 1464 and fn 146; Wright, Once More (n 14) 1126. Fischer’s misuse of causal terminology is particularly striking with his term, ‘dependently sufficient cause’, which he defines as an omission that ‘is insufficient to be a ‘but for’ cause of the result because the other omitted act was necessary for sufficiency’. Fischer (n 1) 279. This is a convoluted and highly misleading way of referring to a lack of strong necessity, which he misdescribes as a lack of ‘independent sufficiency’ and misapplies to situations like Saunders, in which each omission was not a ‘but for’ cause of but was ‘independently sufficient’ (in the lawful guarantee sense that Fischer assumes) for the failure of the braking process and the consequent collision. See ibid 278–79. Stapleton uses without explanation a similarly unhelpful and misleading term, ‘dependent double omissions’, to refer to the omissions at issue in situations like Saunders. Stapleton (n 1) 477–78. 189 190

The NESS Account: Response to Criticisms  321 considerations,194 while also insisting that a theoretical account of causation is necessary to back up these intuitions and to prevent them from leading us astray: Despite the lack of an explicit comprehensive definition of causation, people from time immemorial have shown remarkable agreement in their causal judgments, at least once they are clearly focused on the causal issue rather than on some noncausal inquiry regarding the (most significant for some purpose) cause. In particular, judges and juries, when not confined by incorrect tests or formulas, consistently have demonstrated an ability to make intuitively plausible factual causal determinations. Some scholars rely heavily on this shared yet undefined concept of causation in their writings on causation and responsibility by grounding their arguments on intuitive responses to hypothetical situations. Yet intuitions that are not conjoined with theory in a search for underlying principles are often inadequate for the hard cases and sometimes may mislead even in the easy cases. In these situations in particular we would benefit greatly from elaboration of the concept that, unarticulated and imperfectly understood, underlies the intuitive judgments.195

I continue to believe that the NESS account provides the needed comprehensively applicable account of the concept of causation.

J.  The Proliferation of Causes As some critics have pointed out, the NESS account greatly increases the proliferation of causes.196 The NESS account, like the sine qua non account, recognises acts and omissions as causes through negative as well as positive causation, which dramatically increases the number of causes. Moreover, the NESS account goes beyond the sine qua non account by recognising as causes not only strongly necessary conditions, but also independently strongly sufficient conditions and conditions that were neither strongly necessary nor independently strongly sufficient. Even the sine qua non account is sometimes rejected by judges and others for treating as causes conditions that usually would not be identified as causes in ordinary speech but rather, if thought about at all, would be treated as ‘background conditions’ or as too remote in the causal web to even be acknowledged as background conditions. However, as I noted above,197 this merely reflects a confusion of the causal issue with contextual salience, which in the law is appropriately handled by focusing the causal analysis solely on legally relevant conditions and by applying normative attributable responsibility limitations on the scope or extent of liability for the consequences of those conditions. Depending on the circumstances and purpose of the particular causal enquiry, something not ordinarily treated in common speech as a cause would easily be described as such – eg, oxygen as a necessary condition for a fire in circumstances where oxygen was not expected to be present.198 As most critics of the NESS account acknowledge, the proliferation of causes is not a significant problem in analyses of legal responsibility.199 In law, as well as ordinary life, we   Wright (n 3) 1020.   ibid 1018–19. 196   Collins, Hall and Paul, Introduction (n 9) 25; Fischer (n 1) 290–91; Fumerton and Kress (n 1) 98–99; Moore (n 1) 477–79, 486–95. 197   See text to nn 121–35 above. 198   See Hart and Honoré (n 5) 34–35; Wright (n 3) 1012–14. 199   Fischer (n 1) 289–90, 292; Fumerton and Kress (n 1) 87, 99; Stapleton (n 1) 440, 444, 448–51; Thomson (n 1) 150 fn 9; see Honoré (n 1) 367–69. 194 195

322  Richard W Wright are not interested in determining all of the practically innumerable immediate and remote causes of every event or state of affairs. In law, we are only interested in the possible causes of legally recognised injuries, and we are interested in only a very few of those possible causes: the aspect of the defendant’s conduct that made the defendant subject to potential liability, the negligent aspect (if any) of the plaintiff ’s conduct (which may result in application of the defence of contributory negligence), and highly unexpected intervening necessary conditions or independently lawfully sufficient non-responsible conditions (which generally will result in denials of legal responsibility – misleadingly described as ‘proximate’ or ‘legal’ causation – despite tortious causation).200 Legal responsibility may also be denied for legally relevant conditions that made only a ‘trivial’ contribution to a legally recognised injury201 or that caused a legally recognised injury that was not the result of the actual or anticipated realisation of a risk that made the relevant party’s conduct wrongful.202 Moreover, failures to act to prevent a legally recognised injury generally are not considered to be wrongful (and thus legally relevant), or are considered wrongful only in very limited circumstances, if the person who failed to act had no responsibility for any positive condition in the positive causal process that produced the injury (eg, the no- or limited-duty rules for ‘nonfeasance’).

IV. Conclusion

While I do not claim to have addressed every possible objection to the NESS account, I have attempted to address all those of which I am aware. My consideration of the various objections has reinforced my belief that the NESS account is not merely a very useful tool for identifying singular instances of causation, which almost all of the critics of the NESS account admit, but that it does so precisely because it captures the essence of causation and gives it a comprehensive specification and meaning. I don’t think the former is possible in the absence of the latter.

  Wright (n 5) 1741–44, 1759–74, 1798–1801; Wright, Legal Responsibility (n 9) 1434–78.   See text to n 123 above.   Wright, Legal Responsibility (n 9) 1479–1528.

200 201 202

15 NESS for Beginners CHRIS MILLER

I. Introduction

The NESS (necessary element in a sufficient set) test of causation has attracted the atten­ tion of philosophically minded lawyers ever since its predecessor was first articulated by Hart and Honoré in 1959.1 Today it is probably the version2 developed by Richard Wright which is best known. This chapter examines two criticisms of Wright’s approach: Fischer’s3 suggestion that it is unable to deal adequately with multiple omissions; and the claim by Fumerton and Kress4 that Wright’s treatment of necessity and sufficiency lacks philosoph­ ical rigour. This chapter seeks to show that neither criticism is fatal. The NESS test does not claim to offer a complete account of causation – a task which continues to defeat philoso­ phers. The principal reason why NESS survives these criticisms is that its aims are modest: it offers a coherent approach to attributing cause in situations where ‘but for’ (that is, lawyers’ traditional test of causation) leads only to paradox. In short, NESS can justify an assertion that ‘X caused Y’ in circumstances where ‘overdetermination’ obtains, viz a differ­ ent event, Z, appears to have an equally plausible claim to having been Y’s cause. One important role played by NESS is to reveal a ‘logical skeleton’ which lies beneath the empirical flesh which comprises the ‘facts’ of the case heard in court. The role played by NESS can sometimes be illustrated by considering the components of a simple electrical circuit to show how the necessary elements combine to form a set which is sufficient to produce, for example, light. But the principal value of this heuristic device is that an electri­ cal circuit is clearly devoid of any moral dimension – either the current flows or it does not. Central to Wright’s approach (and contrary to that of many legal scholars in this area) is the assertion that it is possible to construct a causal account of a set of events (leading to a harmful outcome) which is independent of those considerations by which an agent of these events might be deemed legally liable. It is the knot of factual history, not the normative considerations of liability and responsibility, that NESS seeks to untie.   HLA Hart and T Honoré, Causation in the Law (Oxford, Oxford University Press, 1959).   Richard W Wright, ‘Causation in Tort Law’ (1985) 73 California Law Review 1735 [Causation] and later his ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (1988) 73 Iowa Law Review 1001 [Bramble1] and then, ‘Once more into the bramble bush: duty, causal contribution, and the extent of legal responsibility’ (2001) 54 Vanderbilt Law Review 1071 [Bramble2]. 3   David A Fischer, ‘Causation in Fact in Omission Cases’ (1992) Utah Law Review 1335 [Fischer1] and David A Fischer, ‘Insufficient Causes’ (2005-6) 94 Kentucky Law Journal 277 [Fischer2]. 4   Richard Fumerton and Ken Kress, ‘Causation and the Law: Preemption, Lawful Sufficiency and Causal Sufficiency’ (2001) 64 Law & Contemporary Problems 83. 1 2

324  Chris Miller

II.  Establishing the Utility of NESS

A.  Multiple Tortfeasors I believe it is possible to establish the utility of NESS in actual legal cases by first explaining how it can uncover a ‘logical skeleton’ using very simple electrical circuitry such as that depicted in Figure 1.1.

Figure 1.1

B S L

Figure 1.1: A circuit represent­ ing a single sufficient set

Even if some may doubt its power as an explicative device, it must be accepted that such a circuit is devoid of any moral or normative character. The circuits devised here simply illustrate different applications of the laws of current electricity, which can be found in any elementary physics textbook. The amoral character of our circuitry together with a clear recognition as to when additional empirical information is required can, by analogy, illu­ minate the role which NESS can play in the courtroom. We can readily grasp that each of the three components must function (switch S must close, battery B must be charged and light bulb L must be intact) if the circuit is to produce light. We understand that there is a sense in which each is a necessary element in a set {S ∩ B ∩ L} which is sufficient to give light: nothing more is needed. In turn, experience teaches us that the failure of the switch to close {¬S} or the absence or non-functioning of either of the other two elements is sufficient to bring about a state of no light {¬S} ∪ {¬B} ∪ {¬ L}. (Note: ∩ denotes the logical AND operator; ∪ the OR operator; and ¬ denotes negation or, in the context of this chapter, ‘the failure of’ or ‘the absence of’. Note also that we need not, at this stage, to introduce other internal components such as the wiring even though it is another NESS. And we shall defer discussion of what may be labelled ‘external’ requirements, for example, that the laboratory housing an actual circuit is not struck by lightning.) It requires very little elaboration of our battery B to produce an analogue of the ‘two fires’ example – two independent sources of fire simultaneously destroy a house; either could have done so in the absence of the other – which many authors have used to describe causal ‘overdetermination’ by duplication. If the battery in Figure 1.1 is replaced by the two-cell battery in Figure 1.2, then provided both B1 and B2 are adequately charged and connected in parallel, neither is a ‘but for’ cause of illumination but each is a now a necessary element in its own set, viz

InNESS for Beginners  325

1.2

gure 1.3

B1

B2 Figure 1.2: Two sufficient sets

{S ∩ B1 ∩ L} ∪ {S ∩ B2 ∩ L} each of which is sufficient to produce light when the switch S is closed. A further elaboration of the circuit can illustrate a variant of the ‘negligent sportsmen’ paradox: a group of sportsmen negligently and simultaneously fire their guns and one bul­ let injures one of their number.

B1 B2

S Bn

RE Figure 1.3: Multiple sufficient sets

In Figure 1.3, B now consists of two or more cells, connected in parallel; and (in anticipa­ tion of a later discussion on indeterminism) we replace the lamp with a device which kills a laboratory rat. On closing the switch, one short pulse of charge from any cell is sufficient to trigger the device (RE) and the rat is killed. If there is no means of identifying the cell which was the actual source of the charge, then this is equivalent to the absence of ballistics (or any other form of) evidence which incriminates the single ‘guilty’ hunter.

326  Chris Miller By further elaboration and by introducing the time dimension, we could create an ana­ logue of the problem multiple asbestos employers. The battery again consists of n cells connected in parallel and we retain the rat exterminator (RE). But now let the ith cell have its starting point Ti at which it begins to function and becomes available to deliver the voltage needed to activate the rat exterminator. Since the RE operates randomly, there appears, to the observer, to be some hidden variable which determines when and if the RE is activated. T1 T2 T3 Tn E _¦_____¦______¦___________________________________¦_________¦ Only at time E can we examine the rat; if it is dead, we know that at least one cell has discharged at least once (we believe that one pulse of charge alone may be capable of trig­ gering the device). We need this randomness and lack of information to replicate the uncertainty in current understanding of the aetiology of mesothelioma (a form of cancer associated with exposure to asbestos fibres). Each cell now represents an employer (whose control over asbestos is negligent); the charge represents the absorbed asbestos which is implicated, in some as yet unknown way, in the process which leads eventually to the tumour or, in our hypothetical laboratory, to the death of the rat. Note that a cell is not ‘turned off ’ once another becomes operative: it could be that asbestos absorbed during an earlier period of employment plays its part in the tumour-producing process at some later point when the victim is employed elsewhere. We need multiple discharges because the ‘single-fibre’ theory is no more than a theory: two or more fibres may be required or asbes­ tos may be involved at more than one point in the long latency period before the tumour can be diagnosed. With no means of identifying the ‘guilty’ cell, none can pass the ‘but for’ test. Mesothelioma (or any other form of cancer associated with asbestos) does not carry a signature which shows the time(s) at which asbestos was implicated; it may be involved on more than one occasion and, regardless, these notional times are unlikely to be related to the point at which the fibres were absorbed. Before introducing the concepts of risk and indeterminacy, it seems apparent that each cell, along with the lethal device, is a necessary element in one of n sets sufficient to cause the death of the rat. By analogy, asbestos, absorbed in some period of exposure, appears to be a NE within a SS for the eventuation of mesothelioma. If one employer were able to show that his asbestos control measures were not negligent, or if there were a period (of asbestos exposure) when the victim was selfemployed,5 this may affect the assessment of damages in any court case. But this does not affect factual causation: the ‘guilty’ fibre(s) could well have been absorbed in any period of exposure which happens to escape civil liability. In our circuit, no functioning cell can carry a comparable immunity. There is no way of identifying the cell whose discharge did indeed trigger the device. (A cell could be excluded as causal if an empirical test showed that it had lost no charge at all during the experiment – the analogous grounds for immunity in asbes­ tos litigation would be if an employer could demonstrate that his control over asbestos exposure was not negligent.)

5   The point at issue in Barker v Corus (UK) plc (formerly Saint Gobain Pipelines plc); Murray v British Shipbuilders (Hydrodynamics) Ltd; Patterson v Smiths Dock Ltd and others [2006] UKHL 20, [2006] 2 AC 572.

InNESS for Beginners  327

B.  Overdetermination: Cumulative Sources of Harm Nothing hitherto has posed any challenge to Wright’s articulation of NESS. Rather it has served to underline the point he repeatedly stresses, namely, that reality may involve more than one causal pathway, each of which is independently capable of giving rise to harm. However, Fischer6 criticises what he sees as Wright’s overconfidence in what NESS can bring to an understanding of the causal status of elements which are neither necessary nor sufficient. Among his examples is one based on an actual civil case:7 multiple and independent discharges (26 according to Wright)8 combine to pollute a watercourse (and subsequently killed the cattle which drank the polluted water). But how do we attribute legal liability if not all 26 discharges are needed to cause this level of damage? Our electrical circuitry continues to be relevant; but now our battery consists of 26 (one volt) cells con­ nected in series, 25 of which are of type X and one of type Y. We need also to imagine a short interval, after the switch is closed, in which, at random,9 each cell begins to exert its one volt. At some point in this interval, the total voltage will have generated a current which is enough to burn out the filament in the bulb and, with this ‘fuse’ having blown, the circuit is broken. If this occurs with 15 volts, it follows that the type Y cell, along with 10 of the type X, was neither necessary nor sufficient to cause the ‘fuse’ to blow. Common sense would then appear to persuade Fischer that the type Y battery is not causal. Wright’s disa­ greement would stem once again from his insistence that the existence of a multiplicity of sufficient sets should not confuse our understanding of causation. If there is a total of n cells and r are required to burn out the filament, then it is straightforward to calculate10 the number (T) of different combinations of cells which can achieve this outcome. Each of the T combinations represents a sufficient set in which each of the r cells is a necessary elements; with n = 26 and r = 15, T is a very large number (7,726,160). However, we can be sure that the type Y cell appears in 1 in 26 of these combinations and it is therefore, accord­ ing to Wright’s conception of NESS, causal. There is no way of knowing which one, among the very many conceivable NESS sets, was actually instantiated in our circuit. But now imagine that, as a cell begins to exert its voltage and lose its charge, it changes colour from blue to red (and assume further that no cell changes colour after the circuit is broken by the blown fuse). Now the 15 ‘guilty’ cells will be clearly identified by their colour and it will be equally clear whether the type Y cell is among them. Where ‘particularistic’ evidence (equivalent to a cell’s colour change) does exist, cau­ sation is usually straightforward; but there has been considerable debate among legal schol­ ars on the way courts should respond when such evidence is not available. Some argue that in view of the evidential uncertainty resort should be had to probability. Wright’s antipathy   Fischer2 (n 3) 289.   Warren v Parkhurst 92 NYS 725 (NY Sup Ct 1904), aff ’d, 93 NYS 1009 (AD 1905), aff ’d, 78 NE 579 (NY 1906). 8   Bramble2 (n 2) 1107. 9   This randomness plays no more special a role than the random mixing of the different sources of pollution in the stream. 10   T = n! ÷ (r! × (n-r)!) where n! = n × (n-1) × (n-2) . . . . . . . 3 × 2 × 1. In this example, r represents a threshold beyond which pollution-related harm can occur. By positing an equiva­ lent threshold in the harm caused by the cumulative exposure to dust from two – one tortious and the other not – otherwise identical sources, NESS could provide a similar justification for the ruling in Bonnington Castings Ltd v Wardlaw [1956] AC 613 and cognate cases. This justification would apply regardless of the actual contributions of the two sources, and it would avoid reliance upon the notion of ‘material contribution’. 6 7

328  Chris Miller to verdicts based solely on ‘naked statistics’ is made clear in all his writing on causation; less clear is whether, in the polluted stream example, he would require each polluter to pay one twenty-sixth damages in these circumstances or whether he would advocate the imposition of joint and several liability. Before venturing deeper into the bramble bush of probabilistic causation, it is necessary to consider another of Fischer’s criticisms of Wright’s use of NESS, viz pre-emption.

C.  Overdetermination: Pre-emption The example of overdetermination over which Fischer and Wright most clearly disagree is one involving multiple omissions: a road accident ensues after the car rental firm fails to repair the brakes on one of its cars and the person who subsequently rents this car fails to apply these (defective) brakes. Both the rental firm and the driver are negligent but neither can pass the ‘but for’ test. But it seems absurd to hold that neither is a causal agent (as absurd as claiming that neither fire destroyed the house). Wright argues that it was the driver’s failure to apply the brakes which was the cause of the accident since that failure preempted, or rendered irrelevant, the actual state (whether or not defective) of the brakes. Fischer identifies an apparent symmetry in Wright’s earlier text11 and argues that it applies equally well if the failure to repair is taken as causal and pre-emptive of any subsequent application or non-application of the brakes. When dismissing this symmetry and identifying the driver’s non-application of the brake as pre-emptive and causal, Wright relies on a notion of ‘causal sequence or priority’ which, he claims, can be distinguished from ‘mere temporal sequence’. He continues: [t]he causal sequence for the operation of a safeguard is initiated when a person attempts to use the safeguard and then subsequently proceeds, as a result of such attempt, with the activation of the safeguard if the safeguard is present and in proper condition.12

And by way of further explanation, he claims that ‘the activation of the safeguard depends [my emphasis] on someone’s first attempting to use it, so that if no such attempt is made, ‘the (temporally) first omission (the failure to provide a working safeguard) is not causal because it never came into play’.’13 This notion of ‘dependence’ is clearly central and implies an asymmetry. The following scenario attempts to explore further this notion of depend­ ence: four (eighteenth century) conspirators wish to be causally involved in an assassina­ tion: A pours the powder into a musket; B inserts the ball; C ‘cocks’ the musket and hands it to D; D takes aim and fires at the victim. Wright might argue that ‘activation’ of the mus­ ket ‘depends’ on someone (successfully) firing it and, if it is not fired, then any failure to load the musket is not causal because it does not come into play. But it is difficult to avoid circularity when analysing this notion of dependency: if A, B, C, D now refer to the actions of each of the four agents, each action is a NESS (where SS is a set sufficient to complete the   Causation (n 2) 1801.   Bramble2 (n 2) 1128. It is not clear whether Wright is using ‘causal priority’ in the same sense as JL Mackie, who claimed that ‘[t]he relation between cause and effect is commonly taken to be asymmetrical: a cause is related to its effect in some way in which the effect is not related to its cause. We have been using the term causal priority as a name for whatever it is that marks this distinctive direction’ (The cement of the universe: a study of causation (Clarendon, 1974) 160). Wright is also concerned with an asymmetrical relation, but here it is between two puta­ tive causes: the pre-emptor and the pre-empted. 13  ibid. 11 12

InNESS for Beginners  329 assassination) but D must also carry a dependency label (D*) to distinguish it from the others. If instead we concentrate on the sets which are sufficient to bring about the failure of the assassination attempt, then if ¬ (the negation operator) denotes ‘the failure of ’, four immediately present themselves: {¬A} ∪ {¬B} ∪ {¬ C} ∪ {¬ D} Why should a wayward aim {¬D} retain a distinguishing (dependency) label in what is now the (counterfactual) situation of a failed assassination attempt? Far simpler to argue that the only distinguishing characteristic now comes from temporal order: the first to occur is not only sufficient, it is pre-emptive in that it makes any subsequent failure(s) redundant. In the same way, the failure of any one of the three components in the simple circuit (Figure 1.1) is sufficient to prevent illumination. To generalise: in any sequence of multiple failures, the first to be instantiated must be designated as ‘pre-emptive’: its occur­ rence (together with the antecedent conditions) establishes the sufficiency of the set and consigns any subsequent failure to redundancy. Thus, once the battery has failed, the state of the other two components is immaterial to non-achievement of light. (In the ‘Theft, Nonuse, or Misuse of Defective or Missing Safety Devices’ section of his 1985 paper, Wright goes on to argue that the driver’s omission is not causal if his inaction was ‘because he knew [the brakes were] missing or defective’.14 But if the driver concludes that applying the brakes is pointless because they are defective, is that not tantamount to his recognising that he has been pre-empted? Or, more precisely, his capacity to pre-empt (by non-application) has itself been pre-empted by the brakes being defective. And if we share his knowledge of the actual state of the brakes, why should we (or a judicial decisionmaker) come to a different conclusion from the driver? The earliest pre-emptor will make later ones redundant: temporal priority and asymmetry seem to be implicit in any notion of pre-emption.) The usefulness of the NESS test is not diminished by the fact that a credible causal account may need a separate consideration of the time dimension. It was his mistaken belief in the defective brake/non-application causal symmetry which persuaded Fischer to hold that: ‘[t]he temporal sequence of events does not affect the results reached under the NESS test, or the but-for test’.15 In order for the simple circuit to produce light, the three components must simply function, there is no time dimension and no NE in the SS carries any distinguishing label. But that is not true if the outcome requires a particular temporal order of two or more of the NEs in the SS. Successful assassination by musket shot requires four events to adhere to their stated order; it is not enough that these four are NEs con­ joined in an SS. If we accept Wright’s notion of causality sequence or priority, then some, if not all, of the elements {Ei} in a sufficient set must still bear some distinguishing label i=1, 2, 3 but now this suffix specifies the order of the causal sequence or priority. But these labels are not empirically observable, unlike those which distinguish temporal order, and if the latter can offer an adequate explanation of pre-emption, then Occam’s razor would suggest that further complication should be avoided. In the following discussion – of another celebrated16 paradox of overdetermination – distinguishing between two rival causal candidates, the temporal order is again implicit.   Causation (n 2) 1801.   Fischer1 (n 3) 1360. 16   For an account of Jane Stapleton’s disagreement with Wright, see her ‘Perspectives on Causation’ in Jeremy Horder (ed), Oxford Essays in Jurisprudence (Oxford University Press, 2000) 84 fn 59. 14 15

330  Chris Miller

D.  Overdetermination: Interruption Consider the following scenario: One enemy of a desert traveller poisons the water in the traveller’s water keg, a second enemy (not knowing that the water has been poisoned) empties the poisoned water out of the keg, and the victim subsequently dies of dehydration.17

‘But for’ can survive only if we add in some additional information, such as an autopsy which confirms the cause of death as dehydration (and not poisoning or a heart attack, etc). But NESS does not need more information than that entailed in the above: poisoning was not the cause of death because it was never ingested. The poisoning, Wright argues, had nothing to do with the death because it was ‘pre-empted’ by the true cause, viz the denial of the water in the keg, which was sufficient to cause death. For ‘the poisoning set’ to achieve sufficiency would have required the traveller to drink from the poisoned keg which, we know, did not happen. Had the traveller been killed by lightning before drinking the poison, the lightning would be no less pre-emptive, that is, making redundant any other process which could result in his death. (And, it will be argued below, the absence of a dev­ astating lightning strike is a prime example of one of an indefinite number of absences which constitute necessary elements in any sufficient set which attempts to describe any actual occurrence.) Lightning strikes aside, how does death in the desert differ from the defective brakes example? There an accident was, in the absence of some miraculous brake restoration, inevitable. In the desert, it is the interruption (and negation) of one causal chain by another which leads to paradox when applying the ‘but for’ test. Both examples involve overdeter­ mination, but death in the desert is perhaps better described as a case of interruption or supervention rather than pre-emption.18 . . . E1 . . . E2 . . . → H Here E1 and E2 represent two events in a sequence (time flows from left to right) of events of unspecified length culminating in outcome H. By pre-emption, we understand the situ­ ation in which E1 happens to be such that it ‘cancels out’ any subsequent event(s) that could have independently given rise to H. ‘But for’ fails because, in the (counterfactual) absence of E1, E2 remains to establish a set sufficient to cause H. With supervenience or inter­ruption, it is the later event E2 which has the capacity to ‘cancel out’ what has already occurred and to set in train a different path to outcome H.19 Now ‘but for’ fails because, in the (counter­ factual) absence of E2, E1 remains to establish a set sufficient to cause H. Thus duplication – a house burnt down by two simultaneous but independently started fires of equal des­   Bramble2 (n 2) 1115.   The terms ‘early pre-emption’ and ‘late pre-emption’ are sometimes used. The distinction will be familiar to any whist player. If the ace of trumps is the first card to be played then the trick is clearly won and pre-empts (early) any combination of the three remaining cards. If, in another trick, the ace of spades appears to be the win­ ning card until another player (legitimately) trumps it, then this is better understood as ending (late pre-empting) the ace’s supremacy. In both cases, the winning card wins the trick regardless of whether it was played first, second, third or fourth – the order is irrelevant in retrospect. In practice, each player (except the first) considers the pre­ ceding card(s) before choosing hers. 19   The desert traveller scenario illustrates why the consideration of cause in fact must precede the attribution of legal liability. If Enemy 2 were declared insane or had some other immunity from a murder charge and was there­ fore excluded from consideration as a causal agent, this could lead to an erroneous outcome – Enemy1 charged with murder rather than attempted murder – might result. 17 18

InNESS for Beginners  331 tructive power – should therefore be seen as a transitional state, for now we have two inde­ pendent sufficient sets rather than one exhibiting either pre-emption or supervenience. Is it possible to find an analogue of supervention in our electrical circuit without calling upon external effects such as lightning strike? A blown filament in the bulb would immedi­ ately bring about a situation of no light but it also interrupts, and nullifies, the process by which that situation would have arisen with the eventual discharge of the battery. Further contemplation of the circuit in Figure 1.1 should serve to remind us that the simultaneous failure of two or more components is a rare occurrence; it is the more general, non-­ simultaneous occurrence of two acts or omissions which presents the greatest difficulty when unravelling causation. Once we are persuaded to reject ‘but for’ as the appropriate test of causation, we must then be careful to avoid being unduly influenced by counter­ factuals. And NESS can help in that regard – provided we recognise that we may also need to consider the temporal order of the necessary elements which constitute sufficient sets.

III.  Proliferation of NESSs

A.  The Critique of Fumarten and Kress Fumarten and Kress offer a much deeper philosophical critique of Wright’s exposition of NESS. They distinguish five different senses of necessity and sufficiency – formal, analytic, synthetic, lawful, causal – of which the last two are the most important for their critique of NESS. Lawful necessity (or sufficiency) requires ‘empirical investigation’ in order to estab­ lish a law of nature from which necessity or sufficiency can be inferred. They continue: Assuming that it is a law of nature that fire takes place only in the presence of oxygen, then it will be true that both the presence of oxygen is a lawfully necessary condition for the occurrence of fire and that the occurrence of fire is lawfully sufficient for the presence of oxygen.20

Fumarten and Kress then suggest that the ‘most plausible to the concept of necessity on which Wright’s definition relies is logical or analytic necessity; [and] the sense of suffi­ ciency on which his definition relies is lawful sufficiency’.21 In the following section, we shall attempt to demonstrate that Wright’s conception of NESS assumes a notion of lawful necessity and sufficiency. But first it is necessary to consider what Fumarten and Kress see as a separate problem, namely, the proliferation of NESSs. If N represents a complex state of affairs which is ‘lawfully’ sufficient for, say, a fire to occur. The fact B that I am wearing a blue shirt is not a NESS of the fire by virtue of its being a constituent of N. Next consider {¬B ∪ N}; this combination is not lawfully suffi­ cient for the fire: it can obtain without requiring the occurrence of the fire. But if B is true, by adding this fact to {¬B ∪ N}, it follows that N has occurred and since B ∩ {¬B ∪ N} is lawfully sufficient for the fire to occur, the truthful statement about my shirt has become a NESS of the fire (and so could any other similarly truthful statement).22 But as JL Mackie   Fumarten and Kress (n 4) 93.   ibid 94. 22   Alternatively consider {B ∪ ¬B} ∩ N which, if B is true, reduces to {B ∩ N}. It could be that this counterintuitive result is part of what Jaegwon Kim has suggested is a deep-seated ontological problem which arises when truth-functional operators (conjunction, disjunction, negation) are applied to combinations of actual, historical 20 21

332  Chris Miller understood, to construct an ‘absolutely sufficient’ account of any actual historical event, we need to include ‘the whole prior state of the universe’.23 Thus it should come as no surprise that any true statement about the world can appear as a NESS in such an account. In addi­ tion, a truly sufficient account would also have to include the indefinite number of possible events which, by their not happening, have not changed the course of history. To cite Fumarten and Kress’s example: if we apply the NESS test to an instance of harm resulting from a doctor administering a wrong drug, we arguably need to include, as a NESS, the fact a nurse did not spot (and rectify) the error, that another doctor did not spot the error, that Bill Gates did not . . . and so on ad inf. If the first enemy is to succeed in his aim of poison­ ing the desert traveller, we need to exclude, not only the intervention of enemy No 2, but also the possibility of any number of Good Samaritans adding an antidote to the fateful keg. Whenever we postulate the truth of a proposition X then we are also implicitly negating all those conceivable propositions that could establish the falsehood of X. By way of a mun­ dane example: if I report to my wife that ‘I saw Anne on the train this evening’, there is little purpose in adding the riders: that I did not lose my sense of sight this morning; that it was not the Anne who died last year; or that the train was not destroyed by lightning this after­ noon. Clearly the great majority of our daily utterances cannot require us to refute every statement that could contradict what we are saying at that instant. But equally, we have to have an ability to recognise the ‘pre-emptive’ quality of certain events (such as a lightning strike) which we characterise as thwarting what would otherwise seem to be realisable futures. When negotiating the exigencies of everyday life, philosophical introspection is an indul­ gence. Rather than ponder with incredulity the infinite web of contingency that has made the present what it is rather than what it might have been, we assume the existence of boundaries, within space and time, within which we can effectively confine our causal dis­ course.24 In the courtroom, the task is to devise boundaries within which the search for alternative causes (whether by pre-emption or supervenience) of harm can be confined. And, as Wright constantly points out, NESS is needed to illuminate causation only when, through overdetermination, ‘but for’ breaks down. Therefore Fumarten and Kress’s specific claim that Wright’s approach to pre-emption involves ‘vicious conceptual circularity’ demands careful scrutiny.

B.  Lawful and Causal Sufficiency Fumarten and Kress’s argument is developed using the example of a narrow estuary which is blocked when two bridges collapse (see Figure 2, below). According to Wright, the col­ lapse of B2 is, for a ship in position S, irrelevant to its failure to reach the port P: the col­ lapse of B2 has been pre-empted by that of B1. Not so, say Fumarten and Kress: the collapse of B2 (no less than that of B1) is a NESS of the ship’s non-arrival at P. They claim that ‘all we need to establish to reach the conclusion that it was [a NESS] is that it was part of a events; see his ‘Causes and Events’ reproduced in Ernest Sosa and Michael Tooley (eds), Causation (Oxford, Oxford University Press, 1993) 60, 68. 23   JL Mackie ‘Causes and Conditions’ reproduced in Ernest Sosa and Michael Tooley (eds), Causation (Oxford, Oxford University Press, 1993) 33, 41. 24   This point would of course be rejected by proponents of chaos theory.

InNESS for Beginners  333 complex set of conditions that was lawfully sufficient for the ship making it to the relevant port’.25 By way of further explanation, they state: Wright seems to want to distinguish the collapse of the second bridge from the collapse of the first by arguing that the second bridge is not a part of some actual conditions that were lawfully sufficient for the ship failing to reach its destination. . . . It is an actual fact that the second bridge collapsed, and it is an actual fact that the ship in question was below the bridge. These facts, together with other largely negative facts concerning the absence of alternative routes, are lawfully sufficient for the ship not getting to its port.26

Fumarten and Kress appear to be arguing that any fact which formed part of the ‘complex set of conditions’ that were lawfully sufficient for the ship’s non-arrival at P is a NESS. Their point would appear to be another manifestation of the proliferation of NESSs problem discussed above. So, the fact that the ship has a captain is a NESS, and so too is the fact that the captain’s great-grandparents met and had a least one child is another, and so is any other fact concerning the captain’s family tree. The collapse of B2 is a NESS but how can we disregard it as causal, along with the captain’s genealogy as our intuition urges, without falling back on some notion of causal necessity or sufficiency which we ascribe to B1 but not to B2? Firstly, I think our intuition persuades us that because the collapse of B1 forms, what we understand as, an impenetrable barrier for shipping then that fact, together with the prescribed geography of the estuary (viz no other routes to P), is enough to create the spatial boundary within which causation can safely be confined. In other words, the col­ lapse of B1 at the western end of the narrow channel creates, like our simple circuitry, its own causal boundary. But can we support our intuition in this and similar examples with some notion of lawful sufficiency which can avoid circularity? N

S

P

B1 H

B2

Figure 2: Pre-emption: The two bridges problem

The ‘set of conditions’ which obtain in the (Figure 1.1) circuit can hardly be described as complex; more to the point, the circuit forms its own boundary. Moreover, there is no   Fumarten and Kress (n 4) 100.   ibid 101.

25 26

334  Chris Miller shortage of ‘laws’ (thanks to Ohm, Volta, Ampère, etc) to describe current electricity. If we were to use Figure 1.1 to construct an actual circuit in a laboratory, these laws are all that we need to account for the success or failure to achieve light on any given occasion; we do not need to extend discussion to the fact, undeniable though it might be, that an Englishman bought the various components on a Saturday in June. Even if these laws do not currently include one which takes the form: given one component failure, a subsequent failure is not a necessary element of the set that is sufficient to extinguish light, it is not difficult to see how such a law could be established. Imagine an indefinitely long series of experiments in which the three components are made to fail in order but the order varies randomly (viz SBL, LBS, SLB, SBL and so on). The greater the number of occasions on which the light is observed to disappear before the second failure, the more confidence we might place in the ‘law of the sufficiency of the first component failure’. In the same way, we can imagine an observer on a hill H in Figure 2 who records the effect on shipping of an indefinitely long series of randomly occurring bridge collapses. Having never witnessed an arrival in port P of a ship which, on arriving at point S, had at least one collapsed bridge to its east, that observer might feel entitled to propound a ‘law of the sufficiency of the first collapse’. Like that of our circuit, this law is empirical and not particularly complex (although we may also need to stipulate that there is no alternative channel to the port and that the law is confined to ships which cannot become submarines or aeroplanes). More generally, if repeated observations of the consequences of the first event in a temporal sequence of events show such a clear regularity that the use of the term ‘law’ is justified, then it is hard to see why the redundancy of second and subsequent events in a temporal sequence cannot be inferred from such a law. In other words, if it is possible to hypothesise the existence of an empiri­ cally derivable law that holds (supporting our intuition) that the collapse of the first bridge is lawfully sufficient to explain the ship’s non-arrival in port, then we are not obliged to fall back upon some notion of causal sufficiency. Nevertheless, Fumarten and Kress contend that Wright does not adequately distinguish between lawful and causal sufficiency. Their contention that ‘[t]here does not have to be a causal chain leading from X to Y for X to be lawfully sufficient for Y’27 and, especially, their use of the term ‘causal chain’ need careful scrutiny. If they are arguing that a ‘causal chain’ cannot be inferred from frequently repeated observations of Y following X, is this not sim­ ply another rejection of Hume’s notion of causation as regularity? The examples they employ are not helpful. Firstly, ‘[t]he barometer’s falling can be a NESS of the subsequent storm without it being a causally relevant factor for the occurrence of the storm’.28 But why should a pre-scientific mind not assume, after repeated observations of the barometer, that there must be some causal relevance (if not, connection)? Perhaps the weather-god who sends storms reveals their imminence by causing barometers to fall. Replace the phrase ‘weather-god who’ with ‘falling atmospheric pressure which’ and therein lies the extra knowledge which Fumarten and Kress implicitly assume their readers possess if they are to agree that ‘falling barometers cause storms’ is absurd. But without some extra knowledge, not only does correlation fail to establish causation, it cannot show its direction. Falling atmospheric pressure is what epidemiologists term a ‘confounding’ variable: it is registered by falling barometer readings (A) and tends to precede stormy weather (B); thus it explains what can now be seen to be a spurious link between A and B. Philosophers are more likely   ibid.   ibid.

27 28

InNESS for Beginners  335 to recognise this spurious link as an example of what Reichenbach first labelled a ‘conjunc­ tive fork’.29 Their second example is no more compelling: [i]f the sun is casting a five-foot-long shadow on the ground when it is at a forty-five degree angle to a flag pole, we can infer from laws of nature that the flag pole is ten [?] feet high. It does not follow that the shadow is causing the flag pole to be ten feet high.30

Anyone acquainted with basic trigonometry can, from the given data, calculate the height of the flagpole without recourse to any causal concept or law of nature (save, per­ haps, for the observation that light travels in a straight line). This example simply requires us to imagine a right-angled triangle with the other two angles equal (viz 45 degrees); but we readily understand that the height and base of such a triangle are equal. Moreover, that understanding is no less analytic than anything else to be found in Euclid. Clearly, it is easier to give examples of what is not causally relevant than examples where such relevance is not disputed. If reliance on intuition is unavoidable, then intuition is best employed in constructing plausible boundaries. Although he does not employ our termin­ ology, Wright is essentially addressing the boundary problem when he emphasises that: [W]hen applying the NESS test one must always double-check to make sure that the actual conditions that are excluded from the description of the supposedly sufficient set of actual ante­ cedent conditions do not in fact undermine the sufficiency of the described set by preventing the instantiation of one or more of the necessary elements in that set.31

Perhaps Wright has not followed his own advice in the defective brake example. Working backwards from the accident, it is true that the point of the failure to apply the brakes does demarcate a set of conditions which are sufficient to give rise to the accident. But that can­ not justify drawing the boundary (to use our terminology) at that point: if we continue backwards we encounter another condition (the failure to repair) which is no less sufficient for the outcome.32 Equally, working backwards from the death in the desert, a set sufficient for the death appears when the second enemy empties the keg but, again, that does not justify ceasing further enquiry. Among the antecedent conditions, the poison and its pos­ sible role as a pre-emptive factor must obviously be explored. As it happens, there is a very sound reason – no poison could be ingested – for dismissing that possibility. Once NESS is applied, not to idealised examples or to simple electrical circuits, but to the real world there is an indefinitely large number of past events which constitute ‘actual ante­ cedent conditions’, and Wright’s ‘double-check’ can never totally exclude the possibility that a pre-emptive causal factor has been overlooked. Mackie was aware of this problem and addressed it with the notion of a ‘causal field’. To paraphrase the example which Mackie used in regard to singular causal statements: if when we ask the question ‘what caused this man to develop skin cancer?’, we mean ‘why did it develop now rather than twenty years ago?’, then the causal field – within which we can limit our enquiry – is this man’s life   H Reichenbach, The Direction of Time (University of California Press, 1956).   Fumarten and Kress (n 4) 101–02. If the sun is subtending an angle of 45 degrees to a perpendicular flag pole, the shadow and the pole would surely be of the same length. 31   Bramble2 (n 2) 1116, fn 156. 32   Working backwards (or rather, looking westwards) from port P (in Figure 2), a collapsed B2 is clearly suffi­ cient for non-arrival. But again, that does not justify drawing the (spatial) boundary there: another candidate for sufficiency, B1, lies further west (and therefore encountered earlier in any attempted passage from point S). With the hindsight of this example, Wright’s readiness to identify B1 as pre-emptive is hard to reconcile with his earlier insistence that it is the non-application of (defective) brakes that is causal. 29 30

336  Chris Miller history. If, however, we mean ‘why did this man develop cancer whilst other men similarly exposed to radiation did not?’ the causal field is then the class of men thus exposed to radiation. But Mackie makes no promise of certainty: ‘. . . the specification of the field is vague, but this is not a serious obstacle to establishing or using them, either in science or in everyday contexts’.33

C.  NESS and Indeterminacy If the boundary problem cannot be resolved, another point raised by Fumarten and Kress – ‘the NESS test presupposes determinism’ – seems more amenable to a pragmatic approach. The example they employ relies on a device triggered by the decay of a radioactive element. Such decay is inherently indeterministic and can only be explained by the probabilistic laws of quantum mechanics. Exactly how quantum indeterminacy disappears when we observe the macro-phenomena of everyday experience need not concern us here. Moreover, not all indeterminacy originates in quantum mechanics. Let us return to our laboratory and create some epistemic uncertainty in the mind of a student who is there using our simple circuitry to gain an understanding of current elec­ tricity. She has yet to be taught the role played by the battery; in addition, she does not know that a technician is supposed to recharge this battery overnight and then replace it each morning in the simple (Figure 1.1) circuit. But the technician is less than diligent and sometimes forgets to perform this task. After a few weeks, the student realises that, while closing the switch is always necessary to produce light, it is not always sufficient. The stu­ dent is not forced to conclude that the light/no light outcome of closing the switch is inher­ ently indeterministic – she might suspect that the explanation lies in some property of the third component (about which she has yet to learn). Meanwhile, she completes the assign­ ment she was set: to calculate the probability that closing the switch causes the bulb to emit light. In the next semester, she will learn that a charged battery is simply another necessary element in the set that is sufficient to produce light; and she might then realise that the probability she calculated is a measure of the previously hidden factor, viz technician dili­ gence. In view of the previously discussed problems of the proliferation of NESSs and the impossibility of assembling an ‘absolutely sufficient’ account of any actual outcome, the existence of a hidden variable need not prove fatal. Provided that there is little doubt that the tortious act(s) were necessary elements in at least one sufficient set, then, albeit that set is believed to contain some as yet unidentified element(s), the NESS approach can continue to meet the modest aims that Wright claims for it, viz disentangling overdetermination. Returning to our rat exterminator circuit in Figure 1.3, any cell Bi, the switch S and the RE are necessary elements in the ith of n sufficient sets. But that sufficiency also requires a hidden agent, viz whatever it is that makes the discharge from Bi (rather than from Bj) trig­ ger the RE. It could be that the aetiology of mesothelioma is so complex (more than one hidden variable?) that this circuit is a poor analogue. Nevertheless it does at least illustrate the overdetermination that is present in the (all too non-hypothetical) problem of multiple exposures to asbestos. Each exposure forms part of a set of conditions which are potentially sufficient to initiate the tumour; but we have little knowledge of the other factors which, together with absorbed asbestos fibre(s), must be present to complete that sufficiency. In   Mackie (n 23) 42.

33

InNESS for Beginners  337 the continued absence of that knowledge, the language of probability and risk once again come into play. In the United Kingdom, the precise circumstances in which a negligent increase in the risk of injury can incur civil liability have recently been the subject of careful attention by its highest court. Two cases have obliged them to confront the vexed problem of reconciling the fundamental aims of tort with the unavoidable evidential uncertainty which faces any mesothelioma victim who cannot identify, within a number of negligent, occupational exposures to asbestos, the one which ‘caused’ his cancer. And although the House of Lords34 did not explicitly employ the NESS test, their chosen approach can be interpreted as making the increased risk of injury (which has occurred) as a recoverable head of damages and then treating each defendant’s negligence as one necessary element in a set which is independently sufficient for that head.

IV. Conclusion

Wright is unlikely to be found among those who approve of this elevation of increased risk (which has eventuated in harm) into the status of actionable harm. Nevertheless his NESS approach does assist an understanding of the existence of multiple, independent causal pathways to a single instantiation of physical harm, of which the problem of multiple asbestos employers is an all too common example facing courts across the world. NESS was not designed to engage philosophers but to assist jurists when their usually trusty ‘but for’ leads to paradox. The criticisms levelled against NESS seem pedantic when viewed in the context of its modest aims, especially since a complete theory of causation continues to elude philosophers. The problem inherent in the multiple asbestos employers litigation is one of evidential uncertainty. Some may therefore find it difficult to see it as an example of ‘overdetermina­ tion’ and analogous to a multi-celled battery. But such it was because, in the absence of further information, each employer (or more precisely, the asbestos fibres to which his employee was negligently exposed) must be taken as an independently sufficient source of the risk which eventuated in mesothelioma. Therefore it can only be regretted that compensation for many UK mesothelioma victims was delayed until the House of Lords found a way of finessing the conclusion – that none of two or more negligent asbestos employers could be held legally liable – that followed from the ‘but for’ test of necessity. NESS could have readily offered a principled justification for the application of joint and several liability – and it could have done so when these cases were heard at first instance.

34   (n 5) and the earlier multiple asbestos employer litigation Fairchild v Glenhaven Funeral Services Ltd, Fox v Spousal (Midlands) Ltd, Matthews v Associated Portland Cement Manufacturers (1978) Ltd and others [2002] UKHL 22, [2003] 1 AC 32.

16 The MMTS Analysis of Causation HORACIO SPECTOR*

I. Introduction

Causation in moral and legal contexts is generally analysed through two different accounts. One is the conditio sine qua non or ‘but for’ account. This account is usually formulated in terms of counterfactual dependence.1 Thus formulated, it holds that a cause c of an outcome e is an event c such that, had c not occurred, e would not have obtained. Suppose a physician moved by pious motives turns off the life-sustaining equipment of a terminally ill patient. If he had not turned off the equipment, the patient would not have died. The physician’s act is a ‘but for cause’ of the patient’s death. The other account is variously called the nomonological, regularity, or ‘covering law’ theory of causation. Basically, it holds that a cause c of an outcome e is a relevant part of a complex condition that is sufficient, according to a law of nature, for the occurrence of e. The best known version of this account among legal theorists is the so-called INUS/NESS account, originating from Hart and Honoré in their seminal Causation in the Law and developed by John Mackie (INUS).2 Mackie analysed a cause of e as an individual exemplification of what he called an ‘INUS’ condition, that is, an insufficient but non-redundant part of an un-necessary but sufficient condition of e. Richard Wright popularised this analysis in legal circles under the label ‘NESS’.3 According to Wright, a particular condition c was a cause of e if and only if c was a necessary element of a set of antecedent actual conditions that was sufficient for the occurrence of e. A part of a sufficient condition is redundant if without that part this condition is equally sufficient. The antecedent conditions must be restricted to those that are ‘weakly’ necessary for the sufficiency of the set; * Earlier versions of this paper were presented at Universidad Torcuato Di Tella, the University of Aberdeen, Southwestern Law School in Los Angeles and the ‘Ethics Colloquium’ of the Humboldt University in Berlin. I am indebted to my audiences on those occasions for excellent questions and points. I am also grateful to Richard Wright for enriching conversations during the early drafting of the paper, as well as to Eric Blumenson, Eleonora Cresto and Carolina Sartorio for helpful comments. 1   Counterfactual dependence is understood in terms of subjunctive conditionals and ‘possible worlds’ semantics; see: D Lewis, ‘Causation’ and ‘Postscripts to “Causation” ’, in D Lewis (ed), Philosophical Papers Vol II (New York, Oxford University Press, 1986). 2   JL Mackie, The Cement of the Universe: A Study of Causation (Oxford, Clarendon Press, 1980) 60–6; HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Oxford University Press, 1985) 13–22. 3   RW Wright, ‘Causation in Tort Law’ (1985) 73 California Law Review 1735, 1788–91; RW Wright, ‘Causation, Responsibility, Risk, Probability, Naked Statistics, and Proof: Pruning the Bramble Bush by Clarifying the Concepts’ (1988) 73 Iowa Law Review 1001, 1018–34; RW Wright, ‘Once More Into the Bramble Bush: Duty, Causal Contribution, and the Extent of Legal Responsibility’ (2001) 54 Vanderbilt Law Review 1071, 1101–08.

340  Horacio Spector this kind of necessity is called ‘weak’ because it is relative to the sufficiency of the set of antecedent conditions.4 The main advantage of the INUS/NESS account over counterfactual dependence theor­ ies is that it can better explain cases of overdetermination. Two kinds of overdetermination are here relevant. First, unlike counterfactual dependence, the INUS/NESS account is capable of handling cases of ‘concurrent causation’. A condition was an INUS/NESS cause if it was necessary in the circumstances for the sufficiency of an actually sufficient set, even if it was not ‘strongly’ necessary because there were other sufficient sets. The common example is: two careless hunters shoot a person dead simultaneously.5 While each of the shots is not a causal condition of the death in the ‘but for’ sense, both are INUS/NESS causal conditions because each of them is part of a complete instantiation of the antecedent of a causal law that links shooting to dying. Second, the INUS/NESS theory can handle cases of ‘ex ante’ pre-emption, that is, cases in which the conditions corresponding to a certain cause c2 (the ‘backup cause’) will not instantiate until a temporally prior cause c1 failed to bring about the effect. Strevens proposes a clear example: ‘Smith and Jones commit a crime, but if they had not done so the head of the criminal organisation would have sent other members to perform it in their stead, and so it would have been committed anyway’.6 Strevens claims that in this case the requirements of the INUS/NESS test are satisfied with respect to Smith and Jones’s action (ie, cause c1) and that it is irrelevant to apply the test to the backup plan (ie, pre-empted cause c2). In contrast, Smith and Jones’s action is not a ‘but for’ causal condition because, in the closest possible world in which this action is not performed, the crime is still committed, perhaps at the same time and in exactly the same manner. Those advantages notwithstanding, the INUS/NESS account has been criticised on the grounds that it cannot exclude spurious conditions as causes and that it cannot properly resolve cases of pre-emption (‘early’ and ‘late’), as well as cases of ‘alternative causation’, omission and double pre-emption. Wright has tried to respond criticisms by shifting from ‘lawful sufficiency’ to ‘causal sufficiency’, that is, sufficiency grounded on a ‘causal law’.7 He says: ‘a condition was a cause of some consequence if and only if it was part of the complete instantiation of the antecedent of a causal law that links the antecedent and the consequent’.8 The concept of ‘causal law’ is the cornerstone of this new analysis, but critics have retorted that it is precisely this change that renders the account viciously circular.9 Jane Stapleton says that the account can nonetheless be maintained as a helpful algorithm to identify various forms of agent involvement in the law.10 4   T Honoré, ‘Necessary and Sufficient Conditions in Tort Law’, in D Owen (ed), Philosophical Foundations of Tort Law (Oxford, Oxford University Press, 1990) 364. 5   J Stapleton, ‘Choosing What We Mean by “Causation” in the Law’ (2008) 73 Missouri Law Review 438. 6  M Strevens, ‘Mackie Remixed’ in J Keim, M O’Rourke and D Shier (eds), Causation and Explanation (Cambridge, MIT Press, 2007) 44. 7  For the distinction between ‘lawful sufficiency’ and ‘causal sufficiency’, see: R Fumerton and K Kress, ‘Causation and the Law: Preemption, Lawful Sufficiency, and Causal Sufficiency’ (2001) 64:4 Law and Contemporary Problems 83. This distinction is not clear to me. 8   RW Wright, ‘Acts and Omissions as Positive and Negative Causes’ in JW Neyers et al (eds), Emerging Issues in Tort Law (Oxford, Hart Publishing, 2007) 297. 9   Fumerton and Kress, ‘Causation and the Law’ (n 7) 102–05; JJ Thomson, ‘Some reflections on Hart and Honoré, Causation in the Law’, in MH Kramer et al (eds), The Legacy of H.L.A. Hart: Legal, Political, and Moral Philosophy (Oxford, Oxford University Press, 2008), 152–53; MS Moore, Causation and Responsibility: An Essay in Law, Morals, and Metaphysics (Oxford, Oxford University Press, 2009) 495. 10   Stapleton, ‘Choosing What We Mean’ (n 5) 444.

The MMTS Analysis of Causation  341 My purpose in this chapter is not to participate in these disputes. Rather my project is to present a new theory of causation that I call the minimal–maximal three-set (MMTS) analysis. I claim that this analysis is especially apt to clarify issues of causation within moral and legal contexts, and I try to show how this analysis supersedes alternative accounts. Causation is still an important notion in those contexts, mainly when the question is whether we should hold someone responsible for an event, either legally or morally. Causal responsibility of some kind seems to be a necessary condition of both moral and legal responsibility. Other contexts may be dissimilar. For instance, Bertrand Russell maintained almost one hundred years ago that the notion of cause is ‘a relic of a bygone age’ in the scientific language.11 But philosophers of science and scientists alike have continued to use this notion.12 I remain agnostic about whether the language of science needs the concept of cause, but I do think that, if science uses the concept of cause, either in a necessary or optional manner, the MMTS analysis deserves consideration as a hypothesis for scientific contexts too. It is, of course, controversial whether one single notion of cause is used across various scientific contexts and, in any case, I do not make any attempt to explore this issue in this essay. The MMTS analysis is a mixed theory of causation. On the one hand, it is a regularity conception based on causal generalisations. Antecedent conditions in causal generalisations, both positive and negative, are called MMTS conditions. They are causal factors in a generic sense. Yet I take causal relata stricto sensu to be events, that is, discrete, particular changes. Causes are events that instantiate some positive MMTS conditions according to a rule of correspondence that I will introduce below. All concrete causes correspond to abstract causal factors in the generic sense, but the converse does not hold: some causal factors do not correspond to causes in the strict sense; for instance, states of affairs and negative conditions are not causes even if they can be MMTS conditions, that is, causal factors in the generic sense. On the other hand, the intervention of human behaviour in causal processes is explained by a supplementary counterfactual dependence account. On this account, forms of behaviour are not often causes sensu stricto or even causal factors in the generic sense. This is the case with omissions: propositions denoting omissions are generally not MMTS conditions. In brief, agents can condition changes in the world in different ways. I will use the label agent-conditioning to dub all those ways. I believe that agent-conditioning is a Janus-faced phenomenon that must be accounted for by resorting to both regularity and counterfactual dependence views. Agents either make things happen or let things happen. These two behaviours are typically different. Whereas making things is typically a form of causation by events, letting things happen is typically a form of counterfactual determination. My plan is as follows. In the next section, I set out the MMTS analysis. In the following sections (III, IV and V) I will show how the MMTS analysis can respond to the objections raised against the INUS/NESS analysis. In the last section (VI) I will also delineate a counterfactual dependence theory that explains how agents can relate to MMTS conditions in both MMTS and non-MMTS forms. Causal claims in moral and legal contexts are systematically ambiguous. I will disentangle their different senses and define them as precisely as possible.

  B Russell, ‘On the Notion of Cause’ Proceedings of the Aristotelian Society (1913) 13, 1–26.   See eg, M Bunge, Causalidad, el principio de causalidad en la ciencia moderna (Buenos Aires, Eudeba, 1961).

11 12

342  Horacio Spector

II. The Minimal–Maximal Three-Set Analysis

The minimal–maximal three-set (MMTS) analysis can be considered a variety of the regularity or ‘covering law’ theories that tries to solve the problems assailing the INUS/NESS formulation. The most important difference with the INUS/NESS formulation is that the MMTS account adds two sets of propositions: a second set of causal generalisations and a third set of empirical consequences following from the set of antecedent conditions and the set of causal generalisations but not from each of these sets on its own. Thus the MMTS account is formulated in terms of three sets of statements. The first set includes statements that instantiate antecedent conditions in causal generalisations (‘the MMTS conditions’); the second encompasses relevant causal generalisations, and the third set contains empirical statements that follow from the first set with the aid of the second, but not from the first or the second set alone. The introduction of the third set seeks to resolve controversial cases of pre-emption (especially ‘late’), where it might be alleged that the set of initial conditions of the relevant causal generalisation was fully instantiated so that a pre-empted cause was an actual cause (see section IV below). Unlike standard formulations of the INUS/NESS account13, the MMTS analysis avoids the expressions ‘law of nature’ and ‘causal law’ and opts for ‘causal generalisation’. There are two main reasons for this choice. On the one hand, many physical laws are ‘association laws’ that ‘tell how often two qualities or quantities are co-associated’, but cannot be conceived as stating causal relationships.14 It is only causal general propositions that can be appealed to to analyse causation; association laws establish constant universal regularities but not causal relations or causal processes. On the other hand, many cases of causation in the law do not depend on ‘universal laws’ that are true in ‘every physically possible world’ or even in every spaciotemporal configuration of our physical universe. This is the case with respect to social generalisations that only hold within certain historical periods. I agree with Ernest Nagel when he says that ‘the “historically conditioned” character of social phenomena is no inherent obstacle to the formulation of comprehensive transcultural laws’.15 However, as Nagel himself concedes, generalisations stated to explain social phenomena often use vague concepts taken from social practice, and ‘it is difficult to eliminate from them essential reference to matters specific to some particular society (or particular social tradition)’.16 Causal generalisations are essential for the MMTS analysis. The MMTS analysisans is not circular precisely because it uses ‘causal generalisation’ as a meta-linguistic term. That is, whereas the analysandum (‘cause’) denotes a non-linguistic relationship, the analysans refers to a logical relation among sets of propositions. Circularities do not hold across different language orders. In order to set out the MMTS analysis I need to supply a basic conceptual apparatus that includes four concepts: ‘causal generalisation’, ‘satisfaction of a causal generalisation’, ‘proper logical parthood’ and ‘maximal satisfaction of a set of causal generalisations’. The definitions of these concepts are as follows:   See eg Thompson, ‘Some reflections on Hart and Honoré, Causation in the Law’ (n 9) 148.   N Cartwright, How the Laws of Physics Lie (Oxford, Oxford University Press, 1983) 21. 15   E Nagel, The Structure of Science: Problems in the Logic of Scientific Explanation (New York, Harcourt, Brace & World, 1961) 464. 16   Nagel (n 15) 464. Bunge also says that many sociohistorical laws are general rather than universal; see: Bunge (n 12) 283. 13 14

The MMTS Analysis of Causation  343 1. Causal generalisation. A causal generalisation g is a true universal conditional statement that (a) asserts that there is an invariant or robust sequential correlation between empirical facts within a certain domain, and (b) sustains or warrants a technical rule R or entails a causal generalisation g’ that sustains or warrants a technical rule R. Condition (a) requires that a causal generalisation must establish an invariable or robust temporally patterned sequence of facts. By ‘robust’ I mean a sequence of facts that holds with a degree of regularity sufficiently high for the generalisation to be capable of satisfying condition (b). This means that there is a pragmatic encroachment on the definition of condition (a). In turn, condition (b) specifies that a causal generalisation must have actual or potential technical application or entail a causal generalisation that has actual or potential technical application. Though causal generalisations are not technical rules, there is a logical connection between them: for any causal generalisation g there is a technical rule R that is grounded on g or on a causal generalisation g’ that derives from g. Technical rules respond to the general form ‘You can bring about (avoid) y by making x to be the case (not to be the case)’.17 Therefore, technical rules are not hypothetical imperatives or normative requirements but rather technical guides. Here are two examples. Consider the causal generalisation ‘If a substantial dose of arsenic is applied to a human being, then she will die’. For instance, this generalisation sustains the technical rule ‘You can kill a human being by making him ingest a substantial dose of arsenic’. In other cases causal generalisations sustain technical rules that indicate how certain outcomes can be avoided. For instance, the generalisation ‘If an earthquake that measures 5.0 or more on the Richter scale hits a populated area, then a number of people will die’ sustains the technical rule ‘You can diminish or avoid human losses by evacuating a populated area to be hit by an earthquake of the mentioned magnitude’. 2. Satisfaction of a causal generalisation. Let C be a non-empty set of singular empirical propositions true in spatiotemporal region R. C satisfies a causal generalisation g in R iff g together with C entails a non-empty set E of empirical propositions (none of which is entailed by either g or C alone). 3.  Proper logical parthood. Logical parthood is a relation defined for propositional sets. Let S1 and S2 be two sets of propositions: S2 is a logical part of S1 iff S1 entails S2. For example, if S1 is a singleton set including the proposition ‘Caesar died at time t’, and S2 is a singleton including the proposition ‘Caesar died’, S2 is a logical part of S1 because the former proposition entails the latter. Any subset of a propositional set S is a logical part of S, but the converse is not true: a logical part of S need not be a subset of S. (However, a logical part of S is necessarily a subset of the set of logical consequences of S.) Proper logical parthood is then defined in the standard way: S2 is a proper logical part of S1 iff S1 entails S2 but S2 does not entail S1.

17   Condition (b) has affinity with the interventionist conception of causation. This conception was defended by RG Collingwood, An Essay on Metaphysics, revised edn (Oxford, Clarendon Press, 1998) 296–97, as well as by Douglas Gasking, ‘Causation and Recipes’ (1955) 64 Mind 479–87, and GH von Wright, Explanation and Understanding (Ithaca, Cornell University Press, 2004) (first published in 1971). Von Wright quotes Thomas Reid as the classic originator of the view. In recent times other authors have endorsed ‘experimentalist’ or ‘manipulationist’ views of causation. See eg, Cartwright, How the Laws of Physics Lie 36–38; E Flichman, ‘Causación y antropomorfismo’ (1985) 5(2) Análisis Filosófico 37–56; J Pearl, Causality (Cambridge, Cambridge University Press, 2000); H Price, ‘Agency and Probabilistic Causality’ (1991) 42 British Journal for the Philosophy of Science; D Hausman, Causal Asymmetries (Cambridge, Cambridge University Press, 1998); J Woodward, Making Things Happen: A Theory of Causal Explanation (Oxford, Oxford University Press, 2003).

344  Horacio Spector 4. Maximal satisfaction of a set of causal generalisations. Let G be a set of causal generalisations. G can be as large as the causal inquiry demands, depending on the practical context. For instance, in a given case investigators or expert witnesses can take G to be the set of causal generalisations related to ballistics and the action of bullets and projectiles in the human body. Now any set C of true singular empirical propositions maximally satisfies G in spatiotemporal region R iff the set E of empirical propositions that are entailed by G together with C (none of which is entailed by either G or C alone) is not a proper logical part of any set E* of empirical propositions that are entailed by G together with any other set C* of true singular propositions that also satisfy G in R (but are not entailed by either G or C* alone). As I have anticipated, I follow the mainstream philosophical literature on causation holding that in a strict sense causal relata are events, that is, discrete, spatiotemporally located changes.18 Facts and states of affairs are too abstract or too inactive to be the referents of singular terms occurring in causal statements. Thus, it is not the fact that Brutus stabbed Caesar, but rather Brutus’s stabbing Caesar, the cause of Caesar’s death. Whereas descriptions of events designate discrete entities, spatiotemporally located and possessing all their concrete richness and complexity, propositions denote abstract entities whose contours are demarcated by the relevant abstract terms. Therefore, Brutus’s stabbing Caesar is the same event as Brutus’s stabbing Caesar on 15 March 44 BC, but the fact that Brutus stabbed Caesar is not the same fact as the fact that Brutus stabbed Caesar on 15 March 44 BC. And while Brutus’s stabbing Caesar is the same event as Brutus’s stabbing Caesar with a dagger, the fact that Brutus stabbed Caesar is not the same fact as the fact that Brutus stabbed Caesar with a dagger. At the same time, singular causal statements are causal in virtue of their relation to causal laws/generalisations that are canonically formulated in terms of propositions picking out relevant properties of the events causally related. But how do event-descriptions instantiate the abstract, propositional antecedents in causal generalisations? This is a complex question. Donald Davidson’s well-known proposal involves a very cumbersome rendering of causal laws/generalisations in terms of statements universally quantified over events.19 I want to maintain the standard, simple schema of causal laws/generalisations as universal conditional statements: [(x) (Fx ⊃ Gx)]. When the generalisation encompasses relations rather than properties, the standard schema becomes: [(x) (F (x,y) ⊃ G (x,y))]. Therefore, I need a rule of correspondence linking event-descriptions to propositions that allows making singular causal statements instantiate the abstract antecedents of causal generalisations. For this specific purpose I propose a rule of correspondence that may be called Denominalisation. This rules works as follows. I will say that for any event e described by a participial expression, there is a true denominalising proposition pe, that is, a true pro­ position that denominalises the participial description of e into the indicative mood of the corresponding verb. For instance, if the participial description were Brutus’s stabbing Caesar, pe would be Brutus stabbed Caesar. Formally, for any event e described by a participial expression of the form x’s F-ing y, in which ‘x’ and ‘y’ are names, and ‘F’ is a dyadic predicate, there is a true denominalising proposition [F (x, y)]. I will assume that events are truth-makers and propositions are truth-bearers. Accordingly, I will take it that the event e 18  The locus classicus is: D Davidson, Essays on Actions and Events, ‘Essay 7’ (‘Causal Relations’) (New York, Oxford University Press, 1980). 19   ibid, 158.

The MMTS Analysis of Causation  345 described by an event-description of the form x’s F-ing y is the truth-maker of the proposition [F (x, y)]. Simply put, the event e is the truth-maker of the denominalising proposition pe. With the above conceptual apparatus in place, we can say that an event e caused an event e’ iff the following conditions are met for the denominalising proposition pe and the denominalising proposition pe’: (i) there is a set C of true singular empirical propositions that maximally satisfies G in spatiotemporal region R and pe is a member of C; (ii) there is a set E of true singular empirical propositions that includes all the empirical propositions that are entailed by G together with C (none of which is entailed by either G or C alone) and pe’ is a member of E; and (iii)  No proper logical part of C together with G entails E. Let me explain the MMTS conditions in order. Condition (i) I call the maximality condition. It requires the maximality of E, and, indirectly, of C. C must maximally satisfy G in spatiotemporal region R so that E possesses the maximal empirical content that can be deduced from G and C. Unlike the INUS/NESS test, which only appeals to the notion of a minimal sufficient condition, the MMTS analysis introduces the idea of maximality. I call condition (ii) the web condition. It secures that effect e’ is placed within the whole of empirical consequences relative to sets C, G, and E. As we will see, the web condition is critical for allowing MMTS to place e and e’ in the actual causal process, rather than in any pre-empted causal process. The MMTS analysis only picks out actual causes because they alone are correlated with a multitude of empirical consequences that are denoted by E. A pre-empted causal process may match pe, but will generally fail to match other empirical propositions that must also be included in E. Condition (iii) lays down the condition of minimality for C.20 This condition is needed to exclude both redundant and spurious conditions to be causes according to the MMTS analysis. In excluding redundant conditions, it plays a role similar to that played by the ‘internal necessity’ element in the INUS/NESS account. However, as I will show in the next section, my minimality condition is also capable of ruling out spurious conditions. We can put together the three conditions in the following formula: An event e caused an event e’ iff pe is a member of a minimal set C of true singular empirical propositions that maximally satisfies set G in spatiotemporal region R, and pe’ is a member of a set E of true singular propositions that includes all the empirical propositions that are entailed by G together with C (none of which is entailed by either G or C alone).

There might seem that there is a contradiction between the ideas of maximality in (i) and minimality in (iii). This is not so. C must be ‘minimally maximal’, so to speak: it must maximally satisfy set G in spatiotemporal region R but must do so minimally, that is, excluding all redundant and spurious conditions. The basic idea of the MMTS analysis is to pick out the minimal set C of true singular empirical propositions that maximally satisfy G in spatiotemporal region R. If C meets that condition, it entails together with G the 20   For alternative definitions of minimality, see M McDermott, ‘Redundant Causation’ (1995) 46(4) British Journal for the Philosophy of Science, 535; N Hall, ‘Two Concepts of Causation’ in J Collins, N Hall, and LA Paul (eds), Causation and Counterfactuals (Cambridge, MIT Press, 2004) 260; M Ramachandran, ‘A Counterfactual Analysis of Causation’ (1997) 106 Mind 270. Unlike the set-theoretical notions used by these authors, the ‘mereological’ concept introduced in the text also excludes spurious conditions from set C.

346  Horacio Spector maximal set of empirical propositions E. A kind of efficiency animates my account: output maximisation and input minimisation. In simple cases, the MMTS analysis yields the same results as the INUS/NESS account. Consider this case: Lethal remedy. Alfred and Evelyn put bromide, a precipitating agent, into Emily’s regular medicine. The generally innocuous strychnine constituents of the medicine accumulated at the bottom of the bottle forming a lethal concentrate. Emily eventually consumed the fatal dose of strychnine and died from asphyxiation.21

Alfred and Evelyn’s putting bromide into Emily’s medicine caused Emily’s death according to the MMTS account insofar as the denominalising proposition ‘Alfred and Evelyn put bromide into Emily’s medicine’ is a member of a minimal set C of true empirical propositions that maximally satisfy a set of causal generalisations centrally including those concerning the chemistry of precipitation by bromides and the chemistry of strychnine poisoning. In addition, the denominalising proposition ‘Emily died’ is a member of the set of all the empirical propositions that are entailed by C and the mentioned set of chemical generalisations. The INUS/NESS account seems to concord since we can accept that the couple’s action is a non-redundant part of the ‘complete instantiation’ of the antecedent of a causal law (in fact there are more than one) and that Emily’s death is an instantiation of the consequent of that causal law. Yet the MMTS analysis does not necessarily agree with the INUS/NESS account. Consider this case: Deadly wife. In Nazi Germany a woman denounces her liberal husband to the Gestapo as ‘enemy’ of the Third Reich. The husband is condemned to death and executed a few days after.22

There is an indubitable causal relationship between the wife’s denunciation and the husband’s death, but this relationship cannot be understood in terms of a ‘law of nature’ or a ‘causal law’ and, therefore, cannot be captured by the INUS/NESS account, which essentially appeals to those notions. In contrast, the causal connection in this case can easily be captured by the MMTS analysis. In fact, both Lethal remedy and Deadly wife involve causal generalisations in the sense previously defined. No appeal to ‘laws of nature’ or ‘causal laws’ is needed. Deadly wife is a hard case for the INUS/NESS account because there does not seem to be a universal law, either ‘natural’ or ‘causal’, that could establish a universal invariable connection between the wife’s action and the husband’s death. In a different social situation a denunciation of someone as enemy of the government would cause laughter. There is nonetheless a true causal generalisation that links the denunciation with the death of the denunciated person under the special circumstances of Nazi Germany. True, one might try to see this generalisation as a statement derived from a universal law. In order to get that universal law we might try to substitute ‘purely qualitative predicates’ for the predicates ‘Nazi Germany’, ‘Gestapo’ and ‘Third Reich’, which make reference to a particular historical situation. Yet I believe that universal statements resulting from our attempts would be subjected to several exceptions and, therefore, could not really count as universal laws. Eventually we would probably be led to a universal statement formulated in terms of a   This case is taken from Agatha Christie’s famous first novel The Mysterious Affair at Styles.   This hypothetical is adapted from a real post-war German case. See: G Radbruch, ‘Gesetzliches Unrecht und übergesetzliches Recht’ (1946) 1 Süddeutsche Juristenzeitung 105–08. 21 22

The MMTS Analysis of Causation  347 concept like ‘police state’, which could rule out all ‘exceptions’ as falling beyond its range of application. At that point, however, our ‘universal statement’ would really establish a logical or conceptual relationship, rather than a causal, contingent one. Police states are by definition states that kill people under certain conditions. In the following three sections we shall see that the analysis defended in this essay can clearly deal with all the difficult problems presented in the literature, that is, spurious conditions, causal pre-emption, alternative causation, ‘causation’ by omissions, and ‘causation’ by double preventions. These problems can only be treated by the INUS/NESS account, if at all, in ad hoc ways, and I believe that parsimony is an important virtue of philosophical theories.

III.  Spurious Conditions

A serious problem for the INUS/NESS variety of the regularity account is that any condition can figure in an INUS/NESS set S if S also includes a ‘backing’ disjunctive condition that contains a proposition negating that that condition holds and a proposition affirming that an actual causal condition obtains.23 In this way spurious conditions might become causal conditions. Thomson gives the following example.24 Suppose David shoots Charles in the head. Suppose also that David’s shooting was a necessary condition for the set of conditions to be sufficient, in conjunction with the relevant causal generalisations, for Charles’s death. Now David’s shooting Charles is also denoted by the following conditions: (1)  Caesar crossed the Rubicon. (2)  Either Caesar didn’t cross the Rubicon or David shot Charles. If the fact that David shot Charles is a necessary member in a set of conditions that is sufficient for Charles’s death, the conjunction of the facts denoted by propositions (1) and (2) is also a necessary member in a set of conditions that is sufficient for Charles’s death. In fact, given the disjunction in proposition (2), the set cannot be sufficient for Charles’s death (according to the relevant causal generalisation) unless we also include condition (1), so it seems that Caesar’s crossing the Rubicon was a causal condition of Charles’s death under the INUS/NESS account. Strevens holds that, in order to avoid this problem, ‘causal sufficiency ought to be defined, then, so that a set of conditions is causally sufficient for an event e only if the conditions represent a causal process that produces e’. He adds: ‘A set of conditions entailing e represents a causal process producing e, I propose, just in case each step in the entailment corresponds to a strand in the relevant causal web’.25 The problem with this account is that it must give up its goal to provide a non-circular analysis of causation. In fact, this solution is circular as an account of causation in that it employs the expression ‘relevant causal web’. The MMTS analysis supplies a straightforward solution to the problem of spurious conditions. Let’s assume that the proposition ‘David shot Charles’ (pe) is a member of a singleton set C of true propositions that maximally satisfy G in a certain spatiotemporal region. 23   Strevens, ‘Mackie Remixed’ (n 6) 26; Thomson, ‘Some reflections on Hart and Honoré, Causation in the Law’ (n 9) 151–53; Fumerton and Kress, ‘Causation and the Law’ (n 7) 95. 24   Thomson (n 9) 151. 25   Strevens, ‘Mackie Remixed’ (n 6) 28.

348  Horacio Spector The proposition ‘Charles died’ (pe’) is a member of E. Because C together with G entails E, David’s shooting caused Charles’s death. Let C* be a set that contains proposition pe1 ‘Caesar crossed the Rubicon’ and proposition pe2 ‘Either Caesar didn’t cross the Rubicon or David shot Charles’. Because C* together with G also entails E, it might seem that Caesar’s crossing the Rubicon is also a cause of Charles’ death. In fact, if we remove pe1 from C*, C* no longer entails E, so pe1 is necessary for the sufficiency of C*. However, MMTS does not allow Caesar’s crossing the Rubicon to be a cause, which is what we want. Let’s see why. C is a proper logical part of C*, because C* entails C but C does not entail C*. In fact, C does not entail C* because C* also contains the proposition ‘Caesar crossed the Rubicon’. Therefore, there is a set of propositions C, such that (i) C is a proper logical part of C* and (ii) C together with G entails E. (I assume that Caesar’s crossing the Rubicon is empirically irrelevant in relation to G and E.) This shows that C* does not meet the minimality condition. Basically, sufficient sets that contain spurious non-redundant conditions as well as those that contain redundant conditions do not meet the minimality condition. MMTS puts both spurious and redundant conditions on a par.

IV. Pre-emption

We have already said that the INUS/NESS account is able to resolve properly cases of ‘ex ante’ pre-emption in which a ‘backup’ causal process would have been initiated if and only if the actual causal process had not been initiated or successfully completed. However, the INUS/NESS account has a harder time to resolve ‘early’ and ‘late’ pre-emption cases. In ‘early’ pre-emption, both the conditions corresponding to causes C1 and C2 (where C1 is temporally prior to C2) are instantiated, but cause C2 pre-empts cause C1 by cutting off or negating one of the conditions corresponding to cause C1. The problem is that in ‘early’ pre-emption the sufficient set of pre-empted cause C1 is instantiated at a moment prior to the moment in which the effect occurs. In ‘late’ pre-emption cases the pre-empted causal process had not only actually initiated but all its relevant conditions had also been fully instantiated so as to guarantee the occurrence of the effect if not for the intervention of the pre-empting causal process. That is, ‘late’ pre-emption cases are those in which the preempted causal process has been completed except for its effect.26 In these cases the INUS/ NESS theorist seems committed to accepting that the conditions that are necessary in the sufficient set of the pre-empted causal process are INUS/NESS causes, an obviously counterintuitive conclusion. One way out of this conclusion would be for the INUS/NESS theorist to appeal to some notion of physical contiguity between the sufficient set (or some of its members) and the effect. This move would exclude negative conditions as possible causes, a result that some INUS/NESS theorists might not be prepared to accept. Is the MMTS analysis affected by pre-emption? To see how it works in practice, consider some common examples: Final destination 1.  A traveller enters a desert with a water keg. An enemy secretly puts a fatal dose of poison in the water. Before the traveler drinks the water, another person steals the keg thinking it contains pure water. The traveller dies of thirst.27  Moore, Causation and Responsibility: An Essay in Law, Morals, and Metaphysics (n 9) 493.   Hart and Honoré, Causation in the Law (n 2) 239.

26 27

The MMTS Analysis of Causation  349 Final destination 2.  A man, Charles, drinks a fatal dose of poison at a certain time. The poison would normally take 15 minutes to produce Charles’s death. An enemy shoots him dead shortly after. Charles’s death occurs one minute later.28 Final destination 3.  A barge carrying supplies to the port is delayed by a collapsed bridge A that obstructs the waterway. If the barge had not been delayed by bridge A, it would have been delayed anyway by a second collapsed bridge B that also obstructs the waterway.29 Because the MMTS analysis includes the web condition, the effect must always be encompassed in a set of empirical consequences E. I claim that this feature of the MMTS analysis makes its applications fit with our common intuition that pre-empted causes are not actual causes. In Final destination 1 (‘early’ pre-emption), for the poisoning of the water supply to be the cause of the traveller’s death, there must be a set E that maximally includes all the empirical consequences in accordance with the relevant set of causal generalisations concerning the action of poisons in the human body. Therefore, E must include propositions stating the time, the manner and other relevant traits of the traveller’s death. One member of E must say, for instance, that the traveler’s corpse contained residues of poison. For inferring this proposition from C and the relevant set of causal generalisations, C must include the proposition that the traveller drank the poison and this proposition must be true according to the maximality condition. However, this proposition cannot be true, and, therefore, it cannot be a member of C because the keg was emptied before the traveller could drink the water. Final destination 2 (‘late’ pre-emption) is different because the proposition that Charles drank the poison at a certain time was true and, consequently, it can be included within set C. The issue is whether Charles’s drinking of the poison was the cause of his death in accordance with MMTS. Now C together with the relevant set of causal generalisations entails a set E of empirical consequences. Recall that the web condition requires that the proposition ‘Charles died’ be included within set E, and that the definition of maximal satisfaction of G requires as well that E be maximal. For the set E containing the proposition ‘Charles died’ to be maximal, it must also contain propositions that specify the manner and time of Charles’s death. For instance, E must contain the proposition ‘Charles died 15 minutes after drinking the poison’. Now the proposition ‘Charles died 15 minutes after drinking the poison’ cannot be a member of E because it is not true. Nor are true other propositions that presumably are members of E and that refer to occurrences that take place after Charles’s death. For instance, E may include a proposition to the effect that Charles goes into convulsions 10 minutes after the drinking. This proposition is obviously false because Charles is dead by that time. Final destination 3 (‘late’ pre-emption) is controversial for the INUS/NESS account because, depending on how the relevant causal generalisation is framed, it might be thought that the whole gamut of initial conditions of the set C has been fully instantiated as regards both bridge A and bridge B. Of course, this is not the case if the set C is temporally stretched as to capture the physical contact of the ship with the bridge. This contact presumably occurred with respect to bridge A but not with respect to bridge B. Yet if the relevant causal generalisation is framed in such a way that it covers conditions such as the characteristics of the ship, its position, speed and direction, the currents of the river, the collapse of the bridge 28   Wright, ‘Causation in Tort Law’ (n 3) 1773, 1781, 1795; Thomson, ‘Some reflections on Hart and Honoré, Causation in the Law’ (n 9) 151. 29   Hart and Honoré, Causation in the Law (n 2) 250.

350  Horacio Spector and so on, short of any condition that logically entails the physical contact of the ship with the bridge, then it might be alleged that both the collapsed bridge A and the collapsed bridge B are causes of the ship not reaching port on time. Indeed, under this ‘time-bounded’ framing of the causal generalisation, its set C of initial conditions has been fully instantiated even with respect to bridge B. The point is that the success of the INUS/NESS account to discard bridge B as a putative cause relies on how we frame the relevant set of causal generalisations. Needless to say, framing of generalisations is a controversial matter. According to MMTS, the case of the two bridges has a clear resolution. As applied to bridge B, E includes not only the proposition that the ship did not reach port on time but also propositions denoting other facts associated with the ship colliding with bridge B, or being stopped by bridge B, such as the presence of the ship in the nearby of bridge B, or traces of the paint of bridge B on the hull of the ship. Because these other facts denoted by the propositions included in E did not occur, E cannot be true and, therefore, the causal claim with respect to bridge B is false, whereas the causal claim with respect to bridge A is true because both C and E are true relative to bridge A. I have argued so far that, if c1 is the actual cause, as opposed to c2, the relevant set E of empirical consequences will generally be different from the set that we would have if c2 were the actual cause. This claim might be challenged. For example, it might be suggested that a ‘lately’ pre-empted causal process is conceivable that is empirically indistinguishable from the actual process. I cannot imagine an interesting case like this in a moral or legal context, and so I leave its construction to my imaginary challenger. My sense is that, if the actual causal process and the supposedly ‘lately’ pre-empted causal process are empirically equivalent, there is no basis to assert that one causal process has been pre-empted by the other. That is, if two causal processes ran to their respective completions without differentiating themselves at all in empirical terms, I would be inclined to regard them as cases of concurrent causation rather than as cases of pre-emption. In my analysis causal pre-emption is also an empirical notion.

V. Alternative Causation

Cases of alternative causation present problems of set selection for the INUS/NESS account.30 Consider the following cases: Bachelor party.  Dandy’s 10 friends are crazy about his bride, Mary. In the bachelor party each of them puts one drop of poison in Dandy’s wine. Each friend acted on his own (with no previous agreement). Six drops of poison were sufficient to kill Dandy. Dandy drinks the wine and dies. Collegiate decision.  The 10 members of a club’s board expel one member from the club by unanimous vote. Each voter acted on his own (with no previous agreement). A majority of only six votes was necessary according to the club’s rules.31

  ibid 235–53.   Stapleton, ‘Choosing What We Mean’ (n 5) 443; Stapleton’s board has nine members. The case is a version of the so-called ‘leather spray case’, in which the German Federal Supreme Court convicted a company’s board members for intentional bodily harm: 37 BGHSt 106 (1990). The board had failed to recall the dangerous spray from the marketplace. I am indebted to Marcelo Sancinetti for quotation of this case. 30 31

The MMTS Analysis of Causation  351 According to the INUS/NESS account each drop is a cause of Dandy’s death because it is a member of a (possible) six-drop set that is sufficient for Dandy’s death, being each drop necessary for the sufficiency of that set. The same applies to the board’s decision. Each member’s vote is a ‘weakly’ necessary member of a (possible) six-vote set that is sufficient to expel the member. I do not take exception to this conclusion. I agree that each drop and each vote is a causal condition. My problem is with the explanation. It yields two counterintuitive results. First, there are 210 possible six-member sets that are INUS/NESS sets in regard to the relevant outcome (ie, death, expulsion).32 It seems arbitrary to pick out any of those sets to hold one drop/vote to be a causal condition. Second, suppose we pick out set S that contains drops/votes A, B, C, D, E and F. Then each drop/vote is an INUS/NESS causal condition of the outcome. But with respect to S drop/vote H, for instance, is not an INUS/ NESS causal condition. Notice that the INUS/NESS account is formulated in terms of the connective ‘if and only if ’. Therefore, we must turn to a different set S’ that contains H (there are many possible combinations). Now H is not a causal condition with reference to set S and is a causal condition with reference to set S’. This is not a contradiction, but displays a sort of relativity in our causal claims. It seems ad hoc to choose the reference set of initial conditions so that each drop/vote becomes a causal condition. Naturally, an arbitrary choice of the reference set runs afoul of the presumed objectivity of causal claims. According to the MMTS analysis, the set selection problem is avoided because there is one single reference set of initial conditions. The relevant set C of true singular propositions must include the propositions denoting the 10 drops/votes, because the relevant set of causal generalisations, G, must be maximally satisfied. That is, each of the 10 drops/votes is an MMTS causal condition because the corresponding proposition is a member of the minimal set C that is required to generate the maximal set E of empirical consequences. In fact, C maximally satisfies the relevant generalisation when the set E of empirical consequences that are entailed by that generalisation together with C is not a proper logical part of any set E* of empirical consequences that are entailed by that generalisation together with any other set C*. The maximality condition ensures, with respect to Bachelor party, that the time of the death, its manner and other empirical features are included in E. Once maximality is met, we obtain the minimalist antecedent set by excluding all spurious and redundant conditions. Now the supposedly ‘more than needed’ drops/votes must be included in C because some empirical consequences would otherwise be absent from E, in which case G would not be maximally satisfied. This means that the ‘more than needed’ drops/votes are MMTS causal conditions as well. For instance, the additional drops might have speeded up death thus altering time, or they might have modified the quantity or quality of residual substances in the victim’s blood, and so on. So we don’t need the additional drops for ‘sufficing’ death, but we certainly need them to ‘suffice’ other consequences that must be included in the maximal set E. Collegiate decision is more difficult, because the fact that a vote is unanimous as opposed to merely majoritarian does not normally make great empirical differences, but even in this case unanimity must be included in C for some empirical consequences to be present in E. Some of those empirical consequences are, for instance, the emotional reactions on the part of the expelled member and other people in the club, differences in the board proceedings, differences in potential legal actions, and so on. The gist of the web condition is to see every effect as part of a set of empirical consequences, rather than as a free-standing event.

1  2

  210 is the binomial coefficient 10  = 10! = 10 • 9 • 8 • 7 6 (10–6)! 6! (4 • 3 • 2 • 1)

32



352  Horacio Spector

VI. Acts, Omissions and Double Preventions

Consider the following cases: Bad nurse.  A patient has an anaphylactic reaction upon drinking a medication. The nurse fails to give him proper anti-allergic medication, which she could have easily done. The patient dies later on. Bad captain.  A ship captain finds a shipwreck survivor drowning in the sea but fails to throw him a lifebelt, which he could have easily done. The survivor dies later on.

It is clear that the captain and the nurse have both let the survivor and the patient die. Their behaviour is wrongful just as an instance of an omission to aid. But the question now is not whether the behaviour is wrong but whether the nurse and the captain have caused via their omissions the patient’s and the survivor’s deaths. Defenders of the INUS/NESS approach assume that omissions generally are negative causal factors, just as acts are positive causal factors. For instance, Wright associates an omission with the absence of prevention.33 Thus, non-treatment and non-floating-device are negative factors that are ‘weakly’ necessary for the set of relevant causal conditions to be sufficient. On this view, the nurse’s failure to give the patient anti-allergic medication and the captain’s failure to throw the lifebelt are instantiations of those negative conditions and, consequently, INUS/NESS causal conditions of the deaths. This is an erroneous view, however. It is true that the set of NESS/INUS conditions of the allergic patient’s death must include the negative condition that he is not given medication, but the nurse’s failure is not a truth-maker of this negative condition, because someone else could have given the patient proper medication. By the same token, it is true that the set of NESS/INUS conditions of the shipwreck survivor’s death must include the negative condition that he does not have a floating device, but the captain’s failure is not a truth-maker of this negative factor. For instance, a member of the crew could have thrown the drowning person a life-belt. Omissions are not truth-makers of negative causal factors. Whether they are causally sufficient conditions of negative causal factors depends on the circumstances and the relevant causal generalisation. The key to discussing omissions as possible causes is to notice that acts and omissions are forms of behaviour that condition causal processes in different ways. Acts are generally truth-makers or causally sufficient conditions of the positive (or negative) conditions established in the antecedent of a causal generalisation. From the practical viewpoint of the agent, acts function through the application of the technical guides sustained or warranted by causal generalisations. It is only because causal generalisations sustain technical guides that agents can use such generalisations in order to produce outcomes in the world. Technical guides do not exhaust the variety of practical guides that causal generalisations may sustain, but they are essential to understand the connection that there is between causal generalisations and agent-conditioning by acts. Now, whether an act involves a truth-maker or a causally sufficient condition depends on the type of behaviour. Shooting makes true one of the positive conditions (ie, shooting) established in the antecedent of a causal generalisation that relates shooting and death. Alternatively, triggering the gun is a causally sufficient condition of the instantiation of a 33   Wright, ‘Acts and Omissions’ (n 8) 290–92. Wright says: ‘Omissions generally operate as negative causes of some consequence, by precluding the occurrence of a possible preventing cause’ (291).

The MMTS Analysis of Causation  353 positive condition in the antecedent of the generalisation (ie shooting). In the latter example, there are two causal processes: in one process the causal relata are the bodily movement (ie, triggering) and the gun shooting; in the other the causal relata are the shooting and the victim’s death. One causal generalisation relates the shooter’s triggering and the gun’s shooting. The other generalisation links the shooting with the victim’s death. While causation is the general term for both concepts, I introduced the label agent-conditioning to designate the various ways in which agents can relate to the world in making things happen or letting things happen.34 In this example agent-conditioning is just a form of MMTS causation in which one of the conditions in the antecedent of the relevant causal generalisation is a bodily movement of the agent. Accordingly, we can roughly say that a property conceptually required by ‘P does X’ (or ‘P makes X happen’, as something different from ‘P lets X happen’) is that P makes bodily movements that are truth-makers or causally sufficient conditions of the instantiation of a positive condition established in the antecedent of a causal generalisation one consequence of which is X.35 Whereas acts are typically truth-makers or causally sufficient conditions of the instantiation of positive conditions established in the antecedent of a causal generalisation, omissions are ‘but for’ conditions of the instantiation of negative conditions established in the antecedent of a causal generalisation. More precisely, I must restrict my claim to nonpreventive omissions, that is, failures to prevent. In fact, as I will show later on, other omissions may have a different character. My chief assertion in this section is that non-preventive omissions are ‘but for’ conditions of negative MMTS conditions. This is quite different from saying that omissions are MMTS conditions (or NESS/INUS conditions) of negative MMTS conditions. It is also different from asserting that omissions are MMTS conditions (or NESS/INUS conditions) of the outcome of the causal process. The captain lets the survivor die in the sea by omitting to throw him a lifebelt. The causal generalisation leading to the survivor’s death includes in its antecedent a number of negative conditions: the survivor does not have a lifebelt or any other floating device, no other rescue comes in time, and so on. The captain’s omission is neither a truth-maker nor an MMTS condition (or a INUS/NESS condition) of any of those negative conditions. This means that omissions are not negative antecedent conditions in causal generalisations. Omissions are generally forms of behaviour that intervene in the world by allowing certain causal process to run (to completion). As said before, omissions are not truth-makers of negative conditions (save in rare circumstances). They often are not causally sufficient conditions either. Rather, omissions are typically negative ‘but for’ conditions of negative conditions that figure in the antecedent of a causal generalisation. By incurring an omission an agent typically ‘collaborates’ with a causal process. He does not set up the causal process, nor does he change or redirect it. Non-preventive omissions conceptually presuppose that the agent could have counteracted the relevant causal process by negating one of the negative conditions laid down in the antecedent of the causal generalisation. Indeed, the captain could have made the negative condition ‘non-floating-device’ false by throwing the survivor the lifebelt. The nurse could 34   Agent-conditioning in the sense explained must not be confused with agent causation, which refers to the relation of internal volitions and external outcomes; see eg T O’Connor, ‘Agent Causation’ in T O’Connor (ed), Agents, Causes, and Events: Essays on Indeterminism and Free Will (New York, Oxford University Press, 1995) 173– 200; T O’Connor, ‘Why Agent Causation?’ (1996) 24 Philosophical Topics 143–58. 35   This is a variation on a position I defended elsewhere. H Spector, Autonomy and Rights (Oxford, Clarendon Press, 1992) 149.

354  Horacio Spector have made the negative condition ‘non-treatment’ false by giving the patient proper medication. An omission is not a form of ‘negating’ a condition, positive or negative, in the antecedent of a causal generalisation. Instead, an omission typically is a negative ‘but for’ condition of a negative condition. Omissions are generally ‘but for’ ‘causes’, rather than INUS/NESS or MMTS causal conditions. However, the negative conditions in respect to which omissions are ‘but for’ causes are certainly causal conditions of the relevant outcome. That is, ‘non-floating-device’ and ‘non-treatment’ are MMTS conditions of the survivor’s and patient’s deaths.36 We may conclude that there is generally a causal asymmetry between acts and omissions because acts are typically truth-makers or MMTS causal conditions of MMTS positive conditions, that is, of positive conditions in the antecedent of a causal generalisation. (Sometimes, acts are MMTS causal conditions of MMTS negative conditions.) In contrast, omissions are not truth-makers or causally sufficient conditions of negative conditions. Instead, they must typically be analysed under the counterfactual dependence account of causation. Non-preventive omissions are negative ‘but for’ conditions of negative MMTS conditions. One could then expect this account of non-preventive omissions to reproduce a general trait of the counterfactual dependence account, namely, its inability to handle cases of concurrent causation. Indeed, one of the most intricate issues in the literature is how to resolve cases of overdetermination by non-preventive omissions. Consider a wellknown case: Defective non-used brakes. A car renter, Amy, failed to repair defective brakes in a car that he rented to Martin, and Martin failed to use the brakes to avoid running into a pedestrian.37

Unlike other non-preventive omissions, Amy’s and Martin’s omissions are not but-­ conditions of the outcome (ie, the crashing into the pedestrian). If Amy had repaired the brakes, the crash would have nonetheless occurred (the brakes were not used). If Martin had used the brakes, the crash would have anyway occurred (the brakes were defective). However, it is true that the conjunction of Amy’s repairing the brakes and Martin’s applying them would have negated one negative condition in the relevant MMTS causal generalisation (ie, absence of effective braking) and, consequentially, would have prevented the crash. We can state this in this way: (1)  If Amy had repaired the brakes and Martin had applied them, the crash would not have occurred.

Let us assume now that De Morgan’s laws can treat omissions as negations, since this is intuitively plausible in this context.38 By De Morgan’s laws, (1) is equivalent to: 36   While it is common to regard a failure to prevent as a ‘but for’ condition of the relevant effect, or even as a ‘but for’ condition of the causal relation, I was unable to find an account of failures to prevent as ‘but for’ conditions of negative conditions. See, for instance: Stapleton, ‘Choosing What We Mean’ (n 5) 433, 436–37 (she discusses a hypothetical where a gardener fails to provide water to a plant that consequentially dies); Moore, Causation and Responsibility: An Essay in Law, Morals, and Metaphysics (n 9) 351–54, 399–400, 451–52, 478–79 (Moore holds that counterfactual dependence is essential for omission liability, but he denies that ‘but for’ conditions are causes). For an analysis of omissions in terms of counterfactual claims about ‘genuine’ causal processes, see: P Dowe, Physical Causation (Cambridge, Cambridge University Press, 2000) 136–40. 37   Wright, ‘Causation in Tort Law’ (n 3) 1801. 38   De Morgan’s laws say that ‘not-(A and B)’ is equivalent to ‘not-A or not-B’, and that ‘not-(A or B)’ is equivalent to ‘not-A and not-B’. Of course, we can substitute ‘not-A’ for ‘A’, and ‘not-B’ for ‘B’, in which case we obtain, by double negation, the following equivalences: ‘not-(not-A and not-B) iff A or B’, and ‘not-(not-A or not-B) iff A and B’. The latter equivalence can be reversed: ‘A and B iff not-(not-A or not-B)’. This law was used in the text to equate (1) and (2).

The MMTS Analysis of Causation  355 (2)  If the disjunction of Amy’s and Martin’s omissions had not occurred, the crash would not have occurred.

This means that the absence of effective braking and, consequentially, the crash counterfactually depends on the disjunction of Amy’s and Martin’s omissions. Therefore, concurring non-preventive omissions are disjunctive ‘but for’ conditions of MMTS negative conditions (and of the relevant outcomes as well). Because both Amy’s and Martin’s omissions are on a par in regard to their non-preventive relation to the absence of effective braking and crash, there are no reasons to differentiate between both omissions, nor to regard one as pre-emptor of the other. In his monumental Causation and Responsibility,39 Michael Moore holds that a disjunction of omissions cannot be a ‘but for’ condition of an outcome because an outcome cannot counterfactually depend on a disjunction without at the same time depending on each of the disjuncts.40 Moore invokes De Morgan’s laws for this conclusion, but his application of these laws is different from the one made above. If this argument were sound, then it would also apply to the disjunction of omissions as a counterfactual determiner of negative MMTS conditions. The argument is not well-articulated, so I will interpret it in the only way in which it seems intelligible to me. Moore says: ‘If one makes sense of negative and disjoining events, then from not (fire1) and not (fire2), one should be able to infer not (fire1 or fire2), and vice versa’.41 Though Moore formulates the original version of his argument as regards disjunction of events (eg fires), he extends it to omissions in discussing an argument from Phil Dowe.42 The core of his argument seems to be on page 355. There he says: If the negation of this larger event [~(F1 v F2)] is equivalent to the conjunction of the negations of each smaller event [(~F1 . ~F2)], one might well conclude that whatever is not true of each fire considered separately is not true of the larger event that is the disjunction of these fires.

Therefore, for Moore it violates De Morgan’s laws to say ‘that the larger event caused e while neither of the disjuncts making up the larger event caused e’. However, I contend that this is a misapplication of De Morgan’s laws. Indeed, it is possible for a disjunctive event to be a cause of an outcome e even if it is false of each disjunct that it caused e. Let’s assume that it is false of each fire that it caused e. By De Morgan’s laws this entails that it is false that fire1 caused e or that fire2 caused e. But this is not to say that it is not true that fire1 or fire2 caused e. This is so because causal predicates do not distribute over disjunction according to the counterfactual dependence account. That is, (3) (A or B) caused e

is not equivalent to: (4) (A caused e) or (B caused e)

In fact, drawing (4) from (3) is fallacious. We can illustrate this fallacy with a simple example. Let’s accept that Leibniz and Newton independently invented infinitesimal calculus and that no one else was actually working on calculus in the seventeenth century.43   See, generally, Moore, Causation and Responsibility: An Essay in Law, Morals, and Metaphysics (n 9).   Moore (n 9) 355 and 450.   Moore (n 9) 355. One additional problem is that there is a misprint on this page. Moore adds: ‘This means (F1 . ~F2) ≡ ~(F1 v F2)’. This formal representation is mistaken. It should instead say: ‘This means (~F1 . ~F2) ≡ ~(F1 v F2)’. Since this is a misprint, I will take the correct formulation. 42   Moore, (n 9) 450. 43   GG Leibniz and I Newton, El cálculo infinitesimal (Buenos Aires, EUDEBA, 1977). 39 40 41

356  Horacio Spector Accordingly, we can take the following counterfactual as true: (5)  (Newton’s invention or Leibniz’s invention) caused calculus to be known in the seventeenth century.

If causal predicates in the counterfactual dependence account distributed over disjunction, (5) would be equivalent to: (6)  (Newton’s invention caused calculus to be known in the seventeenth century) or (Leibniz’s invention caused calculus to be known in the seventeenth century.)

But (6) is false because each of its disjuncts is false. In effect, in the closest possible world in which Newton did not invent calculus, calculus is still known (in the seventeenth century) through Leibniz’s discovery, and in the closest possible world in which Leibniz did not invent calculus, calculus is still known through Newton’s discovery. Yet (5) is true under our historical assumption because in the closest possible world in which neither Newton nor Leibniz invented calculus, calculus is unknown (in the seventeenth century). Therefore, we must reject the distributiveness of causal predicates (in the counterfactual dependence sense) over disjunction. As long as causal predicates are non-distributive over disjunction, (3) may be true even if (4) is false. Moore’s argument is unsound because he conjoins De Morgan’s laws with an implicit ‘law of distribution of causal predicates over disjunction’ that is not valid. In terms of our prior example, this means that the disjunction of Amy’s and Martin’s omissions can be a ‘but for’ condition whereas each omission on its own is not a ‘but for’ condition. De Morgan’s laws only say that, given that each disjunct is not a ‘but for’ condition, it cannot be true that Amy’s omission is a ‘but for’ condition or Martin’s omission is a ‘but for’ condition. This is compatible with asserting that Amy’s omission or Martin’s omission is a ‘but for’ condition of the non-operation of the brakes. Yet our intuitions tend to blame Martin’s failure more than Amy’s. I speculate that our intuitions are affected by a confusion of normative and empirical claims. The failure to apply the brakes in the face of a pedestrian may be more blameworthy than the failure to repair the brakes (though the latter is also a grave misdeed).44 Empirically, however, concurring non-preventive omissions symmetrically constitute a disjunctive ‘but for’ condition of an MMTS negative condition. This means that the relevant MMTS negative condition (ie, absence of effective braking) counterfactually depends on Amy and Martin’s disjunctive failure. As is obvious, this is totally different from saying that Amy and Martin’s joint failure was a ‘but for’ condition of the relevant negative condition and of the outcome. In other cases, a non-preventive omission can pre-empt a later non-preventive omission. Consider the following example, proposed by Judith Jarvis Thomson: Bad guards. Two guards must take care of a prisoner in a cell. Sally must give him water; Bert must supply bread. Neither fulfilled, but the prisoner dies from dehydration rather than starvation.45

Unlike Defective non-used brakes, this example involves pre-emption rather than concurrent causation. Indeed, in terms of the relevant G whereas Sally’s omission is a ‘but for’ 44  Wright says that Martin’s failure preempts Amy’s: ‘Causation in Tort Law’ (n 3) 1801; Wright ‘Acts and Omissions’ (n 8) 304. From an empirical, non-normative viewpoint, I see no ground to differentiate both failures, which seem symmetrical to me. David Fischer holds the same view; see DA Fischer, ‘Causation in Fact in Omission Cases’ (1992) Utah Law Review 1335, 1349. The fact that both failures occurred at different stages in the sequence makes no difference in terms of the counterfactual dependence account. 45   J J Thomson, ‘Causation: Omissions’, Philosophy and Phenomenological Research LXVI (2003) 81.

The MMTS Analysis of Causation  357 condition of the MMTS negative condition ‘no water’, Bert’s omission is not a ‘but for’ condition of that negative condition. Bert’s omission is a ‘but for’ condition of the negative condition ‘no bread’, but this condition is not a MMTS condition according to the relevant G. In fact, if we tried to include generalisations relative to starvation in G we could not generate a maximal set E of true empirical propositions. We conclude that in most ordinary scenarios, like Bad captain and Bad nurse, omissions are negative ‘but for’ conditions of negative causal conditions. These are the omissions I called ‘non-preventive omissions’. However, there are scenarios in which an omission can be both a negative ‘but for’ condition and a negative MMTS causal condition of a positive causal condition. Consider the following case: Diabolical treadmill. A war prisoner is cheated into a diabolical ergometric treadmill test. The equipment has been modified in such a way that if he stops running he will turn on an electric chair that will electrocute his mate in the next room. He stops running.

The prisoner’s failing to run on the treadmill is a ‘but for’ condition of the chair turning on because if he had continued running on the treadmill, the electric chair would have remained off. It is also an MMTS causal condition of the electric chair turning on because, relative to the relevant set of causal generalisations, C includes the condition that the prisoner stops running, and E includes the chair setting on as a consequence. Notice that, by transitivity of the causation relation, the prisoner’s omission is a negative MMTS causal condition of the mate’s death as well, because there is a further causal generalisation that links the chair turning on and the mate’s electrocution. In extraordinary cases like this, omissions resemble acts because they are MMTS conditions of positive MMTS causal factors and also MMTS conditions of the relevant outcomes, just as acts generally are. There are other scenarios in which omissions are truth-makers of negative MMTS conditions of the outcome, because C must essentially include an omission according to the relevant set of causal generalisations. Consider this case: Impoliteness. Peter made a scornful remark on accent pronunciation in a social meeting in which Bessie was introducing his new boyfriend from abroad. Peter failed to apologise after the meeting. Bessie got resentful.

Suppose this example takes place in a culture where there are social norms that make true a causal generalisation to the effect that if you humiliate someone in a social meeting and then fail to duly apologise, you cause the humiliated person to be resentful. Failure to apologise is not a non-preventive omission in the sense defined above, even though by failing to apologize you fail to prevent the humiliated person’s resentment. Indeed, nonpreventive omissions are not MMTS conditions of the outcome, but only ‘but for’ conditions of negative conditions, because prevention in cases of non-preventive omissions can generally operate through anyone’s behaviour or through non-behavioural processes. In contrast, in this example the relevant causal generalisation essentially refers to the impolite person’s omission. Someone else’s offering his apology would not have prevented Bessie’s resentment; it would have been inappropriate. We have argued that whereas acts involve MMTS causation both in its agent-­conditioning and ordinary causation varieties, omissions generally are negative ‘but for’ causes of negative MMTS causal factors, except in cases like Diabolical treadmill and Impoliteness, where an omission can be causal in a form similar to acts. Just as omissions can sometimes resemble acts, acts can sometimes resemble omissions. Acts resemble omissions when they are

358  Horacio Spector positive ‘but for’ causes of negative MMTS causal factors. This is true of cases where an agent positively removes an obstacle or impediment (eg, a safety device). These actions are called ‘double preventions’ in the philosophical literature. Double preventions are acts that prevent a preventive causal process thus allowing a certain effect to occur. Consider the following cases: Suicidal husbands. Three husbands in a gated community decide to commit suicide pretending an accident for their wives to be able to claim their life-insurance policies. They turn off the circuit breaker, get into the pool, and throw into the water an electrically operated audio equipment. They die by electrocution.46 Air bombing. Suzy’s mission is to bomb an enemy target. Billy is her lone escort in the mission. At t Billy prevents an enemy pilot from destroying Suzy’s fighter at t+1, thus allowing Suzy’s triggering of the bomb at t+2 and the destruction of the enemy target at t+3. The bombing and destruction would not have happened had Billy not prevented the enemy’s prevention of Suzy’s mission.47

It might be thought that double-preventions (ie, pre-emptive preventions) are causal conditions along with the positive causal conditions of the non-prevented causal process. Thus it might seem that the husbands’ turning off of the circuit breaker caused their electrocution just as much as the throwing of the audio equipment into the water. By the same token, it might seem that Billy’s destroying the enemy fighter at t caused the bombing just as much as Suzy’s triggering. The throwing of the audio equipment into the water and Suzy’s triggering were certainly MMTS conditions with regard to the relevant causal generalisations. Is the turning off of the circuit breaker another MMTS condition of the husbands’ death? Is Billy’s destroying the enemy fighter another MMTS condition of the destruction of the enemy target? Obviously they are not. In effect, there is no causal generalisation that includes in its antecedent the turning off of the circuit breaker and death by electrocution in its consequent, just as there is no causal generalisation that includes Billy’s destroying the enemy fighter in the antecedent, the consequent of which is Suzy’s fighter bombing the enemy target. On the other hand, under the counterfactual dependence account it is clear that the husbands’ turning off the circuit breaker allowed the water in the pool to conduct the electricity, and that Billy’s act allowed Suzy’s fighter being in operation on the occasion of triggering. That is, the husbands’ act (setting off the circuit breaker) and Billy’s act were positive ‘but for’ conditions of negative conditions established in the antecedent of the relevant causal generalisations: ‘non-breaking of the circuit’ and ‘non-destruction prior to bombing’. Had the husbands not turned off the circuit breaker, electricity would not have passed through the water. Had Billy not destroyed the enemy airplane, Suzy’s fighter would have been destroyed. Therefore, the husbands and Billy let the relevant causal processes run (to completion). One process resulted in electrocution, the other in destruction of the enemy target. Yet the husbands’ and Billy’s actions were not MMTS causal conditions of those outcomes. Both omissions and double preventions are ‘but for’ conditions of negative causal factors. However, double preventing acts are positive ‘but for’ conditions of the instantiation of 46   This case is taken from: C Piñeiro, Thursday Night Widows, trans M France (London, Bitter Lemon Press, 2009). Similar cases can be found in the legal literature; see eg Moore, Causation and Responsibility: An Essay in Law, Morals, and Metaphysics 62–63 (turning off of respirator and tying up of lifeguard); M Ferrante, ‘Causation in Criminal Responsibility’ (2008) 11.3 New Criminal Law Review 481 (disabling of scuba diving equipment). 47   Hall, ‘Two Concepts of Causation’ (n 20) 241.

The MMTS Analysis of Causation  359 a negative MMTS condition established in the antecedent of a causal generalisation. The relevant negative MMTS condition would have been negated if the double preventer had not acted so as to block the prevention of the causal process. Like omissions double preventions are ‘but for’ conditions related to MMTS causal conditions. The fundamental difference between omissions and double preventing acts is that the former are negative ‘but for’ conditions, whereas the latter are positive ‘but for’ conditions.48

VII. Conclusion

To understand causal claims in moral and legal contexts we must distinguish between ordin­ary causation by events and agent-conditioning. Causes are events that correspond to some positive MMTS conditions according to the relevant set of causal generalisations. Other MMTS conditions are causal factors in a generic sense. Agent-conditioning is a different matter. Acts are generally forms of agent-conditioning that rely on an MMTS relationship between bodily movements and certain MMTS positive conditions. In contrast, non-preventive omissions are generally ‘but for’ conditions of negative conditions. Other forms of agent-conditioning (v.gr., double preventions) can be analysed in terms of various forms of counterfactual dependence related to either positive or negative conditions in MMTS causal relations. Whereas the MMTS account bears on ‘commissive’ agent-­ conditioning (making things happen), counterfactual dependence is the proper account for ‘omissive’ agent-conditioning (letting things happen).

48  Phil Dowe treats omissions and double preventions (which he calls ‘quasi-causation’) as counterfactual claims about ‘genuine causation’. See P Dowe, ‘A Counterfactual Theory of Prevention and “Causation” by Omission’ (2001) 79 (2) Australasian Journal of Philosophy 217. As said in n 36 above, the view defended in this paper is different from Dowe’s because it analyses omissions and double preventions as ‘but for’ conditions of negative causal conditions rather than as ‘but for’ conditions of causal relations.

17 Causing the Behaviour of Others and Other Causal Mixtures RODERICK BAGSHAW

I. Introduction

This chapter investigates the concept of ‘cause’ which ought to be used by tort lawyers when making claims such as that Derek’s wrongful behaviour ‘caused’ Trevor to act in some way, in particular in circumstances where we regard Trevor’s action as ‘voluntary’ rather than ‘coerced’. The central issue is whether a tort lawyer’s enquiry into whether Derek’s wrongful behaviour ‘caused’ Trevor to act in some way ought to be the same as an enquiry into whether Derek’s wrongful behaviour ‘caused’ the kettle to boil or the toaster to burn the toast. If ‘causing’ others to act ought to be different from ‘causing’ machines to work then this will be significant for at least three reasons. First, establishing that a different concept of ‘cause’ ought to be used when discussing whether one person has ‘caused’ another’s behaviour will be helpful for tort lawyers who have to argue cases where it is claimed, for example, that the defendant caused the soprano to break her contract with the impresario or that the defendant university caused the law student to violate the textbook publisher’s copyright. Secondly, if more than one concept of ‘cause’ ought to be used by tort lawyers then this will challenge the views of some ‘causal minimalists’, who recommend that tort lawyers should use a single, simple – ‘minimal’ – concept of ‘cause’. Thirdly, any difference may help to explain why some tort lawyers think there is something special about sequences of events where the ‘voluntary behaviour’ of someone other than the defendant intervenes between a defendant’s wrongful behaviour and a claimant suffering damage:1 for example, where a defendant wrongfully permits overcrowding on its trains and a passenger is injured after a third party takes advantage of the overcrowding to commit a robbery. The reason why these sequences are special may be related to the way in which ‘causing’ others to act is different from ‘causing’ machines to work.

1   Some lawyers think that a special ‘causal doctrine’ applies to such sequences. See, for example, HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 136: ‘the free, deliberate and informed act or omission of a human being, intended to exploit the situation created by defendant, negatives causal connection’.

362  Roderick Bagshaw

II.  What are Disputes About ‘Cause’ Disputes About?

The first step in debating the ‘concept of cause’ must be to clarify what we want to settle. Imagine that two tort lawyers disagree as to whether ‘causing’ others to act is different from ‘causing’ machines to work: what are they disagreeing about? Most lawyers expressly or impliedly adopt the position that such disputes are about language,2 or about the appropriate use of language within the law, or about elements that are used in thinking that are most easily accessible through analysis of the use of language. But, clearly, there can be different sorts of disputes about elements used in thinking or communicating. For example, we can distinguish between disputes about how some word is actually used by some group of people and disputes about how some word ought to be used in order to meet some demand. This chapter involves an investigation of the second sort of dispute; it discusses how tort lawyers ought to understand ‘cause’ and how they ought to use ‘causal language’. But what demands should influence how tort lawyers ought to understand ‘cause’?

A.  Stapleton on Causal Language Professor Jane Stapleton’s writings provide an essential foundation for any discussion of how tort lawyers ought to understand ‘cause’ and how they ought to use ‘causal language’. In her seminal article ‘Choosing what we mean by “Causation” in the Law’,3 and elsewhere,4 she argues that lawyers should choose to use ‘causal language’ only when reporting an enquiry into whether a specific factor was ‘involved’ in the existence of a particular phenomenon, where ‘involvement’ extends to ‘necessity’, ‘duplicate necessity’ and ‘contribution’. She recommends this choice on the basis that it both serves ‘the wide projects of the law’ and is ‘untainted by normative interrogations and controversies’.5 Choosing instead to call factors ‘causes’ in legal contexts only if they are ‘blameworthy’, ‘explanatory’ or ‘abnormal’ would be inferior because the project of identifying how best to reduce the number of deaths and injuries might be hindered.6 And a choice to stipulate that a factor can only count as a ‘cause’ if it is ‘sufficiently relevant’ would be inferior because an enquiry into ‘involvement’ is ‘objective’ in a way that an enquiry into ‘sufficiency’ cannot be.7 Stapleton’s recommendation means that tort lawyers should cease using ‘causal language’, such as ‘the intervention broke the chain of causation’, to explain a conclusion that a defendant should not be liable for a claimant’s injury because a third party’s voluntary behaviour intervened between the wrongdoing and the injury: such interventions do not negative ‘involvement’, so if they prevent liability it is not because they stop the defendant’s wrongdoing from having ‘caused’ the claimant’s injury. Her primary reason for constraining the use of ‘causal language’ in this way is not to extend a defendant’s liability for any 2   eg Lord Hoffmann, ‘Causation’ (2005) 121 LQR 592, 599 regards propositions about what sorts of things can be ‘causes’, if the law so chooses, as ‘observations about language’. 3  J Stapleton, ‘Choosing what we mean by “Causation” in the Law’ (2008) 73 Missouri Law Review 433 [Choosing]. 4   J Stapleton, ‘Causation in the Law’ in H Beebee, C Hitchcock and P Menzies (eds), The Oxford Handbook of Causation (Oxford, Oxford University Press, 2009) [Handbook]. 5   Handbook 746. 6  Choosing 445. 7   Choosing 455.

Causing the Behaviour of Others and Other Causal Mixtures  363 damage which follows the relevant type of intervention by a third party: rather the primary goal is to ensure that once it has been established that a defendant’s wrongdoing was ‘involved’ in such damage any further limits on the scope of liability are transparently justified. Transparent justification is put at risk by loose talk about ‘causation’, especially by references to intuitions and ‘common sense’ about ‘causation’. By insisting that ‘causal language’ should only refer to something which can be described in terms of ‘objective facts’ Stapleton seeks to ensure that any decision-maker has to give an open explanation for the answers to normative controversies which he or she prefers. At this point it is worth distinguishing between two of Stapleton’s proposals: first, that lawyers should use ‘causal language’ to refer to something that can be identified by interrogation of ‘objective facts’ as opposed to ‘normative evaluation’, and second that lawyers should use ‘causal language’ to refer to ‘involvement’. These should be distinguished because adoption of the second proposal would satisfy the first, but adoption of the first does not require adoption of the second. Indeed, as we have noted above, Stapleton relies on a further argument to support this second proposal: the extent to which the adoption of ‘involvement’ will advance ‘the wide projects of the law’. What if this further argument was rejected? Since tort law seems often to deny that a wrongdoer is responsible for damage which follows intervention by a third party’s voluntary behaviour it cannot be taken for granted that holding that the wrongdoer ‘caused’ the damage advances tort law’s purposes. This suggests that even if we accept Stapleton’s first proposal – that lawyers should use ‘causal language’ to refer to something that can be identified by interrogation of ‘objective facts’ – we may not have a sufficient reason for accepting the second unless we can exclude the possibility of some ‘objective’ concept of ‘causation’ which satisfies the purposes of the law even better than ‘involvement’. There are three strategies that someone who wanted to exclude this possibility might pursue. First, they might insist that the types of interventions which it might suit the purposes of tort law to treat as ‘negativing cause’, or ‘breaking the chain of causation’, cannot be defined in terms of ‘objective facts’: the most plausible formulations incorporate evaluative elements, such as that the third party’s behaviour was ‘wholly unreasonable’.8 Secondly, they might argue that even if some ‘objective’ formulation might serve the special ‘purposes of tort law’ it would serve the purposes of law generally less well than ‘involvement’. For example, it would not be satisfactory to have one concept of ‘causation’ to use when deciding if the defendant’s wrongdoing ‘caused’ the harm in a tort case and another to use when deciding in an administrative law case whether it was rational for the regulator to seek to reduce incidence of the damage by controlling the defendant’s behaviour. Thirdly, they might argue that even if an ‘objective’ formulation that would better serve the purposes of law generally could be found, this formulation should not be presented in ‘causal language’ because it would not be a rule about ‘what causes what’ but about something else, such as ‘who should be held responsible for what’.

8   eg Lord Reid famously stated in McKew v Holland & Hannen & Cubitts (Scotland) Ltd [1969] 3 All ER 1621 (HL), 1623, that ‘if the injured man acts unreasonably he cannot hold the defender liable for injury caused by his own unreasonable conduct. His unreasonable conduct is novus actus interveniens. The chain of causation has been broken and what follows must be regarded as caused by his own conduct and not by the defender’s fault or the disability caused by it’.

364  Roderick Bagshaw

B.  The Relationship between ‘Causal Language’ and the Natural World I intend to set the first two strategies aside because of the difficulty of pursuing either to a conclusion: neither can be taken forward without descending into a debate about the ‘general purposes of the law’. The third strategy is of greater importance in clarifying what disputes about ‘causal language’ are disputes about because it suggests that there is some constraint on what lawyers can choose to call ‘causation’ beyond Stapleton’s two demands: that they should choose something which can be defined in terms of ‘objective facts’ and will serve the purposes of law generally. The additional constraint appears to be some external sense of ‘what causes what’. What is the nature of this external sense of ‘what causes what’? If it is no more than how non-lawyers use ‘causal language’, or the elements they use in thinking about the world, then it does not challenge Stapleton’s approach: the cost of any divergence between what non-lawyers currently mean when they use ‘causal language’ and the benefits of adopting ‘involvement’ as the meaning of ‘cause’ for the future could simply become an item on the balance sheet when deciding if adopting ‘involvement’ will best serve the purposes of law generally. The challenge would be significant, however, if the external sense of ‘what causes what’ refers to some feature or features of natural phenomena which form part of the fabric of the natural world. This sort of challenge might be expanded into something like: ‘you cannot encourage people to use the word “cause” in that way because such a way of using the word is inconsistent with the fabric of the natural world’. The challenge would be similar to that which might be mounted against a proposal that the term ‘types of animal’ should be used in future so as to recognise ‘tadpoles’ and ‘frogs’ as different ‘types of animal’ but ‘caterpillars’ and ‘butterflies’ as the same ‘type of animal’: such a use of language is inconsistent with the fabric of the natural world. The significance of the problem raised in the previous paragraph may be illuminated by thinking about the hypothetical proposal that ‘lawyers should only say that a defendant’s wrongdoing “caused” damage in sequences involving the intervention of a third party when the defendant intended his wrongdoing to lead to the third party acting in the way that he did’. There does not seem to be any reason to believe that incorporation of this proposal into the law would require the investigation of matters other than ‘objective facts’, even if a definition of ‘intention’ might be required. Thus Stapleton’s objection to it would have to take the form of a demonstration that using the word ‘cause’ in this way would serve the purposes of the law less well than using it to mean ‘involvement’. That oppositional strategy can be contrasted with that which might be adopted by a person who believes that the fabric of the universe is such that some things ‘bring about’ other things, and that ‘causation’ must be associated with such ‘bringing about’. Such a person might object that it is an error to include ‘intention’ in any account of ‘causation’ simply because all of the natural phenomena which plausibly exemplify ‘bringing about’ are independent of what individual human beings were setting out to achieve. At this stage it is important to distinguish two views that such a person might have as to the nature of the ‘association’ between ‘causation’ and the natural phenomena which plausibly exemplify ‘bringing about’. The first view is that there exists some single natural relation between events, or facts, or some other things, which is ‘causation’. The second view is there are several, distinguishable, natural processes that can plausibly be described as involving ‘one thing bringing about another’ and different views can be held as to which of

Causing the Behaviour of Others and Other Causal Mixtures  365 these ought to be called ‘one thing causing another’. Someone holding this second view might, for example, accept as true descriptions of natural phenomena the statements that ‘when the first billiard ball collided with the second billiard ball there was a transfer of energy and the second ball started to move’, ‘after the dirty scalpel pierced the claimant’s skin there was a transfer of bacteria from the blade into the claimant’s body and the bacteria started to reproduce within the claimant’s body’ and ‘after the defendant noticed the attractive blonde at the next table he transferred his attention and never completed the crossword’. Moreover, they could accept all three statements as examples of ‘causation’ without believing that there is anything which unifies the three types of transfer. To generalise, the person might believe that causal terminology can be used to refer to a range of related natural relations (including ‘source of transferred energy’ and ‘source of transferred bacteria’, and perhaps also ‘reason for transferred attention’) rather than some single specific natural relation. Someone who holds this second view, which I will call ‘pluralist’, can happily admit that there is no single natural relation which is ‘causation’, and need not search for some natural feature which unifies how heating one end of a metal bar ‘causes’ a rise in the temperature of the other end and how telling a wife about her husband’s misdeeds ‘causes’ her to crash her car. To put the matter more informally, a pluralist might believe that the natural relations which can be referred to using ‘causal language’ are related in ways similar to the members of an extended family rather than a class of clones. A pluralist can nonetheless object to any proposal as to how ‘causal language’ should be used which is inconsistent with what we know about the relevant natural processes. Thus a pluralist would object to a proposal that the word ‘cause’ should be used, so it is true to say both ‘by rolling the white ball so that it collided with the stationary black ball Albert caused the black ball to accelerate’ and ‘by rolling the white ball so that it collided with the fastmoving black ball Albert did not cause the black ball to slow down’. And, importantly, the objection would not depend on a demonstration that such a distinction would impede achievement of the goals of the law. Any adherent to the first view, that there is some single natural relation which is ‘causation’, must face the awkward facts that current patterns of usage of ‘causal language’ do not correspond to any single natural relation and that generations of philosophers have been unable to define any such relation. Stapleton is not alone in having concluded that a ‘broad and non-discriminatory concept’ of cause is ‘a philosopher’s myth’.9 Nonetheless, some tort scholars sometimes express themselves as if they believe that a single natural relation exists. Beever, for example, has written that: On the face of it, then, to say that causation is, even in part, a normative or a policy matter is to imply that our judgments or preferences for deciding liability determine the fundamental nature of the universe; as if, were human beings not to exist, or were even just law to be abolished, the fundamental nature of the universe would change. This is, of course, nonsense.10

This passage seems to assume that ‘what causes what’ is part of the ‘fundamental nature of the universe’. But one can accept the truth that how lawyers use language does not alter the ‘laws of physics’ without accepting that there is some single natural relation which is ‘causation’. A pluralist can claim that it is possible to describe the fundamental nature of the universe without employing a concept of ‘causation’ at all, and make only the more moderate demand that if a concept of ‘causation’ is used by lawyers then it should not be designed so   Choosing 439.   A Beever, Rediscovering the Law of Negligence (Oxford, Hart Publishing, 2007) 413.

9 10

366  Roderick Bagshaw as to be inconsistent with what we know about the natural processes that can plausibly be described as involving ‘one thing bringing about another’.

C.  Summary: the Three Demands At the start of this essay I declared that it would consider how tort lawyers ought to understand ‘cause’, or use ‘causal language’. I then identified two demands made by Stapleton, in support of her proposal that in law ‘cause’ should be understood as ‘involvement’: that the legal concept of ‘cause’ should be one that can be described wholly in terms of ‘objective facts’ and that it should serve the purposes of the law generally. I have not rejected either of these demands, but have added a third: that the legal concept of ‘cause’ should not be inconsistent with what we know about the natural processes that can plausibly be described as involving ‘one thing bringing about another’. I have also described a ‘pluralist’ view of these processes, which happens to be the view I regard as most plausible but will not defend in detail.

III.  What is Special About ‘Causing’ Someone to Act? (Part 1)

Should lawyers treat ‘causing’ another person to act as different from ‘causing’ a machine to work? Hart and Honoré state that when lawyers say that Derek ‘caused’ Trevor to behave in some way they use ‘cause’ in a sense which is different from ‘the central type of causation of physical events’.11 Where Derek may have caused Trevor to behave in a particular way by providing reasons12 for behaving in that way Hart and Honoré identify the relevant causal enquiry as asking about matters such as: (i) whether Trevor knew of and understood what Derek said or did; (ii) whether what Derek said or did formed at least part of Trevor’s reasons for behaving in the relevant way; (iii) whether Trevor formed the intention to behave in the relevant way only after Derek’s words or actions; (iv) except in cases of advice, whether Derek intended Trevor to behave in the relevant way.13 This fourfold formulation does not appear to violate Stapleton’s demand that the legal concept of ‘cause’ should be one that can be described wholly in terms of ‘objective facts’: the relevant matters, Trevor’s knowledge, understanding and reasons for acting, and Derek’s intention, are unlikely to require ‘normative evaluation’ as opposed to interrogation of ‘objective’ facts. What, then, could be the objection to the proposal that this non-central, but still objective, ‘concept of causation’ should be used by lawyers when considering such sequences involving the ‘causing’ of another’s behaviour?

  HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 51–52.   We may doubt if there is any essential difference between causing behaviour by providing reasons for it and causing it by providing opportunities. Surely the golden opportunity provided by the handbag left on the passenger seat of the unlocked car provides the thief with a reason for ‘stealing the handbag’ rather than spending the same time seeking a gain through some other project, or settling for an evening without enrichment? By contrast, causing behaviour by diminishing a person’s capacity to reason, for instance by intoxicating him, provoking him or terrifying him, may involve a significantly different mechanism. Cf. HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 59–60. 13   HLA Hart and T Honoré Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 53. 11 12

Causing the Behaviour of Others and Other Causal Mixtures  367 One objection might be that the method which Hart and Honoré used to identify this concept was inferior to a transparent debate as to what formulation should be chosen in order to best serve the law’s purposes. Such an objection could not, of course, rule out in advance the possibility that the law’s purposes might justify using a special concept for this type of sequence. Rather the objection insists that evidence of common usage of such a special concept provides an insufficiently-tested justification for its incorporation in the law. A second objection might be that some elements in Hart and Honoré’s formulation are inconsistent with what we know about the natural processes that can plausibly be described as involving ‘one thing bringing about another’. For example, it might be argued that what we know about how one person’s utterances might bring about another person’s action suggests that no role is played by the state of mind of the person making the utterance: whether Derek’s words ‘brought about’ Trevor’s behaviour does not depend on whether Derek intended to influence Trevor or merely spoke words from a script without understanding their meaning. A third objection might be that a special non-central type of causal concept is simply unnecessary since ordinary ‘involvement’ will deal with this type of sequence perfectly adequately. Having noted these possible objections we should turn to Hart and Honoré’s reasons for distinguishing between the relationship involved when one person ‘causes’ another’s actions by providing reasons for them and ‘the central type of causation of physical events’. Their primary argument seems to be that a claim that some act or statement of Derek provided one of the reasons why Trevor acted in the way which he did on the particular occasion will not necessarily imply any general claim to the effect that such an act or statement would usually be identified by him as a reason for such behaviour, or any claim that the other factors that existed would have been insufficient to persuade him to behave in this way if Derek’s act or statement was absent. For example, Trevor may honestly report that what caused him to choose the lobster was Derek being half an hour late for their date,14 but need not suggest that he has detected any regular relationship between Derek’s punctuality and his selections from restaurant menus, still less a consistent pattern of responses, and need not deny that he usually chooses lobster in this restaurant (for other reasons which may also all have been available on the relevant occasion) when Derek is punctual. One way of summarising the insight on which this argument is based might be that decision-makers like Trevor experience what ‘brings about’ their decisions, and reports of these experiences are not reports of what might have been expected to happen in some not dissimilar world where Derek turned up on time. It may be useful to highlight three features of the type of sequence under discussion. First, the best evidence of whether Derek’s act or statement influenced Trevor will be Trevor’s honest account of what he allowed to influence his decision. Secondly, Trevor’s honest account of what he allowed to influence his decision can be an account of some sort of direct experience and need not make reference to any generalisation. Thirdly, any indirect evidence of the influence of a particular factor on Trevor’s choice as to how he acted will depend on the assumption that Trevor’s ‘pattern of identified factors to choices made’ remains sufficiently constant through time or Trevor’s ‘pattern of identified factors to 14   This example, evidently, does not satisfy the conditions that Hart and Honoré set for a legally correct claim that ‘Derek caused Trevor to choose the lobster’.

368  Roderick Bagshaw choices made’ is similar to the patterns exhibited by other human beings. Clearly such assumptions are less robust when we are dealing with individual human beings on specific occasions than when we are dealing with inanimate systems: it is safer to assume that a rubber ball dropped from a particular height onto a surface with particular properties on Monday will bounce to the same height as the same ball dropped onto the same surface from the same height on Friday, and that a rubber ball from the same production line subjected to the same processes will also bounce to the same height, than to assume that Trevor will behave in the same way on different occasions when he hears a particular utterance, or that Trevor will behave in the same way as Theresa behaved, or as a reasonable person might have been expected to behave. How far might these three features explain the need for a non-central type of enquiry? There seems little doubt that the first and second features suggest that an enquiry into whether Derek’s act or statement was ‘a necessary element for the sufficiency of a sufficient set’ (NESS) risks being under-inclusive. Consider a case where Trevor decides to end his engagement to Charlotte, and before he made this decision Derek carelessly told him that Charlotte’s father was very poor whilst he is actually wealthy. Surely Trevor can report that his misunderstanding of Charlotte’s father’s wealth was one of the factors that ‘caused’ him to end the engagement even if he is unable to divide up the other matters which he took into account in such a way as to show that the misunderstanding was a necessary element in some sufficient set. What Trevor would be reporting would be something like ‘his experience of adding his understanding of Charlotte’s father’s wealth to the catalogue of reasons which collectively proved sufficient’, rather than a contrast with what he believes he would have decided had he known the truth.15 But this risk of under-inclusion does not seem as if it will explain Hart and Honoré’s introduction into the non-central enquiry of a reference to Derek’s intention.

A.  A Problem of Proof? Could there be a further problem, then, that in the absence of evidence of direct experience from Trevor, or in circumstances where Trevor’s interest in the outcome of the enquiry is such that we are reluctant to rely on his uncorroborated testimony, use of the standard enquiry (NESS) may be stymied by a lack of reliable pattern evidence (the third feature)? In other words, if Trevor does not tell us why he chose the lobster then will we regularly be left unable to determine whether Derek’s lateness caused this choice? Lawyers are familiar with ‘proof rules’, such as the doctrine of res ipsa loquitur, which apply when a key fact is difficult to prove directly. Such rules manage the risk of inaccurate conclusions being reached because of the difficultly of direct proof by allowing a party to succeed by only proving some substitute fact, or combination of substitute facts. If there might be a problem with proving indirectly which factors influenced Trevor’s choice then perhaps the ‘non-central’ enquiry is a similar sort of ‘proof rule’? To spell out this suggestion, it is that Hart and Honoré’s non-central test conforms to the pattern of allowing a litigant to succeed where he or she proves the possibility of causation (for example, Trevor knew of Derek’s words before he decided how to act) and an additional condition that 15   As we will see below, Stapleton’s explanation of ‘involvement’ seeks to supplement NESS by a process called ‘disaggregation’ so as to allow it to cope with such situations. Some difficulties with ‘disaggregation’ will be raised below.

Causing the Behaviour of Others and Other Causal Mixtures  369 makes it fair to resolve any doubts against Derek16 despite the absence of reliable direct evid­ence. In this situation the additional condition would be ‘Derek’s intention to persuade Trevor to act’: and proof of this would be what makes it fair to allocate to him the burden of establishing that his words did not have such an effect. Someone who doubts that evidential difficulties are a sufficient reason for using a different concept of ‘cause’ might point out that decisions about whether one thing caused another in wholly inanimate systems also commonly rely on imperfect and potentially unreliable evidence, and in many cases which involve involuntary behaviour courts will be similarly dependent on reports of consciousness by those interested in the outcome: for example, a court may be dependent on a claimant’s report that he was ‘temporarily blinded’ by the undipped headlights and that he would have had time to brake if he had been able to see the obstruction. But, of course, this example may only suggest that alternative causal enquiries based on the possibility of causation and an additional condition making it fair to resolve any doubts in a particular direction may be appropriate in a wider range of cases than those involving ‘causing’ others to act.

B.  The Importance of Trevor’s Choice Despite the fact that it might be possible to defend Hart and Honoré’s non-central test as a ‘proof rule’ there is little doubt that they believed that sequences of the first type are more radically different17 than merely raising problems with the availability and distribution of reliable evidence. Their explanation of the need for a special enquiry also emphasises a further feature: that Trevor chooses whether some act or statement of Derek will influence his behaviour, at least where that behaviour follows a voluntary decision.18 The necessity of Trevor’s choice19 means that Derek’s words or act will always be insufficient to bring about the consequence without the additional subsequent contribution of Trevor. Moreover, since the choice of Trevor will qualify as a cause using the standard enquiry perhaps it is the mixture of potential causes which explains why the law requires a special enquiry.

16   This suggests that if the question ‘did Derek’s words cause Trevor’s voluntary action’ arose in a case where Derek was not a litigant, for example where Trevor sued a newspaper for publishing the defamatory suggestion that his action had been ‘caused’ by Derek’s words and the newspaper sought to show that its story was true, then we would not regard Derek’s intention as important. 17   HLA Hart and T Honoré Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 52, refers to ‘the radical differences which separate “He induced me to do it” from “His blow caused the victim’s death” ’ (emphasis in the original). 18   HLA Hart and T Honoré Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 53 fn 23 suggests that the crucial stage which differs from ‘central causation’ may be between Trevor’s coming to understand what Derek has said and Trevor’s acting because of what he has understood or believed. They say that even if the first stage (coming to understand) is a causal connection of the familiar kind, the second stage (acting because of what he has understood or believed) is not. 19   There may be scope to debate whether Trevor’s choice is a natural feature of ‘voluntary’ decision-making or a pivotal feature in a model of decision-making which is supported for reasons other than its consistency with natural data.

370  Roderick Bagshaw

IV. Causal Mixtures

A.  Stapleton’s Approach to Causal Mixtures How does Stapleton’s specific recommendation as to how lawyers should use ‘causal language’ cope with mixtures of ‘causes’? It will be recalled that her recommendation is that lawyers should choose to use causal language only when reporting the result of an enquiry into whether a specific factor was ‘involved’ in the existence of a particular phenomenon, where ‘involvement’ extends to ‘necessity’, ‘duplicate necessity’ and ‘contribution’. So this enquiry would require us to ask whether Derek’s misinformation was involved in Trevor jilting Charlotte – that is, whether Derek’s misinformation was necessary to Trevor’s decision, or would have been necessary if a duplicate cause had not co-existed, or at least made a contribution to Trevor’s decision. Unless Trevor is unusually mercenary we might expect that, at most, the error as to Charlotte’s father’s wealth was one reason among several supporting his decision to jilt Charlotte. Moreover, as we have already noted, Derek’s misinformation can only ‘cause’ a decision by Trevor if Trevor allows it to carry some weight. In order to decide whether Derek’s misinformation was ‘involved’ in Trevor’s jilting of Charlotte where he had several possible reasons and was in control over which he allowed to influence him we must elucidate Stapleton’s notion of ‘causal contribution’. To do this it will be helpful to begin by considering what counts as ‘causal contribution’ within a physical system.20

B.  Clarifying ‘Causal Contribution’ in a Physical System: the Problem of ‘Disaggregation’ Stapleton recommends that lawyers ought to choose to say that a factor ‘causes’ an injury by way of ‘contribution’ where it is ‘a necessary element for the sufficiency of a sufficient set’ of factors once any excess has been ‘disaggregated’ and left out of account.21 A common example of such ‘causal contribution’ concerns a lake containing fish which will die if the concentration of some chemical in the water exceeds 9 units: A, B and C then simultaneously pollute the water with 8 units, 6 units and 4 units, respectively, and the fish die from chemical poisoning. In this example the ‘disaggregation’ technique identifies all three defendants (A, B and C) as having ‘caused’ the death of the fish by ‘contribution’, on the basis that the amount of chemical each added would have been ‘a necessary element for the sufficiency of a sufficient set’ if we imagined away the excess chemical. One of the reasons Stapleton permits ‘disaggregation’ in identifying ‘causal contribution’ is because she believes it suits the policy of the law to use a relatively inclusive concept of ‘cause’. But even if the legal concept of ‘cause’ ought to be relatively undemanding, it is far from clear that there should be no limits on what types of ‘disaggregation’ are possible. For example, should ‘disaggregation’ be used to identify a factor as a cause of a consequence which occurred before the factor was operative? If A and B added their poison on Wednesday, 20   The shorthand ‘physical system’ is intended to refer to any system where no role is played by mental processes, such as ‘choosing to give weight to a particular factor’. 21   Choosing 476.

Causing the Behaviour of Others and Other Causal Mixtures  371 killing the fish instantly, can C’s addition of 4 further units on Friday be held to have made a ‘contribution’ to the death of the fish by imagining away Wednesday’s excess? Anyone who is inclined to answer this question ‘No, C’s wrong cannot be held to have made a contribution in these circumstances’, must, of course, go on to ask themselves whether they derive their answer from some opinion about what might best serve the policy of the law. The alternative, of course, is that their refusal to identify C’s actions as a ‘cause’ of a pre-existing state of affairs instead tracks our extra-legal understanding of the natural processes by which one thing can ‘bring about’ another: we do not know of any natural mechanism by which what happens at time t+1 can ‘bring about’ what already occurred at time t. The demand of consistency with what we know about natural processes suggests that the ‘disaggregation’ technique must be confined so that C’s actions cannot be described as a ‘cause’ of the death of fish which were already dead. This demand can be made regardless of whether a contrary conclusion (C did ‘cause’ the death of the fish) would help the law to regulate pollution.22 In fact the limits on permissible ‘disaggregation’ may be much more extensive than a prohibition on identifying an action as having ‘caused’ something which pre-dated it. Another useful example, discussed by Stapleton, involves a bridge which is capable of supporting 25 tonnes collapsing at noon whilst a 10-tonne train crosses it, after X has hung 2 tonnes on the bridge at 9 am and Y (unconnected to X) has hung 16 tonnes on it at 10 am. Her conclusion is: There can be no doubt about X’s involvement in the collapse of the bridge: neither the order in which the weights were hung below the bridge nor their state of aggregation at these times are of physical relevance to that collapse.23

One worry about this is the suggestion by Stapleton in her conclusion that X’s act is ‘involved’ in the collapse by ‘contribution’ when Y hung the 16 tonnes before X acted: this seems inconsistent with the key English case of Barnett v Chelsea and Kensington Hospital Management Committee.24 In this case, simplifying the facts somewhat, a doctor negligently failed to examine a workman who presented himself at a hospital casualty department complaining of vomiting. The workman later died from arsenic poisoning. It was held, however, that the doctor was not liable because at the time of the doctor’s negligent failure the workman had already ingested too much arsenic to be saved, given the time that it would have taken a competent doctor to identify the problem and administer an antidote. Can the doctor’s role (not a ‘cause’) in this case be distinguished from X’s role (supposedly ‘involved’ and therefore a ‘cause’) in the collapse of the bridge when Y had already hung 16 tonnes? If Y’s 16 tonnes can be ‘disaggregated’ then why can the doctor’s failure not be identified as causing the death by ‘contribution’ by ‘disaggregating’ the amount of arsenic ingested by the workman so that it was not quite enough to kill him before a reasonably competent doctor administered the antidote? One possible distinction25 might rely on the fact that in Barnett the ingestion of the large quantity of arsenic doomed the workman is some way in which the bridge was not doomed by Y’s 16 tonnes. In the bridge case someone could have spotted the weights placed by Y and 22   It must be doubtful whether anyone would defend using ‘disaggregation’ without a temporal limit in order to identify a cause: anyone who did defend such an approach might end up claiming that by means of ‘disaggregation’ he can show that C ‘caused’ the death of the dinosaurs (as well as the death of the fish). 23  Choosing 475. 24   [1969] 1 QB 428. 25   Raised by Professor Stapleton in discussing this issue with her.

372  Roderick Bagshaw removed them, or someone could have stopped the train, whilst in Barnett we may suppose that by the time that the workman presented himself at the casualty department there was no possibility of any intervention that could have saved him. But this provisional distinction raises further difficulties. First, it seems inconsistent with what Stapleton writes about the case of Hotson v East Berkshire Health Authority.26 Secondly, and more importantly, there is the problem of what ‘no possibility’ might mean. It seems that in Barnett it would not have been impossible to save the workman, but it would have required the doctor to administer the appropriate antidote immediately on the basis of a highly improbable conjecture rather than a confirmatory test. Suppose that we vary the facts further and add that an instant test kit has been developed but is so expensive that it is not used in any casualty department in the United Kingdom. Would any lawyer regard the type of ‘possibility’ which existed in Barnett, or that raised by this new test, as sufficient to lead to a conclusion that the doctor’s negligent failure did ‘cause’ the workman’s death? But, if we conclude from this that ‘no possibility’ must then mean something like ‘the possibility of the claimant avoiding the injury was negligible in the circumstances obtaining at the time’ then, of course, we must revisit the opinion that a contribution can be made by hanging an additional 2 tonnes onto a bridge which already has 16 tonnes hanging on it and a 10 ton train travelling towards it: there must be many sets of circumstances where the 16 tonnes and the oncoming train would make the collapse extremely likely even before the 2 tonnes are added. There are also further reasons for doubting whether the ‘doomed exception’ is a sufficient antidote to the problems with ‘disaggregation’ raised by cases such as Barnett and Hotson. For example, if we vary the facts of Barnett so that the claimant has ingested a vast (and inevitably fatal) quantity of paracetemol, and the doctor’s negligent mis-examination leads to the recommendation that the claimant takes two further paracetemol tablets, then in such circumstances, even if the additional tablets did not accelerate the claimant’s death, it seems less obvious than in Barnett that the doctor’s negligence should be discarded as ‘not contributing’. But could any coherent concept of ‘cause’ treat ‘adding two paracetemols to the mix’ as ‘contributing’ but ‘failing to remove two paracetemols from the mix’ as ‘not-contributing’? It may be useful at this stage to introduce a further group of difficult cases: those where a factor is added to a mixture but the factor’s potential effect depended on a mechanism which did not operate. For example, if we revert to a ‘poisoned fish’ hypothetical, imagine that 10 units of chemical x or 3 units of chemical y will kill the fish, A and B simultaneously add 8 and 5 units of chemical x, respectively, whilst C and D simultaneously add 2 units each of chemical y, but chemical x also has the further property that it neutralises chemical y and renders it harmless for fish. In this new scenario can we say that C and D ‘contributed’ to the death of the fish, even though the chemical they added was fortuitously neutralised, by ‘disaggregating’ the chemical x which made the chemical y innocuous? If the answer to this question is ‘no’, as I believe that it should be,27 then this demonstrates that ‘causal 26   [1987] AC 750 (HL). Discussed in Choosing 475, fn 148. In this case, again simplifying, doctors negligently failed to administer appropriate treatment to a claimant at a time when it was possible, but not probable, that enough blood vessels remained intact to allow him to avoid permanent disability. In the footnote Professor Stapleton contemplates that it might be possible to hold that the failure ‘contributed’ to the permanent disability by disaggregating the injuries that the claimant was suffering from at the time he reached the casualty department. This, however, ignores the fact that at that time he was probably suffering from injuries that would doom him to permanent disability. 27   This belief derives from my understanding of the third demand: the natural mechanism which killed the fish was poisoning by chemical x, and if ‘disaggregation’ was used to claim that chemical y ‘contributed’ to their death then this would be inconsistent with what we know about the natural mechanism.

Causing the Behaviour of Others and Other Causal Mixtures  373 contribution’ requires a far more detailed specification than Stapleton has provided: most likely some additional condition such as ‘disaggregation cannot be used to contradict what we know about the natural processes by which one thing can “bring about” another.’ My claim, then, is that we know that chemical y did not contribute to the death of the fish, and this conclusion does not depend on any view as to optimal legal policy.28

C.  The Natural Process Account of ‘Causal Contribution’ So far I have argued that a satisfactory account of ‘cause by contribution’ will have to include guidance as to when ‘disaggregation’ is impermissible. It may be the case, however, that the best account of ‘cause by contribution’ will not rely on NESS being established following an imaginary ‘disaggregation’, provided that such a ‘disaggregation’ does not lead to a conclusion which contradicts what we know about natural processes. Instead, the best account may appeal directly to what we know about natural processes in order to ‘define’ ‘cause by contribution’. Thus if we revert to the hypothetical case where A, B and C simultaneously pollute water with 8 units, 6 units and 4 units, respectively, of a chemical when 10 units will be sufficient to kill the fish, Stapleton explains the conclusion that C’s wrong contributed to the death of the fish as flowing from C’s wrong being a ‘necessary element for the sufficiency of a sufficient set’ of factors once the pollution caused by A’s and B’s spills is ‘disaggregated’ and the 8 units of excess poison left aside. By contrast, I suggest that the same conclusion can be drawn from the fact that no plausible account of the natural processes involved (how the chemicals came to kill the fish) could present the death of the fish as involving only molecules of poison derived from A’s and B’s spills.29 If we suppose, for example, that the poison operates by being absorbed by the skin of the fish and then destroying cells until the creature dies then it seems wholly implausible that only A’s and B’s poison will have killed cells. Similarly, if a bridge is pulled down by an extra 18 tonnes having been hung beneath it before a 10-tonne train crosses then it would be implausible to give an account suggesting that the unsustainable strain was put on the concrete by only the first 16 tonnes, unless, perhaps the weights were distributed in an unusual manner. By contrast, none of the molecules of arsenic implicated in killing the workman were molecules that the negligent doctor ought to have removed before the workman’s death: only molecules that would have still been in his body if he had received competent medical treatment were involved in killing the workman.30 My suggestion, then, is that the forms of involvement such as ‘being the immediate source of some of the transferred energy which explains the body’s subsequent motion’, ‘being the immediate source of some of the transferred bacteria which explain the wound’s subsequent infection’, ‘being one of the immediate sources of the poison which explains the 28   The best evidence for the extra-legal explanation is that most people seem willing to express an opinion on whether chemical y ‘caused’ the death of the fish without any overt consideration of legal policy. 29   Thus I would draw a different conclusion where the claimant’s dog died after lapping water from the polluted pond and there was evidence that the chemical was so toxic that the first molecule of it that any dog ingested would be immediately fatal. In such a case I am not convinced that A, B and C should each be held to have ‘contributed’ to the death of the claimant’s dog. Whether in such circumstances there should be liability in the absence of causation is, of course, a different question and one which received a positive answer from the House of Lords in Fairchild v Glenhaven [2002] UKHL 22, [2003] 1 AC 32. 30   Hence it may be possible to distinguish the case, raised above, where the negligent doctor prescribes additional paracetemol to a victim already suffering from a paracetemol overdose. Omissions are discussed further below.

374  Roderick Bagshaw subsequent death’ and ‘being one of the immediate sources of the destructive force which explains the subsequent collapse’ are forms of involvement which can be called ‘causal’, and the ‘disaggregation’ technique is no more than an imperfect attempt to describe a single process for identifying these and other forms of ‘playing an active role in an operative mechanism’. To be clear, I do not claim that ‘playing an active role in an operative mechan­ ism’, or ‘active involvement’ is any more than an accommodating label for several relevant forms of relation between events. To echo a previous metaphor, my claim is no more than that the examples of ‘playing an active role in an operative mechanism’ are related like members of an ‘extended family’, rather than as clones.

D.  Objections to the Natural Process Account Two objections must be faced. First, suppose someone objects that describing contribution in this way relies on an illusion, and that all the examples actually involve micro-­acceleration of the injury: if there is more poison in the pond and it operates by being absorbed through the skin then the fish will die more quickly, and similarly the weaker bridge will collapse more quickly, even if the difference in timing is too small to detect using ordinary methods. My answer to this is that even if such an objection can be raised against the fish and bridge examples it is not one which is always available. For example, we could easily modify the fish example so that no fish is capable of absorbing more than a fixed amount of poison per minute no matter what the concentration of the poison in the pond is, where the amount spilled into the pond by the first polluter would have been sufficient to achieve this maximum absorption rate, and, as a result the second defendant’s contribution does not accelerate the absorption or death, though it continues to contribute some of the molecules which in fact are actively involved in destroying the cells in each of the fish. Alternatively, we could imagine a case where to avoid severe injury the claimant must not drink any alcohol, the first defendant adds 10 ml of flavourless alcohol to 100 ml of water in a glass which is to be served to the claimant, and subsequently the second defendant makes a mixture of identical strength, pours away 55ml from the glass and replaces it with 55ml of his mixture. Again, the second defendant’s alcohol molecules will in fact be ingested but this will not lead to any acceleration. In both the modified fish example and the alcohol example I am happy to claim that the second defendant ‘contributes’ to the injury despite the absence of acceleration. The second objection is that the ‘playing an active role in an operative mechanism’ account cannot deal with omissions in a satisfactory way. My response to this is to concede that the objection is valid. But the reason why it is valid is important: omissions simply do not play an active role in operative mechanisms. If, for example, a person ingests a poison and a competent doctor would have administered an antidote long before this poison could cause harm, then even if an incompetent doctor does nothing the natural mechanism which is active in ‘causing’ the harm is the poison alone. The natural account of what caused the death remains the same whether the poisoned man is in the care of an incompetent doctor or on a desert island where there are no medical facilities. When lawyers talk about omissions as ‘causes’ they are generally describing some sort of vicarious physics: they say, for example, that the incompetent doctor’s omission ‘caused’ (vicarious sense) everything which was ‘caused’ (natural sense) by the molecules of poison the doctor should have neutralised.

Causing the Behaviour of Others and Other Causal Mixtures  375 Thus the second objection is valid: a natural mechanism account of ‘contribution’ cannot deal with omissions in a satisfactory way. Ironically, however, the difficulties which vicarious physics faces in cases involving mixtures of omissions may help to demonstrate the utility of the ‘active role’ approach. For example, if we consider the famous hypothetical case where a car crashes into the back of a stationary lorry in circumstances where the driver did not attempt to apply the brakes and a car mechanic had negligently failed to reconnect the brakes after working on the car, then the natural account of the mechanisms involved in the driver’s injury is straightforward: the driver was injured as a result of colliding with a stationary object whilst in a car travelling forward with a particular momentum, etc. In the realm of vicarious physics it is possible to find some people who think that a car with no brakes is doomed and the driver’s failure makes no contribution, and others who think that the absence of connected brakes can only contribute if the brake pedal is pushed so the car mechanic made no contribution, and still others who think that ‘disaggregation’ allows us to conclude that both omissions ‘contributed’. For the purposes of this essay the justification for raising this hypothetical is not to defend any particular outcome, but to suggest that the reason that the dispute cannot be settled is that no answer can be found in what we know about the natural mechanisms by which one thing can bring about another.

E.  Summary: The Natural Process Account I will not attempt to provide a fuller defence of the natural process, or mechanism-based, account of ‘causing by contribution’. To summarise the discussion of causal mixtures so far we should start by recalling that the third demand of any legal concept of ‘cause’ is that it should not be inconsistent with what we know about the natural processes that can plausibly be described as involving ‘one thing bringing about another’. I have argued that an account of ‘causal contribution’ which permits ‘disaggregation’ without limits is likely to disappoint that demand, and have drawn attention to some of the difficulties which might face an attempt to make up for this failing by defining limits on when ‘disaggregation’ is permissible. Such obstacles in the way of a satisfactory ‘disaggregation’ account provide support for a rival account which ‘defines’ ‘causal contribution’ as ‘playing an active role in an operative mechanism’. It should be borne in mind, however that this phrase is no more than a suggestive label for a range of elements in a range of natural mechanisms related to one another in ways similar to the members of an ‘extended family’.

V.  What is Special About ‘Causing’ Someone to Act? (Part 2)

A.  Contribution to Decisions Can this rival account of ‘causing by contribution’ help to elucidate what is special about ‘causing’ someone to act? The clarification of ‘contribution’ in physical systems suggested that in order to decide whether some factor had ‘caused’ damage ‘by contribution’ we need to understand how weights destroy bridges and how molecules of poison kill fish in order to determine whether a particular weight or some particular molecules of poison were ‘actively involved’ in the mechanism. Thus to pursue the same enquiry with respect to

376  Roderick Bagshaw human decision-making we ideally need to understand how one person’s words or actions can play a role in another person’s decisions to act in a particular way. As we noted above, Hart and Honoré’s account emphasises that a factor can only influence a voluntary decision if the decision-maker allows it to ‘have some weight’ as a reason which supports how he or she decides to act. Thus where there is a mixture of potential reasons available to a decision-maker, like Trevor when he is considering whether to jilt Charlotte, we may only be able to determine whether the misinformation that he was given by Derek ‘caused’ his decision (by ‘contribution’) if Trevor tells us. And, of course, kettles and toasters do not have to tell us whether they allowed the pressing of the switch to have an effect: the natural mechanisms which are operative do not include anything equivalent to these choices. Has the previous paragraph settled that ‘causing’ someone to act ought to be treated by tort lawyers as different to ‘causing’ a machine to work? My answer is ‘yes’. But that answer should be qualified by adding that it follows from the general suggestion that a satisfactory account of ‘cause’ ought to be consistent with what we know about ‘natural mechanisms’: we know that human decision-making involves different mechanisms from those incor­ porated in kettles and toasters.

VI. Explaining Intervention Doctrines

How far can the special sense of cause that we use when deciding if one person ‘caused’ another’s behaviour explain the legal doctrine that ‘the free, deliberate and informed act or omission of a human being, intended to exploit the situation created by the defendant, negatives causal connection’?31 I propose to approach this question indirectly, by first considering whether some non-causal reason can explain why tort lawyers deny liability in these circumstances. What non-causal reasons must be assessed?

A. Non-foreseeability One explanation might be that such an intervention may have the effect of making the type of injury suffered by the claimant not reasonably foreseeable.32 But Hart and Honoré insisted that their doctrine was not simply one which identified a sub-category of cases where the eventual injury was ‘unforeseeable’. Moreover, there is no shortage of judicial support for interventions being capable of ‘breaking the chain of causation’ even when they are reasonably foreseeable.33 No doubt some interventions will make the type of injury suffered by the claimant unforeseeable, but foreseeability cannot explain the doctrine on its own.

31   This formulation of the doctrine is found in HLA Hart and T Honoré, Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 136. In Reeves v Commissioner of Police of the Metropolis [2000] AC 360 (HL) Lord Hoffmann described it as an ‘undoubted rule’, but one which is subject to exceptions. 32   Clearly the fact that the precise way in which a third party behaved was not reasonably foreseeable will not in itself make the ‘type of injury’ one which is not reasonably foreseeable. See eg, Philco Radio & Television v J Spurling [1949] 2 All ER 882. 33   A classic example is provided by McKew v Holland & Hannen & Cubitts (Scotland) Ltd [1969] 3 All ER 1621 (HL), 1623.

Causing the Behaviour of Others and Other Causal Mixtures  377

B. Coincidence A second explanation might be the unwillingness of tort lawyers to impose liability for ‘coincidental’ injuries. This explanation is generally thought to ensure that if a taxi driver carelessly drives at a dangerous speed, and as a result reaches and stops at a particular junction just at the moment that a carelessly driven bus ploughs through it colliding with the taxi and injuring its passenger, the taxi driver is not liable for the passenger’s injuries. The precise specification of ‘the coincidence rule’ in current English law is a matter of controversy, but it is probably something like: ‘A defendant whose wrongdoing has caused a claimant’s injury is not liable if the prospect that his behaviour would cause an injury in this particular way played no role in establishing the wrongfulness of this action’. The idea behind this is that if a taxi driver had asked, ‘why is it wrongful for me to speed?’ no tort lawyer would have answered ‘because your vehicle might get hit by a bus whilst waiting at traffic lights’.34 Clearly some intervening ‘voluntary’ acts might be so unconnected with what makes a defendant’s behaviour wrongful as to trigger the coincidence explanation. But I do not think that this can explain all the cases potentially covered by Hart and Honoré’s doctrine. For example, if the defendant punched the claimant to punish him for being such a nuisance to his neighbours and left him dazed, and seeing this another local resident took the opportunity to punch the claimant as further punishment, I doubt that most tort lawyers would label any additional harm inflicted by the second punch as ‘coincidental’, but suspect that many would see this scenario as one where the intervention doctrine ought to be applied. The opinion that the intervention doctrine cannot be explained solely by reference to ‘reasonable foreseeability’ and ‘coincidence’ is also supported by authoritative restatements of it. For example, Clerk & Lindsell refers to the doctrine applying where the ‘voluntary’ intervening act is ‘an event of such impact that it “obliterates” the wrongdoing of the defendant’.35 Or an event which ‘isolates’, ‘insulates’ or ‘eclipses’ the defendant’s conduct.36

C. Fairness An important strand of recent English authority has accepted Stapleton’s view that the intervention doctrine, so far as it exists at all, does not reflect any concept of ‘cause’,37 but is 34   Some authors prefer to explain the rule that relieves defendants from liability for ‘coincidences’ by claiming that it applies where the wrongdoing is not of a sort which will increase the frequency of the type of harm suffered by the claimant: speeding by taxis does not increase the number of claimants injured when stationary taxis are struck by carelessly driven buses. I would argue that this explanation is inferior to that given in the text, however, because I believe that the relieving rule can apply even when the wrongdoing does increase the frequency of the type of harm suffered provided that the increase is not sufficiently great to establish a basis for regarding the defendant’s action as wrongful. For example, imagine that taxis which leave the city to head to the airport between 4.15 pm and 4.30 pm regularly speed in order to be able to pick up passengers off the 5.00 pm flight as return fares, and as a consequence of such speeding tend to be in the vicinity of the junctions near the airport at the same time as a large concentration of buses come the other way: in these circumstances the speeding might well increase the likelihood of a stationary taxi being struck by a negligently driven bus, but I doubt that the speeding driver would be held liable for any injuries his passenger suffered in such an accident. 35   Clerk and Lindsell on Torts 20th edn (London, Sweet & Maxwell, 2010) para 2.101. 36   ibid, para 2.102. 37   Compare, however, Lord Hoffmann’s statement in Gray v Thames Trains [2009] UKHL 33, [54], that the ‘distinction, between causing something and merely providing the occasion for someone else to cause something, is one with which we are very familiar in the law of torts. It is the same principle by which the law normally holds

378  Roderick Bagshaw founded on the ‘unfairness’ of holding the defendant liable for the consequences.38 Thus in Corr v IBC Vehicles Ltd,39 Lord Bingham stated that ‘the rationale of the principle that a novus actus interveniens breaks the chain of causation is fairness’. There are three reasons, however, for being sceptical about the significance of this development. The first reason is that very little has been said in the relevant cases about why it is ‘unfair’ to hold a defendant responsible for harm following an intervention. If ‘fairness’ is often used as no more than a compendious label for a catalogue of unanalysed reasons why a defendant might avoid liability, including ‘non-foreseeability’ and ‘coincidence’, then it can provide only weak support for an argument that none of the reasons corresponds to some natural feature of the ‘causal’ mechanisms concerned. Secondly, some of the cases which have invoked ‘fairness’, including Corr, have done so in order to justify some erosion of or exception to the intervention doctrine, rather than to explain the reasons for its existence. Thirdly, those judges who have sought to explain when the ‘fairness’ rationale will lead to an intervention negativing liability have often resorted to ‘causal language’. For example, in Spencer v Wincanton Holdings Ltd40 Sedley LJ explained that it would be unfair to hold a defendant liable where the intervention meant that further injuries had ‘been in substance brought about by the claimant and not the tortfeasor’. Thus the current judicial fashion for invoking ‘fairness’ should not be regarded as conclusively establishing that the intervention doctrine does not reflect natural features.

D. Intention Those who favour a non-causal explanation might draw further strength from the fact that tort lawyers commonly think that an intervention cannot ‘break the chain of causation’ if the wrongdoer committed the wrong with the intention that a third party should intervene in just the way that he or she did. Thus if tort lawyers hold that a railway company is not liable when it negligently allows its trains to become over-crowded and a third party takes advantage of this situation to steal a wallet, but that the same company is liable if it brings about the same situation with the intention of stimulating such a crime, perhaps to draw attention to one of the consequences of under-investment in the railways, then this suggests that the operative doctrine must be non-causal: a railway’s company’s intentions cannot alter what its actions cause. Can the apparent relevance of ‘intention’ be explained in any way that preserves the plausibility of a ‘causal’ account? One possibility is that ‘intention’, and other indicators of severe culpability, simply distract attention from what it can be said that the wrongdoing has ‘caused’: if the wrongdoer really wanted the claimant to suffer the injury then we feel less need to worry about whether he in fact ‘caused’ it. But this seems insufficient to explain why references to ‘intention’ are so firmly incorporated into the relevant doctrine and its exceptions.

that even though damage would not have occurred but for a tortious act, the defendant is not liable if the immediate cause was the deliberate act of another individual.’ (emphasis added). 38   A key text is the speech of Lord Nicholls in Kuwait Airways Corporation v Iraqi Airways Co (Nos 4 and 5) [2002] UKHL 19, [2002] 2 AC 883, [69]–[70]. 39   Corr v IBC Vehicles Ltd [2008] UKHL 13, [2008] 1 AC 884, [15]. 40   [2009] EWCA Civ 1404, [15] (emphasis added).

Causing the Behaviour of Others and Other Causal Mixtures  379

E.  Causal Potency If none of the most plausible non-causal reasons for limiting liability seems sufficient to explain the intervention doctrine then is some causal account available? In order to establish what a causal account might have to look like it might be helpful to consider the following situation. Suppose that Trevor is walking up a stony beach when he decides to pick up a stone and throw it through Claire’s window. He then does this. Did the presence on the beach of the stone he threw ‘cause’ the breaking of the window? Using the account of ‘contribution’ developed above this question might be rephrased as ‘did the presence of this stone play an active role in the mechanism by which the window was broken?’ The obvious difficulty which this question highlights is that it is hard to describe the mere presence of the stone on the beach as ‘an active role’; the stone’s ‘active role’ seems to owe everything to Trevor’s choice to project it in a particular way. The stone’s ‘inert presence’ on the beach is clearly not wholly irrelevant to the breaking of the window. It was, after all, only by being present on the beach that that stone (eventually) came to strike the window. But it is arguable that ‘inert presence’ does not bear enough resemblance to ‘source of transferred energy’, to be recognised as a member of the extended family: Trevor is the source of the energy.41 One element that might be useful in developing a ‘causal’ explanation of the intervention rule might be the concept of ‘causal potency’. Loosely speaking, the idea is that it is the relative weakness of a defendant’s wrong as a ‘cause’ which is what allows it to be ‘eclipsed’ by the third party’s intervening ‘voluntary’ act, a more potent cause. Thus the mere presence of the stone is a very weak ‘cause’ when set alongside Trevor’s choice. Similarly the intervening wrongdoer’s choice to take advantage of the over-crowding in a train carriage order to commit a theft or a sexual assault might be said to be a far more potent cause of the claimant’s injury than the railway company’s negligence. Is it possible to talk sensibly about measurements of ‘causal potency’, or degrees of causal involvement, or is causation necessarily a bivalent matter (the defendant’s wrong was a cause or it was not)? For example, if it takes 50 units of poison to kill the fish in a pond, and simultaneously A spills 40 units, B spills 30 units and C spills 10 units, is it muddle-headed to say that A was more causally involved in killing the fish than B, who was more involved than C? I do not believe it is muddle-headed. Indeed, most lawyers recognise a doctrine which allows a defendant to avoid liability for an injury that his wrong contributed to if ‘the extent of its contribution is regarded as too small for it to be worthwhile to devote the Law’s resources to it.’42 And any doctrine that requires a measurement of the extent of contribution implies that there can be ‘causal interrogations’ beyond the simple one of ‘involved or not?’ In terms of the law as it applies to intervening ‘voluntary’ acts, however, the important question is whether it is muddle-headed to attempt to measure the relative extent of contribution of a defendant’s wrong and a subsequent voluntary act. For example, if Mrs Webster slips in a badly maintained car park and injures her knee sufficiently badly to require hospital treatment, and at the hospital a negligent doctor chooses to amputate rather than take 41   This suggests that the example involving a stone on a beach may be easier to deal with than a variation where Trevor finds a stone at the top of a tower and decides to push it off so that it will drop onto Claire’s roof. The line between what is ‘inert’ and what is ‘active’ will be considered further below. 42   Choosing 444, fn 20.

380  Roderick Bagshaw the time to assess her injury properly, in circumstances where we know that had the knee been assessed properly she would have made a full recovery within three weeks, can we talk about the extent of contribution of the doctor’s act, relative to the contribution of the original wrong? My view is that we can, perfectly sensibly, make a comparison between the probable state of Mrs Webster’s knee assuming that the doctor treated her competently43 and its actual state after the incompetent treatment, and it is the ratio between the injury assuming competent treatment and the additional injury that provides our assessment of the relative causal potency of the two events. Can ‘causal potency’ meet the demand that ‘causal language’ is only used to refer to matters which can be described in terms of ‘objective facts’? This question is difficult. Where A spills 40 units of poison into the pond and C simultaneously spills 10 units ‘causal potency’ looks as if it is determined by ‘objective fact’ rather than ‘normative evaluation’. The matter is more contentious, however, where A wrongfully leaves his horse-drawn wagon untied in the street, B throws stones at the horses, and C is injured making a heroic attempt to restrain the bolting horses. But even if assessments of ‘causal potency’ in these circumstances seem uncertain this does not establish that they are ‘normative’.

F.  Summary: Explaining the Intervention Doctrine The investigation of the intervention doctrine offers some support to both sides. Those who insist that the doctrine does not reflect any features of the natural mechanisms by which one thing can ‘bring about’ another can point to the incorporation in the doctrine of elements that seem distinctly ‘non-causal’, for example reference to ‘intentions’, and to the current popularity of ‘fairness’ rhetoric. On the other side, however, those unwilling to discount a ‘causal’ explanation can point to the inability of ‘unforeseeability’, ‘coincidence’ and ‘fairness’ to explain the doctrine, and the apparent promise of a concept of ‘causal potency’.

VII. Conclusions

To conclude it may be appropriate to summarise the milestones in my argument. I have argued that a legal concept of ‘cause’ must not be inconsistent with what we know about the natural mechanisms by which one thing can ‘bring about’ another, and that these mechanisms may be related to one another like members of an ‘extended family’ rather than clones. I have sought to demonstrate that where a ‘mixture’ of factors combine to ‘bring about’ an effect the approach of using ‘disaggregation’ to establish ‘causal contribution’ risks identifying factors as ‘causes’ even when that is inconsistent with what we know about the relevant natural mechanisms. This led me to propose an alternative ‘natural mechan­ isms’ approach, centred on the idea of ‘playing an active role in an operative mechanism’, as a rival to the ‘disaggregation’ approach. I then went on to suggest that in situations where one person has said or done something and we are interested in whether this ‘caused’ a second person to act an essential part of the natural mechanism is the decision-maker’s ‘giving of weight’ to the factor. I noted that in situations where there is an abundance of 43   Lawyers will be aware that the comparison is more complicated where more than one course of treatment would be regarded as competent, but this does not alter the general point.

Causing the Behaviour of Others and Other Causal Mixtures  381 reasons which a decision-maker could attach weight to a tort lawyer may be unable to know which reasons ‘contributed’ to the decision unless this is disclosed by the decision-maker. The difference in mechanism explains why lawyers ought to think about ‘causation’ differently when asking whether a defendant ‘caused’ some other person to act and whether he ‘caused’ some machine to work. The final section of the chapter considered whether a ‘causal’ explanation might underpin the ‘intervention doctrine’. I argued that none of the most plausible non-causal explanations provided a satisfactory foundation for the doctrine, and that the concept of ‘causal potency’ might be of particular importance. Together, the attempts to address the situations discussed in this chapter suggest that the ‘causal concepts’ which the law ought to be willing to use should be far richer than ‘causal minimalists’ generally acknowledge.

18 Law’s Approach to Harm in the Context of Scientific Uncertainty: Observations from the Perspectives of Precaution and Indeterminate Causation JOHN PATERSON

I. Introduction

It may be said that law has roles to play in relation to harm ex ante and ex post. Ex ante, it is expected to provide a means of ensuring that substances, technologies and practices that may cause harm will either be prohibited or permitted only to the extent that no (or at least only an acceptable level of) harm will arise.1 Ex post, when harm has occurred, law is expected to determine liability and compensation.2 Ex ante, law operates via legislation and regulation.3 Ex post, law operates via adjudication.4 In performing these roles, law does not of course operate in isolation. The regulation of potentially harmful substances, tech­ nologies and practices depends heavily upon the provision of scientific information, for example, by advisory committees.5 The determination of liability and compensation for harm depends heavily upon the provision of scientific information, in particular by expert witnesses.6 1   Recent high profile cases include the regulation of pharmaceutical drugs. See eg, C Schoonderbeek, ‘Risk Management and Transparency in Pre-clinical Research and Early Clinical Trials: Lessons Learnt from TGN1412’ (2005/06) 8(4) Bio-Science Law Review 151. 2   For a discussion of these issues in relation to environmental harm, see: JH Bates, Liability for Environmental Harm (London, LexisNexis UK, 2004). 3   See generally, J Black, Rules and Regulators (Oxford, Clarendon Press, 1997). 4   J Stapleton, ‘Cause in Fact and the Scope of Liability for Consequences’ (2003) 119 LQR 388. 5   eg, the Commission on Human Medicines whose terms of reference include giving ministers ‘advice with respect to safety, quality or efficacy in relation to medicinal products’, Medicines Act 1968, s 3, as substituted by the Medicines (Advisory Bodies) Regulations 2005, SI 2005/1094. For a discussion of the challenges facing such com­ mittees in the context of the regulation of food standards, see A Turner, ‘Viewpoint: The Story So Far: An Overview of Developments in UK Food Regulation and Associated Advisory Committees’ (1999) 101(4) British Food Journal 274. More recently these challenges have been exposed in the context of the Advisory Council on the Misuse of Drugs: see BP Block, ‘The Nutt Case’ (2009) 173(47) Criminal Law & Justice Weekly 745. 6  J Sanders, ‘Science, Law, and the Expert Witness’ (2009) 72(1) Law & Contemporary Problems 63–96; R Goldberg, Causation and Risk in the Law of Torts: Scientific Evidence and Medicinal Product Liability (Oxford, Hart Publishing 1999). See also their contributions in this volume in ch 2 and 8, respectively.

384  John Paterson These observations open up the question of what it is that law then does with the infor­ mation provided by science. An ungenerous reading would see law as doing little more than putting its imprimatur upon a decision that has actually been made within the realm of science. A moment’s consideration, however, will reveal that matters are seldom this simple. In a regulatory setting, for example, there may be conflicting views among experts about the reliability of the data upon which a risk assessment has been based, leading to very different conclusions about whether a substance should be authorised for use.7 In a court setting, for example, expert witnesses may offer very different interpretations of the same events leading to very different conclusions about liability.8 In other words, it seems clear that whatever law is doing in such circumstances, it is not simply confirming a scientifically agreed position. This observation, however, raises potentially troubling further questions. If science is apparently unable to agree what the ‘right’ answer is in these regulatory or liability situa­ tions, then on what basis does law presume to do so? Surely society relies upon science to determine what is true and what is false?9 If science is unable to do this in a given case, then is law really well placed to intervene? Answering these questions, of course, requires consid­ eration of the different functions that law and science perform within society and an appre­ ciation of what is meant respectively by ‘true’ in science and ‘legal’ in law.10 But these considerations undoubtedly only take the discussion so far. They leave open the issue of how law reaches decisions in situations where science has been unable to.11 A sceptical view might perceive that society’s best means of separating true from false has been supplanted by a second-best alternative, while a more generous reading would point to law’s undoubted abilities in the discovery of facts and the resolution of disputes. Even law’s stoutest defend­ ers, however, must acknowledge concerns about what it is being asked to do in any situa­ tion where it is not simply the case that the scientific evidence is disputed, but rather that science is explicitly and unashamedly unable to offer any evidence at all on the precise point in question. Law confronts such scientific uncertainty in a range of settings. Two of the most prob­ lematical are the following: 1. hypothetical harm – for example, a situation where the regulation of a substance is under consideration, a hypothetical harm has been raised as a reason for non-­authorisation or only tightly-constrained use, but the state of scientific knowledge makes it impossible to decide one way or the other on the basis of the usual risk assessment techniques;12

7   W Weigel, ‘How Well Are Authorities Suited to Prevent Damage to the Environment?’ (1997) 4(1) European Journal of Law and Economics 23. 8  S Haack, ‘Irreconcilable Differences: The Troubled Marriage of Science and Law’ (2009) 72(1) Law & Contemporary Problems 1. 9   N Luhmann, Social Systems (Stanford, California, Stanford University Press, 1996) 55–56, 357–59, 376–77. 10   For Luhmann, law’s function is the stabilisation of normative expectation: see Luhmann (n 9) 373–76. 11   For a discussion of this point in the context of legal theory, see J Paterson, ‘Trans-science, Trans-law and Proceduralisation’ (2003) 12(4) Social and Legal Studies 523; see also ‘Science for Risk Reduction and Sustainable Development: The Role of Law’ in T Beer and A Ismail-Zadeh (eds), Risk Science and Sustainability, NATO Science Series (Dordrecht, Kluwer Academic Publishers, 2003) 63; and ‘Truth or Dare: Expertise and Risk Governance’ in OECD Nuclear Energy Agency Secretariat (ed), Better Integration of Radiation Protection in Modern Society (Paris, OECD/NEA, 2002) 43. 12   In this regard, see especially the contribution by C Cranor in ch 13 of this volume. It is interesting to note that there is an apparent link here with the issue of liability for future harm, although this is beyond the scope of the present chapter. See in this regard the contribution in ch 11 by A Porat and A Stein of this volume.

Law’s Approach to Harm under Uncertainty  385 2. indeterminate causation – for example, a litigation situation where a claimant has undoubtedly suffered harm, a number of defendants have undoubtedly breached the duty they owed to the claimant, but the state of scientific knowledge makes it impossible to decide which of the claimants has caused the harm on the basis of the usual standard of proof.13 It is one thing for law to reach a decision (regulatory or adjudicative) in a situation where experts disagree – it can after all be suggested that law’s ability to hear and test arguments on the basis of well-established techniques makes it well-suited to reach a decision in such cases14 – but it is surely quite another for law to reach a decision in a situation where, on the vital issue, experts have nothing to say. The most obvious response to this situation is of course not to decide. Insofar as science has not yet been able to establish a causal link between a particular substance, technology or practice and a hypothetical harm, it is surely not unreasonable to defer a final decision until such time as science is able to do so or to state that no such link exists. Likewise, inso­ far as science is unable to demonstrate a causal link between a breach of duty and a harm suffered, it is surely not unreasonable to defer a decision until such time as science is able to offer a meaningful view one way or the other. To do otherwise in either of these situations surely asks too much of law: to reach a decision when there is no information or knowledge upon which to base it. There will undoubtedly be occasions when it will be entirely appropriate to behave in this way, that is, to admit that the basis for a decision is simply not present and that accord­ ingly no decision will be taken simply for the sake of deciding. But in many situations with the characteristics just described, not deciding is unlikely to be acceptable. For example, where regulators are asked to decide whether to authorise a particular pesticide for use, the fact that a hypothetical harm has been raised may have to be weighed against the proven benefit that it will bring in terms of increased crop yield.15 Similarly, when a court is asked to decide on liability and compensation in a case characterised by indeterminate causation, the fact that no causal link can be established between the claimant and any one of the defendants may have to be weighed against the sense that justice will not be served if parties who admittedly breached their duty to the claimant will escape liability and the claimant will be left without a remedy.16 In other words, in some cases deciding not to decide would actually constitute a positive choice with clear consequences for the parties concerned. Thus, regulatory and adjudicatory decisions characterised by scientific uncertainty not only involve questions of scientific proof, but also of questions of efficiency and justice, to name but two. This observation can provide some comfort for those who worry about the role of law in the context of scientific uncertainty – efficiency and justice are undoubtedly important fac­ tors to be considered – and yet questions must surely remain when decisions are taken on those grounds in situations where it is not possible to say anything about truth in a way that   This is, of course, the scenario that occupies many of the contributions to this volume.  For a comparison of approaches in this regard in civil and common law jurisdictions, see R Verkerk, ‘Comparative Aspects of Expert Evidence in Civil Litigation’ (2009) 13(3) International Journal of Evidence & Proof 167. 15   For a discussion of cost benefit analysis in regulatory situations, see eg, RW Hahn and RE Litan, ‘Counting Regulatory Benefits and Costs: Lessons for the US and Europe’ (2005) 8(2) Journal of International Economic Law 473. 16   See n 13 and associated text. 13 14

386  John Paterson is otherwise considered to be vitally important whether in regulation or adjudication. Is it possible to be reassured about the quality of decision-making in these situations? What does an examination of actual decisions reveal about the way in which the law copes when scientific evidence is unable to assist? Is it possible to develop some general principles from an examination of such examples or are the circumstances of such cases so particular that generalisation is impossible and unhelpful? In seeking to provide answers to these questions, this chapter will consider two cases: one is a recent case involving the judicial review of a precautionary regulatory decision;17 the other is the leading English case on indeterminate causation. While the chapter so far has discussed regulatory decision-making on one hand and judicial decision-making on the other, there would be a risk in looking at the way in which regulators and judges deal with scientific uncertainty that these are such different settings that little that is meaningful could be said. As a consequence, the discussion of precautionary decision-making consid­ ers the way in which the courts have addressed the presence of scientific uncertainty in such situations. It should be possible, accordingly, to see, first, if there is any commonality between these situations and, secondly, if so, whether it is possible to say anything about any commonality in the response of the judges. In each case the discussion begins (1) with a consideration of the facts, followed by (2) an effort to determine the nature of the scien­ tific uncertainty at the heart of the case, before (3) concluding with an assessment of the way in which the court has responded to that uncertainty. Thereafter, the chapter continues by considering what these cases reveal about the differences and similarities between precautionary decisions and cases of indeterminate causation. Finding that superficial dif­ ferences only serve to mask similarities at a deeper level, in particular with regard to the nature of scientific uncertainty, the question then considered is why the courts in each case appear to display different degrees of willingness to engage with that uncertainty and to reach substantive decisions. In particular, it is suggested that the approach of the courts in the context of indeterminate causation can provide support for a more robust judicial approach in the context of the review of precautionary decision-making. In other words, the fact that in the context of indeterminate causation judges have been at pains to point out that causation is ultimately a legal matter not a scientific one, lends weight to a more sophisticated understanding of the precautionary principle18 – specifically one that demands greater precision in the identification of the situations where that principle is appropriately engaged and in the way in which it is defined. The insistence of the court in the context of indeterminate causation that legal duties cannot be rendered devoid of con­ tent, highlights the possibility that an unwillingness on the part of the court to engage more robustly with the precautionary principle risks rendering it meaningless in practice.

17   This is a decision in which the regulator has been required to take account of the precautionary principle. There are many definitions of this principle, but the one most often referred to is the one contained in the Rio Declaration which states that ‘[w]here there are threats of serious or irreversible damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degrada­ tion’: Report of The United Nations Conference on Environment and Development, Rio de Janeiro, 3–14 June 1992, Annex I, Rio Declaration on Environment and Development, Art 15. 18   See n 17.

Law’s Approach to Harm under Uncertainty  387

II.  Law, Harm and Scientific Uncertainty Ex Ante: Downs v Secretary of State for Environment, Food and Rural Affairs19

A.  The Facts This case involved an application for judicial review from a woman who claimed that she had suffered ill health as a consequence of pesticide spraying over a period of years in fields directly adjacent to her house and garden. The judicial review related in particular to the failure of the Secretary of State to impose a no-spray buffer zone around rural houses and gardens, which the applicant contended was in breach of the duty imposed on him by the Plant Protection Products Directive,20 which had been transposed into UK law by the Plant Protection Products Regulations 2005.21 Described by Collins J as embodying a ‘precautionary approach’, the directive provides at Article 4 that: 1. Member States shall ensure that a plant protection product is not authorised unless . . .

(b) it is established, in the light of current scientific and technical knowledge and shown from appraisal of the dossier provided for in Annex III, that when used in accordance with Article 3(3), and having regard to all normal conditions under which it may be used and to the consequences of its use: . . .



(iv) it has no harmful effect on human or animal health, directly or indirectly (e.g. through drinking water, food or feed) or on groundwater.

Annex III to the directive sets out the requirements for the authorisation of pesticides and provides inter alia in relation to bystander exposure that ‘[s]ufficient information and data must be reported to provide a basis for the selection of appropriate conditions of use, includ­ ing the exclusion of bystanders from treatment areas and separation distances’ and ‘[a]n esti­ mation shall be made, using where available a suitable calculation model in order to permit an evaluation of the bystander exposure likely to arise under the proposed conditions of use’.22 Article 3(3) provides inter alia that ‘Member States shall prescribe that plant protection products must be used properly. Proper use shall include compliance with the conditions established in accordance with Article 4 and specified on the labelling’. In essence, the Secretary of State was confronted with two opposing pieces of expert advice with regard to the implementation of these provisions. On the one hand, the Royal Commission on Environmental Pollution (RCEP) in a report commissioned by the Secretary of State concluded that there was a possible link between exposure to pesticides and ill health, that a more precautionary approach should be adopted, that residents should receive prior notification of spraying on a compulsory as opposed to a voluntary basis and that buffer zones should be introduced.23 On the other hand, the Secretary of State received   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin).   Council Directive 91/414/EEC of 15 July 1991 concerning the placing of plant protection products on the market [1991] OJ L 230. 21   SI 2005/1435. 22   Plant Protection Products Directive, Annex III, Part A, para 7.2.2. 23   Royal Commission on Environmental Pollution, Crop Spraying and the Health of Residents and Bystanders: A Special Report Produced on Request of the Department for Environment, Food and Rural Affairs (London, RCEP, 2005) [RCEP Report] 104 and following. 19 20

388  John Paterson advice from the Advisory Committee on Pesticides (ACP) which rejected these recommen­ dations.24 The Secretary of State – in deciding that no changes were needed to existing arrangements – effectively rejected the RCEP advice and followed that of the ACP.25

B.  The Location and Nature of the Scientific Uncertainty The principal point of disagreement between the RCEP and the ACP relates to the way in which the risk to bystanders is calculated and the measures adopted to protect them. For the RCEP, the standard exposure assessment, while representing ‘a pragmatic approach to a complicated problem’ which is ‘probably conservative and protective in the majority of cases’, is nevertheless ‘based on a gross simplification of real exposure and . . . does not address all of the possible exposure routes’.26 The RCEP lists the problems it perceives with the standard exposure assessment. These include: the fact that no attempt is made to repre­ sent the mechanisms that determine the extent of exposure; that little account is taken of the variability of dispersion conditions encountered in practice; that a narrow definition of bystanders is employed; that it is based on old data, despite significant changes in practice having occurred; and that it assumes good operational practice.27 The RCEP stated that: the present approach may be conservative and protective in its treatment of targets but in view of the absence of any attempt to model the complexity of bystander exposure and the probability of extreme values, we cannot agree that this has been conclusively or transparently demonstrated for the exposure process.28

The RCEP concluded that ‘[a]lthough uncertainty factors are built in . . . they are there to cover issues related to toxicology and do not address the variability of exposure or the uncertainties in exposure assessment’.29 Rejecting the current conservative approach, it called for its urgent replacement ‘by a computational model which is probabilistic, looks at a wider range of possible exposure routes and more robustly reflects worst-case outcomes’.30 It was as a consequence of these uncertainties that the RCEP recommended precautionary measures such as a no-spray buffer zone31 and the compulsory provision of information to residents.32 The ACP, on the other hand, was content that the current approach was adequate. It recognised that scientific uncertainty did exist in relation to, for example, long-term disorders33 and the appropriate limit for bystander exposure,34 but viewed these as minor and of a similar order of magnitude to that existing in related human health risk assess­ 24   Advisory Committee on Pesticides, Crop Spraying and the Health of Residents and Bystanders: A Commentary on the Report Published by the Royal Commission on Environmental Pollution in September 2005, 30 December 2005 [ACP Commentary] 26–34. 25   Department for Environment, Food and Rural Affairs, The Royal Commission on Environmental Pollution Report on Crop Spraying and the Health of Residents and Bystanders – Government Response (London, Defra, 2006) [Government Response]. 26   RCEP Report, para 3.39. 27   ibid, para 3.39. 28   ibid, para 3.44. 29   ibid, para 3.50. 30   ibid, para 3.53. 31   ibid, para 5.86. 32   ibid, paras 6.14, 6.51–6.53. 33   ACP Commentary, para 3.7. 34   ibid, para 3.16.

Law’s Approach to Harm under Uncertainty  389 ment situations.35 It did not deny the need for further data, but did dispute that the current data indicated that there was any problem. As a consequence, while it did not entirely rule out the usefulness of a probabilistic model, it was content with the current simpler approach, provided that it was ‘demonstrably conservative’.36 In regard to the RCEP’s pro­ posed responses it viewed these variously as ‘disproportionate’37 and unenforceable.38 There was, then, a significant disagreement between the two expert bodies offering advice to the minister, but the disagreement was not by any means an unusual one. Faced with conflicting expert advice, the minister chose one over the other, and in choosing not to accept the RCEP’s recommendations and thus to adhere to the status quo he was fully sup­ ported by the conclusions of the ACP.39 The minister furthermore considered that the cost of the RCEP’s recommendations was not justified and that ‘making prior notification to all residents of every spraying event a statutory requirement would be highly bureaucratic’.40

C.  Assessment of the Court’s Response to the Uncertainty At first sight, then, it may appear that this is not a case that should long detain the judge. Insofar as this is a situation where there is conflicting scientific advice, then this would appear to be entirely unexceptional and precisely what a regulatory decision-maker would often expect to have to deal with. The decision may not be an easy one, but the fact that there is scientific advice which he may rely upon provides him with a possible and entirely defensible route to follow. Furthermore, whereas such a decision would not be beyond the scope of judicial review, it would surely only be challengeable in very particular circum­ stances. Indeed, Collins J acknowledges the limited extent to which he could interfere with this sort of decision: There are conflicting views as to the adequacy of the approach adopted: so much is apparent from the report of the RCEP and the response by the ACP. I am not qualified to decide between those views nor is it an appropriate exercise for a judge to undertake on judicial review.41

He does not rule out the possibility that the judge could properly intervene in the event that one view was tainted by Wednesbury irrationality,42 but he is sceptical as to whether that could be established and he notes that ‘we are here at the very fringe of what should properly be the subject of judicial review’.43 He quotes with approval the following recent warning issued by May LJ in a different setting: ‘[t]he scientific judgement is not immune from lawyers’ analysis. But the court must be careful not to substitute its own inexpert view of the science for a tenable scientific opinion’.44 Collins J is thus clear that if all that was being relied on in the application was the minister’s decision as between the two expert

  ibid 3.   ibid, para 3.34.   ibid, para 3.43. 38   ibid, para 6.29. 39   Government Response, especially paras 15, 18, 44 and 58. 40   ACP Commentary, para 121. 41   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin) [38]. 42   Associated Provincial Picture Houses Ltd v Wednesbury Corporation [1947] 2 All ER 680. 43   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin) [38]. 44   R (on the application of British Union for the Abolition of Vivisection) v Secretary of State for the Home Department [2008] EWCA Civ 417 [1]. 35 36 37

390  John Paterson views then he ‘would not be able to find in the claimant’s favour’.45 He adds that “the minis­ ter cannot be said to have erred in law by relying on scientific advice from the ACP in preference to that from the RCEP’46 and that he was entitled to have regard to costs and proportionality in reaching his decision.47 All else equal, then, the minister’s decision will be beyond reproach. And the decision of the judge in this case looks set to follow the general approach that the courts have adopted in similar circumstances, which is to defer to the decision made by the minister or other responsible actor.48 But as Collins J makes clear all else is not equal in his view. The sugges­ tion that the claimant has suffered harm when taken together with the wording of the directive means that this is not a simple case of the minister having to choose between conflicting pieces of expert advice. The directive states in effect that a pesticide can only be authorised and used ‘if it has no harmful effect on human . . . health’. The claimant con­ tended that she had suffered harm. The judge having considered that contention concluded that: There is in my judgement solid evidence produced by the claimant that residents have suffered harm to their health . . . or, at the very least, doubts have reasonably been raised as to the safety of pesticides under the regime which presently exists . . . It is clear that the precautionary principle must apply.49

It was on this basis that Collins J granted the application. In so doing, it is important to be clear about his understanding of the quality of the evidence of harm. Despite the fact that he describes this as ‘solid’, he qualifies that description by adding that ‘[n]one of what is produced establishes the causal link’ and then notes that ‘the establishment of such a link scientifically is a very difficult exercise’.50 The judge goes even further when he acknow­ ledges that ‘what may seem obvious to the victim or to the inexpert observer may in reality not be so’. Notwithstanding these observations, Collins J goes on to find that ‘the medical tests carried out on the claimant provide very powerful reasons for concluding that there has been the necessary cause and effect’.51 Finally, while noting that it is not for the judge in such a case to ‘specify any particular action that [the minister] needs to take’, Collins J does observe that the need to inform residents about pesticide spraying is ‘clear’, that ‘[v]olun­ tary action is not achieving this’ and that ‘there is a strong case for a buffer zone’.52 This may seem to be a very singular decision and one about which serious questions can be asked. It would also appear to have potentially serious implications for precautionary decision-making more generally. It could give rise to a situation where, given a legislative requirement that a precautionary approach should apply, any evidence of harm will be suf­ ficient to justify precautions irrespective of the sort of cost or proportionality considera­ tions that the minister thought important. On the face of it, then, this decision appears to

  Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin) [39].   ibid [62]. 47   ibid [63]. 48   See E Fisher, ‘Is the Precautionary Principle Justiciable?’ (2001) 13(3) Journal of Environmental Law 315. The argument advanced by Fisher is that the principle is not justiciable because of ‘the perception that penetrating review under it is not within the courts’ institutional or constitutional competence’ (at 316). 49   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin) [40]. 50   ibid [65]. 51   ibid [65]. 52   ibid [70]. 45 46

Law’s Approach to Harm under Uncertainty  391 offer ammunition to the critics of the precautionary principle who contend that this is not an appropriate or useful means of dealing with scientific uncertainty.53 The Secretary of State certainly had such concerns, and the Court of Appeal in turn agreed that the approach adopted by Collins J was flawed.54 The initial problem identified by Sullivan LJ (delivering the judgment of the court) with the approach of Collins J is his failure to appreciate the significance of Annex VI to the directive. This annex sets out the ‘Uniform Principles for the Evaluation and Authorisation of Plant Protection Products’.55 These are ‘prescriptive and immensely detailed’56 and insofar as the Secretary of State com­ plies with these then he must be deemed to have fulfilled his obligations under Article 4. It is also the case that a key dimension of the principles is that they are aimed at the harmoni­ sation of authorisation procedures.57 The issue for the respondent becomes one of demon­ strating how the appellant ‘is failing to comply with the “uniform principles” ’.58 In this regard, the respondent argued that (1) there were defects with the bystander exposure model employed by the Secretary of State; (2) the authorisation process did not take account of local effects; and (3) the Secretary of State was not taking action in respect of reports of incidents of harm to human health involving pesticides. In dealing with the first of these points, the court was invited to consider whether the Secretary of State had made a ‘manifest error’ in rejecting one exposure model in favour of the other on the basis that its ability to review such decisions was not limited to considera­ tions of Wednesbury irrationality. Sullivan LJ was not to be drawn on where the boundary between these two concepts (manifest error and Wednesbury irrationality) might lie, but contented himself with reference to authorities supporting the contention that judges should refrain from substituting their assessment for expert views in complex situations.59 In his view, in such cases ‘involving highly technical scientific judgement, the “manifest error” hurdle is a high one’.60 Reviewing the way in which the two expert committees had discussed the issue of the exposure model, the fact that the respondent had been able to put her views to the RCEP and that its views had in turn been considered by the ACP before the Secretary of State reached his decision meant that it was ‘impossible to conclude that there [was] any error, much less any “manifest error” ’.61 Regarding the second argument, the alleged failure to take account of local effects, Sullivan LJ begins by recalling that the point at issue must specifically be what Annex VI required of Member States in relation to such effects.62 In this respect, the ‘obligation is limited to a requirement to take “relevant technical or scientific information” about local 53  For a review of the criticisms levelled at the precautionary principle, see P Sandin, M Peterson M, SO Hansson, C Rude’n and A Juthe, ‘Five charges against the precautionary principle’ (2002) 5(4) Journal of Risk Research 287. 54   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664. 55   The original version of Annex VI was annulled by the ECJ in its judgment of 18 June 1996, Parliament v Council, C-303/94, ECR, I-2943. The reason for this was that the original version referred only to groundwater intended for drinking water and not all groundwater. The new version of Annex VI was introduced by Council Directive 97/57/EC of 22 September 1997 establishing Annex VI to Directive 91/414/EEC concerning the placing of plant protection products on the market. See n 20 above. 56   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664 [26]. 57   ibid [27]. 58   ibid [33]. 59   Specifically Case C-471/OOP(R) Commission v Cambridge Healthcare Supplies Ltd [2001] ECR I-2865. 60   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664 [43]. 61   ibid [53]. 62   ibid [56].

392  John Paterson effects into consideration if such information is reasonably available’63 and Sullivan LJ finds no shortcomings on the part of the Secretary of State in this respect.64 As regards the final argument, the suggestion that the Secretary of State failed to take action in response to reports of possible harm to health arising from pesticides, Sullivan LJ finds that the objection here was in reality that having taken account of these reports the Secretary of State had reached a different conclusion about their significance from that reached by the respondent.65 Referring here to the jurisprudence of the Court of First Instance in relation to the application of the precautionary principle,66 he recalled that whether risks are ‘ “purely hypothetical . . . based on mere hypotheses that have not been scientifically confirmed” . . . is pre-eminently a matter’ for the Secretary of State with the benefit of expert advice.67 In essence, then, the Court of Appeal, in rejecting the points raised by the respondent, has returned to the position prior to Collins J’s judgment where the court essentially defers to the position of the regulatory decision-maker in cases involv­ ing the precautionary principle – with the exception of cases where there is a manifest error or Wednesbury irrationality. Was Collins J simply out of line with the accepted and reasonable position of the judici­ ary or could he be said to have articulated a rather sophisticated argument based on a clearer appreciation of the precautionary principle than is habitually found in many judi­ cial, regulatory or academic discussions? It may well be that Collins J’s judgment rests on more solid ground than appears to be the case at first sight. In order to establish that, it is necessary to have more information about the type of situation that this case represents. First of all, it is necessary to be clear about what the precautionary principle actually requires. It is striking that no definition of the principle appears either in the judgment of Collins J or in that of Sullivan LJ or in the RCEP Report. The ACP Commentary does, however, provide a definition and relies in this regard upon the classic statement of the principle found in the Rio Declaration, viz: ‘Where there are threats of serious or irreversi­ ble damage, lack of full scientific certainty shall not be used as a reason for postponing cost-effective measures to prevent environmental degradation’.68 As the ACP notes, this basic approach has been extended to include harm to health.69 In seeking to understand how this principle could be operationalised, it is useful to begin with David Resnik’s char­ acterisation of the different sorts of situations in which decisions can be made in terms of the information available to the decision-maker:70 1. Decisions under certainty, where the outcomes of different choices are known. For example, the effects of heavy metals on human health are now well-known and understood. Thus, if a decision were made to restrict the emission to the environment of industrial waste containing, for example, lead or mercury, then this would predictably improve (or at least prevent further deterioration in) human health, whereas a decision   ibid [64].   ibid [64]. 65   ibid [69]. 66   Case T-229/04 Kingdom of Sweden v Commission [161] [2007] ECR II-2437. 67   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664 [69]. 68   Report of The United Nations Conference on Environment and Development, Rio de Janeiro, 3–14 June 1992, Annex I, Rio Declaration on Environment and Development, Art15. See n 17 above and associated text. 69   ACP Commentary, Annex 2, p 15. 70   D Resnik, ‘Is the precautionary principle unscientific?’ (2003) 34 Studies in history and philosophy of biological and biomedical sciences 329, 332. 63 64

Law’s Approach to Harm under Uncertainty  393 to allow increased emission of such waste would predictably produce detrimental effects on human health. 2. Decisions under risk, where probabilities can be assigned to the outcomes of different choices and thus methods such as quantitative risk assessment (QRA) can be deployed. For example, where sufficient relevant data in relation to the failure of reinforced con­ crete structures have been gathered over a sufficient period of time, decisions can be made about the design life of a motorway bridge, the required maintenance programme and the risk involved in any ultimate extension of the design life. The risk of failure does not disappear, but it can be assessed in such a way that the degree of confidence is rela­ tively clear and based on data and algorithms which can be interrogated and challenged. 3. Decisions under ignorance, where it is not possible to assign probabilities to the out­ comes of different choices and thus the precautionary principle must be applied. In this situation there are insufficient data to carry out a risk assessment and thus the harm that is feared is essentially hypothetical and the causal connection between the event and the feared consequence has not been established. Utilising this schema, it is possible to locate the Secretary of State’s decision in Downs. Insofar as there was clearly uncertainty about the possible effects of pesticides on bystand­ ers, then this was manifestly not a type 1 decision. Similarly, insofar as the RCEP called for a probabilistic approach and even the ACP admitted that it was impossible on the basis of the current data to assign probabilities, then this was also not a type 2 decision and quanti­ tative risk assessment could not be deployed. Accordingly, the Secretary of State was con­ fronted with a type 3 decision, a decision under ignorance insofar as none of the scientific advice before him was able to assign probabilities to the outcomes of the different choices. The RCEP was explicit about its concerns in this regard, called for data to be collected to allow QRA and in the meantime specified what it regarded as an appropriate precautionary approach. The ACP was nevertheless content that even in the absence of more data to allow QRA the existing approach was conservative. The ACP may, therefore, be said to have con­ fronted the fact that this was a decision under ignorance and to have decided that the cur­ rent approach in terms of the restrictions on the use of pesticides and voluntary notification of bystanders was an appropriately precautionary response. Insofar as this was the case, there must be a question as to the appropriateness of the judge’s decision. On what basis is the judge able to determine that one precautionary approach is better than another? The question calls to mind one of the main criticisms that has been levelled against the precautionary principle, namely that the fact that no probability can be assigned to either the worst-case or the best-case hypothetical scenario means that there is apparently no reason to discount the worst-case scenario and thus no reason not to opt for the most precautionary approach. The critics worry that in practice the precautionary principle can thus reduce to the principle of inaction, with the mere suggestion of harm leading to irrational fear and paralysing potential beneficial technological development.71 It has been pointed out, however, that to understand matters in this way is actually to apply the principle of insufficient reason (where, since it is impossible to say which of a set of possible events is more likely to occur, it is assumed that all are equally probable)72 rather than the   G Charnley, ‘Risk analysis under fire’, (2000) (First Quarter) RISK Newsletter 3.   M Peterson, ‘Transformative decision rules’ (2003) 58 Erkenntnis 71, 71. The principle was first proposed by J Bernoulli in the early eighteenth century. For a discussion, see H-W Sinn, ‘A rehabilitation of the principle of insufficient reason’ (1980) 94(3) Quarterly Journal of Economics, 493. 71 72

394  John Paterson principle of precaution. Given Collins J’s acceptance of the RCEP’s (worst-case) scenario over the ACP’s (best-case) scenario, the question is raised as to whether he has inadvert­ ently substituted insufficient reason for precaution. David Resnik’s proposal for the operationalisation of the precautionary principle, how­ ever, would seem to suggest that this is not the case. In brief he sees this as involving two stages, viz first, in a given decision under ignorance situation, it is necessary to consider whether a hypothetical threat is plausible; and secondly, in so far as a hypothetical threat is deemed plausible, it is necessary to consider whether any proposed response is reasonable. The first question is one that scientists are well placed to answer inasmuch as they are used to considering whether a given hypothesis is plausible prior to committing limited resources to testing it. Well-established criteria for a test of plausibility are available including: coher­ ence, explanatory power, analogy, precedence, precision, and simplicity.73 Insofar as a hypo­ thetical threat meets these criteria it may be deemed to be plausible. It will also be possible to differentiate between different hypothetical threats as more or less plausible utilising these criteria and it is therefore possible to demonstrate that there really is no equivalence with the principle of insufficient reason. Is it then possible to understand Collins J’s decision in Downs as an application of this test of plausibility? It is certainly the case that the judge is careful to explain that he could not simply substitute his own inexpert view for that of the experts, and that he could not intervene where the minister had simply chosen one piece of expert advice over another where there was no evidence of Wednesbury irrationality. There must therefore be another basis for the decision. As was seen above, it is crucial to the decision that the directive required that there be no harm to human health and that the applicant had presented evi­ dence that she had suffered harm. Now, Collins J is again clear that there is no suggestion that a causal link had been established between spraying pesticides in accordance with the currently accepted approach and the harm complained of. He goes even further and notes that proving a link would in any event be ‘ “very difficult’.74 He is nevertheless satisfied that the evidence before him with respect to the spraying and the harm is ‘solid’ or at the very least raises reasonable doubts. It is suggested that this can be read as a finding that the hypothetical link between spraying according to current practice and the harm complained of is plausible and that accordingly the precautionary principle applied. This finding, how­ ever, only takes the argument so far. Recall that the ACP in essence acknowledged that this was a type 3 decision – a decision under ignorance – but nevertheless maintained that the current approach was sufficiently conservative and protective. The question accordingly remains as to the basis on which Collins rejected the ACP’s approach in favour of the RCEP’s. In order to answer this question, it is necessary to look at the second stage of Resnik’s approach to the operationalisation of the precautionary principle. As previously noted, this is to consider whether, insofar as a hypothetical threat is deemed to be plausible, any pro­ posed response is reasonable. Once again, there are well-established criteria for addressing such a question, this time drawn from the realm of law, including: effectiveness, propor­ tionality, realism, efficiency, consistency and non-discrimination.75 Thus, insofar as a pro­ posed response to a plausible hypothetical threat meets these criteria, then it may be   Resnik (n 70) 339.   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin) [49]. 75  Resnik (n 70) 341–42. Resnik’s inspiration here is the European Commission’s Communication on the Precautionary Principle, COM (2000) 1. 73 74

Law’s Approach to Harm under Uncertainty  395 deemed to be reasonable. It will also be possible to differentiate between proposed responses as more or less reasonable utilising these criteria and it is therefore possible to see that there really is no equivalence with the principle of inaction. Can Collins J’s decision be read further as an application of this reasonableness test in comparing the ACP’s satisfaction with the status quo and the RCEP’s desire for greater precaution? Again it can be suggested that this is the case – and indeed this can help to explain the judge’s willingness towards the end of his decision to step beyond what might be regarded as the proper limits of judicial review in offering suggestions as to what form the minister’s decision might ultimately take. In particular, Collins J is impressed by two factors which can be directly related to the suggestions he makes in this regard. First of all, in relation to the disagreement between the RCEP and ACP about compulsory or voluntary arrangements for notifying bystanders about spraying, he notes that regulations already exist which require that beekeepers be given 48 hours’ notice of pesticide spraying.76 As Collins J observes, ‘[i]t is difficult to see why residents should be in a worse position’.77 Secondly, in relation to disagreement between the RCEP and the ACP about a no-spray buffer zone around houses and gardens, the judge notes that such a buffer zone ‘inciden­ tally already exists to avoid spraying too close to watercourses in order to minimise the risk of pesticides entering groundwater’.78 In other words, in determining whether the respec­ tive responses to the plausible hypothetical link between spraying and harm to human health were reasonable, the judge is able to compare them with the responses that are already in place in relation to beekeepers and groundwater. Understood in this way, Collins J’s suggestions as to the possible content of the minister’s ultimate decision appear much less contentious, but rather as moderate proposals based on a coherent application of the precautionary principle. If this reading of the first instance judgment is reasonable, does that mean that the con­ clusions of the Court of Appeal are in question? There is no doubting that that emphasis placed by the latter court on the harmonisation dimension of the directive appears to shift the focus of the debate away from the precautionary principle and on to the overarching objectives of European law and thus could give the impression that the discussion in this chapter, while potentially of academic interest, is nevertheless a legal blind alley. It might be suggested, however, that there are passages in Sullivan LJ’s judgment which raise the possi­ bility that the situation is by no means so cut and dried. It is important to look in particular at those passages where Sullivan LJ considers the way in which Collins J interpreted the precautionary principle. He notes that the judge at first instance, while expressing himself in relatively guarded terms and acknowledging that he was operating at the limits of judicial review, nevertheless considered that there was evi­ dence sufficient to engage the precautionary principle. Sullivan LJ, however, concludes that what Collins J was actually doing was trespassing into territory that was the preserve of the minister and ‘substituting his own evaluation of the available evidence’.79 He points out that this evidence had already been considered by the expert bodies and explicitly taken into 76   Control of Pesticides Regulations 1986, SI 1986/1510, as amended by the Control of Pesticides Regulations 1997, SI 1997/188 reg 6(c)(ii) and sch 4, para 2(1)(g). It is interesting to note that according to the introductory text these regulations were made ‘after consultation with the Advisory Committee on Pesticides’. 77   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin) [61]. 78  Ibid [70], referring apparently, though not explicitly, to the Control of Pesticides Regulations 1986, SI 1986/1510, as amended by the Control of Pesticides Regulations 1997 SI 1997/188 reg 6 and sch 4, para 2(1)(b) and (c), that is, the same regulations that, as noted at n 76 above, were made following consultation with the ACP. 79   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664 [76].

396  John Paterson account by the minister. Thus, he notes that the RCEP found that ‘it is plausible that there could be a link between bystander pesticide exposure and chronic ill health. We find that we are not able to rule out this possibility’.80 He states that the ‘RCEP did not conclude that the evidence showed that local effects were attributable to exposure’.81 Accordingly, he consid­ ers that ‘in respect of both chronic and local effects the RCEP was not saying that they were caused by bystander exposure to pesticides, rather it was saying that the possibility could not be ruled out and that more research was required’.82 While the RCEP then went on to call for a more precautionary approach, Sullivan LJ points to the conclusion reached by the minister on the basis of this and the ACP’s evidence: There is no dispute that some people who have been exposed to pesticides have become ill. The dispute has concerned the causality and underlying basis for these illnesses. On the evidence that we have received we cannot draw firm conclusions on causality . . . The Government believes that being unable to rule out the possibility of a link cannot be considered a basis to support the rec­ ommendation of an urgent need for research . . . Similarly there is no scientific basis for additional precaution beyond the already precautionary approach currently adopted.83

The difficulty here is that both the minister and Sullivan LJ appear to have misunder­ stood both the nuances of the RCEP’s conclusions and the import of the precautionary principle. It is not simply that the RCEP was unable to rule out the possibility of a causal link, as both the minister and the appeal judge concentrate on, but more importantly that that expert body accepted that the hypothesis of a link between exposure and harm was plausible. Therefore, to say that the RCEP had not concluded that the evidence showed that there was a causal link between exposure and harm is to miss the point that such a state­ ment is meaningless in the context of a decision that properly engages the precautionary principle. Where such a causal link has been established, then decisions will either be of type 1 or type 2 using Resnik’s schema, that is, the decision may be taken respectively in the certain knowledge of the outcome or on the basis of standard quantitative risk assessment techniques. In the absence of the establishment of such a causal link, the decision is of type 3 and the question for the minister is precisely to consider the plausibility of the hypothesis that is being proposed. By focusing on the inability to rule out the possibility of a link rather than on the RCEP’s more positive statement that the link is plausible, the minister – and in turn the appeal judge – has arguably sidestepped the question that actually needs to be addressed. The Court of Appeal’s concentration on Annex VI of the directive, namely the Uniform Principles for the Evaluation and Authorisation of Plant Protection Products, needs, however, to be addressed. Is it the case that the overarching objective of harmonisation essentially serves to render the concerns expressed above with the ministerial and judicial treatment of the precautionary principle irrelevant? The answer must surely be no. In discussing the need for the requirements of Annex VI to be complied with in reaching decisions as to the authorisation of plant protection products, the Court of First Instance in Kingdom of Sweden v Commission of the European Communities considered the application of the precautionary principle and observed that it ‘is designed to prevent potential risks. By contrast, purely hypothetical risks, based on mere hypotheses that have not been scien­   RCEP Report, para 2.65, quoted by Sullivan LJ at [79]. Emphasis added by Sullivan LJ.   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664 [80]. 82   ibid [81]. 83   Government Response, paras 17 and 18, quoted by Sullivan LJ at [82]. 80 81

Law’s Approach to Harm under Uncertainty  397 tifically tested cannot be accepted’.84 At first sight, this looks like an attenuation of the prin­ ciple as it has been developed above following Resnik. It is surely the case, however, that for the principle to have any meaningful effect it must come into play precisely where there is such uncertainty that traditional risk assessment techniques are rendered useless by the absence of sufficient data such that probabilities can be assigned. Accordingly, it is precisely where all that is available is a hypothetical link between a substance and a harm that the principle comes into play. If it is to be meaningful, the CFI’s judgment must therefore be read not as requiring that a precautionary approach is justified only when a causal link has been established, but that it is justified when science is able to support the proposition that the hypothesis of a link between a substance and a harm is plausible. Scientific testing in the context of this judgment is thus not required to have established the causal link, but only the plausibility of the hypothesis. While this reading may appear to some to be stretching the clear wording of the CFI’s judgment too far, one need only reflect on what the conse­ quences would be of rejecting it in favour of a more restricted reading: it would be neces­ sary for sufficient data to be collected to move a decision from type 3 to type 2 before a precautionary approach could be adopted, which would effectively defeat the purpose of the principle in the first place.85 To conclude this discussion of an example of judicial decision-making in the context of scientific uncertainty, there is no doubt that, as Collins J himself acknowledged, the case raises questions which lie at the limit of what is properly the concern of judicial review. Even if the above reading of the judgment at first instance and of the judgment on appeal is accepted, it remains the case that it will always be controversial for judges to involve them­ selves in such decisions. That said, however, it can be seen that quite serious questions can be asked about the way in which both the minister and the Court of Appeal have under­ stood the significance of the expert advice of the RCEP. In essence, it is suggested here that they have proceeded on the basis of an incomplete understanding of the precautionary principle. By contrast, while not expressing himself explicitly in the terms used by Resnik, it is a straightforward matter to read Collins J’s judgment as representing a quite sophis­ ticated understanding of the tests of plausibility and reasonableness that constitute Resnik’s operationalisation of the precautionary principle. To that extent, it is not so much that he has substituted his own evaluation of the evidence for that of the minister as that he has pointed out that the minister may well have failed to evaluate the evidence according to the appropriate test. The difficulty with the judgment of the Court of Appeal is that it appears to have proceeded on the basis of the same misunderstanding as that of the minister. The solution to these difficulties lies not in locating the tests of plausibility and reasonableness at the level of judicial review but rather, in the first instance, at the level of the initial min­ isterial decision. Insofar as decision-makers are explicitly directed to consider hypothetical harms on the basis of their plausibility, they will not fall into the trap of rejecting appropri­ ate applications of the precautionary principle on the basis of a misunderstanding of the type of decision they are being asked to make. They will in turn be able to reach a decision on the reasonableness of any proposed response to a plausible hypothetical harm. Judges would then be able to have a clearer and less controversial view of precautionary decisionmaking in the context of applications for judicial review and thus be able to grant such 84   Case T-229/04 Kingdom of Sweden v Commission of the European Communities [161] [2007] ECR II-2437, quoted by Sullivan LJ at [30]. 85   See n 17 above and associated text.

398  John Paterson applications without being suspected of substituting their evaluation of evidence for that of the minister but rather only on the grounds that there has been a shortcoming in the applications of the tests of plausibility and reasonableness. In other words, judges would very much be on the right side of the line that divides substituting their own decision from finding that there are problems with the way in which the decision has been reached. While critics of the precautionary principle will always have fears about any greater opportunity for the judiciary to become involved in such decisions, it may be countered that the approach suggested here serves to ensure first, that the precautionary principle is applied only in those cases where there is genuinely a decision under ignorance; second, that precautionary decisions respect the requirements of practical reason; and, third, that such decisions are not inadvertently taken on the basis of other principles such as that of inaction or of insufficient reason.

III.  Law, Harm and Scientific Uncertainty Ex Post: Fairchild v Glenhaven Funeral Services Ltd and Others86

A.  The Facts This case before the House of Lords in which three conjoined appeals87 were heard dealt with the situation where a party who was suffering (or indeed had died) from mesotheli­ oma had been employed at various times and for various periods by a variety of employers, all of whom had breached the duty of care they owed the appellants to take reasonable care to prevent them inhaling asbestos dust, since there was a known risk that such inhalation might cause mesothelioma. Any other cause of the disease could be ruled out, but the state of scientific knowledge meant that it was impossible to prove, according to the usual stand­ ard of proof, that the mesothelioma had been caused by asbestos dust inhaled while the injured party had been employed by one or another employer or while employed by more than one employer. The appeals had been raised because the Court of Appeal, applying the standard ‘but for’ test,88 had determined that in such circumstances it was impossible for the injured parties to recover damages either from any individual employer or from all of the employers jointly.89 The question for the House of Lords was whether in these very par­ ticular circumstances a modified approach to proof of causation could be justified.

86   Fairchild v Glenhaven Funeral Services Ltd. and others [2002] UKHL 22, [2003] 1 AC 32. Note that the discus­ sion here will not deal with the issue of the apportionment of liability, which is beyond the scope of this chapter, but only with the prior issue of establishing causation in the context of scientific uncertainty. For a discussion of the apportionment of liability dimension in Fairchild, see A Porat and A Stein, ‘Indeterminate Causation and Apportionment of Damages: An Essay on Holtby, Allen, and Fairchild’ (2003) 23(4) OJLS 667. See further Barker v Corus (UK) plc [2006] UKHL 20, [2006] 2 AC 572 and the Compensation Act 2006, s 3. 87  Also Fox v Spousal (Midlands) Ltd. and Matthews v Associated Portland Cement Manufacturers (1978) Ltd and others. 88   See HLA Hart and Tony Honoré, Causation in the Law, 2nd edn (Oxford, Clarendon Press, 1985) 109 and following. 89   [2002] 1 WLR 1052.

Law’s Approach to Harm under Uncertainty  399

B.  The Location and Nature of the Scientific Uncertainty Some indication of the location and nature of the scientific uncertainty in this case is already apparent from the narration of the facts. As Lord Bingham of Cornhill expresses it, ‘[t]he mechanism by which a normal mesothelial cell is transformed into a mesothelioma cell is not known’.90 Medical opinion suggested that a number of changes occur in the pro­ cess of a normal cell becoming malignant. Lord Bingham explains that ‘[a]sbestos acts in at least one of those stages and may (but this is uncertain) act in more than one’.91 Given that people living in urban areas are exposed to and inhale a large number of asbestos fibres without developing the disease, it is apparent that there is a level of exposure that can be tolerated, but what this level might be is ‘not known’.92 While it is accepted that there is a positive correlation between the quantity of dust and fibre inhaled and the risk of develop­ ing the disease, ‘the condition may be caused by a single fibre, or a few fibres, or many fibres: medical opinion holds none of these possibilities to be more probable than any other, and the condition once caused is not aggravated by further exposure’.93 Someone employed sequentially by a variety of employers, each of whom negligently exposed him to asbestos dust, will find it impossible to demonstrate which exposed him to the fibre or fibres which gave rise to the mesothelioma. There is thus ‘no way of identifying, even on a balance of probabilities, the source of the fibre or fibres which initiated the genetic process which cul­ minated in the malignant tumour’.94

C.  Assessment of the Court’s Response to the Uncertainty Confronted with this uncertainty, Lord Bingham proceeds to consider whether any relaxa­ tion of the ‘but for’ test could be justified on the basis of principle, authority, the wider jurisprudence or policy. As regards principle, he identifies the overall object of tort law, which he says ‘is to define cases in which the law may justly hold one party liable to com­ pensate another’.95 The overall aim may thus be said to be that justice must be done. He then considers whether the ‘mechanical application’ of the standard test in the context of scientific uncertainty could be justified96 and concludes in the particular circumstances of the present case (where there is no dispute that all of the defendants had breached their duty and where other causes could be ruled out) that it could not. Far from principle requiring the mechanical application of the test, it would indeed be ‘contrary to principle to insist on application of a rule which appeared . . . to yield unfair results’.97 And he reinforces this point later when he states that ‘the law must be developed coherently, in accordance with principle, so as to serve, even-handedly, the ends of justice’.98 As regards considerations of principle, then, the overall objective of justice may be said to trump   Fairchild v Glenhaven Funeral Services Ltd. and others [2002] UKHL 22, [2003] 1 AC 32 [7]. Emphasis added.   ibid [7]. Emphasis added. 92   ibid [7]. Emphasis added. 93   ibid [7]. Emphasis added. 94   ibid [7]. Emphasis added. 95   ibid [9]. 96   ibid [9]. 97   ibid [13]. 98   ibid [32]. 90 91

400  John Paterson concerns with the strict application of the standard test, albeit where carefully defined con­ ditions are met. Turning to authority, Lord Bingham offers a ‘detailed review’ of the case of McGhee v National Coal Board,99 which he describes as an ‘authority . . . of obvious import­ ance in the present appeal’.100 He draws a number of conclusions from this review, includ­ ing the fact that where a ‘pursuer faced an insuperable problem of proof if the orthodox test of causation was applied, but [where] justice demanded a remedy for the pursuer, a majority of the House adapted the orthodox test to meet the particular case’.101 As regards the wider jurisprudence, Lord Bingham cites judgments from Canada, Australia and the United States, which draw a clear distinction between causation in law and causation vari­ ously in scientific, medical or philosophical terms.102 Lord Bingham finally considers the policy implications of applying the ‘but for’ test in a mechanical way. The fact that the claimant in such a case would be left without a remedy notwithstanding that the various defendants have undoubtedly breached the duty they owed to him would essentially lead to a situation where: an employer exposing his employee to asbestos dust could obtain complete immunity against mes­ othelioma (but not asbestosis) claims by employing only those who had previously been exposed to excessive quantities of asbestos dust. Such a result would reflect no credit on the law.103

There is undoubtedly a possibility that injustice will be done to an employer found liable who has not as a matter of fact exposed the claimant to the asbestos dust that caused the mesothelioma, but Lord Bingham holds that this is ‘heavily outweighed by the injustice of denying redress to a victim’.104 Lord Bingham’s conclusion on the appropriate response of the courts to such problems sets the tone for similar points made by the remaining judges. He states that: in the interests of transparency . . . the courts’ response to the special problem presented by cases such as these should be stated explicitly. I prefer to recognise that the ordinary approach to proof of causation is varied than to resort to the drawing of legal inferences inconsistent with the proven facts.105

Lord Nicholls of Birkenhead echoes this call for linguistic clarity regarding the departure from the usual approach,106 as well as the fact that a value judgment will be involved in weighing the injustice that will be done by different decisions,107 and the need to confine this exceptional approach to appropriate cases.108 Like Lord Bingham, he is also very clear on the nature of the approach – the application of a less stringent test than the standard ‘but for’ test and not the drawing of an inference where none may legitimately be drawn.109 Lord Hoffmann is likewise very clear from the outset of his speech about the precise nature of the problem, namely that it is impossible to assign a probability to one or other of

  [1973] 1 WLR 1.   [2002] UKHL 22, [2003] 1 AC 32 [21].   ibid [21]. 102   ibid [23]–[32]. 103   ibid [33]. 104   ibid [33]. 105   ibid [35]. 106   ibid [36]. 107   ibid [40]. 108   ibid [43]. 109   ibid [45]. 99

100 101

Law’s Approach to Harm under Uncertainty  401 the various possible scenarios.110 He goes on to consider the issues under the headings of principle and authority. With regard to principle, he notes that the approach that is taken to causation in any given situation is dependent upon ‘the purpose of the inquiry’.111 Accordingly, given that in litigation the purpose is to achieve a ‘just solution’, ‘different kinds of case may require different causal requirement rules’.112 These rules must be expli­ cable in terms of justice and fairness.113 Noting the particular circumstances of the present appeals, Lord Hoffmann states that ‘a rule requiring proof of a link between the defendant’s asbestos and the claimant’s disease would . . . empty the duty of content’ and thus that it is open to the court to ‘formulate a different causal requirement in this class of case’.114 He suggests that saying that the claimants’ disease was the result of “significant exposure to asbestos during two employments over a period of eight years’ even if it is impossible to identify the point at which the fatal fibre was inhaled is not only ‘a meaningful causal state­ ment’, it is in fact ‘the only kind of causal statement about the disease’ that a scientific expert would allow in the current state of knowledge. Nothing in logic prevents this from being regarded as sufficient in law; the only question is whether this approach would be ‘just and reasonable and whether the class of cases to which it applies can be sufficiently clearly defined’.115 In answering this question, Lord Hoffmann is clear that in cases meeting the criteria he has laid down, justice, the policy of the common law and morality all sup­ port the proposed approach.116 In his discussion of authority, Lord Hoffmann is also force­ ful on the need for clarity of language in describing what the court is doing in such cases and provides a clarification of the position after McGhee.117 Whereas it may seem that their Lordships in that case treated materially increasing a risk as equivalent to materially con­ tributing to the harm,118 this cannot be the case when the medical experts explicitly refused to say whether one scenario was more likely than another. Rather Lord Hoffmann suggests that what the court meant to say was that in the particular circumstances ‘a breach of duty which materially increased the risk should be treated as if it had materially contributed to the disease’.119 Lord Hutton, while echoing the previous speeches in clarifying the nature of the prob­ lem (the inability to assign probabilities to different scenarios) and also agreeing with them as to the outcome, nevertheless departs from them in his discussion of what a judge may legitimately do in such circumstances. Whereas the speeches discussed so far have been clear that it is not appropriate to draw an inference to bridge an evidentiary gap, Lord Hutton considers that the ‘approach, whereby the layman applying broad common sense draws an inference which the doctors as scientific witnesses are not prepared to draw, is one which is permissible’.120 He also reads the decision of the House of Lords in McGhee in that light.121   ibid [46].   ibid [49]. 112   ibid [52]. 113   ibid [54], [60]. 114   ibid [62]. 115   ibid [62]. 116   ibid [63]. 117   McGhee v National Coal Board [1973] 1 WLR 1. 118   ibid, 42–43 per Lord Reid. 119   [2002] UKHL 22, [2003] 1 AC 32 [65]. 120   ibid [100]. 121   ibid [109]. See also J Stapleton, ‘Lords a’leaping evidentiary gaps’ (2003) 10 Tort Law Review 276. 110 111

402  John Paterson By contrast, Lord Rodger of Earlsferry adopts the approach seen in the earlier speeches. He is particularly clear on the nature of the problem facing the court and what is expected of the court in these circumstances. It is impossible for the claimants to prove a causal link on the balance of probabilities between the breach of duty by any one or more employers and the harm they have suffered ‘[b]ecause of the current state of medical knowledge about the aetiology of mesothelioma’.122 More importantly, Lord Rodger returns to the line of reasoning of the earlier speeches and departed from by Lord Hutton with respect to the role of the judge in such circumstances. Indeed, he is explicit that ‘the proper interpretation of McGhee is critical’.123 While the judge rather than the expert witness is the ultimate arbiter of matters of fact, the judge must nevertheless act according to the evidence. If it is not possible on the basis of undisputed scientific evidence to assign probabilities to different possible causes, then it is not for the judge to draw an inference on the basis of common sense. The judge must be clear about what he or she is doing and not resort to fictions.124 Nor does Lord Rodger evade the impact of the application of a different, less stringent rule of causation. The result may involve ‘rough justice’ with respect to one or more of the employers and this ‘must be faced squarely’.125 He is clear, however, that the ‘opposing potential injustice to claimants should also be addressed squarely’.126 In weighing the alternatives he is influenced by the fact that there was no wrongdoing on the part of the claimants but acknowledged wrongdoing on the part of the defendants. Should any defend­ ants not have caused or contributed to the harm, then this was at best a matter of luck. In addition, Lord Rodger helpfully clarifies the nature of the exception that was established in McGhee: it is hardly a relaxation when all that has been done is to ‘set the requirement of proof at the highest the pursuer could possibly attain’.127 To conclude this discussion of an example of judicial decision-making in the context of scientific uncertainty, the fact that it was impossible to establish causation on the balance of probabilities according to the standard ‘but for’ test, did not preclude a remedy where in the particular circumstances of the case, the claimant had proved all that he could given the state of scientific knowledge and the court was able to establish that the implementation of a modified test was justified on the grounds that justice was thereby served. While (with the exception of Lord Hutton) being clear that the evidentiary gap was not being bridged by way of inference or by judicial intuition or by the application of a common sense test, any of which would have been to ignore the accepted scientific evidence as to the inability to assign probabilities, the House of Lords was able to say that the requirements of causation had been met.128

  ibid [124].   ibid [126]. 124   ibid [150]. 125   ibid [155]. 126   ibid [155]. 127   ibid [155]. 128   It is important to note that the exception to the ‘but for’ test is narrowly drawn in Fairchild and has been held not to apply, for example, in cases of medical negligence. See Gregg v Scott [2005] 2 AC 176. See, further, R Goldberg, ‘Causation and Defences’ in A Grubb, J McHale and J Lang (eds), Principles of Medical Law, 3rd edn (Oxford, Oxford University Press, 2010), paras 6.27–6.28. 122 123

Law’s Approach to Harm under Uncertainty  403

IV. Reading Across Between Precautionary Decisions and Cases of Indeterminate Causation: Differences and Similarities

In the preceding two sections of this chapter, two cases have been considered in which the significant factor that the courts have had to deal with has been the presence of scientific uncertainty of a sort that has raised serious difficulties for the parties seeking a remedy, whether judicial review of a ministerial decision or the establishment of causation in negli­ gence. We now consider what the two cases reveal about the differences and similarities between precautionary decisions and cases of indeterminate causation. On one view, the two cases appear very different. In Downs, it may be said that the decision-maker (and by extension the judge) must deal with the situation where there is a concern that a causal link may ultimately prove to exist between event A (the issue that he or she is asked to decide upon) and event B (a hypothetical harm) but it is currently impossible to say using established quantitative risk assessment techniques whether or not event B would or could arise. In Fairchild, it may be said that the judge is asked to deal with the situation where the existence of event C (a breach of duty) and event D (a harm) can be established but it is currently impossible on the basis of the established legal test to prove the causal connection between them so as to establish liability. In other words, whereas the case involving the precautionary principle (Downs) has to deal with a hypothetical causal link between two events, the case involving indeterminate causation (Fairchild) has to confront the fact that a causal link does exist between two events but that a variety of possible scenarios may be postulated, none of which can be shown to be more or less likely than any other. A closer look, however, reveals a significant similarity. In the first case (Downs), the court is asked to review an administrative decision where the decision-maker is required to apply the precautionary principle. This principle states that where there are threats of serious or irreversible harm, lack of full scientific certainty shall not be used as a reason for postpon­ ing cost-effective measures to prevent that harm.129 Looking at the various situations in which decisions may be made, it becomes clear that the precautionary principle is applica­ ble in those situations where the state of scientific knowledge makes it impossible to assign probabilities to an outcome or outcomes and thus established quantitative risk assessment techniques cannot be used. In other words, the situation is most appropriately character­ ised as a decision under ignorance rather than a decision under risk.130 In the second case, Fairchild, the court is faced with a situation of indeterminate causation where the state of scientific knowledge means that it will be impossible for a tort claimant to establish causa­ tion on the basis of the standard test. This too turns out to be a situation that is character­ ised by the impossibility of assigning probabilities to one or another of a range of possible causation scenarios. If it were possible to establish that the actions of one or more of the employers was more likely than those of the others to have resulted in harm, then a decision could be reached accordingly, but it is precisely impossible to reach any such con­ clusion. It may thus be suggested that this too is a decision under ignorance in the sense that the established methods employed by the court for balancing probabilities have no applica­ tion in this situation. The court is accordingly unable to decide on causation on the balance of probabilities and all else equal the claimant will be left without a remedy.   See n 17 above and associated text.   In the sense of Resnik’s schema discussed previously. See n 70 above and the accompanying text.

129 130

404  John Paterson The magnitude of the difficulty confronting the courts in such situations should not be underestimated and may be gauged by considering the problems posed here also for sci­ ence and philosophy. Even where it is possible to assign probabilities, this does not by any means imply that the decision will be uncontroversial. There are a variety of ways in which the relationship between probability and causation may be conceptualised scientifically, ranging from the modest suggestion that it is no more than a statement about non-causal facts regarding relative frequencies131 through to the complete rejection of the proposition that it is possible to consider causal relations on the basis of probabilities.132 Meanwhile, philosophers are clear about the difficulties they confront in producing an adequate philosophical conceptualisation of indeterminate causation.133 It has been suggested that philosophy in such situations has little option but to understand causation as involv­ ing ‘chance-raising’.134 Insofar as that is an accurate account of the extent of philosophy’s offering in this context, however, it is immediately clear that it finds itself confronting the precise problem of the inability to assign probabilities. The courts themselves, of course, have been at pains especially in the context of indeter­ minate causation cases to point out that the resolution of such problems is ultimately a legal matter, as opposed to a scientific or a philosophical one.135 This fact, however, together with the recognition of the significant similarity that exists at a deep level between pre­ caution and indeterminate causation situations makes the difference in approach all the more striking. Insofar as the presence of scientific uncertainty creates significant difficulties for the aggrieved party in each situation, why should the court be willing to engage sub­ stantively with that problem in one case but not in the other? Whereas Collins J in Downs departed from the traditionally deferential approach of the courts in the context of pre­ cautionary decision-making, the Court of Appeal very much returned to the previous posi­ tion. By contrast, the presence of scientific uncertainty of a similar quality in the context of indeterminate causation in Fairchild was by no means seen by the House of Lords as an obstacle to the reaching of a decision in favour of the claimant. The question that arises, therefore, is whether the two situations are so diverse that the different approach taken by the courts in each is understandable and justified. There is certainly a constitutional argument that could answer this question in the affirmative with­ out further ado. That is, to put it at its simplest, that when it comes to the making of deci­ sions that engage the precautionary principle, the issues at stake are so complex that the courts simply lack the resources as compared to the executive to become substantively involved beyond extreme cases where there is manifest error or Wednesbury irrationality. In other words, precautionary decision-making is constitutionally appropriately the preserve

131   For detail on the nature of the scientific (and more specifically, statistical) techniques relating probabilities to causation, see the contribution by P Dawid in this volume in ch 7. 132   For a discussion, see M Tooley, ‘Probability and causation’ in P Dowe and P Noordhof (eds), Cause and Chance: Causation in an Indeterministic World (London, Routledge, 2004) 77–119. See also the contribution by R Wright in this volume in ch 10. 133   A point noted by Hart and Honoré in the preface to the second edition of Causation in the Law, HLA Hart and T Honoré, Causation in the Law (n 88) xxxvii as they embarked on a response to the philosophical criticisms of their earlier work. See also C Hitchcock, ‘Routes, Processes and Chance-lowering Causes’ in P Dowe and P Noordhof (eds), Cause and Chance: Causation in an Indeterministic World (London, Routledge, 2004) 138. 134   M Ramachandran, ‘Indeterministic Causation and Varieties of Chance-raising’ in P Dowe and P Noordhof (eds), Cause and Chance: Causation in an Indeterministic World (London, Routledge, 2004) 152. See also the con­ tribution by P Dawid in this volume in ch 7. 135   See also Lord Hoffmann in this volume in ch 1.

Law’s Approach to Harm under Uncertainty  405 of the executive and the courts would overstep the proper boundaries of their competences insofar as they became involved in any more detailed consideration of the grounds on which a precautionary decision had been reached.136 It is certainly evident from a reading of the judgments at first instance and on appeal in Downs that this argument was very much present in the minds of the judges. Collins J acknowledged that the case was located at the boundaries of what was properly the sub­ ject of judicial review and explicitly referred to precedents warning judges against substi­ tuting their non-expert views for those of experts.137 The Court of Appeal nevertheless concluded that Collins J had failed to stay on the right side of the line in this regard and had in effect precisely substituted his view for that of the experts when he granted the application.138 It is at this point, however, that reading a precautionary case in conjunction with an indeterminate causation case offers a first suggestion that the judicial approach in one may shed light on that taken in the other. Recall that in both cases the judges have readily acknowledged that the problem they confront is caused by the inability to assign probabili­ ties to one or more scenario. Whereas in Downs, the ultimate decision was that this is a matter that must be dealt with by the executive, in Fairchild the judges had no hesitation in concluding that this was not a barrier to their reaching a decision that the standard ‘but for’ test would not allow. It is true that in Fairchild a wide range of arguments based on princi­ ple, policy, authority, justice and indeed morality were deployed to support the conclusion reached. Of more significance, however, given present concerns was the emphasis placed by the majority of their Lordships on the nature of the exercise in which the judge engages in such circumstances. Specifically, the judge is not drawing an inference to bridge an eviden­ tiary gap where experts have explicitly stated that they are unable to offer any guidance as to probabilities. In other words, he or she is not substituting a non-expert view for that of the experts, but is rather acknowledging that the decision is being taken on the basis of a legal understanding of causation rather than a scientific or philosophical one. This raises intriguing questions about the judicial treatment of precautionary decisionmaking. Insofar as the precautionary principle is properly engaged precisely where it is impossible to reach decisions on the basis of a scientific assessment of risk or the assign­ ment of probabilities, the basis for the decision must be something else. In the ultimate, that basis is a concern to avoid harm which, though hypothetical, can nonetheless not be ruled out and where considerations of, for example, irreversibility are in play. In other words, in such circumstances the decision may be informed by considerations of policy, principle, justice, morality and so on. Such a finding does not automatically lead to the conclusion that judges asked to review precautionary decisions should be at liberty to ride roughshod over legitimately exercised ministerial discretion on the basis that they regard overarching considerations of policy or principle as more important. It does, however, suggest that there is greater scope for judges to consider the way in which precautionary decisions have been made than is allowed by the Court of Appeal in Downs.

136  See E Fisher and R Harding, ‘The Precautionary Principle and Administrative Constitutionalism: the Development of Frameworks for Applying the Precautionary Principle’ in E Fisher, J Jones and R von Schomberg (eds), Implementing the Precautionary Principle: Perspectives and Prospects (Cheltenham, Edward Elgar, 2006). 137   Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin), [38]. 138   Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664, [76].

406  John Paterson Insofar as this is the case, it becomes imperative that there is clarity as to the methodol­ ogy that the judges must follow in reviewing precautionary decisions. It is suggested that such clarity certainly now exists in the context of indeterminate causation following Fairchild. The criteria for the application of this exception to the standard ‘but for’ test have been carefully drawn by the House of Lords. It is further suggested, however, that it is by no means difficult (and would in fact be highly desirable) to achieve similar clarity in the con­ text of the precautionary principle. Collins J took a bold step in that direction in Downs only to have his decision reversed by the Court of Appeal. The latter court may be defended on the grounds that it demonstrated legitimate concern for the limits of judicial review. It may be criticised, however, on the grounds that, in common with the minister, it did not properly appreciate the significance of the precautionary principle. Whether the decision will be appealed to the Supreme Court is unclear at the time of writing, but there would be an opportunity for the court at that stage – at the very least – to make a significant contri­ bution to the degree of precision with which the precautionary principle is understood and applied by the executive and, as a consequence, to ensure that this legal principle has mean­ ing and does not become an empty shell. Insofar as the Supreme Court might need encour­ agement to take such a step, it is suggested that it need look no further than the approach of the Lords in Fairchild. There it was recognised that a legal duty would be ‘emptied of all practical content’139 were the court to decline to develop an exception to the standard ‘but for’ test in the presence of scientific uncertainty. It is surely not stretching the analogy too far to suggest that unless the court is willing to grant the application for judicial review in Downs then the precautionary principle too will be emptied of content.

V. Conclusion

What emerges from this analysis of law’s approach to harm in the context of scientific uncertainty ex ante and ex post is the proposition that there may be a closer relationship between these two situations than is apparent at first sight and thus a stronger foundation for a more robust judicial stance in the review of precautionary decision-making in line with developments in cases of indeterminate causation. It is even possible to suggest that a symmetry exists between situations in which the precautionary principle and the Fairchild exception are respectively engaged. This is best illustrated by the following dia­ gram (Figure 1).140

  [2002] UKHL 22, [2003] 1 AC 32, [155] per Lord Rodger of Earlsferry.   Figure 1’s depiction of the three possible situations for decisions (certainty, risk and ignorance) can be criticised on the basis that there is actually a continuum between certainty and ignorance rather than a series of distinct stages. This is a fair point, but the fact remains that there will be clear examples of situations which are undisputedly decisions under risk or decisions under ignorance. There will also, of course, be situations about which there may be legitimate disagreement as to the appropriate location of the decision, but that fact simply requires decision-makers or judges to be explicit about their reasons for adopting one or another basis for the decision. 139 140

Law’s Approach to Harm under Uncertainty  407 Figure 1: The symmetrical relationship between law’s approaches to harm ex ante and ex post Law’s approach to harm ex ante

Degree of precaution

Sufficiency of data (decision type)

Degree of Indeterminacy

Law’s approach to harm ex post

Unlikely to result in judicial review

Low

High (Certainty)

Low

Unlikely to result in litigation

Quantified Risk Assessment

Medium

Medium (Risk)

Medium

‘but for’ test

Precautionary Principle

High

Low (Ignorance)

High

Fairchild exception

As the diagram shows, it is possible to see the problems confronting law in situations involving precaution or indeterminate causation as fundamentally involving issues of the sufficiency of data, which in turn affects the ability of science to assign probabilities to dif­ ferent propositions. In the case of precautionary decision-making, it is the precise inability to assign probabilities to hypothetical harms that takes the decision out of the realm of risk and into the realm of ignorance and properly engages the precautionary principle rather than quantified risk assessment. The question for the decision-maker thus becomes the appropriate application of the test of the plausibility of the hypothetical harm and, if appropriate, the test of the reasonableness of the response. Similarly, in the case of indeter­ minate causation, it is the inability to assign probabilities to a number of different possible scenarios which means that the standard ‘but for’ test cannot be applied and must be replaced by exceptional criteria to avoid injustice. Insofar as it is now possible to character­ ise this ex post decision as a decision under ignorance that mirrors the ex ante one that engages the precautionary principle, it is also possible to see that the ex ante application of the plausibility and reasonableness tests mirrors the ex post application of the Fairchild criteria. If it is accepted that this symmetry exists, the question then arises as to why there is greater judicial reticence to engage substantively with the ex ante issue than with the ex post one. The answer would appear to be that there is a lack of clarity at the level of the initial application of the precautionary principle by the executive. Greater specification of the tests that properly constitute the operationalisation of the principle would make it easier for judges asked to review such decisions to move beyond a deferential approach and to consider the manner in which they have been taken. In the absence of this sort of change, it is apparent that the principle risks being emptied of content – a fate that the court in Fairchild was keen to ensure did not befall the duty of care owed by a particular group of employers to a particular group of employees.141 This chapter’s message is thus first and foremost that the willingness of judges to con­ front scientific uncertainty in the context of liability should be matched by a willingness to confront it in the context of precautionary decisions. It might also be suggested, however,   [2002] UKHL 22 [2003] 1 AC 32, [155] (Lord Rodger of Earlsferry).

141

408  John Paterson that understanding law’s approach to harm in the context of scientific uncertainty in the terms proposed in this chapter could assist in any future extension of the courts’ willingness to engage substantively in situations of indeterminate causation beyond those currently covered by the Fairchild exception. As Carl Cranor has clearly demonstrated in this volume,142 the number and complexity of those situations is likely only to increase.

  Ch 13.

142

TABLE OF CASES Australia

Adeels Palace Pty Ltd v Moubarak [2009] HCA 48.............................................................244 Allianz Australia Insurance Ltd v GSF Australia Pty Ltd (2005) 221 CLR 568...................243 Amaca Pty Ltd v Ellis [2010] HCA 5................................................... xxvi, 252, 253, 259, 260 Burk v Commonwealth of Australia [2006] VSC 25....................................................249, 250 Burk v Commonwealth of Australia [2008] VSCA 29.................................................249, 250 CAL No 14 Pty Ltd v Scott (2009) 239 CLR 390..................................................................242 Cavenett v Commonwealth of Australia [2005] VSC 333............................................249, 250 Cavenett v Commonwealth of Australia [2007] VSCA 88...........................................249, 250 Chappel v Hart (1998) 195 CLR 232................................................................................7, 255 Commissioner for Government Transport v Adamcik (1961) 106 CLR 292 (Australian HC)................................................................................. xxvi, 245, 246, 247, 259 Dasreef Pty Ltd v Hawchar [2011] HCA 11.................................................................241, 259 Doo v Murphy (unreported, Supreme Court of Queensland, 8 October 1998) (BC 9805163)......................................................................................................................254 Gavalas v Singh (2001) 3 VR 404..........................................................................................257 Gett v Tabet (2009) 254 ALR 504..........................................................................................257 Glenmont Investments Pty Ltd v O’Loughlin (2000) 79 SASR 185....................................254 Lindsay-Field v Three Chimneys Farm Pty Ltd [2010] VSC 436........................................254 Lowns v Woods [1986] Aust Torts Rep 81–376....................................................................255 Malec v JC Hutton Pty Ltd [1990] HCA 20; (1990) 169 CLR 638......................253, 254, 260 Mansini v Martin [1998] QCA 222.......................................................................................254 March v Stramare (E & MH) Pty Ltd (1991) 99 ALR 423 (HCA); [1991] HCA 12; 171 CLR 506.................................................................... 242, 243, 244, 294 Naxakis v Western General Hospital (1999) 197 CLR 269..................................................256 Norris v Blake (by his Tutor Porter) (No 2) (1997) 41 NSWLR 49; (1997) 25 MVR 101............................................................................................................254 R v Danielson and McHardie [1983] 2 NSWLR 733...........................................................250 R v GK [2001] 53 NSWLR 317..............................................................................................250 R v Karger [2001] SASC 64........................................................................... xxvi, 248, 250, 259 R v Mallard [2003] WASCA 296............................................................................................250 R v Parenzee [2007] SASC 143, on appeal [2007] SASC 316.............. xxvi, 245, 247, 248, 259 R v Parker [1912] VLR 152....................................................................................................259 Rufo v Hosking (2004) 61 NSWLR 678; [2004] NSWCA 391.....................................256, 257 Sellars v Adelaide Petroleum NL [1994] HCA 4; (1994) 179 CLR 332.......................254, 260 Seltsam v McGuiness (2000) 49 NSWLR 262; [2000] NSWCA 29................................................................................. xxvi, 250, 251, 252, 259, 260 Seltsam Pty Limited v Ghaleb [2005] NSWCA 208.............................................................254

410  Table of Cases Sullivan v Moody (2001) 207 CLR 562.................................................................................241 Tabet v Gett [2010] HCA 12; (2010) 240 CLR 537.................................... xxvi, 243, 244, 254, 256, 257, 258, 259, 260 Tabet v Mansour [2007] NSWSC 36.....................................................................................257 Travel Compensation Fund v Tambree (2005) 225 CLR 627......................................243, 244 Tszyu v Fightvision Pty Ltd; Fightvision Pty Ltd v Onisforou [1999] NSWCA 323..........254 Wilson v State of Tasmania [1999] TASSC 145....................................................................254 Wright v Commonwealth of Australia [2005] VSCA 309....................................................254

Canada

Athey v Leonati (1996) 3 SCR 458 (Supreme Ct)..........................................................93, 294 Cook v Lewis [1951] SCR 830 (Supreme Court).............................................................59, 75 Duncalf v Capital Health Authority [2009] ABQB 80...........................................................98 Hanke v Resurfice Corp [2007] 1 SCR 333( Supreme Court).... 93 Laferriere v Lawson [1991] 1 SCR 541(Supreme Court).....................................................257 R v C (G) (1996) 110 CCC (3d) 233 (Newfoundland CA).................................................244 R v DD [2000] SCC 43...........................................................................................................244 Snell v Farrell [1990] 2 SCR 311 (Supreme Ct)............................................ xxi, 93, 94, 95, 96, 102, 103, 107, 108, 109

European Union

Commission v Cambridge Healthcare Supplies Ltd, Case C–471/OOP(R), [2001] ECR I–2865.............................................................................................................391 Kingdom of Sweden v Commission, Case T–229/04, [2007] ECR II–2437 (CFI)....................................................................................................................392, 396, 397 Parliament v Council, C–303/94, ECR I–2943, judgment of 18 June 1996 (ECJ)..............391

France

Cour de Cassation Cass Ass Plen Perruche 17 November 2000, Gazette du Palais, 24–25 January 2001; D 2001 Jurisprudence 316............................................................................................. 127–9 Cass civ 27 December 1911: S 1911, 1, 383...........................................................................116 Cass civ 1 14 December 1965, JCP G, 1966, II, 14753, note R Savatier...............................124 Cass civ 1 16 June 1969, JCP G 1970, II, 16402, note R Savatier.........................................117 Cass civ 1 27 January 1970, Bull I no.37, 31, JCP G 1970, II, 16422 note Rabut................124 Cass civ 1 25 November 1971, Bull civ I, no.296..................................................................120 Cass civ 1 27 March 1973, Bull civ I, no.115.........................................................................124

Table of Cases  411 Cass civ 1 17 November 1982, Bull civ I, no.133; JCP 1983, II, 20056, note M Saluden; D 1984, 305, note A Dorsner-Dolivet............................................................124 Cass civ 1 8 April 1986, RTD civ 1987, p 557 obs J Huet.....................................................120 Cass civ 1 13 December 1988, Defrénois 1989, art 34554, no.56 obs J-L Aubert...............120 Cass civ 1 17 February 1993, JCP G. 1994 II, 22226, note A Dorsner-Dolivet............118, 127 Cass civ 1 12 January 1994 and 2 February 1994, JCP G, II, 22294 note Ph Delebecque...114 Cass civ 1 20 June 1995, JCP G, IV, 2010..............................................................................114 Cass civ 1 8 July 1997, Bull Civ I no.239...............................................................................124 Cass civ 1 30 September 1997, Bull civ I no.259; RCA 1997 comm 373.............................121 Cass civ 1 21 October 1997, D aff, 1419................................................................................114 Cass civ 1 12 December 2000, Bull civ I no.323, D 2001 jur p 1650, note C Paulin...114, 127 Cass civ 1 July 2002, D 2002 jur, 2631 note J P Gridel; RTD civ 2002, 821 obs P Jourdain...........................................................................................................................127 Cass civ 1 10 July 2002, Bull Civ I, no.197............................................................................124 Cass civ 1 23 September 2003; Resp civ et assur 2003, chron 28, Ch RadΘ; D 2004, 898, note Y-M SΘrinet and R. Mislawski; JCP G 2003, II, 10179; RLDC 2004, 11, chron S Hocquet-Berg; RTD civ 2004, obs P Jourdain.....................................................124 Cass civ 1 22 May 2008, no.06–18.848..........................................................................126, 174 Cass civ 1 22 May 2008, no.05–10.593..........................................................................126, 174 Cass civ 1 22 May 2008, no.05–20.317..........................................................................126, 174 Cass civ 1 22 May 2008, no.06–10.967..........................................................................126, 174 Cass civ 1 22 May 2008, no.06–14.952..........................................................................126, 174 Cass civ 1 22 May 2008, no.07–17.200..........................................................................126, 174 Cass civ 1)22 January 2009, no.07–16.449, Bull civ 2009, I, no.11; RDC 2009 no.3, 1028, obs O Deshayes.................................................................................................127, 175 Cass civ 1 25 June 2009, no.08–12.781; Bull civ 2009, I, no.41, 308, note P Sargos....127, 174 Cass civ 1 9 July 2009, no.08–11.073............................................................ xxiv, 126, 174, 175 Cass civ 1 24 September 2009, no. 08–10.081 and no.08–16.305, D 2009 AJ 2342, obs Gallmeister; ibid 2010. Pan 49 obs Brun; JCP 2009. 304 obs Mistretta; ibid 383, note Hocquet-Berg; RLDC 2009/65, n 3605, obs Bugnicourt; RLDC 2010/01, n 67–10, note B Parance......................................................................................124 Cass civ 1 24 September 2009, no.08–16.097...............................................................175, 127 Cass civ 1 28 January 2010, no.08–18837.............................................................................124 Cass civ 1 25 November 2010, no.09–16.556............................................................... xxiv, 175 Cass civ 2 4 March 1966, D 1966 somm 110........................................................................119 Cass civ 2 28 November 1984, JCP G 1985, II, 20477 note no.Dejean de la Bâtie..............114 Cass civ 2 4 February 1987, Bull civ II, no.38.......................................................................121 Cass civ 2 7 December 1988, Bull civ II, no.246; RTD civ 1989, 557, obs P Jourdain........119 Cass civ 2 8 February 1989, JCP G 1990, II, 21544 note no.Dejean de la Bâtie; RTD civ 1989, 556, obs P Jourdain............................................................................117, 120 Cass civ 2 2 April 1997, RCA com no.255, note F Leduc.....................................................114 Cass civ 2 27 January 2000, JCP G 2000, I chron 247, obs G Viney ; JCP G II, 10363, note Ph Conte, RCA 2000 comm no.109; RTD civ 2000, 335, obs P Jourdain........117, 127 Cass civ 2 15 June 2000, note G Blanc, D 2001, 886; RTD civ 2000, 849 obs P Jourdain; RCA 2000 comm 292 obs H Groutel.................................................................................114 Cass civ 2 20 June 2002, RCA 2002 comm no.279...............................................................120 Cass civ 2 23 January 2003, D 2003, 2465, note V Depadt-Sebag........................................127

412  Table of Cases Cass civ 2 27 March 2003, Bull civ II, no.76; JCP G 2004, I, 101, no.13, obs G Viney.........................................................................................................................120 Cass civ 2 18 September 2003, D 2004jur p 25 note no.Damas; RTD civ 2004, 108 obs Jourdain RCA 2003 comm 286, obs Groutel, JCP 2004, I, 101 no.18 obs G Viney, JCP G 2004 II 10013 note C Le Tertre; Defrénois 2004 no.11456 obs R Libchaber.........................................................................................................................117 Cass civ 2 20 November 2003, D 2003, 2902........................................................................121 Cass civ 2 11 December 2003, RCA 2004 comm 61.............................................................115 Cass civ 2 17 February 2005, RCA 2005, comm 119............................................................114 Cass civ 2 20 October 2005, Bull civ II no.274, RTD civ 2006, p 122, obs P Jourdain; D 2006 pan 1930 obs P Jourdain.......................................................................................120 Cass civ 3 23 February 1972, Bull civ III, no.130..................................................................125 Cass civ 3 27 June 1972, Bull civ III, no.427.........................................................................124 Cass civ 3 19 February 2003, no.00–13253, RCA 2003, comm no.125, RTD civ 2003, 508.............................................................................................................................120 Cass crim 19 May 1958, Bull crim no.395 D 1958, somm p 149.........................................115 Cass crim 11 January 1961, Bull crim 1961, no.18...............................................................118 Cass crim 14 January 1970, Bull crim 1970, no.23; RTD civ 1970, 574 obs G Durry.........117 Cass crim 14 January 1971, D 1971, p 164, note E Robert...................................................115 Cass crim 8 July 1971, D 1971, p 625, note E Robert...........................................................115 Cass crim 18 October 1995, dr pénal 1996, comm 78, note Véron.....................................115 Cass crim 14 February 1996, Bull crim no.78; Rev sc crim 1996, 856, obs Y Mayaud.......115 Cass crim 23 September 1998, Rev sc crim 1999, 321, obs Y Mayaud................................115 Cass crim 5 October 2004, D IR, 2972. contra Cass crim 15 January 1958 JCP G, 1959, II, 11026, note P Esmein...................................................................................118, 121 Cass com 26 May 1964, Bull civ III, no 271. F Ferrand, no 475...........................................124 Cass soc 7 May 1943, DH 1943, 51........................................................................................119 Cass soc 19 July 1962, Bull civ 1962, IV, no.670...................................................................116

Conseil d’Etat CE 14 October 1966, Marais, (1966) Dalloz Jurisprudence 636..................................121, 122 CE 4 October 1968, Doukakis, [1968] Rec 1100..................................................................116 CE 27 March 1985, Henry (1985).........................................................................................122 CE 1 February 1995, de Bray, (1995) Rec 60.........................................................................122 CE 9 June 1995, Lesprit (1995) AJDA 745............................................................................121 CE 16 June 1999, Tripot Req 177075 (unreported)..............................................................121 CE 9 March 2007, no.267635; JCP A 2007, 2108, note D Jean-Pierre; 2277, note S Carpi-Petit, act 297; JCP G 2007, II, 10142, note A Laude............................127, 176 CE, s. 21, December 2007, Centre Hospitalier de Vienne....................................................124

Cour d’Appel Bourges, 22 January 2009, no.07/01489................................................................................175 Paris 7 July 1989, Courtellemont, Gaz Pal 2, 752 concl Pichot............................................118 Paris 20 November 2003, RCA 2004 comm no.76 obs Ch Radé.........................................120 Paris 25 April 2000, Grillet, Gaz Pal 2001, 1, 478, note F Chabas........................................114 Paris, 9 January 2009 no.04/19067.... 175

Table of Cases  413 Paris, 9 January 2009, no.04/19068.......................................................................................175 Paris, 19 June 2009, no.06/13741...........................................................................................175 Versailles, 28 November 2003 D 2004 jur p 2814, note P Hennion-Jacquet.......................124

Cour administrative d’Appel CAA Marseille, 3 May 2006, Société AXA Courtage, Req 03MA00948...............................121 CAA Nantes, 25 October 1990, Baston, Req 89NT00137....................................................116

Germany

BGH 17 February 1970 (‘Anastasia Decision’), 53 BGHZ 245 (Federal Court of Justice).................................................................................................................................195 BGH 6 October 1998, 1999 NJW 860 (Federal Court of Justice)................................199, 217 ‘Leather spray case’, 37 BGHSt 106 (1990) (Federal Supreme Ct)...............................217, 350

Italy

Cass civ 22nd April 1993, (Sezione Lavoro), 4725 [1993]...................................................217 Cass civ 28th April 1994, (Sezione Terza), 4044 [1994].......................................................217 Cass civ 19th November 1997, (Sezione Lavoro), 11522 [1997].........................................217 Cass civ 6th February 1998, (Sezione Terza), 1286 [1998]...................................................217 Trib Florence 07th January 1999, no 2222, in Resp CP (2000).... 217 Cass 21.01.2000, n. 632..........................................................................................................198 Cass 04.03.2004, n. 4400........................................................................................................198 Cass 05.09.2006, n. 19047......................................................................................................198 Cass 18.04.2007, n. 9238........................................................................................................198 Cass 16.10.2007, n. 21619......................................................................................................198 Cass Civ SU 581 [2008] s 3.9.........................................................................................198, 199 Cass Pen, SU 11.09.2002, n. 30328........................................................................................198

Netherlands

B v Bayer Nederland BV, Hoge Raad 9 October 1992, [1994] Nederlandse Jurispruzentie (NJ) 535 (CJHB).................................................................................................................214

New Zealand

Sew Hoy & Sons Ltd v Coopers & Lybrand (1996) 1 NZLR 392 (NZ CA).........................294

414  Table of Cases United Kingdom

AB v Ministry of Defence [2009] EWHC 1225 (QB).............................................................81 AB v Ministry of Defence [2010] EWCA Civ 1317; (2011) 117 BMLR 101.......152, 193, 194 Alcock v Chief Constable of South Yorkshire Police [1992] 1 AC 310 (HL)..................63, 64 Allen v British Rail Engineering Ltd [2001] PIQR Q101.....................................................191 Allied Maples v Simmons & Simmons [1995] 1 WLR 1602..................................................53 Alphacell Ltd. v Woodward [1972] 2 All E.R. 497............................................................94, 95 Aneco Reinsurance v Johnson & Higgins [2001] UKHL 51; [2001] 2 All ER (Comm) 929.........................................................................................................................52 Ashington Piggeries Ltd v Christopher Hill Ltd [1972] AC 441..........................................149 Associated Provincial Picture Houses Ltd v Wednesbury Corporation [1947] 2 All ER 680......................................................................................... 389, 391, 392, 394, 404 Axa General Insurance Ltd v Lord Advocate (June 2011).........................................57, 88, 91 Axa General Insurance, Petition by [2009] CSOH 57............................................................89 Axa General Insurance, Petition for Judicial Review [2010] CSOH 2................ 82, 88, 89, 90 Baker v Willoughby [1970] AC 467 (HL)...............................................................................67 Barker v Corus (UK) plc [2006] UKHL 20; [2006] 2 AC 572 (HL)......... xxi, xxviii, 8, 28, 29, 30, 31, 33, 34, 36, 38, 39, 42, 57, 58, 61, 62, 63, 64, 65, 66, 67, 68, 69, 71, 74, 75, 77, 78, 79, 80, 81, 82, 83, 90, 191, 192, 193, 194, 214, 226, 227, 252, 326, 398 Barnett v Chelsea and Kensington Hospital Management Committee [1969] 1 QB 428......................................................................................................................371, 372 Blower & others v Edwards & Scottish Widows plc [2010] CSOH 34..................................54 Bonnington Castings Ltd v Wardlaw [1956] AC 613..................................... 66, 154, 191, 327 Bourhill v Young [1943] AC 92 (HL)......................................................................................63 Carter v Basildon and Thurrock University Hospitals NHS Foundation Trust [2007] EWHC 1882 (QB); [2007] LS Law Medical 657...............................................................151 Cartledge v E Jopling & Sons Ltd [1963] AC 758 (HL).........................................................83 Chester v Afshar [2004] UKHL 41; [2005] 1 AC 134.........................................................7, 52 Corr v IBC Vehicles Ltd [2008] UKHL 13; [2008] 1 AC 884...............................................378 Davis & Doherty v Balfour Kilpatrick [2002] EWCA Civ 736.............................................149 Dingley v Chief Constable of Strathclyde Police 1998 SC 548.................... 159, 160, 190, 245 Donoghue v Stevenson [1932] AC 562 (HL)..........................................................................64 Downs v Secretary of State for Environment, Food and Rural Affairs [2008] EWHC 2666 (Admin)......................................... xxx, 387, 389, 390, 394, 395, 403, 404, 405 Dulieu v White & Sons [1901] 2 KB 669 (KBDC).................................................................63 Durham v BAI (in run off) & ors [2009] 2 All ER 26..........................................................194 Environment Agency (formerly National Rivers Authority) v Empress Car Co (Abertillery) Ltd [1999] 2 AC 22...................................................................................6, 7, 9 Fairchild v Glenhaven Funeral Services Ltd, Fox v Spousal (Midlands) Ltd, Matthews v Associated Portland Cement Manufacturers (1978) Ltd and others [2002] EWCA CIv 1881; [2002] 1 WLR 1052 (CA)..............xxv, 59, 60, 65, 193, 398 Fairchild v Glenhaven Funeral Services Ltd, Fox v Spousal (Midlands) Ltd, Matthews v Associated Portland Cement Manufacturers (1978) Ltd and others [2002] UKHL 22; [2003] 1 AC 32.......................................... xx, xxi, xxviii, xxx, 8, 9, 11, 23, 24, 25, 26, 27, 28, 29, 30, 31, 31–40, 57, 58, 59, 61, 62, 63, 65, 66,

Table of Cases  415 67, 68, 69, 70, 74, 75, 77, 78, 79, 80, 81, 82, 86, 90, 91, 95, 96, 129, 192, 193, 194, 226, 243, 244, 252, 294, 337, 373, 398, 399, 400, 401, 402, 403, 404, 405, 406, 407, 408 First Interstate Bank of California v Cohen Arnold [1996] 5 Bank LR 150; 140 Sol Jo LB 12; [1996] PNLR 17.......................................................................................53 Fitzgerald v Lane [1987] 3 WLR 249; [1987] QB 781 (CA).....................................28, 62, 105 Frost (sub nom White) v Chief Constable of South Yorkshire Police [1999] 2 AC 455....................................................................................................................64, 66, 69 Grant v Australian Knitting Mills [1936] AC 85..................................................................209 Gray v Thames Trains [2009] UKHL 33; [2009] 1 AC 1339.... 377 Gregg v Scott [2005] UKHL 2; [2005] 2 AC 176 (HL).........................9, 29, 62, 63, 68, 69, 71, 82, 83, 162, 192, 218, 227, 229, 257, 402 Grieves v FT Everard see Pleural Plaques Litigation, In Re Halifax Life Ltd v DLA Piper Ltd [2009] CSOH 74...............................................................53 Hamilton v Allied Domecq plc [2001] SC 89 (OH); 2006 SC 221 (IH)...............................52 Henderson v The Royal Bank of Scotland [2008] CSOH 146; 2008 SCLR 823.........................................................................................................................51, 52 Hicks v Chief Constable of the South Yorkshire Police [1992] 2 All ER 65 (HL).................83 HM Commissioners of Customs and Excise v Barclays Bank plc [2006] UKHL 28; [2007] 1 AC 181..............................................................................................................63, 74 Holtby v Brigham & Cowan (Hull) Ltd [2000] 3 All ER 421; [2000] ICR 1086 CA (Civ Div).......................................................................................................................26, 191 Hope and Reay v British Nuclear Fuels [1994] 5 Med LR 1........................................188, 189 Hope v BNFL [1994] 5 Med LR 1.........................................................................................149 Hotson v East Berkshire Area Health Authority [1987] 2 WLR 287; [1987] AC 750 (HL).............................................................. 60, 61, 69, 74, 154, 161, 162, 188, 218, 227, 372 Ikarian Reefer, The [1993] 20 FSR 563.................................................................................248 JD v East Berkshire Community Health NHS Trust [2005] UKHL 23; [2005] 2 AC 373 [100]......................................................................................................................61 Johnston v NEI International Combustion Ltd see Pleural Plaques Litigation, In Re Keith v Davidson Chalmers 2004 SC 287; 2003 SCLR 941..............................................51, 55 Kerr v Stiell Facilities Ltd [2009] CSOH 67............................................................................42 Kuwait Airways Corporation v Iraqi Airways Co (Nos 4 and 5) [2002] UKHL 19; [2002] 2 AC 883..............................................................................................................6, 378 Leeds & Holbeck Building Society v Alex Morison & Co (No 2) 2001 SCLR 41.................51 Loveday v Renton [1990] 1 Med LR 117...............................................................149, 188, 190 McFarlane v Tayside Health Board [2000] 2 AC 59 (HL)......................................................74 McGhee v National Coal Board [1972] 3 All ER 1008 (HL); [1973] 1 WLR 1 (HL); 1973 SC (HL) 37; 1973 SLT 14.............................................xxx, 23, 24, 25, 27, 28, 31, 34, 42, 56, 60, 61, 62, 93, 94, 95, 96, 154, 192, 193, 216, 252, 400, 401, 402 McKew v Holland & Hannen & Cubitts (Scotland) Ltd [1969] 3 All ER 1621 (HL).............................................................................................................................363, 376 McLoughlin v O’Brian [1983] 1 AC 410 (HL).......................................................................64 McTear v Imperial Tobacco Limited 2005 2 SC 1............... xxiii, xxiv, xxv, 150, 156, 157, 158, 159, 160, 161, 162, 163, 177, 188 Main v McAndrew Wormald Ltd 1988 SLT 141...................................................................159

416  Table of Cases Mallett v McMonagle [1970] AC 166....................................................................................218 Murray v British Shipbuilders (Hydrodynamics) Ltd..........................................................326 Novartis Grimsby Ltd v Cookson [2007] EWCA Civ 1261...........................................31, 152 Patterson v Smiths Dock Ltd and others [2006] UKHL 20; [2006] 2 AC 572....................326 Performance Cars Ltd v Abraham [1962] 1 QB 33..............................................................191 Perrett v Collins [1998] 2 Lloyd’s Rep 255..............................................................................88 Philco Radio & Television v J Spurling [1949] 2 All ER 882................................................376 Pleural Plaques Litigation, Re [2006] EWCA Civ 27..................................................73, 82, 83 Pleural Plaques Litigation, In Re: Rothwell v Chemical and Insulating Company Ltd; Johnston v NEI International Combustion Ltd; Grieves v F T Everard & Sons [2007] UKHL 39; [2008] 1 AC 281................................... xxvi, 57, 59, 63, 69, 74, 81, 82, 83, 84, 89, 90, 222, 223, 224, 225, 234 Preferred Mortgages v Shanks [2008] CSOH 23....................................................................52 R v Doheny & Adams [1997] 1 Cr App R 369......................................................................250 R v Kennedy (No 2) [2007] UKHL 38; [2008] 1 AC 269......................................... 5, 9, 45, 46 R v Latcha (1999) 104 A Crim R 390....................................................................................250 R v T [2010] EWCA Crim 2439....................................................................................245, 250 R (on the application of British Union for the Abolition of Vivisection) v Secretary of State for the Home Department [2008] EWCA Civ 417..................................................389 Reay v British Nuclear Fuels [1994] 5 Med LR 1..................................................................149 Reeves v Commissioner of Police of the Metropolis [2000] AC 360 (HL).....................6, 376 Rothwell see Pleural Plaques Litigation, In Re Rylands v Fletcher (1868) LR 3 HL 330....................................................................................6 Sanderson v Hull [2008] EWCA Civ 1211; [2009] PIQR P7...........................................31, 62 Sayers & ors v Smithkline Beecham Plc & ors [2006] EWHC 3179....................................194 Sayers and others v Smithkline Beecham plc and others [2007] EWHC 1335 (QB); [2007] All ER (D) 67 (June)...............................................................................................164 Sayers v Smithkline Beecham Plc, Smith Kline & French Laboratories Limited, Merck & Co Inc, Sanofi Pasteur MSD Limited [2007] EWHC 1346 (QB)..............167, 194 Secretary of State for Environment, Food and Rural Affairs v Georgina Downs [2009] EWCA Civ 664.................................................................xxx, 391, 392, 395, 396, 405 Sienkiewicz v Greif (UK) Ltd [2009] EWCA Civ 1159; [2010] QB 370................62, 153, 189 Sienkiewicz v Greif (UK) Ltd [2011] UKSC 10; [2011] 2 WLR 523.....................Preface, xxi, xxiii, xxv, 42, 57, 58, 60, 62, 65, 66, 67, 68, 76, 79, 81, 82, 87, 90, 97, 99, 101, 104, 153, 154, 155, 156, 161, 188, 189, 190, 191, 193, 194, 200, 202, 203, 204, 207, 208, 215, 216, 218, 219, 244, 251, 252 Smith v McNair [2008] CSOH 154.......................................................................................159 South Australia Asset Management Corporation v York Montague Ltd [1997] AC 191.............................................................................................................................52, 53 Spencer v Wincanton Holdings Ltd [2009] EWCA Civ 1404..............................................378 Spring v Guardian Assurance plc [1995] 2 AC 296 (HL).................................................53, 74 Stansbie v Troman [1948] 2 KB 48............................................................................................6 Stapley v Gypsum Mines Ltd [1953] AC 663........................................................................243 Thompson v Smiths Shiprepairers (North Shields) Ltd [1984] 1 All ER 881.....................191 Transco plc v Stockport Metropolitan Borough Council [2004] 2 AC 1................................6 Vadera v Shaw (1998) 45 BMLR 162 (CA)...........................................................................151 Wardlaw v Bonnington Castings 1956 SC (HL) 26; [1956] AC 613...........................24,27, 56

Table of Cases  417 Wardlaw v Farrar [2003] 4 All ER 1358; [2004] Lloyd’s Rep Med 98; [2004] PIQR 19...............................................................................................................................161 White v Jones [1995] 2 AC 207 (HL)......................................................................................74 Wilsher v Essex Area Health Authority [1987] QB 730 (CA)............................ 25, 27, 28, 191 Wilsher v Essex Area Health Authority [1988] 1 All ER 871; [1988] AC 1074 (HL)........................................................... 23, 24, 25, 26, 27, 28, 29, 34, 60, 61, 62, 68, 69, 94, 95, 96, 99, 102, 107, 108, 191, 193 Wright v Dunlop Rubber Co Ltd [1972] 13 KIR 255...........................................................149 X (Minors) v Bedfordshire CC [1995] 2 AC 633 (CA)..........................................................61 XYZ v Schering Health Care Ltd [2002] EWHC 1420; (2002) 70 BMLR 88 (QB)............................................................................................. 152, 153, 154, 188, 190, 194

United States of America

Alberts v Schultz, 975 P2d 1279 (NM 1999)........................................................................233 Althen v Sec’y of Health & Human Services, 418 F 3d 1274 (Fed Cir 2005)............................................................................................. 170, 172, 173, 176 Andreu v Sec’y of Health & Human Services, 569 F 3d 1367 (Fed Cir 2009).....................171 Armstrong World Industries Inc v Aetna Casualty & Surety Co. (1996) 45 Cal.App.4th 1, 37–39 (Cal.Rptr.2d 690)..............................................................................20 Baird v American Medical Optics, 713 A2d 1019 (NJ 1998)...............................................234 Borel v Fibreboard Paper Products Corp, 493 F2d 1076 (5th Cir 1973), cert denied 419 US 869 (1974)...........................................................12, 15, 16, 17, 18, 37, 40 Borg-Warner v Flores, 232 SW3d 765 (Tex 2007)......................................................22, 23, 32 Brown v Board of Education, 347 US 483 (1954)................................................................109 Brown v Superior Court, 751 P2d 470 (Cal 1988).........................................................30, 230 Capizzano v Sec’y of Health & Human Services, 440 F 3d 1317 (Fed Cir 2006)................170 Cedillo v Sec’y of Health & Human Services, 2009 WL 331968 (Fed Cl 12 February 2009), aff’d 89 Fed Cl 158 (2009), aff’d 617 F 3d 1328 (Fed Cir 2010)........165–8, 170–3, 176 Celotex Corp, Matter of, 196 B.R. 973 (Bankr MD Fla 1996)................................................12 Cipollone v Liggett Group Inc (1990) 893 F 2d 541 17 (US Court of Appeals, Third Circuit)................................................................................................................................201 Collins v Eli Lilly Co (1984) 34 NW 2d 37 (Supreme Court of Wisconsin).......................214 Cook v Rockwell Intern Corp 580, F Supp 2d 1071 (D Colo 2006)....................................158 Coughlin v Owens-Illinois Inc, 27 Cal.Rptr. 2d 214 (Cal App 1993)....................................19 Daubert v Merrell Dow Pharmaceuticals Inc, 509 US 579 (1993)..................................................................................xxiv, 164, 171, 177, 178, 247, 259 Daubert v Merrell Dow Pharmaceuticals Inc, 43 F3d 1311 (9th Cir 1995), cert denied 116 SCt 189 (1995) [Daubert II].............................................................151, 215 Day v Boston & Maine Railroad (1902) 52 A 771 (Supreme Ct of Maine)................202, 208 Duncan v Cessna Aircraft Co, 665 SW2d 414 (Tex 1984) (SCt)............................................18 Dwyer v Sec’y of Health & Human Services 03–1202V, 2010 WL 892250 (Fed Cl 12 March 2010).............................................................. 169, 170, 171, 172, 173, 176 Eagle-Picher v Balbos, 578 A2d 228 (Md App, 1992).............................................................23 Erie RR v Tompkins, 304 US 64, 58 SCt 817 (1938)..............................................................16

418  Table of Cases Estep v Sec’y of Health & Human Services, 28 Fed Cl 664 (1993)......................................171 Frye v United States, 293 F 1013 (1923)...............................................................................259 Garcia v Sec’y of Health & Human Services 05–720V, 2008 WL 5068934..........................171 General Electric v Joiner, 522 US 136 (1997).......................................................164, 268, 270 Gibson v American Cyanamid Co, 719 F Supp 2d 1031 (ED Wis 2010)..............................33 Gibson v American Cyanamid Co, 750 F Supp 2d 998 (ED Wis 2010)................................33 Grant v Sec’y of Dep’t of Health & Human Services, 956 F 2d 1144 (Fed Cir 1992)............................................................................................................................165, 171 Gregg v V-J Auto Parts Inc, 943 A2d 216 (Pa 2007).........................................................22, 23 Hazlehurst v Sec’y of Health & Human Services, 03–654V, 2009 WL 332306 (Fed Cl 12 February 2009), aff ’d 88 Fed Cl 473 (2009), aff ’d 604 F 3d 1343 (Fed Cir 2010).........................................................................................158, 165–7, 171–172 Hennessey v Sec’y of Health & Human Services 01–190V, 2010 WL 94560.......................171 Herskovits v Group Health Cooperative of Puget Sound, 664 P 2d 474 (Wash 1983)................................................................................................................227, 233 Hines v Sec’y of Health & Human Services, 940 F 2d 1518 (Fed Cir 1991)........................165 Hodges v Sec’y of Dep’t of Health & Human Services, 9 F3d 958 (Fed Cir 1993).............165 Howard v Wal-Mart Stores Inc (1998) 160 F 3d 358 (US Court of Appeals, Seventh Circuit)..................................................................................................................211 Hymowitz v Eli Lilly & Co, 539 NE2d 1069 (NY 1989)...........................................32, 66, 226 John Crane Inc v Jones, 604 S.E.2d 822 (Ga 2004)..........................................................18, 22 Johnston v United States (1984) 597 F Supp 374 (US District Court, Kansas)..................208 Joint Eastern and Southern Dist Asbestos Litigation, In Re, 129 BR 710 (ED NY 1991), judgment vacated on other grounds 982 F2d 721, opinion modified on rehearing 993 F2d 7 (2nd Cir 1993).....................................................................................................12 Jorgenson v Vener, 616 NW2d 366 (SD 2000)......................................................................233 June v Union Carbide Corp, 577 F3d 1234 (10th Cir 2009)................................................305 Kennedy v Southern California Edison Co, 268 F3d 763 (9th Cir 2001)..............................22 King v Armstrong World Industries Inc, 906 F2d 1022 (5th Cir 1990)................................18 King v Sec’y of Health & Human Services, 03–584V, 2010 WL 892296 (Fed Cl 12 March 2010)........................................................................... 169, 170, 171, 172, 173, 176 Kingston v Chicago & Northwestern Ry Co (1927) 211 NW 913 (Supreme Court of Wisconsin)..........................................................................................................................294 Knudsen v Sec’y of Health & Human Services, 35 F 3d 543 (Fed Cir 1994)......................171 Kramer v Lewisville Memorial Hospital (1993) 858 SW 2d 397 (Supreme Court of Texas)......................................................................................................................217, 218 Kumho Tire Co v Carmichael, 526 US 137(1999)...............................................................164 Landers v East Texas Salt Water Disposal Co, 248 SW 2d 731 (Tex 1952)............................17 Livanovitch v Livanovitch (1926) 131 A 799 (Supreme Court of Vermont)......................200 Lohrmann v Pittsburgh Corning Corp, 782 F2d 1156 (4th Cir 1986)............................17, 23 Marder v GD Searle & Co (1986) 630 F Supp 1087 (US District Court, Maryland), affirmed (1987) 814 F 2d 655 (US Court of Appeals, Fourth Circuit).............................215 Martin v Abbott Labs (1984) 689 P 2d 368 (Supreme Court of Washington)...................214 Matsuyama v Birnbaum, 890 NE 2d 819 (2008) (Supreme Judicial Court of Massachusetts)....................................................................................................................257 Mays v United States, 608 F Supp 1476 (DC Colo 1985), revd on other grounds 806 F2d 976 (10th Cir 1986), cert den 482 US 913............................................................233

Table of Cases  419 Mead v Sec’y of Health & Human Services 03–215V, 2010 WL 892248 (Fed Cl 12 March 2010)................................................................................... 168, 169, 170, 171, 172 Menne v Celotex Corp, 861 F2d 1453 (10th Cir 1988)..........................................................19 Merrell Dow Pharmaceuticals Inc v Havner, 953 SW 2d 706 (Tex 1997)...................151, 155 Methyl Tertiary Butyl Ether (MTBE) Prods Liab Litig, In Re, 175 F Supp 2d 593, 621 (SDNY 2001)..................................................................................................................33 Migues v Fibreboard Corp, 662 F.2d 1182 (5th Cir 1981)...............................................18, 19 Mitchell v Gonzales (1991) 819 P 2d 872 (Supreme Court of California).........................294 Moberly v Sec’y of Health & Human Services, 85 Fed Cl 571 (2009), aff ’d 592 F 3d 1315 (Fed Cir 2010)............................................................................................165, 171 Norfolk & Western Railway Company v Ayres, 538 US 135 (2003).......................................................................... xxv, xxvi, 221, 222, 223, 224, 225, 238 Ortiz v Fibreboard Corp, 527 US 815 (1999).......................................................................238 Pafford v Sec’y of Health & Human Services, 451 F 3d 1352 (Fed Cir 2006)....................................................................................................................172, 173, 176 Perreira v Sec’y of Health & Human Services, 33 F 3d 1375 (Fed Cir 1994)......................169 Rudenauer v Zafiropoulos, 445 Mass 353 (2005).................................................................234 Rutherford v Owens-Illinois Inc, 941 P 2d 1203 (Cal 1997).............xx, 11, 15, 19, 20, 21, 22, 24, 26, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40 Sargent v Massachusetts Accident Co (1940) 29 NE 2d 825 (Supreme Court of Massachusetts)....................................................................................................................201 Scott v Sec’y of Health & Human Services 03–2211V, 2006 WL 2559776..................171, 172 Senn v Merrell-Dow Pharmaceuticals Inc (1988) 751 P 2d 215, fn 1, 222 (Supreme Court of Oregon)................................................................................................................214 Sharpnack v Sec’y of Health & Human Services, 27 Fed Cl 457 (1993)..............................171 Shyface v Sec’y of Health & Human Services, 165 F 3d 1344 (Fed Cir 1999).....................165 Sindell v Abbott Laboratories 607 P 2d 924 (Cal 1980)..... 30, 32, 33, 34, 39, 129, 214, 226, 230 Smith v Washington, 734 NE2d 548 (Ind 2000)...................................................................233 Snyder v Sec’y of Health & Human Services, 01–162V, 2009 WL 332044 (Fed Cl 12 February 2009), aff ’d 88 Fed Cl 706 (2009)........................160, 165, 166, 167, 168, 171, 172 Sterling v Velsicol Chemical Corp 855 F 2d 1188 (US Court of Appeals, Sixth Circuit) (1988)..........................................................................................................216 Stevens v Sec’y of HHS, No 99–594 V, 2001 WL 387418 (Fed Cl 30 March 2001).............165 Summers v Missouri Pacific RR System, 897 F Supp 533 (ED Okla, 1995), aff ’d 132 (10th Cir 1997).....................................................................................................................36 Summers v Tice, 199 P 2d 1 (Cal 1948).................................................. 19, 20, 21, 32, 39, 226 Sumrall v Sec’y of Health & Human Services, 23 Cl Ct 1 (1991)........................................171 Terran v Sec’y of Health & Human Services, 195 F 3d 1302 (Fed Cir 1999)......................171 Thomas ex rel Gramling v Mallett, 285 Wis 2d 236, 701 NW 523 (2005)............................33 United States v Shonubi 895 F Supp 460 (EDNY 1995) 514...............................................101 United States v Shonubi 103 F 3d 1085 (1997) (United States Court of Appeals, Second Circuit)...................................................................................................................208 United States v Shonubi 962 F Supp 370 (1997) (Federal District Court, Eastern District of New York).........................................................................................................208 Vanaman v DAP Inc, 966 A2d 603 (Pa Super, 2009)..............................................................22 Vosburg v Putney (1890) 47 NW 99; (1891) 50 NW 403 (Supreme Court of Wisconsin)......................................................................................................................205

420  Table of Cases Warren v Parkhurst, 92 NYS 725 (NY Sup Ct. 1904), aff ’d 93 NYS 1009 (NY App Div 1905), aff ’d 78 NE 579 (NY 1906).......................................................304, 327 Weisgram v Marley Co, 528 US 440 (2000)..........................................................................164 Wendland v Sparks, 574 NW2d 327 (Iowa, 1998)................................................................233 Ybarra v Spangard, 154 P2d 687 (Cal 1944).........................................................................226

TABLE OF LEGISLATION Australia

Civil Liability Act 2002 (NSW)   s 5D.....................................................................................................................................244   s 5D(1)–(3).........................................................................................................................243   s 5D(4)................................................................................................................................244

European Union

Directive 91/414/EEC (Plant Protection Products).............................................................387   art 3(3)................................................................................................................................387   art 4.............................................................................................................................387, 391   art 4(1)................................................................................................................................387    Annex III, Pt A, para 7.2.2.................................................................................................387    Annex VI.....................................................................................................................391, 396 Directive 97/57/EC.................................................................................................................391

France

Act of 9 April 1898.................................................................................................................116 Act of 30 October 1946..........................................................................................................116 Civil Code (Napoleonic Code)...................................................................... 112, 113, 114, 117   art 1151...............................................................................................................................113   art 1347...............................................................................................................................113   art 1349...............................................................................................................................123   art 1353...............................................................................................................................123   art 1382.......................................................................................................................112, 114    art 1382C civ......................................................................................................................112   art 1383...............................................................................................................................112   art 1384...............................................................................................................................112    art 1384 al 1........................................................................................................................114   art 1385...............................................................................................................................112   art 1386...............................................................................................................................112   art 1386-1–1386-18............................................................................................................116    art 1386-9C civ...................................................................................................................116 Code of Civil Procedure................................................................................................125, 126

422  Table of Legislation   art 145.................................................................................................................................125   art 234.................................................................................................................................125   art 238(3)............................................................................................................................126   art 246.................................................................................................................................125 Code of Criminal Procedure   art 4-1.................................................................................................................................198   arts 371–372.......................................................................................................................198   arts 470–471.......................................................................................................................198 Code de la santé publique, art L 3122-2...............................................................................124 Code of Social Security, arts L 411-1 ff.................................................................................116 Criminal Code 1992    art 121-3 al 4......................................................................................................................115   art 221-6.............................................................................................................................115   art 222-19...........................................................................................................................115 Decree of 13 October 1965....................................................................................................125 Loi no.85–677 of 5 July 1985.................................................................................................116 Loi no.98–389 of 19 May 1998..............................................................................................116 Loi no.2000–647 du 10 juillet 2000.......................................................................................115 Loi no.2002–303 of 4 March 2002................................................................................116, 128    art 102 al 1st.......................................................................................................................124    art L 1142-1 C....................................................................................................................116

Germany

Code of Civil Procedure   s 286....................................................................................................................................195   s 286(1)...............................................................................................................................199 Code of Criminal Procedure.........................................................................................198, 200   s 261....................................................................................................................................199

Italy

Civil Code...............................................................................................................................198 Penal Code..............................................................................................................................198

United Kingdom

Compensation Act 2006................................................................xxi, 30, 58, 65, 77, 79, 81, 90   s 1..........................................................................................................................................77   s 3............................................................................... 30, 42, 75, 77, 80, 81, 90, 192, 226, 398   s 3(1)–(2)..............................................................................................................................79

Table of Legislation  423   s 6............................................................................................................................................8   s 16(3).............................................................................................................................77, 79 Consumer Protection Act 1987   s 2(1)...................................................................................................................................188   s 3........................................................................................................................................153 Damages (Asbestos-related Conditions) (Scotland) Act 2009..........xxi, 57, 58, 84, 89, 90, 91   s 1..........................................................................................................................................89   s 1(1)–(2)........................................................................................................................89, 90   s 1(3).....................................................................................................................................89   s 1(4)...............................................................................................................................89, 90   s 2..........................................................................................................................................89 Limitation Act 1980, s 33.......................................................................................................152 Medicines Act 1968, s 3..........................................................................................................383 Offences against the Person Act 1861.......................................................................................5   s 23................................................................................................................................5, 9, 45 Pneumoconiosis etc. (Workers’ Compensation) Act 1979.....................................................81 Prevention of Harassment Act 1997, s 8(6)............................................................................89 Water Resources Act 1991..........................................................................................................6   s 85(1).................................................................................................................................6, 9

Statutory Instruments Control of Pesticides Regulations 1986 (SI 1986/1510).......................................................395 Control of Pesticides Regulations 1997 (SI 1997/188)   reg 6....................................................................................................................................395   reg 6(c)(ii)..........................................................................................................................395   Sch 4     para 2(1)(b)–(c).............................................................................................................395     para 2(1)(g)....................................................................................................................395 Medicines (Advisory Bodies) Regulations 2005 (SI 2005/1094).........................................383 Plant Protection Products Regulations 2005 (SI 2005/1435)..............................................387

United States of America

2011 Wisconsin Act 2, 2011 SB1..............................................................................................33 California Civ Code, s 1431.2 (West 2007).......................................................................21, 30 California Evidence Code, s 115............................................................................................200 Constitution, art 3, s 2.......................................................................................................15, 16 Fairness in Asbestos Injury Resolution Act.............................................................................13 Federal Employers’ Liability Act 45 USCS....................................................................221, 224   s 51......................................................................................................................................223 Federal Rules of Evidence......................................................................................................171   r 702....................................................................................................................................164 National Childhood Vaccine Injury Act 1986 42 USC................................ xxiv, 150, 164, 165   s 300aa-11(c)(1)(C)(i).......................................................................................................165   s 300aa-11(c)(1)(C)(ii)(I).................................................................................................165   s 300aa-12(d)(2)(B)...........................................................................................................171   s 300aa-13(a)(1)(A)...................................................................................................165, 172   s 300aa-13(b)(1)........................................................................................................172, 173 Restatement (First) of Torts (1934)......................................................................................306 Restatement (Second) of Torts (1965)............................................................................15, 306   s 328D(1)............................................................................................................................219   s 431................................................................................................................................16, 18   s 432(1)...............................................................................................................................294   s 433......................................................................................................................................16   s 433A...................................................................................................................................16   s 433B(2)–(3).....................................................................................................................212 Restatement (Third) of Torts: Apportionment of Liability. (2000)   s 8..........................................................................................................................................38   s 17........................................................................................................................................30

426  Statutory Instruments   s 23........................................................................................................................................15   s 26........................................................................................................................................37   s A18.....................................................................................................................................15   s C18.....................................................................................................................................15   s D18.....................................................................................................................................15   s E18......................................................................................................................................15 Restatement (Third) of Torts: Liability for Physical and Emotional Harm. (2010).................................................................................................................152, 212, 306   s 26.......................................................................................213, 217, 233, 285, 292, 294, 305   s 27......................................................................................................................285, 304, 305   s 28......................................................................................................................213, 215, 216   s 28(a).................................................................................................................................152   s 28(b).................................................................................................................................212   s 36......................................................................................................................................305 Uniform Comparative Fault Act, s 2(b)..................................................................................38 Uniform Model Product Liability Act, Section 111(B)(3).....................................................38

International

European Convention on Human Rights   art 6.......................................................................................................................................89    First Protocol, art 1..............................................................................................................89 Rio Declaration on Environment and Development, art 15...............................................392

index abnormal causes 9, 114, 362 abstract initial base rates xxv, 211–12 academic writings   Canada 93–4   citation 3    counsel, use by 3    France 111, 113, 117–19    Germany 111–12, 117–19    judicial reasoning xix, 3, 4    Scotland xx, 43–4, 46, 55–6 accountability 222, 242, 245 accounting evidence 241 actual causation 204, 206–7, 213, 216–19 additive effects 274–8 adequate causation, theory of 117, 118–19, 121–2 Advisory Commission on Pesticides (ACP) 388–90, 392–5 agent-conditioning xxviii, 341, 353, 359 alternative causation xxviii, 19–20, 32, 212–14, 340, 351–2 American Legal Realists 70–1, 91 analogy, reasoning by 58, 60, 65 anguish and fear, risk of future harm and 221–5, 234 antecedent conditions xxviii, 335–6, 339–44 apportionment 15–17, 21, 31–9, 66–7, 124 asbestos claims 11–40, 57–91    actionable damage 82–3, 86, 89–90   action-based responsibility 72      alternative liability 19–20, 32    analogy, reasoning by 58, 60, 65    anguish and fear 221–5   apportionment 15, 16–17, 21, 31–9, 66–7   asbestosis 14, 15–18, 22 associations xxiii–xxiv, xxv, 150, 155–8, 180–4, 188, 191–2, 342 Atomic Veterans case 192–3 Australia 251–3   Barker v Corus case xxi, 29–34, 38–9, 57–8, 62, 66–70, 74–82, 90–1   Borel v Fibreboard Paper Products case 12, 15–17, 37, 40   Borg-Warner v Flores case 22, 32   Brown v Superior Court case 30    burden of proof 18–21, 62–4    but for test 14–16, 23, 31, 37, 39, 61, 65   cancer 11–40   causal mechanism xx, 14–15, 21, 36–7    common law xxi, 58–60, 63–5, 76–7, 81, 86, 89, 91    comparative responsibility regime 16, 18, 29, 38    compensation xxi, 58, 69, 71–3, 77, 79–90, 337    complex fact patterns 31, 33, 34–5    contributory negligence 16, 21, 28–9

   corrective justice theory xxi, 58, 69–72, 76, 91    creation of risk 72–3    cumulative mechanisms 14, 16–17, 26–7, 37    damages xxi, 15–17, 58, 73, 75, 89    dermatitis 23–5, 28   DES (diethylstilbestrol) cases 30, 32–3    deterrence 71, 72–3    distributive justice 70, 77    divisible injuries 15, 16–19, 26, 34, 37, 79–80    double counting and damages 73      doubling of risk approach xxiii, 65, 75, 151–5, 176,    epidemiological evidence xxiii–xxiv, xvi, 35, 65, 76, 100, 123, 127, 150–62, 168–71, 176–7, 179–84, 186, 188–90, 193–4, 202–4, 206–7, 210, 215, 251–3, 272, 277    evidential gaps 62–3    exceptional categories xxi, 58–61, 63–5, 69, 82    expert evidence 251–3    fair, just and reasonable test 66–7   Fairchild case xxi, xx, 11, 23, 25–33, 36–40, 57–70, 74–82, 90–1    frequency, regularity and proximity test 17–18    future harm, risk of 73, 221–7, 231, 238    general causation 14   Gregg v V-J Auto Parts Inc 22–3    hard cases 59, 63   immunity 326    impossibility criterion 34    increase in risk as form of actionable damage 65, 67–8, 73–6, 79    indeterminate liability 64    indivisible injuries 15, 17–19, 37, 39, 79   injury, risk rather than xx, 11   insurance 84–6    isolated irruptions, statutory rules as 58–9    joint and several liability 15, 29–30, 34, 38–9, 67–9, 78, 80, 82, 337    judicial reasoning 57–91    latency period 14, 16–17    legislation xxi, 57–9, 64–5, 69–70, 77–91   Lohrmann v Pittsburgh Corning Corp case 17–18    loss of chance cases 23–4, 68–9, 73–4, 76    lung cancer 14, 21–2, 31, 35    market share liability 32–3    mass toxic tort cases 12    material contribution 65   McGhee case 23–5, 27–8, 34, 61    medical evidence 251–3    medical negligence 23–4, 68–9    mesothelioma 14, 18–19, 21, 25–30, 34, 60–1, 65–7, 77–82, 85, 326   Migues v Fibreboard Corp case 18–19    multiple causes/agents 14, 26–7, 31, 34–7

428  Index Australia (cont.):    multiple defendants xx, xxviii, 14–16, 26–7, 36–7, 326, 336–7   National Health Service 68–9    negligence 23–4, 59–61, 64, 68–9, 73–4, 82–90   NESS test xxviii, 326, 336–7    non-traditional defendants 13    number of claims 13    outcomes or risks, concern with 73    Parliament, defence to 68–9    pleural plaques 57, 74, 77–90    policy xxi, 57–8, 61–2, 64, 68–71, 74, 76, 83–4, 88, 91    politics 58–9, 61–2, 64, 70–1, 77–91    precautions 25, 29    precedent xxi, 59, 81, 85–6    principle, questions of xxi, 58–9, 68, 71, 80, 86–7    proof xx, 15–24, 36–7, 59–60, 62–5, 68, 88    proportionate liability 66–8, 78–9, 81    proximate cause 23    psychiatric injury or nervous shock 63–4, 66    relative contribution 38–9    retrospectivity 77–8, 83–5    reverse burden of proof 19–21   risk rule xx, 11, 15–40    road accidents 15–16, 19, 37   Rothwell pleural plaques case 57–8, 82–90   Rutherford v Owen-Illinois Inc case xx, 11, 19–24, 26–7, 30–7, 39–40   science     Australia 251–3     expert evidence 251–3      uncertainty, law’s approach to harm in context of xxx, 398–402    Scotland xxi, 58–9, 82–90    similar fact situations, risk rule in 11, 34, 36–7   Sindell v Abbott Laboratories case 30, 32–5   single agent rule 20, 27, 33–4, 36, 61, 66, 69    single hit model 14, 19–20, 23–5, 27, 34–40, 205, 326    smoking 21–2, 31, 35, 252–3    standard of proof 75–6   statistics 251–3    statutory interpretation 58–9, 87    substantial factor test in United States xx, 15–16, 18, 20–3, 32, 37    threshold model 14, 23–4    United Kingdom, risk rule in 23–34, 36–40    United States xx, 11–13, 15–24, 26–7, 30–40    voluntary assumption of responsibility 74–5   Wilsher case case 23, 24–7, 29, 34, 61–2, 69, 74 Australia see scientific and medical expert evidence in causation decisions in Australia autism see Omnibus Autism Proceedings (OAP) test cases in United States balance of probabilities/preponderance of evidence 195, 199–200, 215–20, 399, 402–4 bare risks 228, 231, 236–7, 239 Bayes’ Theorem   abstract initial base rates xxv, 211–12   particularistic evidence, requirement for xxv, 211

  personal chance/statistical chance dichotomy 162–3   product liability cases, scientific evidence in xxiv, 150, 162–3, 177    standard of proof xxv, 211–12    statistics xxiv, 162–3, 177 Becht, A 54–5 Beever, Allan 61, 63, 70–1, 73, 75–6, 97, 365 behaviour of others, causing the 361–81    abnormal causes 362    blameworthy causes 362    breaking the chain of causation 362–3, 376    causal language, use of xxviii, xxix, 361–6, 380    causal minimalists xxix, 361, 281    causal potency xxix, 379–81    cause, meaning of 361–4, 369, 377–8    central type of causation 366–9   choice 369    coincidence 377, 380    common sense 362    contribution xxviii–xxix, 362, 370–6, 379–80    decisions, contribution to 375–6    disaggregation problem xxix, 370–3, 380    disputes about ‘cause’ disputes, meaning of 362–4    doomed exception 371–2    duplicate necessity xxviii, 362, 370     evidence 367–9   experiences 367–8    explanatory causes 362    fairness 377–8, 380    foreseeability 376, 377, 380   generalisations 367   influences 367    intention 364, 366, 368–9, 378, 380    intervention doctrines xxix, 361, 362–4, 372, 376–81   intuition 363    involuntary acts 369    involvement xxviii, 362–6, 373–5, 379    limit of liability xxix, 363, 379    mechanisms which did not operate 372–4    mixtures of causes xxix, 369–75    natural processes xxix, 364–7, 371, 373–5, 380    necessity xxviii, 362, 368, 370, 373–4    negativing causes 363   NESS test 368, 370, 373    non-central types of enquiry xxix, 366–9    normative evaluation 363, 366    objective facts xxviii–xxix, 362–4, 366, 380   omissions 374–5    pattern of identified factors 367–8    physical system, clarification of causal contribution in a 370–3   pluralism 365–6    possibility of causation 368–9, 372   proof 368–9    purpose of the law, serving the 363–4, 366–7    reasons 366, 367, 370, 380–1   res ipsa loquitur 368    sequences 361, 364, 366–9    substitute facts 368    sufficiency 362, 368–70, 373    transparent justification 363

Index  429    vicarious physics 374–5    voluntary acts 361, 362–3, 369, 377, 379–80    what causes what 364–6 behaviour, unpredictability of xxi, 49–55 belief in truth    ex ante causal probabilities xxv, 206–8, 210–13    particularistic evidence, requirement for xxv, 196–7, 208–13    standard of proof xxv, 195–201, 204–13, 220 best explanation, inferences to the 104–6, 108, 262 Beyea, Jan 146 beyond a reasonable doubt 196, 198, 199–202 bias 183–4, 241–2, 248–50 biostatistics and causation in medicinal product liability suits 179–94    associations xxv, 180–4, 188, 191–2    Atomic Veterans case 192–3    attribution of causality 179   bias 183–4    Bradford Hill criteria 181–2, 184, 188–9, 191    burden of proof 188    but for test 191, 193    clinical negligence xxv   confounding 183–4    court, relevance of biostatistics to the 187–94    cumulative injuries 191    differences between medical science and legal approaches 189–94    divisible and indivisible injuries xxv, 193–4    epidemiology, principles of xxiv–xxv, 179–90, 193–4   error 183–4   estimation 184    evidence xxiv–xxv, 179–90, 193   Fairchild case xxv, 192–4    fallibility of scientific approach 185–7    general causation 192    general principles 188–9    increase in risk theory 192    joint and several liability 192    material contribution 191–2   McGhee case 192   McTear case 188   mesothelioma 193–4    multiple causes test 191, 193    null hypothesis 184   observation 179–84    oral contraceptives and strokes 190    plausible causality 191    probability theory 184, 188, 191    proof 190–1, 194    prospective forecasting xxiv, 189–90   regulators xxiv–xxv      retrospective versus prospective forecasting xxiv, 189–90   risk      benefit, assessment of 179–87, 190, 194       definition 183     difference 183      divergent approaches to risk as test of causation 191–3     ratio measures 183

    relative risk 188–9, 192     test xxv, 191–4    scientific method, introduction of 180–2    Seroxat (Paroxetine) 186, 187    significance testing 184    single agent 193    smoking and lung cancer 183    specific (individual) causation 188–9, 192    standard of proof xxv, 191–4    systematic bias 184    TeGenero Incident 187    tests for causation 190–3   Thalidomide 179–80    therapeutic use 179    tick box approach 182    United States 185–6    Vioxx 185–6, 187   Wilsher case 193    withdrawal from market 179, 185 Birks, Peter 108–9 bizarre and unscientific evidence 242, 244–7 bladder cancer and tobacco smoke 152 blameworthy causes 362 Bradford Hill criteria 106, 156–8, 181–2, 184, 188–9, 191, 243 breaking the chain of causation 362–3, 376 Bradford Hill, Austin 106, 156–8, 181–2, 184, 188–9, 191, 251 breast cancer 256, 265–6, 274–8 burden of proof    asbestos 18–21, 62–4    Canada, inference-drawing in fact in 97   France 123    Hepatitis B vaccine xxiv, 204–6    judicial reasoning 7–9    Omnibus Autism Proceedings (OAP) test cases in United States 165    product liability xxiv, 152, 188    scientific fact-finding 97 but for/sine qua non test   administrative law 122    asbestos 14–16, 23, 31, 37, 39, 61, 65    Australia 243    domino effect 117    expert evidence 243   Fairchild case 398–400, 402, 405–5    France 117–18, 120, 122, 129    judicial reasoning 9    material contribution 42, 47    medical evidence 243    multiple causes test 191   NESS test xxviii, 47–8, 285–8, 291–4, 298, 303, 321, 323–4, 326, 330, 337   product liability 158, 191, 193    reverse burden of proof 19–21   science     expert evidence 243      uncertainty, law’s approach to harm in context of 398–400, 402,  405–7        Scotland xx, 41–3, 45, 46–7    standard of proof 218–19    substantial factor test in United States xx, 15, 16, 23

430  Index Canada, inference-drawing in fact in 93–109    academic writings 93–4    burden of proof 97    Canada xxi–xxii, 93–6, 102, 107    cause-in-fact xxi, 93–5, 98, 100–1, 106–9    common sense 94–5, 107    creation of risk 105    emotivism xxii, 104–5, 109    epidemiological research 100    epistemology of fact-finding 109    evidence xxii, 93–5, 100–9    gaps in evidence xxii, 93–5, 103–4, 108–9    generalisations 97, 99–100, 104–5   indeterminancy 108   Inference to the Best Explanation (IBE) 104–6, 108   intuition 108    legal fact-finding 96–9, 109    legal rationality 108   McGhee case 94    medical malpractice xxii, 93–5, 102–4, 107, 109   narratives 105–8    negligence xxi, 93–104, 107, 109    objections xxii, 93–6    probability assessments 100–2    reference classes 99–102    risk and harm, temporal proximity between 106–7    role of inference 101–4    scientific fact-finding 96–102   Snell v Farrell (Canada) xxi–xxii, 93–6, 101–3, 107–9    soft variables 98, 102    statistics xxii, 96–102    United States 105   Wilsher case 94–5, 99, 107–9 cancer see also lung cancer   asbestos 11–40    additive effects 274–8   breast cancer 256, 265–6, 274–8   DES cases 262–3, 265–6, 275    emotional disturbance, leukaemia caused by 245    induction period 262–3    initiation effects 269    latency period 262–3, 276   progression 269    promotion effects 268–70    science, development of 262–3, 265–6, 268–70 Cane, P 81 capacity 6, 205–6, 215 Carbonnier, Jean 112 Carrington, Paul 13 causal language, use of xxviii, xxix, 361–6, 380 causal minimalists xxix, 361, 281 causal models    additive effects 274–8    diseases xxvi, 261–2, 268–81    genetic susceptibility 278–80    immulogical weakening 273–4    multi-generational effects 280    promotion effects 268–70    science, development of xxvi, 261–2, 268–81    single hit model 14, 19, 23–4, 27, 34, 36–7, 39–40, 60, 65, 205, 326

   threshold model 14, 23–4    variable hit models 270–3 causal potency xxix, 379–81 causal sufficiency see sufficiency causally relevant factor account xxvii, 285–90 Causes of Effects (CoE) (token or individual causation) xxii–xxiii, 133–4, 140–5 certainty see uncertainty chance, loss of a see loss of a chance Chapus, R 122 children, disease and 263–8, 270–1, 273, 279 choice 369 circularity    MMTS analysis 340, 342   NESS test xxvii, 285–6, 290, 293, 296, 300–1, 307, 332–3 clear and convincing evidence standard 200–1 Clermont, Kevin 198–9 clinical negligence see medical negligence clinical trials 135–7, 179 Cohen, Jonathan 98–9, 210–12 coincidence 377, 380 collateral effects xxvii, 297 collective action xxvi, 223, 238–9 Collins, John 300–1, 303, 318–19 common law    asbestos xxi, 58–60, 63–5, 76–7, 81, 86, 89, 91    scientific and medical expert evidence in causation decisions in Australia 243–4, 248    standard of proof 195–205, 208, 212, 215–17    statutory interpretation 58–9 common sense 94–5, 107, 243–4, 288, 362, 402 comparative responsibility regime 16, 18, 29, 38 compensation see also damages   asbestos xxi, 58, 69, 71–3, 77, 79–90, 337    central fund 71–2    Compensation Act 2006 xxi, 58, 77, 79–82, 90    corrective justice theory 69, 71–2    double counting 73    France 113–15, 124    future harm, liability for risk of xxvi, 221–39    Omnibus Autism Proceedings (OAP) test cases in United States 165    Pearson Commission 81    pleural plaques 82–90    pneumoconiosis claims 81    politics 58, 79–82    scientific uncertainty, law’s approach to harm in context of 383, 385   Scotland 82–90   undercompensation 73 complexity    asbestos 31, 33, 34–5    Australia 245, 249, 252, 255–6, 259   disease 261    evidence 245, 249, 252, 255–6, 259, 261    experts 245, 249, 252, 255–6, 259    fact patterns 31, 33, 34–5    medical evidence 245, 249, 252, 255–6, 259   NESS test 332–4    product liability cases, scientific evidence in xxiv, 149–50

Index  431    risk rule 31, 33–5   science     development 261      expert evidence 245, 249, 252, 255–6, 259 concurrent causation 340, 356–7 conditional of liability, causation as 111–12 confounding variables 139–40, 146 contract law, France and 113 contrastive causality 134–5 contribution see also material contribution   behaviour of others, causing xxviii–xxix, 362, 370–6, 379–80    cumulative contribution 218    decisions, contribution to 375–6   omissions 374–5    physical system, clarification of causal contribution in a 370–3    relative contribution 38–9    trivial contributions 304–5, 322 contributory negligence 16, 21, 28–9, 120–1 conversion 6–7 corrective justice theory xxi, 58, 69–72, 76, 91 correspondence, denominalisation rule of xxviii, 341, 344 Costello, S 272 costs 228, 230–1, 237 Coumadin 265–6 counterfactuals   decision-making, analysis of xxi, 49–55   hypotheticals 134–5   indeterminacy 49–55    lost chance analysis xxi, 51, 53, 55    misrepresentation xxi, 50    MMTS analysis 339, 341, 354–8   NESS test 50–3, 287–8, 294, 299, 304, 308, 329    road traffic accidents 54–5    scientific and statistical evidence 134–5, 140    Scotland xxi, 49–55 Cranor, Carl 177 Crime   France 115, 122   res ipsa loquitur 220   scientific and medical expert evidence in causation decisions in Australia 242, 260    standard of proof 195–200, 211, 220 cumulative injuries   NESS test 327–8   overdetermination 327–8   product liability 154, 191   standard of proof 218 damages see also compensation    asbestos xxi, 15–17, 58, 73, 75, 89    Australia 243–4, 253–4    backward-looking compensation 232    double counting 73    emotional distress 221–5, 234    expert evidence 243–4, 253–4   Fairchild case 8, 398    forward-looking compensation 232    future harm, liability for risk of 221–39

   immediate claims and future claims, choice between 73    judicial reasoning 3, 8    loss of a chance 218    medical evidence 243–4, 253–4    medical negligence 218   mesothelioma 8   misrepresentation 51    multiple defendants 15    negligence 243–4, 253–4    pain and suffering 221–3, 253–4    probability-based recovery xxvi, 222–3, 226–32    quantum of damages 232–3   science     expert evidence 243–4, 253–4      uncertainty, law’s approach to harm in context of 398   Scotland 51    standard of proof 218    United States 221–4, 233 Danks, David 106 data collection and generation xxii–xxiii, 135–40, 143, 145–6 Daubert framework xxiv, 164, 171, 177–8 Davidson, Donald 344 Dawid, Philip 100 DDT 266 De Morgan, Augustus 354–6 defective products see asbestos claims; product liability cases definition of causation 117–21 definition of cause 5 Deguergue, M 122, 129 Dejean de la Bâtie, N 117, 119 denominalisation rule of correspondence xxviii, 341, 344 dependence theory xxviii, 291, 316–20, 328–9, 339, 341, 354–8 dermatitis 23–5, 28 DES (diethylstilbestrol) cases    asbestos 30, 32–3    cancer 262–3, 265–6, 275   France 124    future harm, liability for risk of 230–1    latency period 262–3    science, development of 262–3, 265–6, 268, 273, 275    standard of proof 214–15 deterrence 71, 72–3, 227–8 diethylstilbestrol see DES (diethylstilbestrol) cases Difference Method 289 disaggregation problem xxix, 370–5, 380 disease and tort law 261–81    acceleration of disease 273    additive effects 274–8    background to development origins 263–8    best explanation, inferences to the 262    biological effects 268–9    biological presumptions 267–8    breast cancer 265–6, 274–8    cancer 262–3, 265–6, 268–70, 274–8    causal models xxvi, 261–2, 268–81    causal relationships 261–2

432  Index disease and tort law (cont.):    children 263–8, 270–1, 273, 279    civil procedure 262–3    complex evidence 261   Coumadin 265–6   DDT 266   DES 262–3, 265–6, 268, 273, 275    development of disease 261–3, 270–3    disquiet about torts 261–2   DNA 279    environmental factors 266, 271–2, 274–8    experimental studies 265–6, 268, 271, 273, 275–8, 280    experts 261, 268–70   exposures 261–3    foetuses 263–8, 270, 273, 275–6, 278–80    general causation 261    genetic susceptibility 278–80    heroin 262, 270–3    immulogical weakening 273–4    induction period 262–3   inferences 261–2   initiation 269    latency periods 261–3, 270–1, 273, 276–7    limitation periods 261–3   mathematics 267    models of causation xxvi, 261–2, 268–81    multi-generational effects 280    oestrogen and breast cancer 274–8    Parkinson’s disease and heroin 262, 270–3    pesticides 271–2, 279   procedure 261–3   progression 269    promotion effects 268–70   science 261–81    smoking and lung cancer 268–70    specific causation 261    thalidomide 264, 265   toxicants 262–78    United Kingdom, breast cancer in 274, 276    United States 261, 274    variable hit models 270–3 distributive justice 70, 77 divisible injuries see also indivisible injuries    asbestos 15, 16–19, 26, 34, 37, 79–80   Fairchild case xxv, 193–4   mesothelioma 193–4    product liability suits, biostatistics in medicinal xxv, 193–4 DNA 248, 250, 279 domino effect 117 doomed exception 371–2 double counting 73 double pre-emption xxviii, 340 double preventions 352–9 doubling of risk approach   asbestos 65, 76   multi-party litigation xxiii, 176   product liability cases, scientific evidence in xxiii, 150, 151–5, 176    standard of proof xxiii, xxv, 202–3, 215–16    United States 151–3

Dowe, Phil 355 driving accidents see road accidents drugs see also DES (diethylstilbestrol) cases; medicines    causing heroin to be administered 5, 9, 45    heroin 5, 9, 45, 262, 270–3    Parkinson’s disease 262, 270–3 Duplication   behaviour of others, causing the xxviii, 362, 370     necessity xxviii, 362, 370     NESS test 42, 47–8, 287, 294, 303–7, 318, 324   Scotland 42, 47–8 Dworkin, Ronald 80–2, 108–9 emotional distress, future harm and 221–5, 234 emotional disturbance, leukaemia caused by 245 emotivism xxii, 104–5, 109 environmental factors 266, 271–2, 274–8 epidemiology   asbestos 65, 76, 251–3    Bradford Hill criteria 251–2    Canada, inference-drawing in fact in 100    doubling of risk 151–5   evidence 65, 76, 150–61, 176–7, 251–3    interpretation xxiii, 150   McTear case xxiii–xxiv, 156–63, 177    Omnibus Autism Proceedings (OAP) test cases in United States 168–71, 173    product liability xxiv–xxv, 150–61, 176–7, 179–90, 193–4    scientific fact-finding 100    standard of proof 205    United States xxiii–xxiv, 151–2, 168–71, 173 epiphenomena (collateral effects of a common cause) xxvii, 297 epistemology   Canada, inference-drawing in fact in 109   facts 7–9, 109   future harm, liability for risk of 227–8, 236   judicial reasoning 7–9 equivalence of conditions, theory of 117–18, 119–21 error   Australia xxvi, 241–2, 247–8, 256, 259–60   bias 184    expert evidence xxvi, 241–2, 247–8, 256, 259–60    hindsight errors xxvi, 259–60    manifest error 391–2, 404–5    medical evidence xxvi, 241–2, 247–8, 256, 259–60    product liability suits, biostatistics in medicinal 183–4   science      expert evidence xxvi, 241–2, 247–8, 256, 259–60      uncertainty, law’s approach to harm in context of 391–2, 404–5 evidence see also burden of proof; evidential gaps; experts; proof; standard of proof   accountability for evidence, counsel making experts 242, 245   accounting evidence 241   admissibility criteria xxvi, 247, 259–60   anguish and fear 224    asbestos 65, 76, 251–3

Index  433    behaviour of others, causing the 367–9    biostatistics xxiv–xxv, 179–90, 193    bizarre and unscientific evidence 242, 244–7    Canada, inference-drawing in fact in xxii, 93–5, 100–9    clear and convincing evidence standard 200–1    complex evidence 245, 249, 252, 255–6, 259    conflicts of evidence 241    epidemiology 65, 76, 150–61, 176–7, 251–3    evaluation, evidence not easily amenable to 242, 250–4    evidential groupings 213    France xxii, 123, 125–6, 129    future harm, liability for risk of 224, 226–9, 237, 239    judicial reasoning 3–4    misunderstood, where facts are 242, 254–6   NESS test 290, 337    novel evidence xxvi, 259–60    Omnibus Autism Proceedings (OAP) test cases in United States 164–71, 173, 177–8    particularistic evidence, requirement for xxv, 196–7, 208–13    product liability xxiv–xxv, 150–3, 165–71, 176–7, 179–90, 193    psychiatric evidence 241–2, 256   relevance 164    reliability 164, 167   science     biostatistics xxiv–xxv, 179–90, 193      harm, law’s approach to 383–8, 401–2, 405      product liability xxiii, xxiv–xxv, 149–90, 193    standard of proof xxv, 202, 207, 209, 213, 219–20    statistics 100, 229–30, 237, 245, 250–3    weight 170–1, 173, 178, 202 evidential gaps   asbestos 62–3    Canada, inference-drawing in fact in xxii, 93–5, 103–4, 108–9    scientific fact-finding 103    scientific uncertainty, law’s approach to harm in context of 401–2, 405   Scotland 49 ex ante causal probabilities xxv, 206–8, 210–13 exceptional categories xxi, 58–61, 63–5, 69, 82, 244 experimental studies   data collection and generation xxii–xxiii, 135–7, 140, 143, 145   disease 265–6, 268, 271, 273, 275–8, 280    natural experiments 206    science, development of 265–6, 268, 271, 273, 275–8, 280    scientific and statistical evidence xxii–xxiii, 135–7, 140, 143, 145    scientific fact-finding 100    standard of proof 205–6    toxicological experimentation 205–6 experts   Australia xxvi, 241–60   disease 261, 268–70    France 122, 125–6    medical evidence xxvi, 241–60

   Omnibus Autism Proceedings (OAP) test cases in United States 165–7, 169, 171    precautionary decision-making 387–91, 394–6    product liability cases, scientific evidence in 151, 157, 159, 165–7, 169, 171   science     Australia xxvi, 241–60     development 261, 268–70      uncertainty, law’s approach to harm in context of 383–4, 387–91,394–6, 405    smoking and lung cancer 268–70 explanatory causes 362 facts   asbestos 31, 33, 34–5    Australia 241–5, 253–7, 259    Canada, inference-drawing in fact in 93–109    cause-in-fact xix, xxi, 3–5, 7–9, 93–5, 98, 100–1, 106–9    complex fact patterns 31, 33, 34–5   epistemology 7–9    experts 241–5, 253–7, 259   Fairchild case 9    filter xix, 5, 7, 9   France 123    historical connection with relevant outcome xix, 4    issue, facts on 195    judicial reasoning xix, 3–5, 7–9    legal fact, causation as xix, 123    legal fact-finding 96–9, 109    medical evidence 241–5, 253–7, 259    misunderstood, where facts are 242, 254–6   NESS test xix, 323    objective facts xxviii–xxix, 362–4, 366, 380    Omnibus Autism Proceedings (OAP) test cases in United States 165, 170–3    part of the history test 4–5, 7, 9    product product liability cases, scientific evidence in 165,    170–3    rules of law xix, 4, 7–8   science      expert evidence 241–5, 253–7, 259     fact-finding 96–102    Scotland 41–2, 45, 46    standard of proof 195    substitute facts 368   truth 195 fairness   behaviour of others, causing the 377–8, 380   fair, just and reasonable test 66–7   scientific uncertainty, law’s approach to harm in context of 401 Fairchild v Glenhaven Funeral Services case   asbestos xx, xxi, 11, 23, 25–33, 36–7, 39–40, 57–70, 74–82, 90–1    balance of probabilities 399, 402–4   Barker v Corus 30–1    but for test 398–400, 402, 405–6    damages 8, 398    divisible and indivisible injuries xxv, 193–4    fact, cause in 9    indeterminate causation xxx , 403–6

434  Index Fairchild v Glenhaven Funeral Services case (cont.):    judicial reasoning 8–9    multiple defendants 8, 398–9, 401   negligence 8   NESS test 8    policy xxi, 58, 61, 65–6, 69–71, 90–1    product liability suits, biostatistics in medicinal xxv, 192–3   risk 8–9   Rutherford v Owen-Illinois Inc case 30–1    scientific and medical expert evidence in causation decisions in Australia 244, 252    scientific uncertainty, law’s approach to harm in context of xxx, 398–408   Scotland 59 false causality 100 Fischer, David 304–6, 317–21, 323, 327–9 foetuses, disease and 263–8, 270, 273, 275–6, 278–80 foreseeability 376, 377, 380 France, causation in 111–29    abnormality standard 114    academic writings 111, 113, 117–19    action of things 114    active role 114    adequate causation, theory of 117, 118–19, 121–2    administrative courts xxii, 111, 115–16, 121–2, 128–9   apportionment 124    basic requirement of liability 112–16    bespoke regimes 116    burden of proof 123    but for test 117–18, 120, 122, 129    case law 120–2    certain and direct, causation as being 113    Civil Code 112–14, 117, 123    civil courts, autonomous approach and xxii, 111, 120–1    civil justice system, functioning of xxii, 111    civil law 112–14, 117–26, 129    compensation 113–15, 124    conditional of liability, causation as 111–12    contract law 113    contributory negligence 120–1    Criminal Code 115    criminal law 115, 122    damages 111, 116    defective products 116    definition of causation 117–22   DES (diethylstilbestrol) cases 124    document-based approach 125    domino effect 117    equivalence of conditions, theory of 117–18, 119–21    evidence xxii, 123, 125–6, 129    experts 122, 125–6    extent of liability 122    extra-contractual liability 111    fault-based liability 114–16, 129    general causation xxiv, 123, 127, 174–6    German jurists 111–12, 117–19    Hepatitis B vaccine xxiv, 123–4, 126–7, 174–6    indirect causation 115

   indivisible injury 117    intime conviction 195, 199   involvement 116   judiciary     policy 127–8      role and function xxii, 111, 121–9    legal causation 123, 127    legal fact, causation as 123    loss of a chance 124    margin of discretion 122    medical negligence 116, 117–18, 121, 124, 125, 128    medicines xxiv, 173–8   necessity 114    negligence 119–21, 124, 125, 128    no-fault liability 114–16    objective probability, theory of 118   obligation de moyens 114   obligations de résultat 114    policy xxii, 111, 127–9    presumptive reasoning xxii, 114, 123–4, 127, 174    probability 119, 123    procedural issues xxii, 122–7    product liability cases, scientific evidence in xxiii,    xxiv, 150, 173–8    proof 114, 123–4    proximate temporal relationships 174–6    public/private law divide xxii, 111    road traffic accidents 116      scientific causation xxiv, 127, 150, 173–8    serious, precise and concordant presumptions 174–5, 178    social security 116    specific causation 123, 174    standard of proof 195, 198–9, 217    statistics 126, 174    strict liability 114    systemic differences 124–7   threshold 122    work-related accidents 116 free will 5–6, 309 frequency, regularity and proximity test 17–18 Fumerton, R 289, 295–6, 310, 323 future harm, liability for risk of 221–39    anguish and fear 221–5, 234    asbestos 73, 221–7, 231, 238    backward-looking compensation 232    bare risks 228, 231, 236–7, 239    collective action xxvi, 223, 238–9    costs 228, 230–1, 237   damages 221–39   DES cases 230–1   deterrence 227–8    diluted funds prospect 223, 225, 238    emotional distress 221–5, 234    epistemology 227–8, 236    evidence 224, 226–9, 237, 239    forward-looking compensation 232    increase in risk 221    indeterminate causation 226, 228–31    insolvency, risk of 223, 238   insurance 235–6

Index  435   Johnstone v NEI International Combustion (UK) xxvi, 222, 223–5    joint and several liability 226    limitation periods 234    loss of a chance 227    mathematical probabilities 230    medical negligence 226–7, 229   mesothelioma 221    minimum threshold 228–9, 231, 237, 239    negligence 222, 225–7, 234–5   Norfolk & Western Railway Co v Ayres case xxv–xxvi, 221–5, 238    objections 222–3, 236–8    pain and suffering, damages for 221–3    paradigmatic cases 226–7    parasitic accountability doctrine 222    past harm 226–33    pleural plaques 222, 224–5    policy 221, 224    postponement and immediate recovery, choice between xxvi, 223, 233–8    preponderance of evidence 228–9, 231, 237, 239    probability-based recovery (PRP) xxvi, 222–3, 226–39    quantum of damages 232–3    recurrent wrongdoers 230–1    redundancy objection 237    social benefits 236    stand-alone harm 222, 225, 234    statistical evidence 229–30, 237    statutory funds 238    substantial risk 222, 237–8    unidentifiable victims 227, 230–1    unidentifiable wrongdoers 226, 229    United Kingdom 221–7, 239    United States 221–6, 230, 233, 238–9    victim migration xxvi, 238    wait and see approach xxvi, 223, 233–8 gatecrasher hypothesis 98, 99, 210, 212 Geistfeld, Mark 213 general causation   asbestos 14   diseases 261    France xxiv, 123, 127, 174–6    Hepatitis B vaccine xxiv    Omnibus Autism Proceedings (OAP) test cases in United States 164, 166–73    product liability xxiv, 150, 152, 156–7, 160, 164, 166–76, 192    proof xxiii– xxiv, 14, 123, 127, 152, 158–9, 174–8, 205    science, development of 261    standard of proof 205, 215 generalisations   behaviour of others, causing the 367   Canada, inference-drawing in fact in 97, 99–100, 104–5   legislation 58    MMTS analysis xxviii, 341–54, 357–9   NESS test 288, 290–2, 297–304, 307–8    standard of proof xxv, 205–9

generic sense, causal factors in a xxviii, 341, 359 Germany 111–12, 117–19, 200, 217 Goldberg, Richard 23, 29,32,97,125–6, 149–52, 154,159, 162, 177, 188, 207,209, 214, 383, 402 Green, Michael D xi, 18, 19, 22, 30, 33–36, 39, 75–6,149, 152, 210, 215 Greenland, Sander 133, 141, 143, 145–6, 151, 156, 162, 170, 206 Gregg v V-J Auto Parts Inc 22–3 Hall, Ned 300–1 hard cases 59, 63 harm in context of scientific uncertainty, law’s approach to 383–408    Advisory Commission on Pesticides (ACP) 388–90, 392–5    asbestos xxx, 398–402   authority 401    balance of probabilities 399, 402–4    but for test 398–400, 402, 405–7    bystander exposure to pesticides, risk of 387–9, 393, 395–6    common sense 402    compensation 383, 385    Court of Appeal xxix–xxx, 391–2, 395–8, 404–6   damages 398   Downs v Secretary of State case xxx, 387–98, 403–5   efficiency 385–6    evidence 383–8, 401–2, 405    ex ante harm 383, 407    ex post harm 383, 407   executive 404–7    experts 383–4, 387–91, 394–6, 405   Fairchild case xxx, 398–408   fairness 401    hypothetical harm 384–5, 394–8, 403, 407    ignorance, decisions under xxx, 393, 398, 403    inaction, principle of 393, 395    indeterminate causation xxix–xxx, 385, 386, 403–8    insufficient reason, principle of 393–4   intuition 402    judicial decision-making 385–98, 402–6    judicial review xxix–xxx, 386–98, 405    justice 385–6, 399–401   legislation 383    liability 383–5, 407–8    location and nature of scientific uncertainty 388–9, 399    manifest error 391–2, 404–5   McGhee case 400–2    mesothelioma xxx, 398–402    multiple defendants 398–9, 401    pesticides 385, 386–98    Plants Protection Products Directive 386–98    plausibility, test of 394, 397–8   policy 401    precautionary decision-making xxix–xxx, 386–98, 403–7    precautionary principle xxx, 386, 387, 391–7, 403–7   principle 401    probabilities 400–1, 403–4, 407   proportionality 390–1  

436  Index harm in context of scientific uncertainty, law’s approach to (cont.):    quantitative risk assessment 393, 396, 403, 407    reasonableness test 394–5, 397–8, 407   regulation 383–98    risk assessment 384, 387–8, 393, 396–7, 403, 407    risk, decisions under 393    Royal Commission on Environmental Pollution (RCEP) 387–90, 392–7    standard of proof 385, 399, 402–4   sufficiency 407   truth 385–6   Wednesbury irrationality 389, 391–2, 403–4 Hart, HLA xxvii, xxviii, xxix, 5, 9, 41, 43, 106, 108, 285–94, 298, 307, 323, 339, 366–9, 376–7 Hastie, Reid 104–6 Heindel, Jerry 271–2 Hepatitis B vaccine xxiv, 123–4, 126–7, 150, 174–6, 204–6 heterodox or iconoclastic theories xxvi, 245–8, 259–60 hindsight errors xxvi, 259–60 historical connection with relevant outcome xix, 4 HIV 246–7 Hogg, Martin 86–7 Honoré, Tony xxvii, xxviii, xxix, 5, 9, 41, 43, 106, 108, 285–94, 298, 307, 323, 339, 366–9, 376–7 Hume, David 334 hypotheticals 134–5, 384–5, 394–8, 403, 407 iconoclastic or heterodox theories xxvi, 245–8, 259–60 ignorance, decisions under xxx, 393, 398, 403 immediate recovery see postponement and immediate recovery, choice between immunity 326 impossibility 34 inaction, principle of 393, 395 increase in risk   asbestos 65, 67–8, 73–6, 79    future harm, liability for risk of 221    material contribution 192   NESS test 337    product liability suits, biostatistics in medicinal 192 indeterminacy   asbestos 64    Canada, inference-drawing in fact in 108    counterfactual analysis 49–55    default rule 53–4   Fairchild case xxx, 403–6    future harm, liability for risk of 226, 228–31    human behaviour, unpredictability of 49–51    mapping cases of indeterminate causation xxi, 48–9    nature of problem 49–51     NESS test xxvii, 50–1, 53, 309–11, 326, 336–7    potentially indeterminate behaviour of defendant 52–3    potentially indeterminate behaviour of pursuer 52    precautionary decision-making 386–98

   scientific uncertainty, law’s approach to harm in context of xxix–xxx, 385, 386, 403–8    Scotland 42–3, 48–55 indirect causation 115 indistinguishable alternative tortious claims 212–14 individual causation see specific causation indivisible injuries see also divisible injuries    asbestos 15, 17–19, 37, 39, 79   Fairchild case xxv, 193–4   France 117    mesothelioma 18–19, 193–4    product liability suits, biostatistics in medicinal xxv, 193–4   Scotland 42 inferences    academic writings 93–4    Best Explanation, Inferences to the 104–6, 108, 262    burden of proof 97    Canada xxi–xxii, 93–109    cause-in-fact xxi, 93–5, 98, 100–1, 106–9    common sense 94–5, 107    creation of risk 105   diseases 261–2    emotivism xxii, 104–5, 109    epidemiological research 100    epistemology of fact-finding 109    evidence xxii, 93–5, 100–9    gaps in evidence xxii, 93–5, 103–4, 108–9    generalisations 97, 99–100, 104–5   indeterminancy 108   Inference to the Best Explanation (IBE) 104–6, 108, 262   intuition 108    legal fact-finding 96–9, 109    legal rationality 108   McGhee case 94    medical malpractice xxii, 93–5, 102–4, 107, 109   narratives 105–8    negligence xxi, 93–104, 107, 109   NESS test 291    objections xxii, 93–6    probability assessments 100–2    reference classes 99–102   res ipsa loquitur 219–20    risk and harm, temporal proximity between 106–7    role of inference 101–4    science, development of 261–2    scientific fact-finding 96–102   Snell v Farrell (Canada) xxi–xxii, 93–6, 101–3, 107–9    soft variables 98, 102    standard of proof 214, 219–20    statistics xxii, 96–102    United States 105   Wilsher case 94–5, 99, 107–9 initial base rates xxv, 211–12 insolvency, risk of 223, 238 insufficient but necessary part of an unnecessary but sufficient     condition test see INUS test insufficient reason, principle of 393–4 insurance 7, 68, 73, 78, 84–6, 128, 235–7

Index  437 intention 3–4, 364, 366, 368–9, 378, 380 intervention    behaviour of others, causing the xxix, 361, 362–4, 372, 376–81    breaking the chain of causation 362–3, 376    causal potency xxix, 379–81    coincidence 377, 380    fairness 377–8, 380    foreseeability 376, 380    intention 378, 380    judicial reasoning 5–7, 9    limit of liability xxix, 363, 379 intime conviction standard 195, 199, 205 intuition    behaviour of others, causing the 363    Canada, inference-drawing in fact in 108    MMTS analysis 356   NESS test 306, 320–1, 335    scientific uncertainty, law’s approach to harm in context of xxix 402 INUS (insufficient but necessary part of an unnecessary but sufficient condition) test xxvii, xxviii, 285–90, 295, 339–42, 346–52 involvement xxviii, 46–7, 116, 362–6, 373–5, 379 irrationality 389, 391–2, 403–4 isolated irruptions, statutory rules as 58–9 Italy 197–8 joint and several liability    asbestos 15, 29–30, 34, 38–9, 337    future harm, liability for risk of 226    judicial reasoning 8    mesothelioma 30, 34   NESS test xxviii, 3337    product liability suits, biostatistics in medicinal 192    standard of proof 212, 214–15    United States 38–9 Jourdain, P 121 judicial reasoning xix–xx, 3–9    abnormal events 9    academic writings xix, 3, 4   asbestos 57–91    attribution of liability 3    binary system 7–8    burden of proof 7–9    but for test 9   capacity 6    causal relationship requirement xix–xx, 7, 9    common sense 402   conversion 6–7    counsel, academic writings and 3    damages 3, 8    definition of cause 5    deliberate acts of others 5–7, 9    deviations from ordinary course of events 5   epistemology 7–9   evidence 3–4    extraordinary occurrences 5    fact, cause in xix, 3–5, 7–9   Fairchild case 8–9    filter xix, 5, 7, 9    France xxii, 111, 121–9

   free will 5–6    hard cases 59    heroin to be administered, causing 5, 9    historical connection with relevant outcome xix, 4   intuition 402    joint and several liability 8    language xix, 4–5    legal causation xix, 3, 5    medical negligence 7, 9    mesothelioma, damages for 8    moral responsibility 5–6   murder 3–4    negligence 3–4, 6, 8–9   NESS test xix, 3, 4–5, 9    part of the history test 7, 9    payments, entitling people to 3    policy and purpose xix, 70–1   precautions 5–6    real causation 5, 7   risk 8–9    rules of law xix, 4, 7–8    scientific uncertainty, law’s approach to harm in context of 385–98, 402–6    Scotland 43–6, 49, 53, 55–6    standard of proof 8    statutory interpretation xix, 4    strict liability 6–7    third persons, intervention of 5–7, 9    two-stage test xix, 3, 4–5, 9    voluntary human acts 5–6    warn, duty to 7 judicial review xxix–xxx, 386–98, 405 juries    hindsight errors xxvi, 259–60   instructions 200–2    junk science, use of xxiv, 164    scientific and medical expert evidence in causation decisions in Australia xxvi, 244, 250, 259–60    standard of proof 200–2, 211, 220    United States 200–2    victims, sympathy for xxvi, 245, 260 justice   asbestos xxi, 58, 69–72, 76–7, 91    corrective justice theory xxi, 58, 69–72, 76, 91    distributive justice 70, 77    product liability cases, scientific evidence in xxiii, 176    scientific uncertainty, law’s approach to harm in context of 385–6, 399–401 Khoury, Lara 95–6, 99, 124–6 Klar, Lewis 95–6, 103 Klein, Andy 75 Koch, Robert 180 Kortenkamp, Andreas 274–6 Kress, K 289, 295–6, 310, 323 Landrigan, P 266 Language   use of causal xxviii, xxix, 361–6, 380   rules, of xix, 4–5 latency period 14, 16–17, 261–3, 270–1, 273, 276–7

438  Index legal causation   fact, cause in xix–xx, 3–5, 43, 45   France 123, 127   judicial reasoning xix, 3, 5   scientific and statistical evidence 134   Scotland 41, 43, 45 legal malpractice 218 Legal Services Commission (LSC) xxiii, 176 legislation and asbestos claims 64–5, 77–91    ad hoc legislation, role of xxi, 58    analogy, reasoning by 58, 65   Barker v Corus case 77–9    binding precedent xxi    common law 58–9    corrective justice theory 70    exceptional categories 69   generalisation 58    isolated irruptions, as 58–9    pleural plaques xxi, 82–90    policy xxi, 57, 70, 77–9, 91    politics 58–9, 79–81    principle, questions of xxi, 58–9, 80 leukaemia caused by emotional disturbance 245 Liebniz, Gottfried 355–6 limit of liability xxix, 363, 379 limitation periods 234, 261–3 loss of a chance    asbestos 23–4, 68–9, 73–4, 86    counterfactual analysis xxi, 51, 53, 55   damages 218   France 124    future harm, liability for risk of 227   Hotson case 218    medical negligence xxv, 68–9, 74, 124, 217–18, 227    scientific and medical expert evidence in causation decisions in Australia xxvi, 244, 254, 256–9    standard of proof xxv, 217–18 lung cancer    asbestos14, 21–2, 31, 35    product liability 156–63, 177, 183    risk rule 31    scientific and statistical evidence 134–5, 137    smoking 21–2, 31, 35, 134–5, 137, 156–63, 177, 183 Mackie, John xxvii, xxviii, 285–97, 309–11, 331–2, 335–6, 339 mapping cases of indeterminate causation xxi, 48–9 market share liability 32–3 Marteau, P 117 material contribution   asbestos 65    but for test 42, 47   dermatitis 192    increase in risk 192    product liability suits, biostatistics in medicinal 191–2    scientific and medical expert evidence in causation decisions in Australia 242–3, 244    Scotland 42, 47 mathematics 199, 201, 209–13, 230, 267 McBride, WG 179–80 medical evidence in Australia xxvi, 241–60

medical negligence   asbestos 23–4, 68–9, 74   damages 218    France 116, 117–18, 121, 124, 125, 128    future harm, liability for risk of 226–7, 229   Hotson case 218    judicial reasoning 7, 9    loss of a chance xxv, 68–9, 74, 124, 217–18, 227    product liability suits, biostatistics in medicinal xxv    warn, duty to 9 medicines see also DES (diethylstilbestrol) cases    biostatistics xxiv–xxv, 179–94    Canada, inference-drawing in fact in xxii, 93–5, 102–4, 107, 109   Coumadin 265–6    France xxiv, 173–8    oral contraceptives and strokes 151, 152–4, 190    product liability suits xxiv–xxv, 150, 156, 164–94      Seroxat (Paroxetine) 186, 187    TeGenero Incident 187    Thalidomide 179–80, 264, 265    Vioxx 185–6, 187    withdrawal from market 179, 185 mesothelioma    asbestos 14, 18–19, 21, 25–30, 34, 60–1, 65–7, 77–82, 85, 337   damages 8    future harm, liability for risk of 221    joint and several liability 30, 34    judicial reasoning 8   NESS test 326, 336–7    product liability suits, biostatistics in medicinal 193–4    scientific uncertainty, law’s approach to harm in context of xxx, 398–402    single hit model 19 Migues v Fibreboard Corp case 18–19 Mill, John Stuart 286, 289, 311–12, 315 Miller, Chris 50, 54–5, 157–8 minimal-maximal 3-set analysis of causation see MMTS (minimal-maximal 3-set) analysis of causation minimalists xxix, 48–9, 361, 281 miscarriages of justice xxvi, 241–2, 247 misrepresentation 50–5 mistake see error misunderstood facts 242, 254–6 mixtures of causes xxix, 369–75 MMR vaccines xxiv, 150, 164–73 MMTS (minimal-maximal 3-set) analysis of causation 339–59   acts and omissions 352–9   agent-conditioning xxviii, 341, 353, 359   alternative causation xxviii, 340, 351–2   antecedent conditions xxviii, 339–44    association laws 342    but for test 339–40, 353, 355–9    causal sufficiency xxviii, 339–59    circularity 340, 342    concurrent causation 340, 356–7    counterfactual dependence 339, 341, 354–8

Index  439    denominalisation rule of correspondence xxviii, 341, 344    dependence theory xxviii, 339, 341, 354–8    double preemption xxviii, 340    double preventions 352–9    empirical consequences, set of xxviii, 342–6, 348–51, 357    generalisations, set of causal xxviii, 341–54, 357–9    generic sense, causal factors in a xxviii, 341, 359    internal necessity element 345   intuition 356   INUS test xxviii, 339–42, 346–52    lawful sufficiency 340   legal context xxviii, 339, 341    maximal satisfaction of a set of causal generalisations 342, 344–6, 351    minimality conditions 345–8   moral context xxviii, 339, 341    nature, law of 346    negative factors xxviii, 352–9   NESS test xxviii, 339–42, 346–52    nomonological, regularity or covering law of causation xxviii, 339, 341–2    omissions xxviii, 341, 347, 352–9    overdetermination 340, 354    positive conditions xxviii, 352–4, 357–9    preemption xxviii, 340, 342, 345, 348–50, 355–7   preventions 352–9    proper logical parthood 342, 343    redundant conditions 339–40, 345–6, 348    satisfaction of causal generalisations 342, 343   science 341    social phenomena, character of 342    spurious conditions, exclusions of xxviii, 340, 345, 347–8    universal laws 342, 346–7    weak sufficiency 339–40 models of causation see causal models monotonicity 142–3 Moore, Michael 292–3, 299–304, 312–14, 355–6 Morabio, A 182 morality xxviii, 5–6, 323–4, 339, 341 more probable than not standard 198, 199–202, 215–16, 219 multi-party litigation xxiii, 176 multiple causes/agents   asbestos 14, 26–7, 31, 34–7   but for test 191   product liability suits, biostatistics in medicinal 191, 193   risk rule 31, 36–7   scientific and medical expert evidence in causation decisions in Australia     242–3 multiple defendants   asbestos xx, xxviii, 14–16, 26–7, 36–7, 326, 336–7   damages 15   Fairchild case 8, 398–9, 401   NESS test xxviii, 324–6, 336–7    scientific uncertainty, law’s approach to harm in context of 398–9, 401 murder 3–4

Nagel, Ernest 342 National Health Service (NHS) 68–9 National Vaccine Injury Compensation Program xxiv, 176–8 natural causation xxvii, 285–322 natural experiments 206 natural processes xxix, 364–7, 371, 373–5, 380 nature, law of 346 necessary element in a sufficient set test see NESS test of causation necessity   behaviour of others, causing the xxviii, 362, 368, 370, 373–4   France 114   internal necessity element 345   INUS test xxvii, xxviii, 285–90, 295, 339–42, 346–52   lawful necessity/sufficiency 331–6   NESS test xxvii–xxviii, 285–307, 323, 331   Scotland 41–2, 47   spurious necessity, creation of xxvii, 295–6   weak necessity 286–8, 291, 295–6 negativing causes xxvii, xxviii, 311–22, 352–9, 363 negligence see also medical negligence   asbestos 16, 21, 28–9, 59–61, 64, 68, 73–4, 82–90    Australia 242–3, 253–4, 256–9    Canada, inference-drawing in fact in xxi, 93–104, 107, 109    contributory negligence 16, 21, 28–9, 120–1    damages 243–4, 253–4    experts 242–3, 253–4, 256–9   Fairchild case 8–9    France 119–21, 124, 125, 128    future harm, liability for risk of 222, 225–7, 234–5    judicial reasoning 3–4, 6, 8–9    legal malpractice 218    medical evidence 242–3, 253–4, 256–9   misrepresentation 50–5   NESS test xxviii, 326, 336–7    pleural plaques 82–90    product liability cases, scientific evidence in 150   science      expert evidence 242–3, 253–4, 256–9     fact-finding 96–101 nervous shock 63–4, 66 NESS test of causation 285–322, 323–37    antecedent conditions 289–90, 299, 301–4, 335–6    behaviour of others, causing the 368, 370, 373    but for test xxviii, 47–8, 285–6, 288, 291–4, 298, 303, 321–6, 330, 337    causal directionality, problem of xxvii, 296    causal sufficiency xxvi, 289–311, 320–1, 323–7, 332–6    causally relevant factor account xxvii, 285–90    circularity of NESS xxvii, 285–6, 290, 293, 296, 300–1, 307, 332–3      collateral effects of a common cause xxvii, 297    common sense 288    complex conditions 332–4    conditions not or not proven to be strongly necessary or sufficient xxvii    counterfactuals 50–3, 287–8, 294, 299, 304, 308, 329    criticisms xxvii–xxviii, 285–337

440  Index NESS test of causation (cont.):    cumulative sources of harm 327–8    dependency 291, 316–20, 328–9   Difference Method 289    duplicative causation 47–8, 287, 294, 303–7, 318, 324    empirical relation, causation as an 286    epiphenomena (collateral effects of a common cause) xxvii, 297    evidence 290, 337    facts xix, 323   Fairchild case 9    free will 309    generalisations 288, 290–2, 297–304, 307–8    Hart and Honoré’s causally relevant factor account xxvii, 285–90    human decisions and actions as causes, treatment of xxvii   immunity 326    increase in proliferation of causes xxvii, 304, 321–2    increase in risk 337    indeterminancy xxvii, 50–1, 53, 309–11, 326, 336–7   inferences 291   interruption 330–1    intuition 306, 320–1, 335   INUS account xxvii, 285–90, 295    involvement concept 306–7    joint and several liability xxviii, 337    judicial reasoning xix, 3, 4–5, 9    lawful sufficiency 289, 291–2, 295, 297–8, 331–6    logical skeletons, uncovering 323–4    mesothelioma 326, 336–7    MMTS analysis xxviii, 339–42, 346–52    moral or normative character 323–4    multiple asbestos employers’ litigation xxviii, 326, 336–7    multiple defendants xxviii, 324–6, 336–7    multiple omissions 323, 328    natural causation 285–322   necessity xxvii–xxviii, 285–307, 323, 331    negligence xxviii, 326, 336–7    omissions xxvii–xxviii, 311–21, 323, 328    overdetermined causation xxvii, 287, 292–3, 309, 316–24, 327–31, 337    positive causation 312–19    pre-emption xxvii, 287, 294, 297–303, 309, 311, 316, 318–19, 328–32    prevention causal process 308–9    probabilities xxvii, 309–11, 327–8, 336    proliferation of causes 292, 304, 321–2, 331–7    reasons as causes 307–9    road accidents 47–8    Scotland 43, 46–51, 53, 56    single fibre theory 286, 289–91, 296, 307, 326    spurious necessity, creation of xxvii, 295–6   statistics 309–10    strong sufficiency 286–9, 291–2, 297–8, 303, 307, 310–11, 315    successional nature of causation 289–90    sufficiency xxvii–xxviii, 323, 331

     strong sufficiency 286–9, 291–2, 297–8, 303, 307, 310–11, 315      weak sufficiency 286–8, 291, 295–6    trivial contributions 304–5, 322    utility of NESS, establishment of 324–31    weak sufficiency 286–8, 291, 295–6 Newton, Isaac 355–6 no-fault liability 114–16 nomonological, regularity or covering law of causation xxviii, 339, 341–2 Northwick Park clinical trial disaster 187 novel evidence xxvi, 259–60 novus actus interveniens 362–3, 376 Nussbaum, Martha 98 objectivity   behaviour of others, causing the xxviii–xxix, 362–4, 366, 380   facts xxviii–xxix, 362–4, 366, 380   France 118   probability, theory of objective 118   scientific and medical expert evidence in causation decisions in Australia     241–2, 248–50 Observations   product liability suits, biostatistics in medicinal 179–84   pure observation xxiii, 137   scientific and statistical evidence 137–40, 143, 146   scientific fact-finding 100 oestrogen 274–8 see also DES (diethylstilbestrol) omissions   behaviour of others, causing the 374–5   contribution 374–5   MMTS analysis xxviii, 341, 347, 352–9    multiple omissions 323, 328    negative omissions 311–21   NESS test xxvii–xxviii, 311–21, 323, 328   overdetermination 316–21 Omnibus Autism Proceedings (OAP) test cases in United States xxiv, 164–73    assessment of scientific testimony of proof 164    burden of proof 165   compensation 165   Daubert framework 164, 171, 177–8    epidemiological evidence 168–71, 173    evidence 164–71, 173, 177–8    experts 165–7, 169, 171    fact, causation in 165, 170–3    general causation 164, 166–73    implications of test cases 170–3    junk science, juries and xxiv, 164    MMR vaccines xxiv, 150, 164–73   National Childhood Vaccine Injury Act of 1986 xxiv, 150, 164–5    petitioner’s first theory 164–7   probability 172–3   proof 164–5      proximate temporal relationship 171–3    specific causation 171–3    standard of proof 172    Table Injury 165

Index  441    thimerosal containing vaccines xxiv, 150, 164–6, 168–70    United States 150 oral contraceptives and strokes 151, 152–4, 190 overdetermination    cumulative sources of harm 327–8   interruption 330–1    MMTS analysis 340, 354   NESS test xxvii, 287, 292–3, 309, 316–21, 323, 324, 327–31, 337    omissions 311–21, 354   pre-emption 328–9   Scotland 42–3 Paillet, M 122 pain and suffering, damages for 221–3, 253–4 parasitic accountability doctrine 222 Parkinson’s disease 262, 270–3 Paroxetine 186, 187 part of the history test 4–5, 7, 9 particularistic evidence, requirement for xxv, 196–7, 208–13 Paul, LA 300–1 Pearl, Judea 106, 146 Pécresse, Valérie 129 Pennington, Nancy 104–6 Perera, F 278–9 personal chance/statistical chance dichotomy 161–3 pesticides 271–2, 279, 385, 386–98 plausibility 191, 394, 397–8 pleural plaques    actionable damage 82–3, 86, 89–90    anguish and fear, future harm and 222, 224–5    asbestos 57, 74, 77–90   compensation 82–90    future harm, liability for risk of 222, 224–5   insurance 84–6   legislation 82–90   negligence 82–90    policy 83–4, 88   politics 82–90    principle, questions of 86–7   retrospectivity 83–5   Scotland 82–90 pluralism 365–6 pneumoconiosis claims 81 policy    American Legal Realists 70–1, 91   asbestos xxi, 57–8, 61–2, 64, 68–71, 74, 76, 83–4, 88, 91    Australia 241, 243    corrective justice theory xxi, 58, 69–71    experts 241, 243   Fairchild case xxi, 58, 61, 65–6, 69–71, 90–1    France xxii, 111, 127–9    future harm, liability for risk of 221, 224    judicial reasoning xix, 70–1    legislation xxi, 57, 70, 77–9, 91    medical evidence 241, 243    pleural plaques 83–4, 88   politics 70–1    psychiatric injury 64

  science     expert evidence 241, 243      uncertainty, law’s approach to harm in context of 401    standard of proof 204 politics    asbestos 58–9, 61–2, 64, 70–1, 77–91   Barker v Corus case 77–82    compensation 58, 79–82    corrective justice theory 70–1    legislation 58–9, 79–81    pleural plaques 82–90   policy 70–1 positive causation, NESS test and 312–19 postponement and immediate recovery, choice between xxvi, 223, 233–8    bare risks 237    case for liability 234–6    collective action xxvi, 223, 238–9   costs 237    insolvency, risk of 223, 238   insurance 235–7    limitation periods 234    minimum threshold objection 237   objections 236–8    redundancy objection 237    statistical evidence 237    statutory funds 238 post-traumatic stress disorder (PTSD) 249–50 potency xxix, 379–81 Pothier, RJ 113 precautions    asbestos 25, 29    experts 387–91, 394–6    indeterminate causation 386–98    judicial reasoning 5–6    judicial review 386–98, 405    precautionary principle xxx, 386, 387, 391–7, 403–7   regulation 386–98    scientific uncertainty, law’s approach to harm in context of xxix–xxx,    386–98, 403–7 precedent xxi, 17, 30, 59, 64, 78, 81, 85–6, 253, 405 pre-emption   double pre-emption xxviii, 340     MMTS analysis xxviii, 340, 342, 345, 348–50, 355–7   NESS test xxvii, 287, 294, 297–303, 309, 311, 316, 318–19, 328–32   overdetermination 328–9   Scotland 48 preponderance of evidence 195, 199–200, 215–20, 399, 402–4 presumptive reasoning xxii, 114, 123–4, 127 prevention   MMTS analysis 352–9   NESS test 308–9   omissions 352–9 prima facie presumptions 219–20 principles   asbestos xxi, 58–9, 68, 71, 80, 86–7    legislation xxi, 58–9, 80    pleural plaques 86–7

442  Index probabilities see also balance of probabilities/ preponderance of evidence    belief 197, 200, 204   Canada, inference-drawing in fact in 100–2    damages xxvi, 222–3, 226–32    estimates of probability of causation 141–2    ex ante causal probabilities xxv, 206–8, 210–13    France 119, 123    future harm, liability for risk of xxvi, 222–3, 226–39    mathematical probabilities 199, 201, 209–13    more probable than not standard 198, 199–202, 215–16, 219   NESS test xxvii, 309–11, 327–8, 336    objections 222–3, 236    objective probability, theory of 118    Omnibus Autism Proceedings (OAP) test cases in United States 172–3    product liability suits, biostatistics in medicinal 184, 188, 191   quantum 232–3   refinements 228–31   science      evidence xxiii, 133, 140–3, 146–7     fact-finding 100–1      uncertainty, law’s approach to harm in context of 400–1, 403–4, 407    standard of proof 196–7, 199–213, 215–19    statistics xxiii, xxv, 133, 140–3, 146–7, 196–204, 207, 209–20 product liability cases 149–78 see also asbestos claims    access to justice xxiii, 176    admissibility xxiv, 152, 164, 177–8    associations xxiii–xxiv, 150, 155–8, 180–4, 188, 191–2    Atomic Veterans case 192–3    attribution of causality 179    Bayes’ Theorem xxiv, 150, 162–3, 177   bias 183–4   biostatistics 149–78    bladder cancer and tobacco smoke 152    Bradford Hill criteria 181–2, 184, 188–9, 191    burden of proof xxiv, 152, 188    but for test 158, 191, 193    complex cases xxiv, 149–50     confounding 183–4    court, relevance of biostatistics to the 187–94      cumulative causes 154, 191   Daubert framework xxiv, 164, 171, 177–8    divisible and indivisible injuries xxv, 193–4    doubling of risk of injury xxiii, 150, 151–5, 176    drugs xxiv, 150, 156, 164–78    epidemiological evidence xxiv–xxv, 150–61, 176–7, 179–90, 193–4   error 183–4   estimations 184   evidence 149–78    experts 151, 157, 159, 165–7, 169, 171    fact, causation in 165, 170–3   Fairchild case xxv, 192–3    fallibility of scientific approach 185–7    France xxiii, xxiv, 116, 150, 173–6, 178    gatekeeping role xxiv, 164, 178

   general causation xxiii– xxiv, 150, 152, 156–7, 160, 164, 166–76, 192    general principles 188–9   Gregg v Scott 162    Hepatitis B vaccine xxiv, 150, 174–6   Hotson case 162    increase in risk theory 192   interpretation 150    joint and several liability 192    junk science by juries, use of xxiv, 164    legal approaches and medical science, differences between 189–94    Legal Services Commission xxiii, 176    lung cancer and smoking 156–63, 177    material contribution 191–2   McGhee case 154, 192   McTear case xxiii–xxiv, 150, 156, 158–63, 177, 188    medicinal cases xxv, 149–78   mesothelioma 193–4    multi-party litigation xxiii, 176    multiple causes test 191, 193   National Vaccine Injury Compensation Program xxiv, 176–8   negligence 150    null hypothesis 184   observation 179–84    Omnibus Autism Proceedings (OAP) test cases in United States xxiv, 150, 164–73    oral contraceptives and strokes, link between 151, 152–4, 190    personal chance/statistical chance dichotomy 161–3, 191    probability theory 184, 188, 191    proof 149–50, 155, 164–5, 190–1, 194    prospective forecasting xxiv, 189–90    proximate temporal relationship 171–6    public funding xxiii, 176    retrospective versus prospective forecasting xxiv, 189–90   risk      benefit, assessment of 179–87, 190, 194       definition 183     difference 183      divergent approaches to risk as test of causation 191–3     ratio measures 183     relative risk 188–9, 192       test xxv, 191–4    scientific evidence 149–78    scientific method, introduction of 180–2    Scotland xxiii–xxiv, 150, 156, 157, 159–60, 177    serious, precise and concordant presumptions 174–5, 178    Seroxat (Paroxetine) 186, 187    significance testing 184    single agent 193    smoking 152, 156–61, 177, 183    specific causation xxiii–xxiv , 150, 152, 156–7, 160, 171–4, 188–9, 192    standard of proof xxiii–xxiv, xxv, 150–63, 172, 176–7, 191–4    statistics xxiv, 151, 159–63, 174, 177

Index  443   sufficiency of scientific evidence 150, 152, 184    TeGenero Incident 187    tests for causation 190–3    therapeutic use 179   Thalidomide 179–80   threshold 152    tick box approach 182    United Kingdom xxiii, 150–63, 176    United States xxiii, xxiv, 150, 151–2, 164–73, 176, 185–6   vaccines xxiv, 150, 164–77    Vioxx 185–6, 187   Wilsher case 193    withdrawal from market 179, 185 proliferation of causes 292, 304, 321–2, 331–7 proof see also burden of proof; standard of proof    asbestos xx, 15–24, 36–7, 59–60, 62–5, 68, 88    behaviour of others, causing the 368–9   France 114, 123–4      general causation xxiii– xxiv, 14, 123, 127, 152, 158–9, 174–8, 205    Omnibus Autism Proceedings (OAP) test cases in United States 164–5    product liability 149–50, 155, 164–5, 190–1, 194   res ipsa loquitur 368    specific causation xxiii–xxiv, 123, 150, 152, 156, 205, 215–16 proportionality 66–8, 78–9, 81 prospective forecasting xxiv, 189–90 proximate cause 23, 43, 188 psychiatric evidence 245, 249–50, 253–4 psychiatric injury or nervous shock 63–4, 66 public funding xxiii, 176 public/private law divide xxii, 111 Putman, Hilary 98 Ramazzini, Bernardino 275 real causation 5, 7 reasonableness   fair, just and reasonable test 66–7   reasonable certainty standard 216   scientific uncertainty, law’s approach to harm in context of 394–5, 397–8, 407 reasons 307–9, 366, 367, 370, 380–1 recurrent wrongdoers 230–1 redundancy 237, 339–40, 345–6, 348 regularity   asbestos 17–18   frequency, regularity and proximity test 17–18    nomonological, regularity or covering law of causation xxviii, 339, 341–2 regulation xxiv–xxv, 383–98 Reichenbach, H 335 relative contribution 38–9 res ipsa loquitur xxv, 219–20, 368 Resnik, David xxx, 392, 394, 396–7 retrospectivity xxiv, 77–8, 83–4, 189–90 risk see also increase in risk   asbestos xx, 11, 15–40, 72–3   assessment of risk 384, 387–8, 393, 396–7, 403, 407   bare risk 228, 231, 236–7, 239   benefit, assessment of 179–87, 190, 194  

  Canada, inference-drawing in fact in 106–7    complex fact patterns 31, 33–5   definition 183   difference 183    divergent approaches to risk as test of causation 191–3    doubling of risk xxiii, 65, 76, 150, 151–5, 176    expansion xx, 11, 22, 31–3   Fairchild case 9    future harm 221–39    injury rather than risk xx, 11    insolvency 223, 238    judicial reasoning 8–9    limitations 32–6, 39    lung cancer 31   multi-agent fact situations 31, 36–7    outcomes or risks, concern with 73    product liability suits, biostatistics in medicinal xxv, 179–94    quantitative risk assessment 393, 396, 403, 407    ratio measures 183    relative risk 188–9, 192    risk rule xx, 11, 15–40    scientific uncertainty, law’s approach to harm in context of 384, 387–8, 393,396–7, 403, 407   scope 11    similar fact situations 11    temporal proximity between risk and harm 106–7    United Kingdom 23–34, 36–40    United States 15–24, 30–40 road accidents   asbestos 15–16, 19, 37    counterfactual analysis 54–5   France 116   indeterminancy 54–5   NESS test 47–8    Scotland 47–8, 54–5 Robins, James M 143, 145–6 Royal Commission on Environmental Pollution (RCEP) 387–90, 392–7 Rubin, Donald B 146 rules of law xix, 4, 7–8 Russell, Bertrand 341 Rutter, Michael 170 Sainte-Rose, M 129 Schaffer, Jonathan 302, 306, 314 Schardein, James 263, 267 Schroeder, Christopher 72 Schwartz, Gary 71 science see also scientific and medical expert evidence in causation decisions in Australia; scientific and statistical evidence in assessing causality, role of   biostatistics and causation in medicinal product liability suits 179–94   burden of proof 97   Canada, inference-drawing in fact in 96–102    cause-in-fact 98, 100–1   epidemiological research 100      evidence xxiii, xxiv, 103, 149–78    experimental studies 100

444  Index science (cont.):   fact-finding 96–102    false causality 100   France 127    gatecrasher paradox 98, 99    generalisations 97, 99–100    harm in context of scientific uncertainty, law’s approach to 383–408    legal fact-finding 96–9    MMTS analysis 341   negligence 96–101    observational studies 100    probability assessments 100–1    product liability xxiii, xxiv, 149–78    reference classes 99–102    soft variables 98, 102   statistics 96–102 scientific and medical expert evidence in causation decisions in Australia xxvi, 241–60   accountability for evidence, counsel making experts 242, 245   accounting evidence 241   admissibility criteria xxvi, 247, 259–60   asbestos 251–3   bias 241–2, 248–50      bizarre and unscientific evidence 242, 244–7    Bradford Hill criteria 251–2    but for test 243    civil proceedings 242, 260    common law 243–4, 248    common sense approach 243–4    complex evidence 245, 249, 252, 255–6, 259    conflicts of evidence 241    criminal proceedings 242, 260    damages 243–4, 253–4   DNA 248, 250    emotional disturbance, leukaemia caused by 245    epidemiological evidence 251–3   errors xxvi, 241–2, 247–8, 256, 259–60    evaluation, evidence not easily amenable to 242, 250–4    exceptional circumstances 244   facts      fact-finding 241, 245, 253, 259      factual causation 242–4, 257, 259     misunderstood, where 242, 254–6   Fairchild case 244, 252    hindsight errors xxvi, 259–60    HIV, causation of 246–7    iconoclastic or heterodox theories xxvi, 245–8, 259–60    juries xxvi, 244–5, 250, 259–60    loss of chance reasoning xxvi, 244, 254, 256–9    material contribution 242–3, 244    miscarriages of justice xxvi, 241–2, 247    misunderstood, where facts are 242, 254–6    multiple factors 242–3    negligence 242–3, 253–4, 256–9   New South Wales 243–4    novel evidence xxvi, 259–60    objectivity, lack of 241–2, 248–50      pain and suffering, damages for 253–4

   policy 241, 243    post-traumatic stress disorder 249–50    psychiatric evidence 241–2, 256    psychiatric harm 245, 249–50, 253–4    standard of proof 244, 257    statistical evidence 245, 250–3    victims, sympathy for xxvi, 245, 260 scientific and statistical evidence in assessing causality, role of 133–47    aspirin as cure for headaches 134, 135–6, 140–5    Berkeley graduate admissions, gender and xxiii, 138–9    black box approach 136–7, 143, 145    Causes of Effects (CoE) (token or individual causation) xxii–xxiii, 133–4, 140–5    cautionary tales 138–40    conditional probability 140–1    confounding variables 139–40, 146    contrastive causality 134–5    coronary artery disease xxiii, 138, 140    counterfactuals 134–5, 140    data collection and generation xxii–xxiii, 135–40, 143, 145–6   Effects of Causes (EOC) (type or general causation) xxii–xxiii, 133–7, 145–6    estimates of probability of causation 141–2    experimental studies xxii–xxiii, 135–7, 140, 143, 145    hormone replacement therapy and coronary artery disease xxiii, 138, 140   hypotheticals 134–5    legal causation 134   monotonicity 142–3    observational studies 137–40, 143, 146    probability of causation xxiii, 133, 140–3, 146–7    pure observation xxiii, 137    randomised controlled clinical trials 135–7    role of science 143–5    smoking and lung cancer 134–5, 137    varieties of causal questions 133–5 Scotland    academic writings xx, 43–4, 46, 55–6    actionable injury, nature of xxi    appropriate test 44    asbestos xxi, 58–9, 82–90   Barker v Corus 42    but for test xx, 41–3, 45, 46–8    causal uncertainty xxi, 48–55   causation simpliciter 45   compensation 82–90    content of causal enquiry 44–7    context-specific understanding xx, 44    counterfactual analysis of decision-making xxi, 49–55    current position 41–4   damages 51   delict 41–56    doctrinal perspective on causation xx, xxi, 41–56    duplicate causation 42, 47–8    evidentiary gap 49    fact, causation in 41–2, 45, 46    factual causation 41–3, 45, 46   Fairchild case 59

Index  445    future doctrinal development 44–55    heroin, administration of 45    human behaviour, unpredictability of xxi, 49–55    indeterminacy 42–3, 48–55    indivisible injuries 42   insurance 84    ‘involvement’, use of word 46–7    judiciary 43–6, 49, 53, 55–6   Kennedy objection 45–6    legal causation 41, 43, 45    legally specific content of causation 44    mapping cases of indeterminate causation xxi, 48–9    material contribution test 42, 47   McGhee v NCB case 42   McTear case 150    minimally sufficient conditions 48–9   misrepresentation 50–5    necessity 41–2, 47    negligence 49, 50–5   NESS test xx, 43, 46–51, 53, 56    new test 47–8    normative issues 45–6    overdetermined outcomes 42–3    part of causation, what is 44    pleural plaques 82–90    potentially indeterminate behaviour of defendant 52–3    potentially indeterminate behaviour of pursuer 52    pre-emptive causes 48    product liability cases, scientific evidence in xxiii–xxiv, 150, 156, 157, 159–60, 177    proximate cause 43    risk creation 42    road accidents 47–8, 54–5    Scottish Parliament 82–90    significant or important causes 43    sole cause 42    statutory interpretation 45–6   taxonomy      mapping types of indeterminate causation xxi, 48–9     production xxi    terminology, reform of xx–xxi, 44–7    uncertainty cases 48–55 Selikoff, Irving 12 Seroxat (Paroxetine) 186, 187 Sherwin, Emily 198–9 significance testing 184 significant or important causes 43 Sindell v Abbott Laboratories case 30, 32–5 sine qua non test see but for/sine qua non test single agent theory xx, 20, 27, 33–6, 39–40, 42, 61–3, 66, 69, 193 single hit model 14, 19–20, 23–5, 27, 34–40, 205, 326 smoking    asbestos 21–2, 31, 35, 252–3    expert evidence 268–70    lung cancer 21–2, 31, 35, 134–5, 137, 156–63, 177, 183, 268–79    product liability 152, 156–61, 177, 183   science     development 268–70

     evidence 134–5, 137, 152, 156–61, 177    standard of proof 201, 207    statistical evidence 134–5, 137 Snow, John 181–2 social phenomena, character of 342 specific causation    Causes of Effects (CoE) xxii–xxiii, 133–4, 140–5   diseases 261    France 123, 174    Omnibus Autism Proceedings (OAP) test cases in United States 171–3    product liability 150, 152, 156–7, 160, 171–4, 188–9, 192    proof xxiii–xxiv, 123, 150, 152, 156, 205, 215–16    science, development of 261    standard of proof 2–6, 215 spurious necessity, creation of xxvii, 295–6 stand-alone harm 222, 225, 234 standard of proof 195–220    abstract initial base rates xxv, 211–12    actual causation 204, 206–7, 213, 216–19    alternative tortious claims 212–14    Australia 244, 257    balance of probabilities/preponderance of evidence 195, 199–200, 215–20      scientific uncertainty, law’s approach to harm in context of 399, 402–4     United States 200    Bayes’ Theorem xxv, 211–12    belief in truth xxv, 195–201, 204–13, 220    beyond a reasonable doubt 196, 198, 199–202    but for test 218–19    causal capacity 205–6, 215    civil law 195–201, 204–5, 208, 212, 215–17    clear and convincing evidence standard 200–1    common law 195–205, 208, 212, 215–17    concrete conditions 205    convictions or belief 195–7    criminal law 195–200, 211, 220    cumulative contribution 218   damages 218   DES (diethylstilbestrol) cases 214–15    doubling of risk xxiii, xxv, 202–3, 215–16   epidemiology 205    evidence xxv, 202, 207, 209, 213, 219–20    evidential groupings 213    ex ante causal probabilities xxv, 206–8, 210–13   experimentation 205–6    experts 244, 257    facts in issue 195    France 195, 198–9, 217    full proof 197–8    gatecrasher hypothesis 210, 212    general causation 205, 215    generalisations xxv, 205–9    Germany 200, 217    indistinguishable alternative tortious claims 212–14    inferences 214, 219–20    initial base rates xxv, 211–12    intime conviction standard 195, 199, 205   Italy 197–8    joint and several liability 212, 214–15

446  Index standard of proof (cont.):    judicial reasoning 8    juries 200–2, 211, 220    legal malpractice cases 218    lost chances in medical malpractice cases xxv, 217–18    mathematical probabilities 199, 201, 209–13    medical evidence 244, 257    medical malpractice 216–19    more probable than not standard 198, 199–202, 215–16, 219    natural experiments 206   negligence 216–20    Omnibus Autism Proceedings (OAP) test cases in United States 172    particularistic evidence, requirement for xxv, 196–7, 208–13   policy 204    prima facie presumptions 219–20    probability 196–7, 199–213, 215–19    problems and paradoxes 212–20    product liability xxiii–xxv, 150–63, 172, 176–7, 191–4    reasonable certainty standard 216   res ipsa loquitur xxv, 219–20   science      expert evidence 244, 257      uncertainty, law’s approach to harm in context of 385, 399, 402–4   Sienkiewicz v Grief (UK) case xxv, 155–6, 202, 216    singular instance of causation 205    smoking, cancer caused by 201, 207    specific causation 205, 215    statistical probabilities xxv, 196–7, 199–204, 207, 209–20    toxic exposures 215–16    toxicological experimentation 205–6    truth of facts in issue 195    United States 200–2, 213–18   vaccines 214    virtual certainty 196    weight of evidence 202 Stapleton, Jane xix, xxviii, xxix, 4, 7, 14–15, 22–4, 43–6, 54, 292, 298–9, 305–7, 319–20, 340, 362–6, 370–3, 377 statistics    50% xxv, 216–17, 220   asbestos 251–3    Australia 245, 250–3    Bayes’ Theorem xxiv, 162–3, 177    biostatistics xxiv–xxv, 179–94    Bradford-Hill criteria 251–2    Canada, inference-drawing in fact in xxii, 96–102    evidence 100, 133–47, 229–30, 237, 245, 250–3    experts 245, 250–3    France 126, 174    future harm, liability for risk of 229–30, 237   Germany 217   McTear case xxiv, 150, 159–63, 177, 188    medical evidence 245, 250–3    medicinal product liability suits xxiv–xxv, 179–94     NESS test 309–10

   personal chance/statistical chance dichotomy 161–3    postponement and immediate recovery, choice between 237    probabilities xxv, 196–7, 199–204, 207, 209–20    product liability cases, scientific evidence in xxiv, 151, 159–63, 174, 177   science     expert evidence 245, 250–3     fact-finding 96–102    standard of proof xxv, 196–7, 199–204, 207, 209–20 Stevens, Robert 70–1, 73–6, 292 Strevens, Michael 295, 340, 347 strict liability 6–7, 114 substantial factor test in United States xx, 15–16, 18, 20–3, 32, 37 substitute facts 368 successional nature of causation 289–90 sufficiency   behaviour of others, causing the 362, 368–70, 373   insufficient reason, principle of 393–4   INUS test xxvii, xxviii, 285–90, 295, 339–42, 346–52   lawful necessity/sufficiency 289, 291–2, 295, 297–8, 331–6   MMTS analysis xxviii, 339–59   NESS test xxvi–xxviii, 286–311, 315, 320–1, 323–7, 331–6   product liability cases, scientific evidence in 150, 152   science     evidence 150, 152     product liability 150, 152     uncertainty, law’s approach to harm in context of 407   strong sufficiency 286–9, 291–2, 297–8, 303, 307, 310–11, 315   weak sufficiency 339–40 systematic bias 184 Tarrible, J 112 Taruffo, Michele 197 taxonomy    mapping types of indeterminate causation xxi, 48–9   production xxi TeGenero incident 187 temporal relationships   Omnibus Autism Proceedings (OAP) test cases in United States 171–3   product liability cases, scientific evidence in 171–6    proximate temporal relationships 171–6 Thalidomide 179–80, 264, 265 Thomson, Judith Jarvis 347, 356 Tian, Jin 146 time limits 234, 261–3 Träger, L 118 Tribe, Laurence 97–8, 102 trivial contributions 304–5, 322 truth   belief in truth xxv, 195–201, 204–13, 220    ex ante causal probabilities xxv, 206–8, 210–13    facts in issue 195    particularistic evidence, requirement for xxv, 196–7, 208–13

Index  447    scientific uncertainty, law’s approach to harm in context of 385–6    standard of proof 195 uncertainty see also harm in context of scientific uncertainty, law’s approach to    case study 49–55    certain and direct, causation as being 113   France 113    reasonable certainty standard 216    Scotland xxi, 48–55    virtual certainty 196 unidentifiable victims 227, 230–1 unidentifiable wrongdoers 226, 229 United States    American Legal Realists 70–1, 91   asbestos xx, 11–13, 15–24, 26–7, 30–40    balance of probabilities 200   Borel v Fibreboard Paper Products case 12, 15–17, 37, 40   Borg-Warner v Flores case 22, 32   Brown v Superior Court case 30    Canada, inference-drawing in fact in 105    comparative responsibility regime 16, 18, 29, 38    damages 221–4, 233   Daubert framework xxiv, 164, 171, 177–8   DES cases 30, 32–3    diseases 261, 274    doubling of risk 151–3    emotional distress 221–4   epidemiology 151–2    future harm, liability for risk of 221–6, 230, 233, 238–9   Gregg v V-J Auto Parts Inc 22–3    joint and several liability 38–9   juries 200–2   Lohrmann v Pittsburgh Corning Corp case 17–18   Migues v Fibreboard Corp case 18–19    Omnibus Autism Proceedings (OAP) test cases in xxiv, 150, 164–73    product liability xxiii, xxiv, 150, 151–2, 164–73, 176, 185–6   proportionality 66–7    risk rule 15–24, 30–40   Rutherford v Owen-Illinois Inc case xx, 11, 19–24, 26–7, 30–7, 39–40

  Science     development of 261, 274     evidence 150, 152   Sindell v Abbott Laboratories case 30, 32–5    standard of proof 200–2, 213–18   states 38–9    substantial factor test xx, 15–16, 18, 20–3, 32, 37    Vioxx 185–6, 187 universal laws 342, 346–7 unpredictability of human behaviour xxi, 49–55 vaccines   France 123–4, 126–7    Hepatitis B vaccine 123–4, 126–7    MMR vaccines xxiv, 150, 164–73   National Vaccine Injury Compensation Program xxiv, 176–8    Omnibus Autism Proceedings (OAP) test cases in United States xxiv, 150, 164–73    product liability cases, scientific evidence in xxiv, 150, 164–77    standard of proof 214    thimerosal containing vaccines xxiv, 150, 164–6, 168–70 variable hit models 270–3 Verma, TS 106 victim migration xxvi, 238 Viney, G 121 virtual certainty 196 voluntary assumption of responsibility 74–5 von Bar, L 118 von Büri, M 117 von Kries, J 118 von Liszt, F 117 wait and see approach xxvi, 223, 233–8 Waldron, Jeremy 80 warn, duty to 7 weak necessity 286–8, 291, 295–6 weak sufficiency 339–40 Wednesbury irrationality 389, 391–2, 403–4 Weinrib, Ernest J 71–2 Williams, Glanville 99 Woodruff, TJ 279 Wright, Richard xxvii–xxviii, 46–7, 50, 96, 99, 102, 105, 323, 327–37, 339, 341, 352