Almost every system of tort law subscribes to a version of the following rule: “Proof of Causation Rule (PCR): D cannot be held liable for losses arising from C's injury unless C can prove (to the relevant standard) that D's conduct was a cause of C's injury.” But consider the following case considered in Dr. Steel's book: “Two hunters. D1, D2 and C are on a hunting trip. D1 and D2 negligently fire in the direction of C. One bullet strikes C but it cannot be determined, due to the simultaneity of the shots, and the similarity of the bullets used, which of D1 and D2 caused C's injury” (p. 139).

Since C cannot prove either that D1’s conduct was a cause of his injury or that D2’s conduct was a cause of his injury, it seems to follow from the proof of causation rule that C’s claims against both defendants should fail. This feels like the wrong result. After all, C is provably entitled to compensation from someone; it seems odd to deprive him of it just because two people acted wrongfully towards him rather than one.

Faced with such cases, many legal systems have introduced limited exceptions to the proof of causation rule. Steel's stated aim in Proof of Causation in Tort Law is to compare the development of these exceptions in English, French and German law and examine to what extent they can be justified. His conclusion is that, although the proof of causation rule should continue to apply in general, exceptions to it can be justified in cases, like the example of the two hunters, wherein the only barrier to a finding of liability is uncertainty either over which defendant's conduct caused the claimant's harm or over which claimant(s) the defendant's conduct injured. But, in many ways, Steel's book delivers more than advertised: along the way, we also find clear and engaging discussions of a variety of comparative and normative topics, from the epistemology of statistical evidence (pp. 85–101) to the law on “material contribution” (pp. 239–47), which help to guide the reader through some notoriously thorny conceptual terrain. I would enthusiastically recommend this book to anyone with a practical or scholarly interest in tort law, causation or legal epistemology.

In chapters 1–3, Steel argues that the common law version of the proof of causation rule – according to which D should be held liable for C’s losses only if it is more probable than not, conditional on the right kind of available evidence, that D’s conduct was a cause of C’s injury – is generally justified. He starts by arguing that causation, as a matter of substantive law, matters to liability, basically because one is under no moral duty to compensate someone for losses for which one is not responsible, and one is not responsible for something one did not cause (pp. 105–11). Steel then provides a “limited defence” of the common law “balance of probability rule” over the more demanding standards of proof in French and German law (according to which the claimant must convince the judge of the truth of a proposition, so that the judge achieves “a level of certainty which silences doubts for practical purposes even if not eliminating them fully”), on the grounds that the balance of probability rule strikes a better balance between the defendant's and claimant's interests (pp. 127–29). The balance of probability rule is silent on what to do when there is perfect equality between the probability of the defendant's and the claimant's allegations; in such cases, Steel argues that the defendant's allegations should be accepted, because interfering with an agent's individual liberty by forcing her to pay for a loss for which she is not responsible (a “false positive”) is more unjust, ceritus paribus, than depriving an agent of compensation to which he is entitled (a “false negative”).

In the example of the two hunters, it cannot be established, of either defendant, that she is responsible for C’s losses on the balance of probability, because each defendant is no more than 50% likely to have caused C’s injury on the evidence available. Yet, despite this, Steel argues that a finding of liability against both defendants can be justified. He proposes a number of arguments for this conclusion. The first is what he calls “prevented claim theory”. The idea is that, although C cannot prove that D1’s conduct was a cause of her injury, he can prove that D1 either caused the injury or caused him to be unable to recover from D2 in respect of the injury. Either way, D1 provably caused a loss equal in value to C’s injury. The same goes for D2. So D1 and D2 can both be held fully liable, on ordinary principles, for C’s losses. So conceived, the desired outcome in the two hunters’ example is no exception to the proof of causation rule after all.

There is something fishy about this argument, on the face of it. According to Steel's prevented claim theory, D1 is liable by virtue of having either caused C’s injury or precluded a finding of liability in respect of C’s injury against D2. But suppose D1 caused the injury; then D2 cannot be held liable in virtue of having precluded a finding of liability against D1, since D1, by assumption, is liable. Or suppose D1 precluded a finding of liability against D2; then D2 cannot be held liable in virtue of having caused C’s injury, since D1, by assumption, precluded such a finding. Either way, if D1 is liable, D2 is not. By parallel reasoning, if D2 is liable, D1 is not. So the prevented claim theory cannot justify a finding of liability against both defendants: the success of a claim against one undermines the basis of any claim against the other.

Steel thinks that he can avoid this problem by specifying more precisely the thing of which the non-causative defendant deprived C (pp. 182–83). Suppose D1 caused the injury. Since the prevented claim theory holds both defendants liable for C’s losses, D2, as a matter of fact, did not prevent C from claiming from D1. But she did prevent C from claiming from D1 on certain grounds, namely the grounds that D1 caused the injury. That is why we hold D2 liable, according to Steel: because her actions prevented the specific wrong D1 committed against C (namely wrongfully injuring her) from being corrected. But it is not so clear, at least to me, why merely restricting the grounds on which liability against another defendant can be imposed – as opposed to precluding a finding of liability against that defendant altogether – is something by virtue of which a duty to compensate can arise. In other words, it is not clear to me that C’s being required to sue D1 on one set of grounds rather than another can really be described as a loss at all, much less one equal in value to the injury itself. To what extent is C worse off as a result of D2’s actions, if she would have recovered her losses from D1 either way?

Steel's second argument for holding both defendants liable in the two hunters’ example is what he calls the “relative injustice argument” (pp. 185–89). According to this argument, dispensing with the proof of causation rule in cases like that of the two hunters, by either reversing the burden of proof or relaxing the standard of proof on the issue of whether the defendants causally contributed to the injury, is justified on the grounds that it decreases the total amount of expected injustice. Steel's explanation for this is subtle. Notice first that, regardless of the outcome in the two hunters’ example, it is certain that someone will suffer an injustice: either a “false positive” if both defendants are held liable or a “false negative” if neither is. As we saw above, Steel thinks that false negatives are preferable to false positives, ceritus paribus. It would seem to follow, therefore, that dispensing with the proof of causation rule in the two hunters’ example would lead to more injustice, not less. But there is more to be said, for ceritus is not quite paribus. If C’s claims against both defendants were to succeed, D1 and D2 would each have a 50% chance of being held liable for a loss they did not cause, whereas, if the claims were to fail, C would have a 100% chance of being deprived of compensation to which he is entitled. In other words, the “risk” of a false positive is “spread out” over two people, D1 and D2, whereas the “risk” of a false negative is “concentrated” in just one person, C. Steel seems to think that the former outcome is preferable because there are pro tanto reasons to distribute, not just costs, but also mere risks of costs, in the fairest possible way: in other words, he endorses a version of the so-called “identified person bias” (see e.g. I. Glenn Cohen, N. Daniels and N. Eyal (eds.), Identified Versus Statistical Lives: An Interdisciplinary Perspective (Oxford 2015)). This is a controversial view in ethics, however, and one would expect a more detailed defence of it given the role it plays in Steel's argument. It also seems in tension with what Steel says in chapter 5 about exceptional rules that allow claimants to recover for the mere loss of a chance of avoiding injury. Although he agrees that such losses are actionable in theory, Steel insists that they are “relatively minor” (p. 366) compared with the injury itself (so that, for example, the damages awardable in respect of a loss of a 40% chance of avoiding an injury should be much less than 40% of the damages awardable in respect of the injury itself). If mere risks of harm have little value in and of themselves, it is not clear why one should place much weight on legal outcomes which distribute them more fairly.

It is important to see these arguments in context, however. One thing Steel's book brings out particularly well is just how hard it is to justify limited exceptions to the proof of causation rule. Many common arguments for dispensing with the proof of causation rule in certain cases – for example, that it would increase deterrence (pp. 269–75) or reflect the grossness of the defendant's misconduct (pp. 200–06) – straightforwardly apply in a much wider range of cases than those in which exceptions to the proof of causation rule are typically allowed, with the result that many legal systems are simply inconsistent (English law in particular is so “exemplary in its inconsistency” (p. 370) that Steel at times struggles to conceal his frustration with it). The primary virtue of the arguments Steel considers (including the prevented claim theory and the relative injustice argument) is that they allow us to “contain” exceptions to the proof of causation rule within a clearly delineated class of cases, without undermining the normative foundations of the rule more generally. But one might naturally wonder whether dispensing with the proof of causation rule – as opposed, say, to introducing a no-fault state-sponsored compensation scheme – is really the best way of satisfying the desire to ensure that C is compensated in cases such as that of the two hunters, especially given how persuasively Steel defends the proof of causation rule as a general rule. As Steel himself points out (p. 371), legal systems have historically been most disposed to dispense with the proof of causation rule in situations where the compensation available through alternative sources is relatively minimal. It is not clear how much motivation to change the proof of causation rule would remain if victims had reliable access to compensation outside the tort system.

These minor points notwithstanding, Proof of Causation in Tort Law is a timely and important contribution to a foundational debate – that concerning the nature, justification and proper scope of the proof of causation rule in tort law. At the very least, Steel's comprehensive scholarship and razor-sharp analysis promises to bring some much-needed clarity to an area of the law in dire need of reform. I hope and expect it will be widely read.