1. Concept and purpose
Casum sentit dominus. Occurrence of damage alone will not trigger liability. As long as the law of torts/delict rests conceptually on individual responsibility, liability requires everywhere—including in EU private law—that the damage can be attributed to the person held liable, or, more precisely, attributed to this person’s conduct, or to the conduct of another for whom he is responsible, or to an object which he has to keep under his control. This attribution is usually analysed under the catch-all heading of ‘causation’, irrespective of whether the individual claim is contractual, quasi-contractual or extra-contractual in nature; this entry will mainly deal with claims of the latter kind.
Attribution is not based on an objective, extra-judicial concept, but is the result of normative considerations. Rather than falling back on cause and effect as they are understood in natural science or in philosophy, every legal system has had to develop a specific legal notion of causation. Remarkably enough, even codified systems do not tackle this task by legislation. Instead, it is incumbent on courts and legal literature to decide borderline cases and flesh out general rules and principles (but see the comparatively detailed rules in the draft for a new Austrian law of delict of 2005/07 and the Principles of European Tort Law (PETL)).
This process has brought about a dichotomy that can be found in almost all European legal systems and also in EU private law. On the one hand, it is asked whether a particular conduct can actually be the cause of harm (‘factual’ or ‘natural’ causation). On the other hand, a whole range of criteria aims at limiting this all too wide set of possible causes. In order to express that these criteria are of a normative kind, it is advisable to speak not of causation (eg ‘legal’ causation), but of scope of liability. Comprehensive causation theories unite both aspects in one notion, but in doing so may run the risk of concealing the normative reasons for a decision behind seemingly objective and descriptive terms. A unified concept is also adopted by Art VI.-4:101(1) of the Draft Common Frame of Reference (DCFR) (and the identical provision found in Art 4:101(1) PEL Liab Dam of the Study Group on a European Civil Code) that reads like a general clause: ‘A person causes legally relevant damage to another if the damage is to be regarded as a consequence of that person’s conduct or the source of danger for which that person is responsible.’
A further complication arises from the fact that the causal nexus is almost inevitably intertwined with the other elements of delictual liability which it connects. An obvious link exists with the requirements of fault or breach of a duty of care, respectively. This is particularly apparent in Roman law, where culpa was the dominant factor in liability. The Roman jurists had not yet developed a specific doctrine of causation; the nexus between a person’s conduct and the damage was established by an interpretation of the operative verb in the pertinent rule establishing liability (eg occidere for killing somebody). It was not until the 19th century that modern general theories of causation had evolved. Under the French general clause of delictual liability, the requirement of faute is of such paramount importance that normative considerations limiting liability are sometimes made here rather than in the context of causation. And the first enquiry in the English tort of negligence is which duty of care the person to be held liable has breached, and only in a second step is it asked whether the breach of this duty has caused the damage complained of. Connections also exist between causation and the notion of damage. This can be illustrated by the example of loss of a chance following medical malpractice (perte d’une chance). If the loss of a chance itself constitutes the compensable and actionable damage (as in France; see also Art 7.4.3(2) UNIDROIT PICC), causation is definitely present and proportional damages can be granted. If, however, one regards the deterioration of health as the damage, one has to ask whether this result has actually been caused by the mistake or whether it would also have occurred even without the mistake.
Despite this possible overlap, causation and scope of liability remain indispensable elements of any concept of liability. Even in a ‘flexible system’ (such as the PETL), both cannot completely recede into the background in favour of the other prerequisites.
2. First level of enquiry: factual causation
a) Problems to be regulated and trends of international legal development
All European legal systems begin every causal enquiry—albeit often impliedly—by applying a so-called conditio sine qua non-test or ‘but for’ test; in Scotland, this test is sometimes called causa sine qua non without any difference in substance. The test’s question is whether, but for the defendant’s conduct, the harmful result would still have occurred. However, this seemingly simple formula entails a number of problems: (i) Despite its ostensibly objective character, even the ‘but for’ test cannot be applied, at least in borderline cases, without normative considerations based on experience and probabilities. The process of mentally eliminating one element in the sequence of events necessarily requires hypothetical thinking. This becomes particularly obvious in cases where the defendant’s conduct was an omission. The ‘but for’ test can capture these cases only in a modified version: an omission is the natural cause of a harmful result if the action that the defendant failed to take would have prevented it. (ii) Similarly, the ‘but for’ test reaches its limits in cases of ‘psychological causation’. Despite possible differences in detail as to the exact prerequisites, all legal systems accept at least in principle that a causal nexus can be predicated on a psychological influence. How a person would have reacted but for this influence can never be ascertained with scientific certainty but rather necessitates a hypothetical evaluation. (iii) The simple ‘but for’ test breaks down if a harmful result is over-determined, ie if one of two or more independent and simultaneous events would have sufficed to bring about the harmful result. Yet over-determination of a result is nowhere regarded as a reason to deny causation. Causation doctrine has developed the so-called ‘NESS test’ (‘necessary element of a sufficient set’) to conceptualize this finding: in the first step, this test eliminates possible causes from the sequence of events, and only in the second step does it ask whether the event in question was a necessary element in a sufficient causal chain. (iv) Finally, difficulties arise in cases that are brought together in the category of ‘hypothetical causation’. These cases have in common that a subsequent event would have brought about the same damage. The subsequent event can consist in the conduct of another person, or it can just as well be a hazard or a circumstance in the victim’s sphere (eg a special predisposition). Here, both the first and the second event can be mentally eliminated and the damage would remain. Liability can thus only be justified in a normative way; the attribution problem in these cases concerns the scope of liability and, in a way, also the notion of damage. It is hardly possible to provide a general answer since the issue is controversial not only internationally but also within national legal systems. The predominant view seems to be that liability for a clearly defined damage that has already occurred as a consequence of the first event remains unaffected. If, however, the damage is of a continuous nature (eg consequential harm arising from the loss of use of a damaged object), the second event is taken into account: a hazardous event or an event from the victim’s sphere may extinguish liability from this point onward. Delictual conduct of a third party may—similar to the case of an over-determined result—trigger that party’s liability together with that of the first tortfeasor. However, when it comes to details, many questions remain: how are the damages to be assessed (eg destruction of an object in general or because of a specific event)? When can one say that a particular loss no longer continues (eg reduction of earning capacity as a single event or continuous loss of income)? Do the same rules apply to personal injury and to damage to property (because one can attach a reduced value to property after the first event)? If the second event increases the already existent damage by aggravating it or adding further harm, the first tortfeasor will only be liable for the increase if it can be attributed to his conduct under the general rules.
Generally, it is the victim’s task to prove causation. But, at least in theory, the standard of proof may vary from a balance of probabilities (‘more likely than not’; eg England, Scotland) along different shades up to the conviction of the judge without reasonable doubt (eg Germany, Portugal). Nevertheless, it is widely assumed that these different standards will hardly ever lead to different results in terms of the practical outcome of a case. Some legal systems help the victim in certain cases by allowing him to rely on prima facie evidence or by reversing the burden of proof where he is unable to meet the required high standard of proof. The effects of standard of proof and burden of proof are felt mostly in cases where causation is uncertain because of evidentiary gaps. Cases of liability for merely possible causation are known to many legal systems. In this scenario, every rule-maker will have to choose between an ‘all-or-nothing’ approach (no liability or full liability despite merely possible causation) or partial, ie pro rata, liability depending on the level of probability (proportional liability). The paradigmatic example is the case of alternative tortfeasors. A logical albeit unsatisfactory solution would be the one adopted by Swiss law: since none of the acts in question can be proven to have caused the damage, the victim does not receive any damages. Many legal systems (Austria, Germany, Italy, the Netherlands) follow the opposite approach and hold—for instance by means of a rebuttable presumption—all potential tortfeasors jointly and severally liable if the conduct of each of them fulfils all other requirements of liability and would, taken as such, have sufficed to bring about the entire harm; as a consequence, the risk of recourse is borne by the potential tortfeasors. This approach often leads to liability being denied where the harm could also have been caused by chance. A third route would be to hold each potential tortfeasor liable for a share of the entire loss that corresponds to the degree of probability with which he has caused the loss. Cases of loss of a chance can be analysed from a similar point of view. Especially in mass torts or long-term risks (eg pharmaceutical liability or asbestos cases), it becomes apparent that the evolution of sensible rules is not yet complete. ‘External’ approaches such as compensation funds or insurance models may serve as alternatives to avoid evidentiary difficulties.
b) Approaches in the European model rules on liability
The ‘but for’ test can be read at best impliedly into the general attribution rule of the DCFR quoted above. This rule is supplemented by two provisions on multiple tortfeasors (Arts VI.-4:102 f DCFR; Arts 4:102 f PEL Liab Dam). If they have collaborated, all potential tortfeasors are taken to have caused the damage. In the case of alternative tortfeasors, each potential tortfeasor is rebuttably presumed to have caused the damage; if he cannot rebut the presumption, he and the other potential tortfeasors are held jointly and severally liable.
In Arts 3:101 to 3:106 PETL, by contrast, one finds unusually detailed and partly innovative rules on factual causation. By way of introduction, these rules expressly spell out the ‘but for’ test, supplemented by a provision on over-determined damage to the effect that where each of a number of activities would have caused the damage at one and the same time, each activity is regarded as a cause of the victim’s damage. What follows is a whole range of specific rules for cases where the ‘but for’ test in itself is not sufficient because it cannot establish why a certain conduct was to be treated as a cause or because causation cannot be established with certainty. Regarding hypothetical causation, the PETL distinguish between definite and irreversible damage on the one hand and continuous damage on the other. In the first case, the subsequent event is only taken into account to the extent that it aggravates the damage or leads to additional damage, or if it has influence on the calculation of damages (eg by reducing the victim’s life expectancy). In the case of continuous damage, both events are regarded as causal for the continuous damage occurring from the time of the subsequent event. Where causation is uncertain, the PETL opt for a system of proportional liability depending on the likelihood with which an activity has caused the damage. This system is applied not only to the classical case of alternative tortfeasors but also to alternative victims and thus provides a model to cope with mass torts. Uncertain causes in the victim’s sphere (including hazard) reduce his claim depending on the likelihood with which they have caused the damage.
c) EU private law
In all contexts of liability, the EU courts choose, expressly or impliedly, the conditio sine qua non test as their starting point of any causal enquiry. EU legislation sometimes mentions causation as a prerequisite, but it does not provide a detailed definition or description of the concept. The Product Liability Directive (Dir 85/374), for instance, requires that a causal relationship between the defect of the product and the damage be established and adds that liability is not affected by a causal contribution on the part of a third party but may be reduced in the case of contributory fault on the part of the victim himself. Another important example are the rudimentary provisions in Art 4 of the Environmental Liability Directive (Dir 2004/35).
3. Second level of enquiry: scope of liability
The ‘but for’ test yields a multitude of equivalent causes; it can even be used to identify the victim’s conduct as the source of the damage. What is therefore required at the second level of enquiry is a normative limitation in order to establish whether a particular harmful result is suite directe (France) and zurechenbar (Germany), or whether it is ‘too remote’ (England). In all national legal systems, in the European model rules as well as in EU private law, it is on this level of enquiry where the nexus between an event and a harmful result is actually established. Even in Belgium where it is often claimed that the conditio sine qua non test is the only element of causal enquiry, in reality the courts cannot do without this second step.
a) Criteria to be found in national legal systems
Despite the broad consensus that this second step of enquiry is necessary, no predominant European approach can be discerned from the impressive diversity and variety of criteria offered. Yet differences in the results reached are much less marked than the differences in the conceptual framework. The following criteria, none of which can be regarded as always decisive or taking precedence, stand out and at the same time reflect a tendency away from generalizing attribution theories. (i) Probability. Generally, it is assumed that consequences will not be attributed to an event if it would, in the natural course of events, seem irrelevant and would only under the most peculiar and entirely improbable circumstances lead to the damage complained of. In German law, this is captured by the ‘adequacy’ test; similar considerations can be found in many other legal systems, eg in Austria, Portugal and Switzerland as well as, with only slight differences, in France. (ii) Scope of the rule (relativité aquilienne; Schutzzweck der Norm). The damage complained of must be covered by the protective purpose of the legal norm that has been violated or the duty that has been breached (eg England, Germany, the Netherlands). (iii) Foreseeability. This criterion dominates English case law, but it can also be found in other legal systems. It also has resonances in the adequacy test. The assessment whether a consequence has been reasonably foreseeable leaves considerable room for normative evaluations, eg regarding the perspective or the reference point of the evaluation. (iv) Degree of fault. Consequences that have been caused intentionally are always attributable. If the tortfeasor’s conduct was grossly negligent, there is a widespread tendency to accept attribution or at least to help the victim by alleviating his burden of proof. (v) Alternative lawful conduct (rechtmäßiges Alternativverhalten). Generally, attribution will have to be denied if the tortfeasor can show that, even if he had not acted wrongfully or breached a duty of care, the victim would have suffered the same damage. However, there are exceptions to this defence depending on the importance and content of the duty breached (eg the right to self-determination of a patient who has not been properly informed of the risks associated with his medical treatment). (vi) Type of damage and nature of liability (see expressly Art 6:98 Burgerlijk Wetboek (BW)). Broadly speaking, personal injuries are more likely to be treated as attributable than damage to property or pure economic loss. As far as personal injuries are concerned, even a victim’s special predisposition to injury (and thus an improbable course of events) cannot discharge the tortfeasor (so-called ‘eggshell skull rule’: the tortfeasor has to take the victim as he finds him). Differences may also exist between liability for fault and strict liability; ultimately, some statutes establishing strict liability have been devised for the very purpose of catering for improbable harm (eg in the field of nuclear energy). (vii) ‘Breaking the chain of causation’. This rather unfortunate label is attached to cases where a third party, the victim himself, or a hazard interferes with the sequence of events that the tortfeasor has ignited. Solutions to individual cases may vary between national legal systems. But, what is more, the whole attempt to subsume these cases under a category of their own has proven to be unhelpful and misleading. Rather, the effects of the supervening event on the attribution of the damage to the tortfeasor should be assessed according to the general rules. Conduct of a third party, for example, does not constitute an ‘interruption’ of the causal nexus if this conduct is nothing more than the realization of a risk created by the tortfeasor. If a third party’s intervention was intentional or unforeseeable, attribution may have to be denied—unless, of course, prevention of this conduct was precisely the duty that the tortfeasor had breached. Regarding interference by the victim, there is a significant overlap with the rules on contributory negligence (contributory conduct or activity). Finally, force majeure or acts of God are in many places regarded as a reason to deny attribution (eg France, Switzerland). (viii) Occasionally, courts confine themselves to policy considerations to answer questions of attribution, for instance to encourage a generally desirable behaviour such as assistance to accident victims.
b) Approaches in the European model rules on liability
Both sets of model rules leave substantial leeway for normative considerations. Apart from enshrining the eggshell skull rule in Art VI.-4:101(2), the provisions of the DCFR do not provide any specific guidance or indicia as to when a harmful result is attributable. Instead, all the necessary evaluation has to take place under the notion of ‘consequence’. Again, the PETL adopt a more comprehensive regulatory technique; compared to their detailed rules on factual causation, however, their approach, which is based on the idea of a flexible system, appears rather vague and difficult to apply. In a provision entitled ‘scope of liability’ (Art 3:201), the PETL give a non-exhaustive list of factors to be taken into account. These factors are also familiar to national legal systems, ie foreseeability, the nature and the value of the protected interests, the basis of liability, the extent of the ordinary risks of life, and the protective purpose of the rule that has been violated.
c) EU private law
The EU courts have not adopted the approach of one of the Member States, but instead apply a mix of criteria on a case-by-case basis to establish a ‘direct causal link’. This approach is partly criticized as unclear. Among these criteria, the main focus is on the notions of foreseeability and directness (or, put negatively, remoteness). Directness, in turn, appears as a cloak for normative considerations; the evaluation may also take into account the protective purpose of the rule that has been violated. Judgments in which the direct causal link has been critical mostly concern cases of pure economic loss; here, attribution is accepted rather restrictively. Difficult problems of attribution that have only partly been solved may ensue if the conduct of both the EU and a Member State has contributed to the damage (eg cooperation of EU institutions and national authorities in applying EU law).
4. Uniform law
Conventions on issues of liability also merely presuppose that causation is a prerequisite of liability, but do not define what is required for a causal link to be present. This can be seen, for instance, in the Paris Convention on Third Party Liability in the Field of Nuclear Energy, in the International Convention on Civil Liability for Oil Pollution Damage (marine pollution (compensation)) and in the Montreal Convention (air transportation (contractual liability)).
Herbert LA Hart and Tony Honoré, Causation in the Law (2nd edn, 1985); Reinhard Zimmermann, The Law of Obligations (1996), 988 ff; Christian von Bar, The Common European Law of Torts, vol II (2000) nn 411 ff; Jaap Spier (ed), Unification of Tort Law: Causation (2000); Jane Stapleton, ‘Legal Cause: Cause-in-fact and the Scope of Liability for Consequences’ (2001) 54 Vanderbilt Law Review 941; Wolfgang Wurmnest, Grundzüge eines europäischen Haftungsrechts: Eine rechtsvergleichende Untersuchung des Gemeinschaftsrechts (2003) 159 ff; Lars Klöhn, ‘“Wertende Kausalität” im Spiegel von Rechtsvergleichung, Rechtsdogmatik und Rechtsökonomik’ (2006) 105 Zeitschrift für vergleichende Rechtswissenschaft 455; Martin Schmidt-Kessel, Reform des Schadenersatzrechts, vol I (2006) 72-3, 76, 90, 144 ff; Cees van Dam, European Tort Law (2006) nos 1101 ff; Luboš Tichý (ed), Causation in Law (2007); Bénédict Winiger, Helmut Koziol, Bernhard A Koch and Reinhard Zimmermann (eds), Digest of European Tort Law, vol I: Essential Cases on Natural Causation (2007); Helmut Koziol and Reiner Schulze (eds), Tort Law of the European Community (2008) 47 ff, 335 ff, 354-5, 378 ff, 413 ff, 465 ff, 498 ff, 533 ff, 543 ff, 569 ff.