You are on page 1of 25

Risk, Harm, and Responsibility

University Press Scholarship Online

Oxford Scholarship Online

The Philosophical Foundations of Tort Law


David G. Owen

Print publication date: 1997


Print ISBN-13: 9780198265795
Published to Oxford Scholarship Online: March 2012
DOI: 10.1093/acprof:oso/9780198265795.001.0001

Risk, Harm, and Responsibility


Stephen R. Perry

DOI:10.1093/acprof:oso/9780198265795.003.0015

Abstract and Keywords


Since risk is one of the central concepts in modern tort law, any theory of torts rooted in
notions of interpersonal justice and individual responsibility must get to grips with the
questions presented in this chapter. With a view to contributing to a responsibility-based
theory of tort law, this theory advances and defends those questions. It offers a general
analysis of risk that draws on work in the philosophy of probability and suggests that
there are two main conceptions of risk. The chapter also considers the thesis that
subjecting another person to risk is a form of harm distinct from any physical harm the
person might suffer. This thesis is most plausible if it is understood as presupposing the
objective conception of risk. Even understood in that way, however, the thesis cannot be
accepted; risk does not, it argues, constitute harm in itself. The chapter then considers
moral responsibility for physical damage.
Keywords: risk, tort law, justice, responsibility, philosophy, probability, harm

Page 1 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


What is risk? Does subjecting another person to riskor reducing his or her chance of
avoiding an adverse physical outcomeconstitute a distinct form of harm? What is the
basis of moral responsibility for physical harm that was unintentionally caused by risky
conduct? Because risk is one of the central concepts in modern tort law, any theory of
torts rooted in notions of interpersonal justice and individual responsibility must come to
grips with the first and third of these questions. In the light of a recent academic and
judicial trend to characterize risk as harm, a response to the second question should be
of interest as well. In this essay, with a view to contributing to a responsibility-based
theory of tort law, I advance and defend answers to all three questions.
In section I, I offer a general analysis of risk that draws on work in the philosophy of
probability. I suggest that there are two main conceptions of risk, one objective in
character and the other epistemic, and that both are, in different ways, plausible
candidates for explaining the moral significance that we attribute to risk. In section II, I
consider the thesis that subjecting another person to risk is a form of harm distinct from
any physical harm the person might suffer. This thesis is most plausible if it is understood
as presupposing the objective conception of risk. Even understood in that way, however,
the thesis cannot be accepted; risk does not, I argue, constitute harm in itself.
In section III, I consider moral responsibility for physical damage: under what
circumstances should an agent whose risky conduct caused unintentional physical harm
to another person be held morally responsible for that harm? One common argument,
which I call the libertarian argument, suggests that persons should be held morally
responsible for the materialization of any objective risk that their conduct creates. The
libertarian argument would support a standard of absolute liability in tort law, thus
requiring that liability be imposed simply for causing harm. Both the argument and the
standard of absolute liability must, however, be rejected. Moral responsibility for
physical harm should be grounded in a notion of epistemic rather than objective risk. This
understanding of responsibility, which, borrowing a term from Tony Honor,1 call
outcome-responsibility, (p.322) is consistent both with the negligence standard and with
risk-based standards of strict liability such as the rule for ultrahazardous activities.

I. The Nature of Risk


First, what is a risk? In ordinary language conduct is typically said to be risky when it
gives rise to a chance of a bad outcome of some kind. The concept thus involves two main
elements: first, a notion of chance or probability, and second, a notion of harm. The pretheoretical concept of risk presumably does not define the relationship between these
two elements in a precise way, but a natural move is to stipulate that a risk is the
mathematical expectation of harm, i.e., the product of the probability of occur-rence of the
harm and the magnitude of the harm were it to occur (P H). For present purposes the
notion of harm can be treated as unproblematic.1 Let me loosely define harm as a setback
to an interest, and an interest as an aspect of personal well-being. We may assume that
independent moral arguments will determine which interests are of sufficient importance
to warrant protection by the law of torts; the concept of risk can then be applied to
potential interferences with the interests those arguments single out. The troublesome

Page 2 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


aspect of the concept of risk, and the one that plays the more intrinsic role in shaping its
meaning, is not harm but the element of probability. The idea of risk can apply to many
different types of harm or bad outcome, but what we mean by risk is fundamentally
determined by the understanding of probability we take the concept to presuppose.
Philosophers have developed several quite different theoretical accounts of probability.2
These theories are sometimes treated as competing and mutually incompatible, but the
predominant contemporary view tends to regard probability in a pluralist light. More
particularly, it has become commonplace to recognize that two distinct but non-conflicting
conceptions of probability must both be recognized: one is concerned with prob-abilities
that in some way have objective existence, while the other characterizes probabilities in
terms of the current state of our knowledge and beliefs.3 I shall call these the objective
and epistemic conceptions, respectively. Ian Hacking has written that the concept of
probability has possessed this essential duality ever since it began to emerge in its
modern (p.323) form in the seventeenth century.4 Both the objective and the epistemic
conceptions of probability can be given a number of different interpretations. To explain
what these are, and to show why particular interpretations are particularly apposite for
understanding the moral significance of risk, it will be necessary to take a short detour
through the byways of the philosophy of probability.
To begin with the objective conception, people who speak of probability in an objective
sense usually have in mind one or the other of two ideas. The first is the understanding of
probability that emerges from the so-called relative-frequency theory of probability.5 A
probability is, on this view, just the stable relative frequency that may be exhibited by the
occurrences of a given attribute within a series of events or objects in the physical world,
where those events or objects can in an appropriate way be characterized as similar
(e.g., the relative-frequency of .5 associated with the attribute of coming up heads in a
series of similar events, namely, the flips of a fair coin). The formal definition of a
probability is usually given as the limit of a frequency within an infinite sequence. Despite
this reliance on the notion of infinity, the relative-frequency theory regards probabilities
as just a certain kind of empirical fact. The second idea people tend to have in mind when
they speak of objective probability is concerned with the operation of indeterministic
causal processes, such as those associated with quantum mechanics. If a particular
outcome, such as the final resting place of a billiard ball, is completely determined by the
preceding state of the physical universe, then the relevant causal process is
deterministic. If a particular outcome, say the time at which a given uranium atom decays,
is not so determined, then the causal process is indeterministic. There is no further,
perhaps unknown, causal mechanism that lies behind a statement of the probability of
decay. On this view a probability is again a certain kind of empirical fact, but one that has
to do with the nature of causal processes.
These two senses of objective probability are related. To see the connection, first recall
that a probability in the relative-frequency sense is defined with respect to a certain
reference class of actual or possible events. The probability that a given coin will come up
heads, for example, is defined relative to the reference class of flips of that coin. The

Page 3 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


causal processes underlying the events in the reference class could, however, be either
deterministic or indeterministic. Assume that the relative frequency with which a given
coin comes up heads is .5. If the causal processes involved are deterministic, then it will in
principle be possible to predict whether a given coin-flip will come up heads or tails, and
hence to partition the reference class in advance into two sub-classes, one consisting of all
head-events and the (p.324) other of all tail-events. So far as the first sub-class is
concerned, the probability of the coin coming up heads is one, whereas for the second
sub-class the probability is zero. If, on the other hand, the causal processes are indeterministic, then it will not be possible to partition the reference class in this way. In an
important sense the probability statement expresses an ultimate fact about the universe.
But whichever of these possibilities holds, we can still meaningfully speak of the relative
frequency, with respect to the original reference class of all flips of the coin, of its coming
up heads. Independently of whether the relevant causal processes are deterministic or
indeterministic, the relative frequency of .5 with which the coin comes up heads is an
objective property of the physical world. Objective probability in the indeterministic
sense is thus a special case of objective probability in the relative-frequency sense.
Of course, the other side of this coin (so to speak) is the interesting observation that
objective probability can exist in a deterministic universe; probability need not be viewed
as merely an epistemic phenomenon that emerges only because we can never have
complete knowledge of deterministic processes. Even if the preceding state of the
universe determines the outcome of any given coin flip, the statement that there is a
probability of .5 that the coin will come up heads expresses a fact about the physical
world. As a practical matter, of course, we cannot predict whether a coin will come up
heads or tails. The result of any given coin flip can thus be characterized as random, and
this is so whether the underlying causal processes are deterministic or indeterministic.
The distinction between determinism and indeterminism turns on predictability in
principle, not predictability in practice. In practice, not only might we be unable to predict
particular outcomes, we might not know whether it was in principle even possible to
predict outcomes, i.e., we might not know whether the causal processes were or were
not deterministic. And even if we knew that outcomes were predictable in principle, this
would not necessarily be particularly useful or significant information. (It would not, for
example, have any effect on the various purposes for which we flip coins.) Moreover the
distinction between predictability in principle and unpredictability is by no means crystal
clear.6 Taken together, these points suggest that we should treat relative frequency
rather than indeterminacy as the most basic understanding of objective probability. From
now on I shall accordingly employ the term objective probability to mean the relativefrequency conception of probability. Reference will be made to the special case of
indeterminacy as and when necessary.
(p.325) Epistemic probability, like objective probability, can also be given different
interpretations. For purposes of this essay, however, which is primarily concerned with
risk regarded as a moral phenomenon, I shall assume that epistemic probability is best
understood in terms of what I shall call the reasonableness account. The reasonableness
account is based on the following two assumptions: first, that objective probabilities in the

Page 4 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


relative-frequency sense exist; and secondly, that human beings possess intersubjectively valid standards of inductive reasoning and rational belief that are at least
pragmatically justified and that permit them to discover or estimate those relative
frequencies. Epistemic probability involves, according to the reasonableness account, a
relation between evidentiary premises, on the one hand, and a conclusion about the way
the world is, on the other. The evidence will consist, often, of partial observations of
relative frequency within a sequence of (possible) objects or events that cannot be
observed in its entirety, but sometimes it will take a more indirect form. The conclusion
will be an estimate of relative frequency within the total sequence.7 Epistemic probability
judgments are thus estimates of objective probabilities that are based on, and relative to,
a given body of evidence.8
According to the reasonableness account, epistemic probability judgements are capable
of expressing knowledge about the world, although that knowledge will oftenperhaps
alwaysbe imperfect or incomplete. The standards of reasoning assumed to be
appropriate for justifying these judgements are based on the modes of probabilistic and
inductive reasoning, both formal and informal, that human beings actually employ. There
are a number of ways in which these standards might be characterized: one might, for
example, formulate precise inductive principles that operate in conjunction with formal
statistical techniques and the probability calculus; or one might seek to discover an
informed consensus about the probability of a given type of event; or one might look to
the probability judgements of a reasonable or representative person. Different
characterizations might be suited to different purposes. Thus a formal characterization,
employing sophisticated statistical techniques, might be particularly appropriate for
scientific inquiry, whereas an informal characterization, based on the intuitive probability
judgements of a reasonable person, might be more suitable for determining moral
responsibility. A characterization along the latter (p.326) lines is, not coincidentally,
reminiscent of the understanding of risk to be found in tort law.
I do not attempt in this chapter to establish the truth of the two assumptions on which the
reasonableness account reststhe existence of objective probabilities, and the availability
of inter-subjectively valid methods for estimating thembut they must be taken to be
true if a conception of risk based on epistemic probability is to have moral significance.
The first assumption ensures that epistemic judgements of risk are concerned with real
possibilities of harm, while the second ensures that those possibilities are capable of being
estimated by modes of reasoning that both the person who creates a risk and the person
who is subjected to it can accept. If the first assumption were not true we could still
formulate a non-objective conception of probability, but as the discussion below of radical
subjectivism illustrates, it seems unlikely that we would be left with a morally significant
concept of risk. If, on the other hand, the second assumption were not true, there would
be no forms of common reasoning about risk, and little or no agreement about what risks
existed and how serious they were. This would make it difficult, if not impossible, to
formulate common standards of conduct relating to risky behavior (whatever that term
might mean in such a world).

Page 5 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


There are other possible interpretations of epistemic probability besides the
reasonableness account. In the remainder of this section I would like briefly to describe
the two most influential of these interpretations, and to indicate why they are not capable
of explaining or illuminating the moral significance that risk holds for human beings.
Readers who do not wish to explore probability theory any further may safely take a
faster but less scenic route and skip to section II.
The first of the two alternative interpretations of epistemic probability is suggested by
radical versions of the subjectivist (or personalist, or Bayesian) theory of probability that
has its roots in the work of F. P. Ramsey.9 On the general subjectivist view, a judgement
of probability measures a particular individuals actual degree of belief or confidence in a
given proposition. The measure of degree of belief is usually understood behavioristically, by reference to the odds that the individual would be willing to accept within a
coherent series of bets. Coherence is understood as conformity to the standard
probability calculus. Radical versions of subjectivism, which give rise to the particular
interpretation of epistemic probability with which we are now concerned, hold that
coherence is the only constraint on rational belief. Radical subjectivists maintain that we
can speak meaningfully (p.327) only of individual assessments of probability; there is,
they assert, no such thing as probability in an objective sense.10 This view thus places
very minimal restrictions on what will count as rational belief: probability judgements can
permissibly vary quite widely from person to person.
The second alternative interpretation of epistemic probability, which in some respects
resembles the reasonableness account, is grounded in the so-called logical-relation (or a
priori) theory of probability.11 The essential idea is that probability is a logical relation
between a conclusion or hypothesis about the way things are, on the one hand, and
premises that set out evidence relevant to that hypothesis, on the other. Probability is
taken to be a measure of the degree of rational belief in (or confirmation of) the
hypothesis. The relation between premises and conclusion is understood as logical rather
than psychological in character, and to that extent is taken to be an objective matter.12 At
the same time, however, probability statements are always relative to given evidence
only. By treating probability statements as logical truths, this view places quite stringent
restrictions on what will count as rational belief.
Neither the radical subjectivist nor the logical-relation understanding of epistemic
probability seems able to explain why we attribute the moral significance to risk that we
do. The radical subjectivist account assumes there is no objective conception of
probability that extends beyond the weak constraints of the coherence requirement. On
this view there is no general inter-subjective basis for distinguishing among risks, and
hence no basis for holding someone morally responsible for risk creation. If I think that
you are doing something that poses a high risk of harm, whereas you think that the risk is
low or non-existent, no determinative argument (apart from considerations of coherence)
could be brought to bear by either of us to show that the other was wrong. Risk thus
understood would lack moral significance, because it would have no systematic and
agreed-upon relation to the physical world in which consequences follow from actions.

Page 6 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


Risk would be an indication of the betting odds a given person would be willing to accept
in an objectively unpredictable world, and nothing more. Perhaps, on a radical
subjectivist understanding of probability, an individuals assessment of the risks his or
her action poses for others might be relevant (p.328) to an inquiry into whether he or
she had a culpable state of mind. But since in a radically subjectivist world the notion of
acting in order to bring about an effect might well have a meaning very different from
the one we now attribute to it, even this is far from clear.
Of course radical subjectivists recognize that, for various reasons, individuals
probability judgements often converge, but they regard this as a reflection of
psychological rather than physical regularities.13 A consistent and rigorously developed
radical subjectivist understanding of probability would in fact seem to have no room for
our ordinary concept of physical causality, since that concept presupposes regularities in
the physical world. Of course nothing in this discussion of radical subjectivism establishes
that among the properties of the world are objective probabilities in the relativefrequency sense. The claim is simply that unless there are such objective probabilities,
and unless our epistemic conception of probability is systematically related to them, the
epistemic conception will lack moral significance.14
Turning now to the logical-relation understanding of epistemic probability, there are at
least two reasons why it is likely to prove problematic from a moral point of view. The first
concerns the fact that, on this understanding, probability judgements are not empirically
verifiable statements: they are regarded, rather, as expressing an a priori, logical relation
among propositions. As in the case of the radical subjectivist account, but for different
reasons, the logical-relation understanding does not regard probability judgements as
attempts to estimate relative frequencies in the physical world. As before, it is difficult to
attribute moral significance to an account of epistemic probability that lacks this feature;
logical truth cannot, by itself, have any moral import. Various attempts have been made
by logical-relation theorists to show how and why a priori probability judgements can tell
us something about the physical world, but we need not decide here whether any of
these have succeeded. For present purposes it suffices to point out that if a systematic
link between a priori judgements and empirically verifiable relative frequencies can be
established, it is that link that will matter in an account of the moral significance of risk and
not the judgements supposed logical basis. In other words it does not matter, from a
moral perspective, whether the sense of epistemic probability judgements must
ultimately be explained by reference to a logical relation among propositions; the
reasonableness account, which resembles the logical-relation (p.329) theory in
regarding probability as involving a relation between evidentiary premises and a
conclusion about the way the world is, can simply remain agnostic on this question. What
ultimately matters, from a moral point of view, is whether probability judgements are
generally able, under specifiable circumstances, to serve as reliable estimates of relative
frequencies in the physical world.
The second reason why the logical-relation understanding of epistemic probability is
unlikely to be able to account for the moral significance of risk is that the logical relations

Page 7 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


the theory posits do not seem to be readily accessible to human understanding. Carnap,
for example, proposed a very complex, language-dependent scheme of inductive logic
that does not seem practically applicable by ordinary human beings except to very simple
(and highly simplified) problems of probability. Difficult practical and philosophical
questions also arise in specifying the formal language in which the scheme of inductive
logic is to be formulated. Even if we were able to assert with some confidence that there
is some logical relation, representing a uniquely correct probability judgement, between
a given body of evidence and a given conclusion about how things might be, this fact
would be of little moral significance if human beings did not generally have the capacity to
reason in the requisite manner.15
I have argued in this section that the epistemic conception of probability, understood in
terms of the reasonableness account, can plausibly be regarded as one possible point of
departure for explaining why we attribute moral significance to risk creation. But I have
not claimed that it is the only such point of departure. Since the reasonableness account
makes essential reference to the objective conception of risk, it is plausible to think that
the latter is morally significant in its own right. Perhaps it is the only morally significant
conception of risk, and lends the epistemic conception whatever plausibility it has in this
regard by means of the conceptual link between them. Thus in discussing whether risk is
a distinct form of harm, and in examining the basis of moral responsibility for physical
harm caused by risky conduct, it is important to ensure that neither objective nor
epistemic probability is overlooked. As we shall see, the thesis that risk is a distinct form
of harm is best understood in terms of objective risk, whereas moral responsibility for
physical harm should be understood as being based on epistemic risk.

(p.330) II. Is Risk a Harm in Itself?


A number of writers and judges have suggested in recent years that subjecting another
person to a risk of physical harm, or, alternatively, reducing his or her chance of avoiding
an adverse physical outcome, is itself a form of harm that should be compensable in tort
law.16 It will be convenient to discuss this issue with reference to a concrete set of facts.
The English case of Hotson v. East Berkshire Area Health Authority,17 in which a claim for
lost chance was put forward, is quite suitable for this purpose.
The facts in Hotson were as follows. The plaintiff suffered a fracture of his left femoral
epiphysis, which the defendant health authority negligently failed to diagnose. Although
the injury was correctly diagnosed and treated five days later, the plaintiff eventually
developed a seriously disabling condition known as avascular necrosis. The immediate
cause of this condition was a reduction in the supply of blood reaching the epiphysis. This
reduction could have been caused either by the rupture at the time of the fracture of a
sufficiently high proportion of the blood vessels supplying the epiphysis, or by
subsequent occlusion by pressure within the joint during the five-day period before
proper treatment was administered. The trial judge found as a fact that at the time the
plaintiff first arrived at the defendants hospital, there was a .75 risk that the avascular
necrosis could not be avoided by proper treatment.18 He also found that by the end of
the five-day delay the avascular necrosis was inevitable, i.e., the probability of its

Page 8 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


occurrence was one. On these figures it is clear that, if the standard of proof is the
balance of probabilities, the defendants negligent diagnosis should not be treated as a
cause of the plaintiffs avascular necrosis. But the plaintiff argued that the defendant had
caused him to lose a .25 chance of avoiding avascular necrosis, and that this lost chance
should itself be treated as damage compensable in tort. This argument was accepted by
the trial judge (p.331) and by the English Court of Appeal, but it was ultimately rejected
by the House of Lords.
The .75 figure represents the background risk, at the time the plaintiff first arrived at the
hospital, of his contracting avascular necrosis. The .25 figure represents the augmentation
of the risk brought about by the defendants negligent conduct. Although one can speak,
as courts and legal commentators tend to do, of the reduced or lost chance of avoiding
the adverse physical outcome, I shall instead refer to this augmented probability as the
agent-imposed risk. If the adverse physical outcome materializes, it is possible to use the
antecedent risk figures to compute the probability that the outcome was caused by the
agents action. In Hotson the figures for antecedent risk and ex post probability of
causation coincide, since the avascular necrosis had become inevitable by the time
proper treatment was finally administered. But this is a special circumstance that will not
always be present.19 Our concern here will always be with the antecedent agent-imposed
risk, rather than with the ex post probability that the agent caused the physical outcome.
While physical harm was, on the special facts of Hotson, inevitable, this will not always be
the case. It seems clear, moreover, that the status of agent-imposed risk as harm in itself
should not depend on whether the threatened physical harm has materialized (so long, at
least, as the risk has not passed). On an appropriate set of facts, a Hotson-type claim could
thus be made even in the absence of physical harm.
The plaintiffs argument in Hotson, then, was that the agent-imposed risk was itself
damage of a kind distinct from the physical damage (in Hotson, the avascular necrosis).
While remaining neutral for the time being on whether this argument is sound, let me
refer to the novel form of harm that the plaintiff was alleged to have suffered in Hotson as
risk damage. To assess the claim that risk damage is a true form of damage, we begin by
asking whether it is epistemic or objective risk that is in issue here. To this end it will be
helpful to consider the characterization of the lost chance of avoiding avascular necrosis
in our terms, the agent-imposed riskthat was accepted in Hotson by the House of
Lords:
[Counsel] who appeared for the plaintiffsaid that in relation to the facts of this case
as found by the judge what was meant by a chance was that if 100 people had
suffered the same injury as the plaintiff 75 of them would have developed
avascular necrosis of the whole femoral head and 25 would not. This, he said, was
an asset possessed by the plaintiff when he arrived at the authoritys hospital.It
was this (p.332) asset which [counsel] submits the plaintiff lost in consequence of
the negligent failure of the authority to diagnose his injury properly.20
This characterization of agent-imposed risk is clearly reminiscent of the objective
conception of probability, which is concerned with the relative frequency of an attribute
Page 9 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


within a reference class. Here the relevant attribute is the occurrence of avascular
necrosis, and the reference class is the set of (possible) persons who have suffered the
same injury as the plaintiff. But since the risk figures are simply the trial judges
estimate of the true relative frequencies, they are best regarded as instances of epistemic, not objective, risk. If the estimates turned out to be accurate, epis-temic and
objective risk would coincide; the relative frequency of avascular necrosis in the
reference class would be .25 (objective probability), and that would be a fact of which we
had knowledge (epistemic probability). Of course the figure for agent-imposed risk that
comes to be accepted in a case like Hotson might not coincide with the true relative
frequency of harm. This might happen either because a mistake in inductive reasoning
occurred or because there was insufficient evidence to ensure a reliable probability
judgement. In the case of a mistake in inductive reasoning the accepted figure for agentimposed risk would not, strictly speaking, be an epistemic probability, since our definition
of epistemic probability assumes compliance with the modes of reasoning sanctioned by
the reasonableness account. For present purposes, however, there is no need to
distinguish between these two ways in which an accepted risk figure might differ from the
objective risk.
The trial judge in Hotson concluded that, on the facts of that case, the agent-imposed risk
was .25. As it happens, there is some reason to think that his estimate might have been
subject to at least one of the two forms of error just discussed.21 Suppose for a moment
that the mistake was a particularly egregious one, and that a delay of five days has no
significant effect on the objective frequency with which injuries of the kind suffered by
the plaintiff are treatable. The objective agent-imposed risk would then be zero. If this
fact subsequently came to light, the claim that the defendant had caused risk damage to
the plaintiff would no longer be plausible. This strongly suggests that risk damage must
be understood in terms of objective, not epistemic, risk. That is hardly a surprising
conclusion, since harm involves interference with some aspect of well-being, and such
interference can ordinarily be expected to have objective existence in the physical world.
In this respect physical damage is a paradigm of harm. But the point that risk damage
must be understood in terms of objective risk is nonetheless worth emphasizing, for the
following reason. Objective risk cannot be directly observed. We must always rely on our
best estimate of the objective (p.333) risk, using whatever evidence is to hand. We
necessarily operate, in other words, with the notion of epistemic risk. While it is true that
only objective risk can plausibly be thought to constitute a form of damage, the fact that
we are limited to observing objective risk through the lens of epistemic risk gives rise to
the ever-present danger of confusing the two.
This brings us to the central question we must answer: is agent-imposed risk,
understood in objective and not epistemic terms, properly regarded as a form of damage
in its own right? For purposes of illustration let us assume that, contrary to what in fact
appears to have been the case, the trial judges figures in Hotson of .75 for the
background risk and .25 for the agent-imposed risk were accurate estimates of the
objective risks. Can the agent-imposed risk of .25 plausibly be characterized as a distinct
form of damage, distinguishable from the physical damage that ultimately occurred in the

Page 10 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


form of avascular necrosis? Given the assumption that the relevant figures represent
objective and not just epistemic risks, it follows that seventy-five of 100 persons who had
suffered the same injury as the plaintiff (to use the language of Lord Mackay in the
House of Lords) would have developed avascular necrosis even if proper treatment had
been administered right away, whereas twenty-five persons would have responded to
treatment. These relative frequencies must be regarded as objective facts about the
physical world which would hold even if we had no knowledge of them.
Our main concern, in determining whether risk damage is a true form of damage, is with
the definition of the reference class in terms of which the pertinent relative frequencies
are expressed. The reference class employed in Hotson was a hypothetical group of 100
representative persons who were assumed to have suffered the same injury as the
plaintiff. This obviously cannot mean that the injuries in the sample group are to be
regarded as precisely the same in every respect. What must be intended, rather, is that
the 100 injuries all fall under some general description of physiological states that has not
been spelled out in detail but that could, if necessary, be given a more explicit formulation
by reference to the plaintiffs condition at the time he was injured. Since the information
that could be obtained about his condition will, inevitably, be in at least some respects
incomplete, this description of physiological states will necessarily cover a range of cases
that in various respects differ from one another. For example, since it was not known in
Hotson what proportion of the blood vessels supplying the plaintiffs epiphysis had been
ruptured at the time of the original accident, the relevant physiological description will
subsume a variety of cases in which the extent of damage to the blood vessels ranges
from relatively minimal to relatively severe. These variations in the extent of injury
sustained will be reflected in the sample of 100 representative persons assumed to have
suffered the same injury as the plaintiff.
(p.334) Suppose now that it was possible systematically to distinguish among different
degrees of blood-vessel damage in cases of injury similar to the one sustained by the
plaintiff in Hotson. Assuming for the moment that there is some known and more or less
constant threshold of degree of damage that distinguishes treatable from untreatable
injuries, it would then be possible to partition the reference class of 100 persons into the
twenty-five who were treatable and the seventy-five who were not. We would also know
into which sub-class the plaintiff fell. If he fell into the treatable sub-class, then it would be
possible to say that the defendant, in negligently misdiagnosing his injury and sending him
away without proper medical care, had causally contributed to the development of
avascular necrosis. If, on the other hand, the plaintiff fell into the sub-class of persons
whose injury was untreatable, it would be possible to say that the defendant had not
causally contributed to the onset of avascular necrosis. The claim that the plaintiff
suffered risk damage at the hands of the defendant must, if it is to be generally
defensible, be capable of surviving such a partitioning of the reference class. The partition
does not, after all, call into question the original risk figures, which have been assumed to
refer to objective probabilities. The only difference between the state of affairs that
obtained in Hotson and the circumstance we are now envisaging is that we are supposing
we have sufficient knowledge to distinguish the treatable from the untreatable cases. But

Page 11 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


the claim that the plaintiff suffered risk damage must surely be inde-pendent of
possession of knowledge of this kind.
In fact, the claim that the plaintiff suffered risk damage begins to look quite implausible
when the treatable cases can in practice be distinguished from the untreatable cases. If
the plaintiffs injury was treatable, then the defendant caused him physical damage; there
is no reason to say that it also caused him another, separate harm that takes the form of
risk damage. If, on the other hand, the plaintiffs injury was not treatable, then there is no
reason to say that the defendant caused him any damage, whether we call it physical
damage or risk damage; we can no longer plausibly maintain that he lost a chance of
avoiding an adverse physical outcome. As we just saw, however, nothing should turn on
whether we have the knowledge to distinguish in practice between the treatable and the
untreatable cases. If the causal processes involved are deterministic then a distinction
can be drawn in prin-ciple between the two categories of case, whether we have the
knowledge to do so in practice or not, and so long as such a distinction is possible in
principle, it makes no sense to claim that the plaintiff suffered, at the hands of the
defendant, a peculiar, non-physical injury of the kind I have labelled risk damage. The
assertion that the plaintiff suffered risk damage is simply a fiction employed to get around
the practical difficulties involved in distin-guishing the treatable from the untreatable
cases, i.e., the difficulties in establishing causation with respect to the physical harm.
(p.335) I have assumed, for the purposes of this part of the discussion, that the original
risk figures in Hotson of .75 background risk and .25 agent-imposed risk involved
objective probabilities. It is important to recognize, however, that these probabilities, like
all objective probabilities, are relative frequencies that have been defined with respect to
a certain class of persons: they are properties of those classes and not of particular
individuals like the plaintiff. But it is individuals who are said to suffer risk damage, not
classes of persons, and therein lies the fallacy of the claim that risk damage constitutes
injury in its own right. The Hotson plaintiffs lost chance of avoiding avascular necrosis,
because it cannot be construed as an aspect of the well-being of an individual, cannot
constitute an interest of the kind tort law is concerned to protect. To come at the same
point from a different direction, notice that the plaintiff cannot be said to belong to a
single, canonically designated reference class in terms of which the relevant risks must be
defined. Even with respect to the single incident in Hotson, the plaintiff belongs to an
indefinitely large number of reference classes. Suppose, for example, that the plaintiffs
injury was untreatable. Then he belongs to the reference class of persons with
untreatable injuries, and this is so whether we can identify those persons or not. The
probability of developing avascular necrosis even with proper treatment is, within this
reference class, one, a state of affairs that is quite compatible with the fact that within the
original reference class, consisting of persons with injuries vaguely similar to the plaintiffs,
the probability of developing avascular necrosis even if proper treatment were
administered is only .75.
Other possibilities can easily be imagined. Earlier I made the simplifying assumption that
there is some more or less constant threshold of degree of blood-vessel damage that

Page 12 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


distinguishes the treatable injuries from the untreatable within the general category of
injuries vaguely similar to the one the plaintiff in Hotson originally sustained. Let me now
relax this assumption and suppose instead that the threshold varies with a number of
different factors, some of which are known to us and some of which are unknown. Then
various combinations of these factors will define different reference classes, although of
course we can only refer to those classes that are defined in terms of factors we happen
to know about. Suppose, for example, that two of the factors known to us are body weight
and bone structure. Then we could define a reference class consisting of persons who
suffer an injury of the same general kind as the plaintiff and who also have a bone
structure and body weight similar to his. This class, which would presumably overlap
both the class of treatable injuries and the class of untreatable injuries, would have
associated with it an objective probability of inevitably developing avascular necrosis (i.e.,
of developing the disease even if given proper treatment) that we could expect to be
different from any of the other probabilities so far considered. Relaxing the assumption
(p.336) of a constant threshold does not, of course, affect our earlier conclusion that
someone in the position of the plaintiff in Hotson does not suffer any harm that can
plausibly be characterized as risk damage, since we are still supposing that the relevant
causal processes are deterministic. So long as this supposition holds, the treatable
injuries will be distinguishable in principle from the untreatable, and this is so even
though the factors that determine into which category a particular injury falls have now
been assumed to be relatively numerous and complex.
The conclusion thus far, then, is that if the processes that caused or might in the future
cause physical harm are deterministic, then there is no basis for saying that a person who
has been put at risk by another of suffering such harm has, just by reason of being put at
risk, sustained damage distinct in kind from the physical harm. In other words, in cases
where determinism holds, there is no such thing as risk damage. The basis for saying that
a person is at risk of suffering a certain type of physical harm is that he or she belongs to
a particular reference class with which is associated a known objective probability of harm
of that type. But there is nothing mag-ical about the particular reference class selected:
generally it will simply be the narrowest class, given the current state of our knowledge,
for which we are able to determine with some degree of accuracy the relative frequency
of the type of harm in question. If we were in a position to describe the objective
probabilities associated with a still narrower reference class to which we knew the person
in question also belonged, presumably we would do so. Moreover we know, given the
assumption of determinism, that it is in principle possible to partition any such reference
class into two subclasses, one for which the probability of physical harm is one and one
for which the probability is zero. Thus, there is simply no ground for the claim that the
objective probability of harm, associated with the particular reference class that we are
currently in a position to describe, itself constitutes a distinct form of damage.
The natural question to ask at this point is, of course, how the analysis changes if the
causal processes are not wholly deterministic. There does not seem to me to be any
simple answer to this question, and for present purposes I will limit myself to a few
observations. First, if a particular objective probability of physical harm reflects an

Page 13 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


indeterministic causal process, so that the reference class with respect to which that
probability is defined cannot be further partitioned, conduct that has the effect of placing
someone within the reference class in question might perhaps be characterized as having
caused a distinct form of damage. The question is not straightforward, but conduct that
places someones well-being at the mercy of an indeterministic roll of the dice seems
distinguishable from the situations discussed above, where the assumption was, in effect,
that the risk-imposer would either definitely cause physical harm or would definitely not
cause (p.337) physical harm, but because of imperfect knowledge we would not be in a
position to say which of these two states of affairs in fact obtained. In the indeterministic
case there seems to be a true detrimental shift in position that is simply not present in the
deterministic case (unless, of course, true physical harm in fact is caused).
The second observation is that, while there may thus be a sense in which risk-imposition
does constitute damage in itself, the advocates of the risk damage thesis have not
distinguished between the deterministic and indeterministic cases. Moreover in many of
the fact situations in which risk damage has been alleged, the causal processes at work
seem more likely to have been deterministic than indeterministic in character. This is true
of Hotson, for example, where the House of Lords made the very plausible assumption
that at the time the plaintiff arrived at the hospital either enough blood vessels were still
intact to make his injury treatable, or enough had been destroyed to make avascular
necrosis inevitable.22
The third and final observation I would make on the topic of indetermin-ism is that just
because the causal processes underlying a given type of event are, in part or at some
level, indeterministic, it does not follow that any risk associated with such an event should
be treated as damage in itself. It seems clear, for example, that true indeterminism exists
at the level of sub-atomic particles, but there is some reason to think that random
deviations in different directions compensate for one another, in effect washing
indeterminism out of the system at the macroscopic level. If that were the case then
inter-actions among what J. L. Austin called medium-sized dry goods would, for all intents
and purposes, be deterministic in character.23 There is clearly much more to be said
about the relationship between causal indeterminism and the thesis that risk-imposition is
damage in itself, but these observations should at least engender caution. Even given our
knowledge that indeterminism exists in the world, we generally do not have reason to
treat the objective probabilities we encounter on an everyday basis as anything other
than incomplete descriptions of essentially deterministic events.24
To repeat, my main conclusion in this section is that, at least in the case of events involving
deterministic causal processes, there is no such thing as risk damage.25 It is important,
however, to emphasize the relatively (p.338) modest nature of this conclusion. I have
argued that risk does not generally constitute damage, but not that there is no such thing
as risk: probability in both its epistemic and objective senses is a meaningful concept, and
risk is straightforwardly defined in terms of probability. Nor have I argued that liability in
tort should not be imposed for risk creation as such. To support that conclusion a further
premise is required, to the effect that tort liability should only be imposed in order to

Page 14 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


compensate for damage caused (in a but-for or NESS sense). I am generally prepared to
accept this claim, or something like it, but I have not argued for it here, and the question
can hardly be regarded as uncontroversial. Such a premise would be rejected by many
deterrence theorists, for example, including in particular many who accept an economic
interpretation of tort law.26
Finally, I have not argued that there should not be recovery in tort even in a case like
Hotson itself. I have simply argued that the lost chance of .25 of not developing avascular
necrosis should not be regarded as damage in its own right. There could, however, be
psychological harm that occurs when the plaintiff in a case like Hotson discovers that he
or she has not (p.339) received proper treatment, and perhaps this should be legally
compensable.27 More importantly, there is a very plausible argument that the lost
opportunity to receive proper medical treatment is a distinct form of harm in its own
right.28 When the plaintiff in Hotson left the hospital after the initial misdiagnosis, he was
relying on the defendants supposed skill and competence in diagnosing and treating leg
injuries, and if he had had any reason to doubt their competence he presumably would
have gone immediately to different doctors. As a result of this reliance he lost the
opportunity to receive proper treatment at a time when there was, according to our best
present knowledge, at least some chance, however small, that he might have materially
benefited from it. The .25 lost chance of avoiding avascular necrosis might then be
regarded, not as damage in itself, but as evidence bearing on the valuation of the lost
opportunity.29
It is, in my view, no coincidence that the tort cases in which the risk damage thesis has
won any degree of judicial acceptance have all been medical malpractice actions where it
was plausible to think that there had been both reliance by the plaintiff and consequent
lost opportunity. We must at least consider the possibility that the approving judges really
had in mind the notion of lost opportunity, not lost chance or risk, or, perhaps more
plausibly in the majority of cases, that they simply failed to distinguish between these two
notions. So far as I am aware, there is no case in which risk has been treated as damage
in itself outside the context of a pre-existing con-sensual relationship involving reliance. If
risk creation truly is a form of damage in its own right, it should be present even in a
stranger case involving, say, a risk of disease arising from exposure to a toxic element.
But the claim that risk is a distinct form of damage appears not to be advanced on facts
such as these.

III. Risk and Responsibility for Physical Harm


If, as was argued in the preceding section, risk cannot be regarded as harm in its own
right, then it is natural to think that moral responsibility for risky behavior should take the
form of responsibility for the physical harm, if any, that was unintentionally caused by the
conduct in question (assuming, of course, that the harm falls within the risk the conduct
created). We might also consider the possibility that someone should be held morally
(p.340) responsible for simply engaging in the risky behavior, quite apart from whether
harm in fact ensues, but since our present concern is with tort theory rather than
criminal theory I shall limit discussion to responsibility for physical harm that an agents

Page 15 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


conduct has actually caused. Since it will not be possible in the present chapter to
consider this issue at any length, I shall concentrate on clarifying the proper
understanding of risk within a theory of responsibility for causing harm, drawing on the
account of risk set out in section I. We must consider two possibilities: first, that
responsibility for physical harm is concerned with conduct that was risky in the objective
sense, and secondly, that such responsibility is concerned with conduct that was risky in
the epistemic sense.
Objective probability, and hence objective risk, is defined in terms of the relative
frequency of an attribute within a reference class. Risk in this sense exists whether we
know about it or not. Responsibility for harm resulting from conduct that was risky in the
objective sense cannot, therefore, depend on antecedent knowledge or foreseeability.
Responsibility of this kind, if it exists, is equivalent to responsibility determined solely by
the fact that ones conduct caused harm, since if a particular action causes harm it must
fall within one or more reference classes for which, whether or not we are in a position to
describe those classes, the relative frequency of the relevant type of harm is greater
than zero. Translated into tort terms, such responsibility would amount to a form of strict
indeed, absoluteliability. Absolute liability is often described as a matter of acting at
ones own peril or risk; as the preceding discussion makes clear, it is objective risk that is
at issue here.
The strongest and most influential argument for absolute liability has often been put
forward by libertarians, and it fits together in a natural way with libertarian political
philosophy.30 I shall accordingly call it the libertarian argument. The essential idea is that if
I choose to act in the world, I am both fully entitled to whatever gains I may make and
fully responsible for whatever harms I may suffer or cause. The claim about gains leads
to the standard libertarian thesis that forced redistribution is in general ille-gitimate. The
claim about harm is the basis of the libertarian defence of a standard of absolute liability in
tort law. Both claims are premised on a certain understanding of the moral significance of
action. I have a choice about (p.341) whether to become active in the world, and if I
choose action over passivity then all subsequent consequences, both good and bad, are
appropriately chalked up to my moral ledger and no one elses. Thus other persons are
not entitled to share in whatever gains may accrue to me as a result of my activity,31 but
by the same token I cannot force passive by-standers to absorb any loss my activities
may create. In economic jargon, agents are properly compelled to internalize the losses
they cause others. In distinguishing as it does between activity and passivity, the
libertarian argument thus supposes that moral consequences, both good and bad, attach
not just to a decision to act in this or that way, but to a decision to act tout court.
In The Common Law Holmes criticized what is recognizably a version of the libertarian
argument, which he formulated in the following way:
Every man, it is said, has an absolute right to his person, and so forth, free from
detriment at the hands of his neighbors. In the cases put, the plaintiff has done
nothing; the defendant, on the other hand, has chosen to act. As between the two,
the party whose voluntary conduct has caused the damage should suffer, rather
Page 16 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


than one who has had no share in producing it.32
The main elements of the libertarian argument, including in particular the distinction
between activity and passivity, are clearly present here. But Holmes rejected the
argument, not just on grounds of policypolicy meaning, for Holmes, utilitarian
considerationsbut also because it offend[ed] the sense of justice.33 The initial premise
of Holmes counter-argument is that the reason for making the performance of an act a
requirement of liability in tort is to ensure that the defendant should have made a
choice.34 This premise is, of course, also one of the starting-points of the libertarian
argument. But then Holmes continues: the only possible purpose of introducing this
moral element is to make the power of avoiding the evil complained of a condition of
liability. There is no such power where the evil cannot be foreseen.35 At this point
Holmes is clearly introducing a different understanding of the moral significance of action.
It is not choice as such that has moral significance, but rather choice accompanied by the
power to avoid the resulting consequence. The capacity to avoid the result, rather than
the mere fact of having voluntarily chosen to act, is what gives rise to moral responsibility
for harm.
But why should we accept Holmes claim that it is avoidability of the consequence, rather
than the choice to act as such, that gives rise to moral responsibility for physical harm?
Holmes himself has little to say on this (p.342) point, but he drops a hint, almost in
passing, in his subsequent discussion of the argument from policy: A man need not, it is
true, do this or that act,the term act implies a choice,but he must act somehow.36
This sug-gests that Holmes rejects, at least so far as injurers are concerned, the
distinction presupposed by the libertarian argument between activity and passivity. Thus
we might reconstruct his counter-argument in the following way. The libertarian
argument supposes that moral significance attaches to a decision to act as such, because
it supposes that persons have a true choice concerning whether to be active or passive
beings. But in fact there is no such choice: the human condition is such that everyone
must choose to act in some way or other.37 Given that everyone must be active, no moral
consequences can attach to action per se. Something more is therefore required to
ground responsibility for causing harm, and that further requirement is most plausibly
thought to be a capacity to avoid the harm. But there can be no such capacity unless the
agent is capable of foreseeing the result. As Holmes famously put it, [a] choice which
entails a concealed consequence is as to that consequence no choice.38
It is possible to take this line of thought further. The libertarian argument treats injurers
as active and victims as passive, in each case by choice. Holmes responds that injurers
are necessarily active, not active by choice, because activity is inevitable. But if that is so,
then victims must be characterized as active and not just injurers. This is, in essence, the
point made by Coase when he wrote that [i]f we are to discuss the problem [of harmful
interactions between two persons] in terms of causation, both parties cause the
damage.39 The libertarian argument assumes that one person is the cause of harm to
another, where the image is of a dominant, active party acting upon a subordinate,
passive party. Coase was responding to a similar image embodied in the received

Page 17 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


economic wisdom that social costs should be internalized to the activity that caused them.
In addition to arguing that, from an economic point of view social costs are, under certain
cir-cumstances, best dealt with by private market transactions, Coase also maintained
that this image of one party acting upon the other is profoundly misleading and should be
rejected. Instead, he suggested, we must see the harm in question as having been
caused by the interaction of two active partiesin other words, as the product of two
activities, and not just one. Thus the injury that results when a motorist runs down a
pedestrian should not be regarded as having been caused by one person acting
unilaterally (p.343) upon another, but rather as the upshot of two decisions to act: the
motorists decision to drive when, where and how she did, and the pedestrians decision
to walk when, where and how he did.
The image associated with the libertarian argument of one, active party who exerts force
against another, passive party is often buttressed by an analogy with property rights, and
in particular rights in land.40 Robert Nozicks metaphor of a line or boundary that
circumscribes an area in moral space around an individual draws attention to this
analogy in a particularly vivid way.41 A violation of individual rights is equated with a
border-crossing, and this notion is equated in turn with, essentially, causing harm. But
the analogy with a physical border must ultimately be cashed out in terms of the
distinction between an active injurer and a passive victim, and both the Holmesian and
Coasean arguments call this distinction into question. In addition, the Coasean argument
undermines the claim that one can pick out one of the parties to a harmful interaction as
the cause of the harm.
I have argued elsewhere for an understanding of harmful interaction and causation that
is consistent with the views of both Holmes and Coase,42 and I need not repeat that
discussion here. The point I wish to emphasize for present purposes is this. The
libertarian argument, if it were valid, would justify a form of moral responsibility that was
triggered by the materialization of an objective risk. But in making foreseeability the
touchstone of responsibility for causing harm, Holmes is implicitly invoking a notion of
epistemic risk. Foreseeability is a function of (possibly incomplete or only partially
accurate) knowledge of relative frequencies, not of relative frequencies tout court. As
Holmes makes clear, the key moral concept here is avoidability: the point of making
foreseeability a requirement of responsibility for physical harm is that an agent is unable
to avoid harm unless he or she can foresee it. In the absence of foreseeability, the harm
is simply the unfortunate upshot of an interaction between two persons. There is no
reason in justice to shift the loss from where it fell, since there is no basis for morally
attributing the harm to one party rather than the other.
The reasonableness account of epistemic probability maintains that such probabilities are
estimations of relative frequency that have been made in accordance with accepted
standards of inductive reasoning and rational belief. Epistemic probability judgements, in
so far as they presuppose inter-subjectively valid standards of reasoning, can be
characterized as objective. But it is natural to say that avoidability, as a moral concept,
must look to a subjective capacity to avoid harm, and hence to a subjective capacity to

Page 18 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


(p.344) foresee it. How can this point be reconciled with the objectivity of the
reasonableness account, and, indeed, with the similarly objective understanding of
foreseeability in negligence law (reasonable foreseeability, in the language of tort)? This
issue cannot be examined in detail here, but the brief answer is that we are dealing with a
number of different conceptions of the objective/subjective distinction.
The notion of reasonable foreseeability in tort law is objective in its content: the laws
determination of which risks must be foreseen by an agent is based upon, among other
things, an appropriate interpretation of the reasonableness account of epistemic
probability (more particularly, an interpretation expressed in terms of the judgements of
an ordinary or reasonable person). The standard of due care in negligence law is similarly
objective in its content: the precautions the law requires in a given situation of risk are
determined by what a reasonable person, rather than the individual defendant, thinks
should be done. But norms are almost always objective so far as their content is
concerned; it is not up to the individual agent to decide what constitutes, say, murder.
The notions of reasonable foreseeability and due care are also objective in another sense,
which is that neither requires advertence: the individual agent need not be aware, on the
particular occasion, either that he is creating unreasonable risks or taking inadequate
pre-cautions. This is, of course, an important form of the objective/subjective distinction
in criminal law.
But objectivity regarding content and advertence is perfectly compatible with subjectivity
regarding capacity, at least so long as we understand capacity in a certain way. According
to the view I have in mind, which I cannot defend here but only sketch in very
abbreviated form, capacity to foresee and avoid harm is appropriately understood as a
general ability that the individual who caused a given injury ordinarily succeeds in exercising in other, similar situations.43 Whether the person was capable of foreseeing and
avoiding the harm on the particular occasion may or may not be a meaningful question,
but either way it is the general capacity that matters so far as moral responsibility is
concerned. The necessary assumption, then, is that most people, across a wide variety of
circumstances, have the general capacity to foresee and avoid harm that the law
attributes to the ordinary or reasonable person. The result is a form of responsibility
for unintentional harm that I call, borrowing a term from (p.345) Tony Honor,
outcome-responsibility.44 The key moral concept that underpins such responsibility is, as
Holmes suggested, avoidability. It can be contrasted with the central form of
responsibility in criminal law, for which the label culpability is appropriate: here the key
moral concept is an intention to harm, or at least a subjective awareness that harm will or
might occur. Obviously these conceptions of responsibility can overlap, since there is no
reason to deny that a person who is aware of the risk he or she is creating is outcomeresponsible for the resulting harm if the risk materializes. But the two conceptions of
responsibility are nonetheless distinct: outcome-responsibility is, among other things,
generally less blameworthy than culpability.
Outcome-responsibility, like the libertarian conception of absolute liability considered
earlier, focuses on responsibility for harmful outcomes rather than on responsibility for

Page 19 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


action as such. But it differs from the libertarian conception in that it does not
automatically call for an agent to whom responsibility for harm is being attributed to pay
compensation. Outcome-responsibility affects reasons for action in a more diffuse fashion:
in a given situation it may require the agent to apologize, for example, or to obtain
assistance. It also constitutes the basis for a moral obligation to compensate, but in order
actually to establish such an obligation further argument is necessary. In general, then,
outcome-responsi-bility can serve as the moral foundation in tort law for either a
negligence standard of care, or for a danger-based standard of strict liability such as the
rule in Rylands v. Fletcher or the American rule for ultrahazardous activities, since both
kinds of liability presuppose reasonable foreseeabil-ity. But the considerations that bear
on the choice between these liability rules, as well as on their more specific formulation,
are matters that must be left to another occasion.

IV. Conclusion
Risk is a central concept in both tort law and tort theory. In this essay I have argued that
it is important to distinguish between an objective and an epis-temic conception of risk.
The former is grounded in the standard relative-frequency account of probability, while
the latter is concerned with the evidentiary basis for judgements or estimations of
relative frequency. The commonly held thesis that risk is harm in itself is most plausibly
explicated in terms of the objective conception of risk. But even under this favorable
interpretation that thesis must, I have argued, be rejected. Moral responsibility for the
consequences of risky conduct should be understood not as responsibility for creating a
risk, but as responsibility for the unintentional physical harm, if any, that ensues. Such
outcome-responsibility can, however, (p.346) only plausibly be attributed to a person if
his or her conduct was risky in the epistemic and not just in the objective sense. This
suggests in turn that within a theory of tort law based on notions of interpersonal justice
and individual responsibility, foreseeability of harm ought to be considered a necessary
condition of liability.
Notes:
(1 ) For purposes of this chapter I treat the terms harm, injury, and damage as
synonymous. In some contexts, however, it is appropriate to assign them distinct
meanings. See Stephen R. Perry, The Moral Foundations of Tort Law, 11 IOWA L. REV.
449, 498 (1992).
(2) For an excellent overview, see ROY WEATHERFORD , PHILOSOPHICAL
FOUNDATIONS OF PROBABILITY THEORY (1982).
(3) See, e.g. RUDOLF CARNAP , LOGICAL FOUNDATIONS OF PROBABILITY (2d ed.
1962), 1951; DAVID LEWIS , A Subjectivists Guide to Objective Chance, in 2
PHILOSOPHICAL PAPERS (1986) 83.
(4) IAN HACKING , THE EMERGENCE OF PROBABILITY (1975), 1223.
(5) See particularly RICHARD VON MISES , PROBABILITY, STATISTICS, AND TRUTH
(1957); HANS REICHENBACH , THE THEORY OF PROBABILITY (1949).

Page 20 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


(6) See Joseph Ford, How Random is a Coin Toss?, 36 PHYSICS TODAY 40 (April 1983).
Ford maintains that a coin toss is predictable in the sense that it is describable by a finite
algorithm. But it is unpredictable in the sense that the description of the initial condition
requires, because chaotic processes are involved, an infinite amount of information.
(7) It should be noted that even those theorists who maintain that a relative-frequency
theory is the only defensible account of probability must still acknowledge the
distinctionbetween an actual frequency in the physical world and our estimate or
approximation of that frequency. Cf. WEATHERFORD , supra, note 2, at 1312, 142, 177.
This distinction creates sufficient logical room to permitin fact, to requirethe
development of an epistemic conception of probability along the lines of the
reasonableness account.
(8) Presumably one could also regard the resulting probability judgement as an
expression of the degree of rational belief that is warranted in the conclusion that the
property whoserelative frequency is being considered will be instantiated in a given case.
(9) F. P. Ramsey, Truth and Probability, in FOUNDATIONS: ESSAYS IN PHILOSOPHY,
LOGIC, MATHEMATICS, AND ECONOMICS (D.H. Mellor (ed.), 1978), 58. For versions
of radical subjectivism, see 1 BRUNO DE FINETTI , THEORY OF PROBABILITY: A
CRITICAL INTRODUCTORY TREATMENT (1974); LEONARD SAVAGE , THE
FOUNDATIONS OF STATISTICS (1954), 2768.
(10) De Finetti, one of the leading radical subjectivists, formulated his central thesis in
upper case letters: PROBABILITY DOES NOT EXIST: DE FINETTI , supra, note 9, at x.
(11 ) See JOHN M. KEYNES , A TREATISE ON PROBABILITY (1921); CARNAP , supra,
note 3. The three most influential general theories of probability are the relativefrequency theory, the subjectivist theory, and the logical-relation theory.
(12) It should be emphasized that probability statements are not regarded as objective in
the sense presupposed by a relative-frequency theory, i.e., they are not understood as
expressingfacts about the physical world. They are, rather, supposed to be objective in
an a priori, non-empirical sense: for each possible set of evidentiary premises, whether
true or not, and for eachconclusion about the way things might be that can be formulated
in the language, there is alogically correct probability judgement that in principle is
specifiable a priori.
(13) WEATHERFORD , supra, note 2, at 236.
(14) It bears mention that some subjectivists, while claiming that we do not necessarily
have knowledge of objective probabilities, do nonetheless accept that there are
pragmatic standardsfor assessing equally coherent probability judgements as better or
worse. These standardsInvolve a proper respect for frequencies, arising from a proper
respect for induction: Simon Blackburn, Opinions and Chances, in PROSPECTS FOR
PRAGMATISM: ESSAYS IN MEMORY OF F. P.RAMSEY (D. H. Mellor (ed.), 1980), 175,

Page 21 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


195.
(15) Keynes, the other main proponent of a logical relation understanding of probability,
took a very different approach from that of Carnap: see Keynes, supra, note 11. Keynes
held, perhaps under the influence of G. E. Moore, that we have the capacity directly to
perceive or intuit probability relations. This led him to maintain, among other things, that
some probabilities are non-numerical and some non-comparative. This approach
obviously ties probability judgements much more closely to actual human capacities, but
it does so at the cost of rendering problematic the thesis that such judgements are logical
in character.
(16) See Joseph H. King, Jr., Causation, Valuation and Chance in Personal Injury Torts
Involving Preexisting Conditions and Future Consequences, 90 YALE L.J. 1353 (1981);
Richard W. Wright, Causation in Tort Law, 73 CAL. L. REV. 1735, 181416 (1985);
WILLIAM M. LANDES and RICHARD A. POSNER , THE ECONOMIC STRUCTURE OF
TORT LAW (1987), 263; Jane Stapleton, The Gist of Negligence, pt. 2, 104 L.Q.R. 389
(1988); Herskovits v. Group Health Coop, of Puget Sound, 664 P.2d 474 (Wash. 1983)
(Pearson, J., concurring); DeBurkarte v. Louvar, 393 N.W.2d 131 (Iowa 1986); Hotson v.
East Berkshire Area Health Authority [1987] 2 W.L.R.287, revd. [1987] A.C. 750;
Lawson v. Laferriere, R.J.Q. 27 (Quebec 1989), revd, [1991] 1S.C.R. 541 (Can.).
(17) [1987] 2 W.L.R. 287, revd. [1987] A.C. 750. I have criticized the thesis that risk
constitutes harm in itself, along lines similar to the argument presented in this section, in
Stephen R. Perry, Protected Interests and Undertakings in the Law of Negligence, 42 U.
TORONTO L.J. 247, 25262 (1992).
(18) For the sake of convenience in exposition I shall henceforth use the terms risk and
probability more or less interchangeably. Given our earlier definition of risk as the
mathematical expectation of harm (P H), this amounts to an assumption that H = 1.
(19) In general, if A is the agent-imposed risk and B is the background risk, the ex post
probability that the agent caused the outcome is A/(A + B). The ex post probability that a
factor contributing to the background risk caused the outcome is B/(A + B). In Hotson
the figures for antecedent risk coincide with those for ex post probability of causation
because A + B = 1.
(20) Hotson [1987] 1 A.C. at 783, per Lord Mackay.
(21 ) Id. at 785.
(22) Hotson [1987] 1 A.C. at 785, per Lord Mackay.
(23) Hans Reichenbach, one of the main proponents of the relative frequency theory of
probability, held a thesis along these lines: see HANS REICHENBACH , ATOM AND
COSMOS: THE WORLDOF MODERN PHYSICS (1933), 278; WEATHERFORD , supra,
note 2, at 181.

Page 22 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


(24) It is worth remarking that the basic tests of causation in tort law, namely, the butfortest and its more sophisticated variant, the NESS test, both appear to presuppose
determinism. This is implicit in the notion of necessity that each test in different ways
incorporates. For discussion of the NESS test (Necessary Element of a Sufficient Set),
see Wright, supra, note 16.
(25) Mention should be made of Sindell v. Abbott Lab., 607 P.2d 924 (Cal. 1980), since
thereis a sense in which that case imposed liability for risk creation (as measured by
market share).
The underlying rationale in Sindell, however, was to ensure that each of a group of
tortfeasors pays out compensatory damages equivalent to the total amount of tortious
physical harm that that tortfeasor can be expected to have caused in a defined category
of case. Since risk is not being treated as damage in its own right, but rather as an
indicator of total physical harm in a category of cases, the rationale of the case is not called
into question by my arguments here. It is worth noting that the total-physical-harm
rationale for market share liability was rejected in Hymowitz v. Eli Lilly & Co., 539 N.E.2d
1069 (N.Y. 1989), where a justification based on culpability was accepted instead. For
that reason Hymowitz is a more problematic decision than Sindell. A variant on the totalphysical-harm rationale of Sindell was suggested by Weinstein, C.J., in In re Agent
Orange Product Liability Litigation, 597 F. Supp. 740, 8339 (E.D.N.Y. 1984). Given the
impossibility of sorting out which individuals had contracted cancer from Agent Orange
and which from background risks, the proposal was, in essence, that all cancer victims in
the relevant group might be able to recover a share of their damages proportional to the
increase in total risk created by the agent-imposed risk. The amount the defendant would
have to pay out should be roughly equivalent to the physical harm it had actually caused.
(26) That risk should as a matter of corrective justice be compensable as such, even in the
absence of physical harm, is the ultimate conclusion of a debate between Christopher
Schroeder and Ken Simons. See Christopher H. Schroeder, Corrective Justice and
Liability for Increasing Risks, 37 UCLA L. REV. 439 (1990); Kenneth W. Simons,
Corrective Justice and Liability for Risk-Creation: A Comment, 38 UCLA L. REV. 113
(1990); Christopher H. Schroeder, Corrective Justice, Liability for Risks, and Tort Law,
38 UCLA L. REV. 143 (1990). Neither Schroeder nor Simons explicitly states that risk is
harm, but without this premise the conclusion that risk is compensable on corrective
justice grounds is not very plausible. Glen Robinson also holds that risk should be
compensable, but his discussion is ambiguous as to whether risk is properly treated as
harm in itself. Glen O. Robinson, Probabilistic Causation and Compensation for Tortious
Risk, 14 J. LEGAL STUD. 779 (1985). LANDES and POSNER , supra, note 16, at 263,
expressly maintain that risk can be treated as a form of harm, but on the deterrence
theory they defend this premise does not appear to be required in order to argue for
the compensability of risk as such. Landes and Posners discussion of probability is
confused, in my view, by a failure to distinguish objective from epistemic probability. See
particularly id at 2349.

Page 23 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


(27) See Lawson v. Laferriere, [1991] 1 S.C.R. 541, 558 (Can.); Ayers v. Township of
Jackson, 525 A.2d 287, 295 (N.J. 1987).
(28) See further, Perry, supra, note 17, at 30816.
(29) It also seems reasonable to think that medical expenses incurred to avoid or
minimize aknown, agent-imposed risk of suffering physical harm in the future should be
recoverable intort. Liability for pure economic loss, which as a general matter gives rise
to difficulties in neg-ligence law, is not problematic in this context. See Perry, supra, note
17, at 2679.
(30) A version of the libertarian argument was put forward by Baron Bramwell, a 19th
century English judge known for his laissez-faire ideology, in Bamford v. Turnley, 122 E.
R. 27, 33 (Ex. Ch., 1862). In his early work in tort theory Richard Epstein advanced an
argument very similar to Bramwells. See Richard A. Epstein, A Theory of Strict Liability,
2 J. LEGAL STUD. 151, 1578 (1973); Richard A. Epstein, Intentional Harms, 4 J. LEGAL
STUD. 391, 398 (1975). For discussion of Epsteins version of the argument, see Stephen
R. Perry, The Impossibility of General Strict Liability, 1 CAN. J.L. & JURIS. 147, 14859
(1988). The libertarian argument seems also to be implicitly presupposed by Robert
Nozick, since he assumes that an obligation to compensate arises simply from causing
harm. See ROBERT NOZICK , ANARCHY, STATE, AND UTOPIA (1974), 71.
(31 ) The status attributed by the libertarian argument to positive externalities, i.e.,
benefitsthat flow from the agents conduct but that accrue to other persons, is not
entirely clear.
(32) OLIVER WENDELL HOLMES, JR., THE COMMON LAW (1881), 84.
(33) Id. at 96.
(34) Id. at 95.
(35) Id.
(36) OLIVER WENDELL HOLMES, JR., THE COMMON LAW (1881), at 95.
(37) A variant of the argument, which Holmes would probably not accept, would add
thateven a decision to remain motionless is, precisely because it is a voluntary decision,
morallyequivalent to an act in the usual sense, understood as involving muscular activity.
For Holmes, [a]n act is always a voluntary muscular contraction, and nothing else: id. at
91.
(38) HOLMES , supra, note 32, at 94.
(39) R. H. Coase, The Problem of Social Cost, 3 J.L. & ECON. 1, 13 (1960).
(40) Cf. Richard A. Epstein, Automobile No-Fault Plans: A Second Look at First Principles,

Page 24 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

Risk, Harm, and Responsibility


13 CREIOHTON L. REV. 769, 775 (1980).
(41 ) NOZICK , supra, note 30, at 57.
(42) Perry supra, note 30, at 1549; Perry, supra, note 1, at 4617.
(43) Cf Tony Honore, Can and Cant, 73 MIND 463 (1964); Tony Honore, Responsibility
and Luck, 104 L.Q.R. 530, 54952 (1988) [hereinafter Responsibility]; DANIEL C.
DENNETT , ELBOWROOM: THE VARIETIES OF FREE WILL WORTH WANTING (1984),
ch.6.
(44) Honor Responsibility, supra, note 43. My understanding of outcome-responsibility
differs in several important respects from Honors. See further, Perry, supra, note 1, at
488512; Stephen R. Perry, Loss, Agency, and Responsibility for Outcomes: Three
Conceptions of Corrective Justice, in TORT THEORY (Ken Cooper-Stephenson & Elaine
Gibson (eds.), 1993),24, 3847.

Access brought to you by: Pontificia Universidad Catolica


del Peru (PUCP)

Page 25 of 25
PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2015.
All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a
monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: Pontificia
Universidad Catolica del Peru (PUCP); date: 30 April 2015

You might also like