Tony Honore-Responsibility and Fault (1999) - Ch2,5

!
2"
RESPONSIBILITY AND LUCK
THE MORAL BASIS OF STRICT LIABILITY
This essay1 touches on both civil and common law, but its central
theme is a very broad one: that being responsible in law and in ordi-
nary life is not the same thing as being at fault or to blame. It starts
from the well-known rule of the tort of negligence that requires
everyone to attain an objective standard of care and competence; and
then moves towards a theory of responsibility in which the contrast
between fault and strict liability, though not obliterated, is blunted.
The argument is in three sections. As it is a little complicated, I
summarise it at the start. The first section argues that the objective
standard of competence, though purporting to be based wholly on
fault, really imposes a form of strict liability on those who suffer from
unavoidable shortcomings. In the absence of a moral justification for
strict liability it cannot be supported.
To justify strict liability we must first show why people should
sometimes bear the risk of bad luck, including in bad luck such mis-
fortunes as being stupid or clumsy. The second section argues that to
bear the risk of bad luck is inherent in the basic form of responsibil-
ity in any society, which I term outcome responsibility. Outcome
responsibility means being responsible for the good and harm we
bring about by what we do. By allocating credit for the good out-
comes of actions and discredit for bad ones, society imposes outcome
responsibility; though often the rewards it attaches and, outside the
law, the sanctions it imposes are informal and vague. Under a system
of outcome responsibility we are forced, if we want to keep our social
account in balance, to make what amounts to a series of bets on our
choices and their outcomes. Provided we have a minimum capacity
1. A version of the 13th Blackstone lecture delivered in Oxford under the auspices of
Pembroke College on 8 May 1988. First published in 104 Law Quarterly Review
(1988) pp.530–53.
14
for choosing and acting, we win the bets and get credit for good out-
comes more than we lose them and incur discredit for bad ones. We
have to take the risk of harmful outcomes that may be sheer bad luck
and not our fault; but that does not make the system unfair to people
who are likely to be winners overall.
Seen in this light civil liability in law, whether strict or based on
fault, can be defended on the ground that it specifies an extra sanc-
tion to be imposed on a person who has anyhow lost a bet and will in
consequence incur discredit. The main role of legal liability is to rein-
force our basic outcome responsibility with formal sanctions such as
compensation or punishment. One ground for legal liability, fault, is
present when the person’s conduct not only has a bad outcome but
displays a bad disposition. Another, which leads to strict liability, is
present when the activity that has a bad outcome is specially danger-
ous to others. In practice these grounds often overlap.
Can the system of outcome responsibility itself be justified or
shown to be inescapable? Outcome responsibility is, I contend,
inescapable because it is the counterpart and at the same time a con-
stituent of our personal identity and character. We could not dispense
with outcome responsibility without ceasing to be persons.
The third section deals with capacity and freedom. Outcome
responsibility, though inescapable, can fairly be imposed only on
those who possess a sufficient general capacity for decision and
action. Capacity can for this purpose be tested by asking what a given
person normally achieves when he tries. It has, however, often been
thought that responsibility for a particular action requires something
further; a capacity on the part of the person to have acted differently
given all the factors, external and internal, that were present on the
occasion in question. But this sort of anti-deterministic capacity can-
not exist. So my theory of responsibility, though it does not require
determinism to be true, is compatible with it. The difference between
fault and strict liability in this respect is merely that a person guilty of
fault must have, besides a general capacity for decision and action, the
ability to succeed most of the time in doing the sort of thing that
would on this occasion have averted the harm. A person held strictly
liable must have the same general capacity but need not have this spe-
cial ability.
15
RESPONSIBILITY AND FAULT
1. N EGLIGENCE : T HE O BJECTIVE S TANDARD

OF C OMPETENCE
The starting point of the argument is one that at first sight has noth-
ing to do with strict liability. It is the familiar rule of the law of neg-
ligence that everyone is bound to reach an objective standard of care
and competence. One corollary is that a person is liable for sub-
standard conduct, or is prejudiced by it, even if he acts as he does
because of a shortcoming that he cannot help. A stupid lorry-driver
fails to slow down in the fog and has an accident. One of normal
intelligence would have slowed down and avoided it. An impetuous
pedestrian dashes across the street and is run over. A more equable
person would have waited and walked across safely. An inexperienced
surgeon cuts an artery that with greater experience he would have
known how to avoid. In view of the objective theory of negligence,
their respective shortcomings (stupidity, impetuousness, perhaps also
inexperience) make them legally liable or reduce their claims.2
Some explanation of the terms employed may be useful.
“Negligence” refers to civil liability, in whatever system, for unin-
tended harm, where this is caused by someone’s failure to meet the
required standard of competence. The liability may be in tort or
delict, for breach of contract, under a special statute or under more
than one of these at once. Negligence, the common law term, is here
used in a transsystematic way, as equivalent to Roman culpa, German
Fahrlässigkeit, and French faute de négligence. Lack of competence
may stem from a number of factors. The person concerned, for
whom the term “shortcomer” is to hand, may not have the qualities,
physical, intellectual or emotional, needed to attain the standard set
for the task in question. He may lack the proper makeup. He may
suffer from a defect of character, temperament or physique, limited
intelligence, accident proneness, bad co-ordination, slow reactions.
On the other hand he may possess the physical and mental equip-
ment but lack education or training. Or he may have had all these
but, as a child, lack maturity, or, as a novice, experience. The short-
comer’s lack of competence may thus stem from a defect of physique
2. For inexperience the issue is debated: Nettleship v. Weston [1971] 2 QB 691; Wilsher
v. Essex Health Authority [1987] QB 730.
16
or character, or a deficiency in intelligence, learning or experience.

Shortcoming comprises these defects and deficiencies.
The objective theory of negligence, which is orthodox in the lead-
ing systems of law,3 requires people to display the same competence
as a hypothetical model person. To bring in a model person is to
translate a normative standard into a hypothetical descriptive stan-
dard. The model is variously depicted in different legal cultures. He
or she is seen as a diligent father, a reasonable man/woman, an
abstract type,4 or a careful and conscientious member of the class in
question.5 The class may, according to the case, be that of doctor, dri-
ver, company director, air pilot,6 pedestrian etc. The classification
depends on the type of person expected to undertake the task perfor-
mance of which goes wrong (operating, driving, running a business,
crossing the street), and to display care or skill in doing so.7 The
objective standard has, on the orthodox view, to be met both in
avoiding harm to others and in protecting oneself.8 Failure to meet it
may either make the injurer liable to pay damages as defendant or
may bar the injured, in whole or in part, from recovering damages as
plaintiff.
Such is the general law. It applies, however, only to persons who
have capacity, a notion distinct from competence. The incapable are
not held to it. In most systems a child is regarded as wholly or partly
3. E.g. F.V. Harper, F.I. James and O.S. Gray, The Law of Torts (2nd ed. 1986) vol. III
p.103f.; A. Weill and F. Terré, Droit Civil des Obligations (4th ed. 1986) s.628;
E. Deutsch, Haftungsrecht: Allgemeine Lehre (1976) p.268f. In Germany the objective
theory is statutory. BGB s.276 lays down that a person subject to a duty acts negli-
gently is he fails to take account of the care required by (social) intercourse: die im
Verkehr erforderliche Sorgfalt ausser acht lässt.
4. As French lawyers put it.
5. The German formulation.
6. Viz. an ordinary air pilot, not a pilot “having the same training and experience as Fred
Heath”: Heath v. Swift Wings Inc. 40 NC App. 158, 252 SE 2d 526, 529 (1979).
7. This is the answer to Nipperdey’s objection that the class selected might equally be that
of the phlegmatic, cholerics, alcoholics, etc. Staudinger, BGB s.276 no.18. The class
test is not alien to English law: The Lady Gwendolen [1965] P. 294 (reasonable
shipowner); Philips v. Whiteley [1938] 1 All ER 566 (reasonable jeweller); Wilsher v.
Essex Health Authority [1987] QB 730 (duty of doctor relates to post which he occu-
pies, not his own qualifications and experience).
8. Some have argued that the incapable should, in protecting themselves, be judged by a
more subjective standard (Harper, James and Gray, above n.3, vol. III pp.462–4);
H.H. Terry, 29 Harvard Law Rev. (1914–5) 47; Dellwo v. Pearson 259 Minn. 452,
107 NW 2d 859, 862–3 (1961), rejected in Dunn v. Teti 280 Pa. Super 399, 421 Atl.
2d 782, 784 (1980).
17
incapable, in some the insane,9 in some the elderly.10 The incapable

may be exempt altogether, or required to meet a standard adjusted to
their limitations. Then there are special cases where even a person of
full capacity has only to do as much as he can,11 or as much as he does
when occupied with his own affairs.12 The objective standard gives
way to a subjective or mixed standard.
This outline of the law provides the background to what follows.
Historically negligence, like intentional wrongdoing, has been
regarded as a species of fault. Nearly all writers continue so to regard
it, and the equivalent terms in other systems (culpa, Fahrlässigkeit,
etc.) all connote fault. According to a long-standing opinion, which
provisionally I adopt, fault is imputed only to a person who could
have controlled the situation in which he was placed but failed to do
so. Only someone who could in the circumstances have acted otherwise
is morally responsible, and, if necessary, blamed for his action.13 This
view is open to challenge; and on a narrow interpretation of “could”
or a wide interpretation of “circumstance” I shall argue that it should
be rejected.14 But there is no doubt that the “could have done other-
wise” test appeals powerfully to both lawyers and philosophers. Apply
that test to the law of negligence, and a person is at fault only when
he could observe the objective standard, but fails to do so. So the
incompetent shortcomer who fails to come up to standard is not at
fault and should not be held guilty of negligence. There is linguisti-
cally an exception when the incompetence is due to a defect, or as it
is often called, “fault” of character. But though on a long term view it
9. Some cases allow mental subnormality to reduce the standard expected: e.g. Miller v.
Trinity Medical Center 260 NW 2d 4, 6–7 (1977).
10. Elderly plaintiff: e.g. Johnson v. St Paul City R Co 67 Minn. 260, 69 NW 900,901
(1897).
11. E.g. because he is confronted with a situation that is not of his choosing, like a fire
which through no fault of his breaks out on his land: Goldman v. Hargrave [1967] 1
AC 645. This duty to do what you can (eigenmögliche Sorgfalt: Deutsch, above n.3,
at p.280), which in German law applies to matters such as the award of a solatium
for death, injury or loss of liberty (BGB s.847 Schmerzengeld) is to be distinguished
from the degree of care which the agent, who may have fallen into a slack routine,
normally gives to his own affairs (below n.12).
12. The civil law diligentia quam in suis rebus, which applied to transactions like gratu-
itous deposit, where it seemed unreasonable to demand that the depositee should take
greater care of his friend’s property than his own. BGB s.277 and commentaries pro-
vide modern examples.
13. E.g. A.J. Ayer, Philosophical Essays (1954) p.27.
14. Below pp.32–8.
18
may be possible to correct some faults of character, or to avoid acquir-

ing them in the first place, in any immediate or short-term context a
faulty character is something with which the person concerned has to
cope much as he has to cope with a faulty engine. Moreover, the law
treats as fault other sorts of incompetence due to, say, stupidity or
clumsiness, which the person clearly cannot correct in time, and per-
haps not ever.15
Can this be justified? On one line of argument the shortcomer is not
morally to blame but reasons of policy justify us in treating him as
responsible. A consequentialist may argue that it is right to impose
objective liability because to do so maximises either average or mini-
mum utility, wealth, satisfaction of preferences or some other good.
The rule of objective liability tends to increase economic efficiency, or
to redistribute resources in favour of accident victims in a desirable
way. But though these may be reasons for sustaining the rule, if it can
independently be shown to be fair, what justification is there for pur-
suing these aims by imposing liability on those who are not at fault?
After all, in criminal law we think it is only those who have consciously
made a wrong choice whom we can properly treat, within limits, as fair
game: as instruments of general deterrence, who need not be regarded
solely as ends in themselves but in part as means to enhance the secu-
rity of others.16 Why should it be different in civil law? It is true that
the sanction in civil law is to pay money, not go to prison. The stigma
is less. But we do not think that in criminal law the innocent who lack
mens rea should be fined rather than imprisoned. We think they should
be acquitted.
Another pragmatic argument for the objective standard concerns
proof. It is hard to prove that the person concerned has not done as
much to avoid harming himself or others as he could have done.
Failure to meet the objective standard is easier to prove, since we have
by experience an intuitive idea of how to apply that standard. We
reckon to know how a reasonable motorist drives a car but, unless we
15. Harper, James and Gray, above n.3, vol. III p.104; H. and L. Mazeaud and A. Tunc,
Responsabilité Civile (6th ed. 1965) vol. I ss.418f, distinguish social and moral from
legal fault, the latter including cases that I would regard as coming within strict lia-
bility, since the mark of fault is that it reflects adversely on the character or disposi-
tion of the agent.
16. There are of course those who deny that to treat the guilty as means in this way is
morally permissible, or who advance (unconvincing) arguments to the effect that
punishment does not involve this.
19
know his driving very well, find it almost impossible to fix the stan-
dard of which a particular motorist is capable.
It is true that the objective standard makes proof easier and so saves
trouble and cost. In applying any normative standard it is convenient
to be guided by conventional practices, though these are clearly not
immune from criticism. Practice may be slack, as in the case of those
pilots who allegedly navigate car ferries without verifying that the
doors are closed. But often practice serves to settle, at least provision-
ally, the content of the objective standard.17
The difficulty of proof has some weight in support of the objective
standard, but by itself hardly provides a compelling reason for it.
Moreover, in criminal law the difficulty of proving intention, often
considerable, does not let in proof of a standard or model intention
instead of the actual intention of the defendant.
Another argument for objective liability is that any unfairness there
may be in holding the shortcomer to an objective standard is minor
in extent. Civil liability is often covered by third-party insurance, for
example, motor insurance. A person who is insured suffers less than
if he had to pay from his own pocket. So, it is said, in these cases the
objective standard does not impose much hardship. But the objective
standard existed in the ancient world and continued in force for
many centuries before third-party insurance became a widespread,
though certainly not universal, practice. Has the justification for it
sprung into being in the industrial age? Moreover, the insurance argu-
ment, which applies only to those walks of life in which insurance is
the standard or required practice, merely displaces the question. Why
should a person who commits no fault be burdened with insurance
premiums?
The various consequentialist arguments of social and legal policy
so far adduced seem not to provide an adequate reason for imposing
objective liability in the absence of a moral justification of a non-con-
sequentialist sort. The arguments they provide are supplementary
rather than basic. What is in issue is the justice of objective liability,
not merely its expediency. Is it fair to the individuals concerned to
hold them responsible for their shortcomings? The good conse-
quences may be necessary, but hardly sufficient reasons for doing so.
17. For a similar point in regard to the standard for children, see the good discussion in
Tyler v. Weed 285 Mich. 460, 280 NW 827 (1938).
20
What justifications of this more ambitious sort are on offer?

Perhaps the Roman jurist Gaius was the first to rationalise the objec-
tive theory of unintentional fault, a theory which is an invention of
law, not philosophy.18 He starts from the Case of the Weak Muleteer:
“A muleteer who lacks the skill to restrain his mules, so that they crush a
slave, is ordinarily said to be liable for fault. The same applies if he is too
weak to restrain them. Nor does it appear unjust to count infirmity as
fault, since no one should undertake a task when he knows or should
know that his infirmity will make its execution dangerous to others”.19
To rebut the argument that the muleteer whose lack of skill or
strength causes harm is not at fault, since, given his infirmity, he
could not avoid the harm, Gaius argues that the muleteer was at fault
at an earlier stage.20 It was this prior fault, he implies, which consisted
in undertaking a task for which he was not equipped, that made the
later accident inevitable. It was a fault because the muleteer knew or
should have known of his own incompetence before he took the
mules on the road.
Is this argument persuasive? If the muleteer knew he was incom-
petent fault was indeed, by our ordinary criteria of judgement, pre-
sent. He knowingly courted danger for himself and others. But if he
did not appreciate his own incompetence, what is the force of the
assertion that he “should have known” of it? Perhaps he was too
obtuse to realise his own infirmity, and no one pointed it out. In his
own eyes he was an ace muleteer. If he then has to pay for the acci-
dent he is paying for his defective makeup, not for fault.
Like Gaius, we can generalise the muleteer case. It is not true that
a person who undertakes a task for which he is not up to scratch need
be at fault in failing to realise the fact. Lack of skill often goes hand
in hand with lack of the nous to recognise one’s incompetence. Prior
fault can serve to justify only some of the cases in which legal systems
impose liability for lack of competence. Can some other theory jus-
tify the rest?
18. D. Daube, Roman Law: Linguistic, Social and Philosophical Aspects (1969) pp.131,
151f. It is true that Aristotle set up the model of a zealous person (spoudaios), and pro-
vided a theory of how people are ultimately responsible for their characters. But to
expound and seek to justify the objective standard is to take a further step.
19. Digest 9.2.8.1 (Gaius 7 ed. prov.). Essentially the same argument appears in Wilsher
v. Essex Health Authority [1987] QB 730, 777C per Browne-Wilkinson VC.
20. A species of introductory or disabling fault (einleitender Fahrlässigkeit): Deutsch,
above n.3, at p.280.
21
One possibility is promising. At times the person whose conduct is

sub-standard has made an express or implied promise to reach the
model standard. The surgeon may put himself forward as expert in
heart by-passes, the lorry driver as capable of driving a snow-plough
in bad weather. It may then be legitimate to infer that they are
promising to perform up to their assumed expertise when they oper-
ate on X or clear Y’s land. But they need not and usually do not
promise to reach the objective standard. Thus, most newly qualified
doctors do not pretend to years of practice. On the whole profes-
sional people allow others to assume in them certain skills, but make
no specific claim, and fight shy of any warranty. Promising is still
more remote from the pure tort situation like a road accident
between parties who had no previous dealings with one another. A
motorist is no less liable for bad driving if he hangs L-plates on his car
or, as is not uncommon, plasters on his rear window a notice draw-
ing attention to his incompetence.
Promising, then, will not do as a general moral basis for objective
liability in negligence. What of justified reliance? It is to this that
Blackstone resorted to explain the doctor’s liability for malpractice:21
“For it hath been solemnly resolved that mala praxis is a great misde-
meanour and offence at common law, whether it be for curiosity and
experiment, or by neglect; because it breaks the trust which the party had
placed in his physician, and tends to the patent’s destruction.”
But not every patient trusts his doctor. Some have doubts, but see no
alternative. Nor do all motorists inspire the confidence of other road
users. A car zigzags down the road towards me. I think the driver
incompetent or drunk. But if by his bad driving he runs me down it
is no defence to my action for damages that I did not think him com-
petent. I rely on the competence of motorists in general, but need not
suppose any particular motorist to be up to scratch in order to hold
him to the objective standard.
In any case the crucial question is not whether in general people
rely on others to meet the objective standard of competence but
whether they are entitled to do so. In law they are, but on what moral
basis? For the objective standard of competence imposes a form of
strict liability on that minority of shortcomers who cannot attain it;
21. W. Blackstone, Commentaries (1st ed. 1769) vol. 3 p.122.

22
and to philosophers and many lawyers strict liability is a stumbling

block. Nagel says that it “may have its legal uses but seems irrational
as a moral position”.22
2. R ESPONSIBILITY FOR O UTCOMES

AND S TRICT L IABILITY
On the contrary, given certain conditions, strict liability can be

morally defended. For present purposes liability is strict when it
attaches to us by virtue of our conduct and its outcome alone, irre-
spective of fault.23 It falls typically on those who pursue permissible
but dangerous activities: storing explosives, running nuclear power
stations, keeping wild animals, marketing drugs or other dangerous
products, and, in France and Germany, driving a car. In common law
systems an employer is strictly liable for the employee’s torts in the
course of employment, though it is also usually necessary to show
that the employee is at fault. In such cases the law does not forbid
people to undertake the activity or make the decision to employ
Jones, because the activity or decision can have value. To forbid it
would be illiberal. But society’s consent is subject to the entrepreneur
taking on himself, within certain limits, the risk that the decision or
activity will misfire.24
The effect of strict liability is therefore to put an activity or deci-
sion at the risk of the agent or decision-maker as regards all but highly
exceptional sequences of events. It does so irrespective of whether the
persons storing the explosives, taking on Jones as foreman, or selling
the dangerous drugs has taken reasonable care to see that what he
does will not cause harm.
One point of imposing strict liability is to dispense with the need
to prove fault on the part of people who are in fact at fault. If the
activity that entails strict liability goes wrong, it generally goes wrong
through someone’s fault. But in some cases strict liability penalises
22. T. Nagel, “Moral Luck”: Mortal Questions (1979) pp.24–38, at p.31.

23. Insurance is not on this definition a form of strict liability, since an insurer is not
liable for the outcome of his own conduct.
24. The risk imposed on the entrepreneur does not normally extend to harm that would
not have occurred in the absence of the dangerous activity or decision but that is
attributable to force majeure, Act of God, and similar “external” causes.
23
not fault but bad luck, which in this context includes, besides ordi-
nary accidents, the bad luck of being saddled with shortcomings.25
The objective standard of liability in negligence has a like dual effect.
Often it merely makes it easy to prove negligence on the part of some-
one who is in fact at fault. But sometimes it penalizes the bad luck of
those who suffer from shortcomings. The principle involved in
imposing ordinary strict liability, say for storing explosives, and in
applying the objective standard of negligence is at bottom the same.
Most of those held liable will be at fault but a minority will not. Thus
in Britain a motorist who has an accident because, though he scraped
through the test, he is too clumsy or stupid to drive properly can suf-
fer, directly or through his insurer, for his bad luck. So under a dif-
ferent legal theory can his opposite number in France or Germany
who is held liable for a road accident even though no fault on his part
can be proved.
Holmes, in a classic passage of The Common Law, points to this
combination of bad luck and blame. He says,
“If a man is born hasty and awkward, is always having accidents and hurt-
ing himself or his neighbours, no doubt his congenital defects will be
allowed for in the courts of Heaven, but his slips are no less troublesome
to his neighbours than if they sprang from guilty neglect.”
Hence
“the law considers . . . what would be blameworthy in the average man,
the man of ordinary intelligence and prudence, and determines liability
by that. If we fall below the level in those gifts, it is our misfortune.”
Holmes gives no morally convincing reason for holding the acci-
dent-prone man liable for his bad luck. The member of the awkward
squad is not to blame for his defect. In the courts of Heaven, accord-
ingly, he goes scot-free, but on earth his neighbours insist that he be
held liable. True, but by what right? It is certainly bad luck to fall
below a decent level in the gifts needed for social intercourse, but why
should this misfortune entail legal liability?
Any principle that can justify responsibility for bad luck must be
fair. If it is to be fair, it must entail that when we bear the risk of bad
luck we also benefit if our luck is good. Allocation according to luck
25. Including “faults” of character and other shortcomings, which, viewed from a certain
point of view, are the outcome of a cosmic lottery.
24
must cut both ways. So the system will be fair only if there are situa-
tions in which we implicitly bet on the outcomes of our actions.
Do such situations exist? Imagine that when we reach a decision to
do X rather than Y – let us say to attempt a U-turn rather than to go
on to the next roundabout – we are choosing to put our money on X
and its outcome rather than Y and its outcome. When we opt for the
U-turn rather than the roundabout, we implicitly bet that we will get
to our destination quicker by making the U-turn. Our decision for
U-turn rather than roundabout will be like a decision to put money
on L’Escargot rather than Red Rum to win the Grand National. But
we will not be like ordinary punters but rather jockeys who, contrary
to existing practice, are allowed to back a horse and ride on it so that
they can influence the result of their bet. Thus, when we choose X
(say the U-turn), the bet we make is to be analysed as follows. We bet
we can do X (the U-turn) and that X will have the more favourable
outcome (getting there quicker). In calculating the odds for achiev-
ing the favourable outcome we have to discount the chance that we
may not be able to do X or that the outcome of X, if we do it, will not
be what we predict. Thus, we may not manage the U-turn; we may
instead cause an accident. Or, we may manage it but find we were
misinformed about the route, so that it would have been quicker to
go on to the roundabout anyhow.
One difference between an implicit bet on outcomes and an ordi-
nary wager concerns the stake and the winnings. In an ordinary bet
we know the amount of the stake and often the potential winnings in
advance. In implicit bets on the outcome of our actions, on the other
hand, we do not precisely know the stake and winnings in advance;
only that they will be proportionate to the outcome. The terms of the
bets we make with other members of our community (and indirectly
with ourselves) when we choose X rather than Y is that if we succeed
and have guessed the outcome right we receive credit for it. If we
manage the U-turn and get to our destination quicker we get credit
for that success. But if we botch it, have an accident, or mistake the
route, that is chalked up against us. This remains true even if the
botch or miscalculation is not our fault, though of course it generally
is. How much responsibility in terms of credit or debit accrues to us
– how big the stake and winnings are – depends on how important
the successful or botched outcome is in the eyes of others. We cannot
tell the precise amounts in advance, but usually have a rough idea
25
what they are likely to be. But whether the outcome is foreseeable or
not, we can of course easily miscalculate. Despite the uncertainties,
over a span of time more outcomes are likely to redound to our credit
than to our debit, so that we are not permanently saddled with a los-
ing ticket. Risk and benefit go hand in hand. Ubi emolumentum ibi
onus.26
The betting system that I have outlined seems to me to correspond
to our experience of everyday life. Our actions impinge on others,
who resent it if the effects on them are harmful. So we are, unless
indifferent to the reaction of others, forced to gamble in order to
acquire credit and esteem and avoid exposure to discredit and resent-
ment. Even by opting for inaction we cannot avoid putting out
money on the table, because, as members of a community, we know
that we will be held responsible for whatever good or ill we cause,
including whatever harm our inaction produces in those cases where
omissions count as causes. To choose and execute a course of conduct
is to bet on your skill and judgement of the probabilities. Choosing
is inescapably betting.
If this suggestion is correct, we live under a system by which a com-
munity allocates responsibility according to outcomes, and we are
consequently forced to make bets on those outcomes. It is, I believe,
no exaggeration to say that this dual system colours, often uncon-
sciously, everything we do. Not only are actions and outcomes con-
ventionally allocated to people but we and others are entitled to insist
that they should be so allocated. Any moral or legal theory must
therefore be condemned as inadequate unless it justifies the allocation
of responsibility according to outcomes or shows that the system rests
on a mistake.
Given certain conditions outcome allocation can be defended as fair.
The necessary conditions are that the system must in its operation be
impartial, reciprocal and over a period, beneficial. It must apply impar-
tially to all those who possess a minimum capacity for reasoned choice
and action. It must be reciprocal in that each such person is entitled to
apply it to others and they to him. It must work so as to entitle each
person to potential benefits that are likely on the whole to outweigh the
detriments to which it subjects him. This makes it unfair to apply the
system to the incapable, for whom there is no likely surplus of benefit
26. Justinian, Institutes 1.17; Digest 17.2.55; 50.17.10.

26
over detriment. But for the capable the three conditions are normally
satisfied. All those who possess a minimum capacity stand to profit
from the system of outcome allocation most of the time and, if there is
a minority of permanent losers, they teeter on the edge of incapacity.27
The system of outcome allocation and outcome responsibility,
which it entails, can therefore in certain conditions be defended as
fair. Outcome responsibility automatically settles the bets made by
people who by choosing a course of action implicitly put their money
on the result. In any event outcome responsibility is, I contend, the
basic type of responsibility in a community: more fundamental than
either moral responsibility as generally understood, which requires
fault, or legal responsibility, which requires either fault or special
danger.
The analogy with a form of gambling in which we mostly win but
sometime lose helps to explain and in part to justify not merely out-
come responsibility but strict liability. In certain areas of life where
there is a special risk that what we do will have a harmful outcome,
society insists on pressing our responsibility for outcomes to its lim-
its. Our fellow citizens insist that we pay up in terms of compensa-
tion rather than just apologise or ring for the ambulance or comfort
the injured.
The justification of strict liability therefore depends in part on the
fairness of outcome responsibility. Strict liability is one species of
enhanced responsibility for outcomes. This does not entail that
whenever a harmful outcome is properly allocated to someone, this
justifies imposing on him a strict liability to compensate for that out-
come. Contrary to what Epstein28 and possibly Fletcher29 at times
seem to suggest, responsibility for a harmful outcome should not
automatically involve a legal duty to compensate. An extra element is
needed to ground the legal sanction. Sometimes the extra element is
fault. Fault discloses a hostile or uncooperative disposition on the
part of the harm doer, who either positively intends harm or disre-
gards a known risk of it. For strict liability, the extra element is usu-
ally that the conduct of the harm doer carries with it a special risk of
harm of the sort that has in fact come about. In that event, a conse-
27. We need to rethink the way in which to apply notions of responsibility to them, tak-
ing due account of their need to be seen as genuine persons.
28. R. Epstein, “A theory of strict liability”: 2 Journal of Legal Studies (1973) 151.
29. G.P. Fletcher, “Fairness and utility in tort theory”, 85 Harvard Law Rev. (1972) 537.
27
quentialist argument for avoiding serious harm reinforces the non-

consequentialist arguments for imposing outcome responsibility.
Legal theorists tend to ignore this combination because they view
strict liability in isolation. They think of it as a burden imposed on a
person who, not being provably at fault, should not morally be held
responsible for the harm resulting from his conduct. But this is to
overlook that outside the law people of full capacity stand to win
more than they lose from the system of outcome allocation. We gen-
erally succeed in doing what we set about doing and get credit for it;
only occasionally do we lose out. Of course the law of strict liability
is solely concerned with the debit side of the account. It makes those
who have incurred discredit pay compensation, even without fault.
Yet in a wider perspective, strict liability when the shortcomer is not
at fault, must be weighed not merely against a preponderance of suc-
cessful outcomes, but against its obverse: the occasions on which by
good luck our ill-judged actions have a happy outcome and we unde-
servedly escape discredit or even manage to obtain unmerited credit.
Viewed in this light, strict liability merely serves to surcharge on
grounds of social policy the debit side of an account that is in most
instances comfortably in credit.
Though strict liability attaches when we choose a course of action
to which special risks attach, that does not make it a species of fault
liability. The person who chooses to pursue a dangerous activity may
have good reasons for doing so; his conduct may in no way evince an
objectionable attitude to others. He may not, when he decides to
store explosives or employ Jones as foreman, appreciate the risks he is
running. He may think the chances of fire negligible and Jones thor-
oughly competent. Does the fact that he may be strictly liable for a
choice made in partial ignorance destroy the moral argument in
favour of strict liability for risky activities? That depends.
Presumably a person should not be strictly liable if he did not
understand the system by which in a society (1) people are held
accountable for the good or bad outcomes of their actions and (2) a
heavier potential liability attaches to dangerous activities than to
relatively safe ones. Strict liability, like fault liability, should require
the appropriate capacity. But there is a distinction between not
understanding the system and not appreciating the risk involved
in a particular choice, for example to employ Jones or store explo-
sives.
28
Strict liability and fault liability run parallel in this respect. Those
who do not understand outcome allocation and the difference
between right and wrong (or lawful and unlawful) are generally held
incapable of fault, but in order to be guilty of fault they need not
know that a particular action is wrongful. On this analysis, virtually
everyone who possesses the general capacity needed for outcome
responsibility also possesses the capacity to be legally responsible both
on the basis of fault and strict liability.
A further point that bears on the fairness of outcome allocation,
and so of strict liability, is that neither is something that the person
subject to them has chosen. Does that make these institutions
morally dubious? We have never made a social contract providing for
outcome allocation, still less for strict liability. They bind us willy-
nilly.
It is true that we have no choice in the matter. But outcome allo-
cation can be defended on grounds deeper than the overall balance of
benefit over burden; and so, in its wake, can strict liability. For out-
come allocation is crucial to our identity as persons; and, unless we
were persons who possessed an identity, the question of whether it
was fair to subject us to responsibility could not arise. If actions and
outcomes were not ascribed to us on the basis of our bodily move-
ments and their mental accompaniments, we could have no continu-
ing history or character.30 There would indeed be bodies and,
associated with them, minds. Each would posses a certain continuity.
They could be labelled A, B, C. But having decided nothing and
done nothing these entities would hardly be people.
In the real world, fortunately, human bodily movements and their
mental accompaniments are with some exceptions interpreted as
actions and decisions. They are ascribed to authors, who accordingly
count as persons; and it is by virtue of these ascriptions that each of
us has a history, an identity and a character. But there is a price to be
paid for being a person. As the counterpart of this status we are
responsible for our actions and their consequences, and sometimes
this responsibility exposes us to legal sanctions. To ascribe person-
hood and responsibility to people in this way is to apply normative
principles. It is not merely that others attribute to us an identity and
a character, but that we are entitled to claim them for ourselves and
30. H.L.A. Hart and T. Honoré, Causation in the Law (2nd ed. 1985) lxxx–lxxxi.
29
to ascribe them to others. Others in turn not only hold us responsible

for our actions and their outcomes, but are entitled to do so. Of
course the balance between personhood and responsibility cannot,
any more than the system of outcome allocation, be said to rest on a
social contract. We have never decided to assume responsibility in
exchange for the gift of personal identity. Both are natural in the sense
that we can neither choose them nor give them up.31 Considered as
a bargain, the exchange would not even, properly speaking, be in our
interest; for to be responsible is part of what it means to be a person
and hence to have interests. But the normative principles involved
may be regarded in a pre-moral sense as well-founded, since they
embody a balance between identity and responsibility.
Such are the normative arguments for allocation according to out-
come and, as a corollary, for strict legal liability for the harmful
upshot of risky conduct. In practice most ordinary people endorse the
former and most lawyers the latter, though either might be hard put
to say why. Virtually no one inside or outside the law believes that
fault and desert are the sole basis of responsibility.32 In their off-duty
moments even those philosophers and theologians who in theory
cleave to fault alone assign credit and discredit for actions and their
outcomes in cases where blame and praise are not in point. Take a
non-moral example: the contrasting fortunes of X and Y, two foot-
ballers playing in a needle match. X miskicks but a gust of wind car-
ries the ball into the opposing goal. He is credited with a goal, but not
praised for scoring it. It would be better, of course, had he been skil-
ful as well as lucky, for he would then both be credited with the goal
and praised for scoring it. Y aims a skilful shot at goal, but this time
a gust of wind diverts it. He is praised for his good shot but not cred-
ited with a goal. It would be still worse for him had his shot been a
bad one. X is lucky, Y unlucky; but it is the outcome of their actions,
not what they deserve, that primarily determines credit or its absence.
Desert merely increases or diminishes credit or discredit. Take a legal
example. I fire at my rival intending to kill him. It would be murder
if I succeeded, but I miss. I am guilty only of an attempt. If fault is to
be judged by disposition my fault is as great as if I had hit him, but
my responsibility is less. Now for an extra-legal example. If purely by
31. Cf. P. Strawson, 48 Proceedings of the British Academy (1962) 1, 24 (on induction).
32. “Forgive us our sins both voluntary and involuntary” (Orthodox liturgy).
30
your fault in darting out into the road I run you over, I must stop,
send for the ambulance and give you what help I can in the mean-
time. My responsibility is not as great as if I had been at fault. It may
not be legal: that depends on the applicable system of law. It may not
be moral in the sense that I am morally responsible for the accident
itself. But, just because I have hurt you, I am responsible, and by
virtue of that responsibility bound to take certain steps. Indeed,
unless I am wholly insensitive, I shall feel and express regret for the
harm I have done. For it is a myth that fault and desert are essential
to responsibility. They serve rather to increase the credit or discredit
for the outcome of our behaviour that we incur in any event.
It is only this primary outcome responsibility that can explain why
we (rightly) judge murder more severely than attempted murder and
causing death by dangerous driving more severely than dangerous
driving. It is said that morally the harmful outcome makes no differ-
ence; and indeed the difference between causing death by dangerous
driving and mere dangerous driving, like the difference between aim-
ing a good shot at goal and scoring a goal, is causal, a matter of out-
comes. On a narrow view of morality the cases are not morally
distinguishable. For allocation according to outcomes is not alloca-
tion to according to effort, talent or disposition. A good outcome can
sometimes be achieved with less effort than a bad outcome, and by a
person with less talent and a worse character. Outcome allocation is
allocation according to results, whether they constitute achievements
or botches. But it does not follow that the system of allocation
according to result, in contrast with its application to individual
instances, lacks a moral or pre-moral basis. The person concerned,
though he cannot be sure what the outcome of his action will be, has
chosen to act in the knowledge that he will be credited or debited
with whatever it turns out to be. Moreover, we cannot opt out of the
system by which we obtain credit for favourable outcomes; and so we
cannot slough off the burden of discredit either. Finally, it is out-
comes that in the long run make us what we are.
31
3. C APACITY AND F REEDOM
It seems possible, then, to justify both outcome responsibility and the

liability, strict or fault based, that legal systems superimpose on it, if
those held responsible have the proper capacity and make the relevant
choice. To be outcome responsible they must understand both the
system of allocation according to outcomes and how in practice to
use causal notions to settle what counts as the outcome of an action.
The relevant choice is to choose to act knowing that the outcome of
the action will be attributed to the agent.
To be responsible for fault, morally or legally, the person concerned
must in addition understand right and wrong (or lawful and unlaw-
ful) and the system of allocation based on that. The relevant choice is
the decision to act, in the knowledge of how that distinction is
employed, either with a view to doing harm or in disregard of a
known risk of harm. To have the capacity for strict liability the per-
son must understand that a special responsibility attaches to socially
dangerous activities. The relevant choice is the decision in the light of
that knowledge to embark on the activity in question.
Thus explained, the two forms of legal liability are seen to be
species of the genus outcome responsibility. The main difference
between them is that when someone is at fault he should by standards
known to him have acted differently, whereas when he is held to strict
liability this need not be the case. But the fact that the person held
liable has violated a known standard, though it may increase his
responsibility, does not create it.
But is it not a condition of every form of responsibility that the per-
son responsible could have acted differently? The capacities so far
listed have been cognitive: the ability to understand cause and conse-
quence, right and wrong, the notion of a dangerous activity, and the
systems of allocating responsibility based on these factors. Since choice
is admittedly a condition of responsibility, does it not follow that the
person held responsible must also possess the cognitive capacity to
choose and to control his conduct in accordance with his choice?
The capacity to reach decisions and act on them must on any view
be a condition of responsibility. But how are we to fix what precise
capacity is needed and test its presence on a given occasion?
Intuitively we think of ourselves, in one mood, as responsible for an
32
outcome only when we could have done something different that

would have brought about a different outcome. Philosophers and
lawyers have without great success debated how to analyse this capac-
ity.33 In a different mood, however, we are prepared to bypass the
supposed capacity to do otherwise. In that alternative mood we think
of ourselves as responsible even when our conduct is determined by
internal factors such as limited intelligence or bad temper that are,
either in the short or long run, beyond our control. These internal
factors, unlike external constraints, do not then seem to relieve us of
responsibility. Yet we cannot, at any rate when in their grip, do other-
wise. So we catch ourselves believing, perhaps not at one and the
same moment, that the ability to do otherwise both is and is not
a condition of responsibility.34 That is unacceptable. If we find
ourselves embracing contradictory beliefs, it is time to abandon or
modify one of them.
The need to do this, and a hint of the best way to set about it,
emerges if we reflect on how lawyers have tried to solve puzzles about
the capacity of children. What standard of care does the law of negli-
gence expect of a child who crosses a busy street or plays a game with
a stick? It seems attractive to opt, not for an adult standard, nor a uni-
form standard for all children, but for whatever standard the particu-
lar child is capable of. But how can we discover what this is? How are
we to judge, say, the capacity of a child of eight who knows that he
must take care not to put his companion’s eye out with a stick, but is
in the excitement of the game bad at keeping the danger in mind?
Was he at that moment capable of thinking about his playmate’s eye
and acting accordingly? Even supposing that in the excitement of the
game the average child of eight remains careful to avoid poking his
friend’s eye out, perhaps this particular child is immature for his age.
So in setting the standard for children, some jurisdictions lay down
that account must be taken not only of the child’s age but of his expe-
rience and maturity.35 Assume, then, that most children of the expe-
rience and maturity of this child would have avoided putting their
friend’s eye out. That still does not show that this child could have
33. E.g. B. Aune, 27 Analysis (1967) pp.191–5; K. Lehrer, 29 Analysis (1968) pp.29–32.
34. The oscillation is forcefully brought out by T. Nagel, The View from Nowhere (1986)
ch.7.
35. E.g. Mack v. Davis 76 Ill. App 2d 88, 221 NE 2d 121,126 (1966): age, experience
and intelligence.
33
avoided the accident if “could have avoided” refers to a capacity exer-

cisable in all the conditions of the incident. For this child, relatively
mature though he was, may have been distracted at the crucial
moment by, say, a sudden shout, that would not have distracted an
average child of the same maturity. Do we then go on to ask whether
the average child of the same age, experience and maturity who was
distracted by a shout would have put his playmate’s eye out? Nor is
there any good reason to stop there. We can go on to add detail upon
detail. But as the required standard moves asymptotically towards
that of the particular child in all the concrete conditions of the case,
it ceases to be a standard and becomes an ever more detailed descrip-
tion of the untoward event.
What this example rawly exposes is that the question whether the
child was “capable” of doing otherwise, and so was responsible, is not
a plain question of fact. Knowing more about the child does not set-
tle the answer. On the contrary, what we need to know in order to
determine responsibility is whether a child of a certain degree of, say,
maturity is expected to behave with that degree of maturity, not in the
sense that he will predictably behave in that way on a given occasion,
but that he takes the risk of behaviour that displays a lesser degree of
maturity.36 And while it is no accident that the word “expected” lends
itself to both a descriptive and a normative interpretation, the ques-
tion “what is to be expected of him?” is meant to combine elements
of both. The usual behaviour of the child or similar children bears on
the answer but does not determine it. A child who in general pos-
sesses a certain degree of maturity does not on every occasion display
that maturity; and it is a gratuitous assumption that he could do so.
Hence if a child’s general level of maturity determines what law or
morality requires of him, that is because we think it fair to judge him
by the standard that he usually manages to attain. But even if we
adopt this so-called “subjective” standard, really intermediate
between the objective and the subjective, we still impose on the child
the risk that in a particular instance he may not come up to it and
may in the special conditions of the case be unable to do so.
Does it follow that we have to abandon “capacity to act otherwise”
as a condition of responsibility? No, for we can give it a more plausi-
36. Esser-Schmidt, Schuldrecht (6th ed. 1984) vol. I p.377: the object of the law of oblig-
ations is to divide zones of risk from one another.
34
ble interpretation. The capacity to act otherwise has hitherto been

taken to be a capacity exercisable in the particular circumstances in
which the person concerned finds himself. Legal statements of the objec-
tive standard of care, for example, all refer to the care to be expected
of a model person in the circumstances. None set a standard to be
complied with regardless.
The word “circumstance” invites a distinction between external
and internal factors.37 The most obvious circumstances are things
outside us, such as place, time, weather, noise, social pressure. The
person involved (let us call him or her the agent) must take account
of them in deciding what to do but is not expected to change them.
He is not responsible for them, for they are not part of his being, but
part of the scene on which he has to act. The most obvious factors
that are not circumstances are things that are inside the agent or part
of his makeup, like being stupid or in a bad temper. The person con-
cerned must either overcome these or face the consequences that
ensue. It follows that what counts as a circumstance is often not
determined by crude geography, inside or outside. It can be a nor-
mative question: is it good policy to treat such-and-such a factor, let
us say inexperience, as something for which the person concerned is
responsible? Thus, there is a case for treating some aspects of a per-
son’s physical makeup, though in one sense internal to him, as cir-
cumstances of which he must take account but which he cannot be
expected to alter. Someone who is too short to see over a wall must
take account of his disability by walking round the side or, if that is
not possible, fetching a ladder.38 He is not to be treated as if he could
see over the wall unaided.
But the process of externalising the elements that go into a person’s
makeup cannot be carried beyond a certain point without obliterat-
ing the distinction between internal factors for which a person is
responsible and circumstances for which he is not. The boundary can
shift but it cannot be wiped out without eliminating the person along
with the responsibility.39 For given the totality of internal factors –
the person’s makeup and his physical, mental and emotional state at
37. H. and L. Mazeaud and A. Tunc, above n.15, vol. 1 s.431.

38. Cf. Mahan v. State to Use of Carr, 172 Md. 373, 191 Atl. 575, 579–80 (1937); and
for deafness Otterbeck v. Lamb 85 Nev. 456, 456 Pac. 2d 855 (1969).
39. “Dans la vie sociale on répond de son être”: Mazeaud and Tunc, above n.15, vol. I
p.490.
35
the time and place in question – it is implausible to suppose that on

a given occasion he can act otherwise than he does. If I am in a filthy
temper I cannot at the same time conform to a model of calm delib-
eration. If to be responsible the person concerned must have been
able to act otherwise in the circumstances, then not all internal fac-
tors can count as circumstances.
Legal systems recognise this. Though they vary in what they regard
as internal factors in relation to which the actor is at risk, all agree that
he should be saddled with some such factors that are not, therefore,
to be accounted circumstances. Thus, everyone must bear the risk of
his own bad temper or irritability; but it is disputed whether inexpe-
rience is to be treated as a circumstance rather than a deficiency for
which the inexperienced person must take the rap.40 If a surgeon is
sued for negligence is it relevant that he is newly qualified and has not
performed the operation before? The issue is a normative one on
which opinions legitimately differ.41
Though there is often doubt where best to draw the line between
internal factors and circumstances, both our everyday judgements
and the law of negligence assume that it must be drawn somewhere.
A driver must drive reasonably in the circumstances, where circum-
stances include the slippery road and the dim light but not fatigue or
a lapse of concentration. But why is a lapse of concentration not
accounted a circumstance? Could one not argue that a driver whose
concentration lapses need only conform to the model of a driver,
otherwise competent, whose concentration has lapsed? Why is the
unfortunate driver blamed for his lapse and held to be negligent in
law? Every driver has some lapses of concentration.
To put it generally, in no activity or walk of life can people consis-
tently maintain the high standard of skill and care required by law
without variation. Indeed that is to understate the matter. Empirical
studies show, for example, a high rate of error on the part of even
competent drivers.42 It is not only the stupid and clumsy who are
40. Above n.2.

41. On the view taken here it is not unfair to hold the inexperienced surgeon, etc., to the
standard of an experienced surgeon. He can in the normal course of events look for-
ward to a much longer period of professional practice, after he has gained the neces-
sary experience, than the relatively short period spent acquiring it, so that, taking his
professional practice as a whole, he is likely to be an overall winner.
42. U.S. Department of Transportation: Automobile Insurance and Compensation Study
1970 pp.177–8: “In Washington D.C. a ‘good’ driver viz. one without an accident
36
incapable of meeting the objective standard of care on the road. The

intelligent and adroit cannot consistently meet it either. So if a vigi-
lant and skilful driver has a lapse of attention from which an accident
results, and we say that, unlike the driver with slow reactions, he
could have swerved in time, our ground for saying this cannot be a
supposed capacity to remain alert every moment that he is at the
wheel. We must instead be thinking of his general capacity as a dri-
ver. To ascribe responsibility to a competent driver whose attention
has lapsed we need not believe that given all the elements internal and
external of the concrete situation he could, despite his lapse, have
swerved and avoided the accident.
If then it is a condition of responsibility that the person to be held
responsible could have acted differently in the circumstances, circum-
stances must be distinguished from those internal factors for which the
agent is on normative grounds to be held at risk. We then have a choice.
Either we abandon the notion that responsibility depends on the abil-
ity to do otherwise in the circumstances. Or we retain the notion, but
construe the required capacity as a general ability to perform the sort of
action that would in the instant case have led to a different outcome.
This general ability need not have been exercisable in all the concrete
conditions, external and internal, of the case.
The second view is preferable since it enables us to retain in mod-
ified form our common sense belief in the importance of capacity to
act otherwise as an element in responsibility. On this view the capac-
ity to remain alert, when alertness would have avoided the accident,
does not refer to the possibility of someone’s remaining in a steady
state of alertness for an indefinite period but rather to an ability to
remain alert in normal conditions most of the time. So, though we
are responsible for lack of alertness, it is not a condition of our
responsibility that we should be able to exercise our capacity to
remain alert on every occasion when it is sought to hold us respons-
ible. To construe capacity in this way as general capacity is not to
embrace determinism as a world view, but to make the point that, if
law and morals require for responsibility an invariant capacity that
can be exercised on every occasion, no one will consistently be
responsible for his conduct.
within the preceding five years commits on average, in five minutes of driving, at least
nine errors of different kinds.”
37
That is not an acceptable conclusion. Moreover, we are not driven

to it. We can take the view that, even if human conduct is determined
by the factors, internal and external, present on a given occasion –
which we have at present no way of deciding – an agent can properly
be judged on the basis of his general capacity.
The analysis of capacity here advocated is not new. I put it forward
more than twenty years ago.43 The theme of my paper, in the idiom
of the age, was the distinction between “can general” and “can par-
ticular”. A golfer can (general) hole a six-foot putt if he possesses the
general capacity to do so; and he possesses that capacity if it is usually
the case that when he tries he succeeds. It is compatible with his pos-
sessing this general capacity that on a given occasion he cannot (par-
ticular) hole a putt of six feet or less. If he tries to hole a particular
putt and does not succeed that shows that he could not (particular)
do so. It will nevertheless be true that he could (general) have holed
it, since he possessed the resources physical and mental to succeed.
But internal factors, such as fatigue or lack of concentration, may
have prevented him exercising his capacities on this occasion.
Though reflected in verbal usage, the point is normative rather
than linguistic. The reason for allotting responsibility to people on
the basis of their general capacity is that they then stand to win most
of the time, because it is true by definition that, when they try, they
usually perform up to their ability. It is for this reason that in many
contexts we rightly insist in both law and morals on fault as a condi-
tion of responsibility. The fault system penalises those whose conduct
displays a bad disposition, but at the same time affords people fair
opportunities of escaping sanctions for botches and bad outcomes
most of the time. It provides an incentive, often internal as well as
external, to take advantage of these opportunities. Therein lies its jus-
tification, not in the unfounded assumption that, given the external
and internal constraints present on a given occasion, the person at
fault could (particular) have done otherwise than he did.
Thus to judge people according to their general capacity neither
rules out nor requires determinism. General capacities can be mea-
sured by how people generally perform when they try to execute a
given type of action, like shutting the door or crossing the street or
43. “Can and Can’t”, 73 Mind (1964) 463–479, reprinted, below p.144. Endorsed by
D. Dennett, Elbow Room (1984) pp.147–8.
38
holing a six-foot putt. Whether their choices and performances are

determined by pre-existing factors has no obvious bearing on the fair-
ness of judging them on this basis. But such judgements sit most
comfortably with the view that, while strict determinism may be
false, a person’s character, aims and general capacity plus the circum-
stances of which he is aware normally determine his choice; his choice
normally determines his conduct; and his conduct plus the circum-
stances in which he acts normally determine the outcome. If we are
to be responsible for the outcomes of our actions, we are likely to find
these loosely framed hypotheses reassuring rather than alarming.
How else would we be able to act in accordance with our character
and general capacities or to achieve our aims, including the reflexive
aim of changing, to a limited extent, our own character and capaci-
ties?
Even if our conduct is incompletely determined, the freedom we
value is independent of this indeterminacy. Freedom is, in this con-
text, a notion partly descriptive and partly normative. External or
(better) circumstantial freedom depends on the absence of oppressive
constraints that result in our conduct failing to reflect our character
and aims. Conduct in the face of such oppressive constraints is not
rated as fully voluntary, and our responsibility for it is correspond-
ingly diminished. That we are unfree and so not the true authors of
actions undertaken in conditions of oppressive constraint is plainly a
normative notion, though it presupposes a certain descriptive back-
ground. It is the same with the internal aspects of freedom. Our inter-
nal freedom is circumscribed by our general capacities for
understanding and action. That people who possess adequate general
capacities are the authors and originating cause of their conduct and
its outcome is, once again, a normative conception that presupposes
a certain descriptive background. We are, consistently with this
notion, specially free when the conduct and its outcome lies within
our specific capacities, so that we can plan to bring about the out-
come with some confidence. Actions undertaken in these conditions
afford the prime example of fully voluntary conduct.
39
4. C ONCLUSION
The argument has been complex but its core is simple. Our respon-
sibility for what we do and for its outcome is inseparable from our
status as persons. We cannot disclaim outcome responsibility without
undermining that status; and outcome responsibility is therefore
more fundamental than moral and legal responsibility, which are
species of it. But the fairness of holding someone responsible outside
or inside the law depends on their possessing a general capacity for
decision and action such that, under the system of bets into which
society forces them, they stand over a span of time to win more than
they lose. The system is not a fair one to apply to those whose limited
capacities make them consistent losers. Given a sufficient general
capacity, however, it involves no great extension of principle for the
law to impose strict liability for risky activities alongside fault liabil-
ity for conduct that discloses an uncooperative disposition. To bear
the risk of bad luck is inseparable from being a choosing person.
40
!5"
NECESSARY AND SUFFICIENT
CONDITIONS IN TORT LAW 1
Argument about causation inside and outside the law is often con-
cerned with the question: must a cause be a necessary condition, or a
sufficient condition of a result, or must it be a necessary element in a set
of conditions jointly sufficient to produce the result? This essay sup-
ports the third view both outside the law and inside it, whenever a
sequence of physical events is in issue. A different but related idea can
be used to explain reasons for human action, “causing” or inducing
people to act rather than causing things to happen. There is, I con-
tend, no special legal meaning of causation and the related concepts
such as inducement.
On the other hand, law alone determines when causal connection
must be proved and when it can be dispensed with in legal contexts.
When causal connection must be proved, law also settles what must
be shown to have caused what. This is specially important in tort law.
Tort law generally imposes fault liability on people who by their
wrongful conduct cause harm to others; but sometimes it imposes
strict liability on people who engage in activities which, though not
wrongful, cause others harm. The causal inquiry takes a different
form according as fault or strict liability is in issue.
Tort lawyers have traditionally held the view that, whatever the
meaning of causal connection, the way to test whether it exists in a
given case is to ask whether in the circumstances the harmful result
would have occurred in the absence of the wrongful act. This is to
adopt the but-for test, by which causal connection exists between
condition and consequence whenever, but for the condition, the con-
sequence would not in the circumstances have occurred. The but-for
1. First published in Philosophical Foundations of Tort Law (ed. David G. Owen, Oxford 1995)
pp.363–85.
94
NECESSARY AND SUFFICIENT CONDITIONS IN TORT LAW
test is also called the test of strong necessity.2 It often gives what is
intuitively the right answer, but sometimes fails to do so. This test is
meant to exclude those factors that had no impact on a particular
course of events. But sometimes it cuts out factors that did have an
impact. It serves to exclude many irrelevant conditions, but also
excludes others that are relevant.3 Thus in some cases of over deter-
mination – cases where two or more independent wrongful acts alone
would have sufficed to bring about the harm – the but-for test leads
to the dubious conclusion that neither act caused the harm. So, if the
but-for test were applied to such a case, the victim could recover from
neither wrongdoer. As a way around this and other difficulties, courts
have developed an alternative test of causal connection, for use in tort
cases when the but-for test seems to give the wrong answer. This alter-
native test asks whether the condition in question was a “substantial”
factor in producing the result. But this device, which allows judges
and juries to follow their intuitions, is purely pragmatic4 and leaves
the theoretical problem untouched. The substantial factor notion is
indefinite, if not indeterminate, and it is difficult to see how it might
be fitted into a coherent theory of causation.
1. T HE C AUSATION S TORY : M ODERN E FFORTS TO

U NRAVEL ITS M YSTERIES
In 1959 Herbert Hart and I proposed a thesis that we thought would

explain the basis of and justify our intuitions in certain unusual cases,
such as the over-determination situation, as well as straightforward
ones.5 We argued that, to be a cause of an event, a prior event must
be shown to be a causally relevant condition of that event. Whether
something more is needed to turn a causally relevant condition into
a cause may for the moment be left aside; for now let it be assumed
2. On strong and weak necessity and sufficiency see John L. Mackie, The Cement of the
Universe: A Study of Causation pp.39–40, 60–6, 126–7 (1974); Richard W. Wright,
“Causation, Responsibility, Risk, Probability, Naked Statistics and Proof: Pruning the
Bramble Bush by Clarifying the Concepts” [hereafter Pruning] 73 Iowa Law Rev. 1001,
1020 (1988).
3. Wright, Pruning, above n.2, at p.1022.
4. Id. at pp.1018–19.
5. H.L.A. Hart and A.M. Honoré, Causation in the Law (1959).
95
(contrary to our view) that every causally relevant condition is a

cause. Our theory was that an event such as a prior wrongful act is
causally relevant only if it is a necessary element in a set of conditions
that is together sufficient to produce the consequence.
“A condition may be necessary just in the sense that it is one of a set of
conditions jointly sufficient for the production of the consequence. It is
necessary because it is required to complete this set”.6
At least that was true, we said, of physical sequences.7 This weak sense
of “necessary” (necessary to that particular set of conditions) can be
contrasted with the strong necessity inherent in the but-for test. On
the other hand our sense of “sufficient” is a strong one. If the set of
conditions sufficient to produce the consequence occurred, the con-
sequence occurred; and if the consequence had not occurred, the set
of conditions would not have occurred either.8
In 1965 John Mackie had applied our idea to causal regularities –
causal generalizations rather than specific events.9 He combined with
it the doctrine of the plurality of causes. This latter theory postulates
that on different occasions certain types of events – say death or road
accidents – can have different causes. We discover causal regularities,
and ultimately scientific causal laws, by assembling sets of conditions.
Ideally these should be such that we can say of a condition that it is a
necessary10 member of a set such that, when all the members of the
set are present, the consequence invariably follows. The conditions
are jointly sufficient for the result but, in view of the plurality of
causes, they may not actually be necessary for it, since there may be
alternative sets of conditions that will also produce the same conse-
quence.11 Mackie called such a condition – a necessary member of a
set jointly sufficient (but unnecessary) to produce a given type of
6. Id. at p.106. A different, though related account was needed for interpersonal transactions.
7. H.L.A. Hart and Tony Honoré, Causation in the Law (2nd ed. 1985) pp.51–61, 125. See
below, nn.59–61 and accompanying text.
8. This is to apply the reverse but-for test. Wright, Pruning, above n.2, at p.1021 n.108.
Richard Wright termed a condition of this sort, necessary in a weak sense but sufficient in
a strong sense, a “NESS” condition. He derived it from the NESS test, which is meant to
replace the but-for test: id. at 1019. See below, nn.16 and 17 and accompanying text.
9. Mackie, above n.2, at pp.59–63 (building to some extent on Konrad Marc-Wogau, “On
Historical Explanation”, 28 Theoria 213–33 (1962)).
10. Mackie says insufficient (on its own), which is obvious, but also non-redundant, i.e. nec-
essary as an element in that particular jointly sufficient set of conditions: id. p.62.
11. I have modified his terminology so far as the use of “factor” and “condition” are concerned.
96
result – an INUS (Insufficient but Necessary part of an Unnecessary

but Sufficient) condition.
For example, there may be one causal regularity, by which a driver’s
not keeping a proper lookout plus certain other conditions is suffi-
cient to bring about a road accident, and another causal regularity, by
which going too fast plus a different set of conditions is sufficient for
a road accident. If so, not keeping a proper lookout and going too fast
are both INUS conditions of road accidents, though, belonging to
different sets,12 neither condition is necessary to produce a road acci-
dent.
It is worth noting that sets of conditions of this sort generally have
as one of their members the absence of counteracting or frustrating
conditions. A certain dose of strychnine will, given certain bodily
conditions, result in death, but only (1) in the absence of an antidote
and (2) in the absence of some other cause of death intervening
before the poison takes effect. The strength of the dose, the bodily
conditions and the absence of an antidote, the absence of some other
intervening cause of death etc. are all INUS members of the set of
conditions that is together sufficient to result in death.
With some hesitation13 Mackie concluded that though this analy-
sis of causal regularities as sets of INUS conditions often applies not
only to types of event but to specific events,14 it does not always do
so.15 He thought that at least sometimes we can say of a specific
instance of a condition, such as not keeping a proper lookout, that it
caused a collision, though we are not thereby committed to any reg-
ularity that would make not keeping a lookout an INUS condition of
road accidents. It was not just that we and scientific experts are unable
to frame a satisfactory generalization combining speed, position of
the victim, visibility, road condition etc. in the right proportions.
Rather we sometimes conclude that, but for the lack of a proper look-
out, this collision would not have occurred without believing even in
principle that this was an instance of a regularity concerning speed,
position, road condition and the other factors. “A singular causal
12. They could also, of course, belong to the same set.

13. Mackie, above n.2, at pp.48–50 adduces powerful arguments for a cause as something nec-
essary and sufficient in the circumstances for its effect. Cf. Wright, Pruning, above n.2, at
pp.1028–9.
14. Mackie, above n.2, at pp.65–6.
15. Id. pp.40–58.
97
statement need not imply even the vaguest generalization”.16 If in a

specific case we can back our conviction up by pointing to a set of
INUS conditions, so much the better; but that is not essential.
In 1985 Richard Wright in an important paper dissented from
Mackie’s view on this point and, propounding the NESS (Necessary
Element of a Sufficient Set) test, elaborated the idea that Hart and I
had originally put forward.17 His version is only marginally different
from ours. According to Wright
“a particular condition was a cause of (contributed to) a specific result if
and only if it was a necessary element of a set of antecedent actual condi-
tions that was sufficient for the occurrence of the result”.18
This perhaps differs from our view in one respect. Wright seems to
require that all the conditions that belong to the set be antecedent to
the result, whereas it appears to me that, to be sufficient for the result,
some conditions at least can be of the sort that must persist until the
result occurs. An example is the icy condition of the road as a factor
in a road accident.
This difference of view, if it is one, does not affect the value of the
NESS test, which is widely supported by tort theorists, and not only
by them.19 Nevertheless Mackie is not alone in defending the but-for
notion as an analysis of singular causal statements. In German crim-
inal law Friedrich Toepel has recently published a monograph on
crimes of negligence, which clearly have much in common with tort
actions. Toepel supports the but-for test and dissents from the dom-
inant view among German scholars, which rejects it.20
The controversy is clearly not over. In sections 2 and 3, I explore
some aspects of causal connection that are common to the NESS and
but-for theories and others that divide them. The analysis focuses in
particular on how far causation in tort law depends on concepts that
16. Mackie, above n.2, at pp. 77–8, effectively criticized by Wright, Pruning, above n.2, at
pp.1031–4.
17. Richard Wright, “Causation in Tort Law” 73 Cal. Law Rev. 1735 (1985) (especially
pp.1788–1813) [hereafter Causation]; Wright, Pruning, above n.2, at pp.1018–44.
18. Wright, Pruning, above n.2, at p.1019 n.98.
19. See Wright, id. p.1019 n.98. For a criminal law example see Ingeborg Puppe, “Der Erfolg
und seine kausale Erklärung im Strafrecht”, 92 ZStW 863, 867f (1980); “Die Beziehung
zwischen Sorgfaltswidrigkeit und Erfolg bei den Fahrlässigkeitsdelikten” 99 ZStW 595–6
(1987).
20. F. Toepel, Kausalität und Pflichtwidrigkeitszusammenhang beim fahrlässigen Erfolgsdelikt
(1992).
98
apply outside the law, and how far it reflects normative considera-
tions, legal or moral.
2. C OMMON E LEMENTS IN THE NESS AND B UT -F OR T HEORIES
Some points are, or should be, common ground between the NESS
and but-for theories. First, each of these theories embraces both a
search for the meaning of “causally relevant condition” and a test by
which we can tell whether a condition is causally relevant. That is,
they are both at times semantic and at times heuristic. The impor-
tance of the distinction is that we may believe that a condition was a
cause of a specific event because it was a member of a set of condi-
tions jointly sufficient to bring it about and yet not be able to say
exactly what those conditions were. We may therefore mean by cause
a condition of that sort, and yet be satisfied in a given instance that it
was a cause on the basis of the rough uniformities observed in every-
day life that lead us to think that the presence of the condition made
a difference to the outcome. We may do this though we could not
specify the INUS conditions that are exemplified in the case in
hand.21 No one will deny that the but-for test has in many instances
a heuristic value: it often provides a quick way of testing the existence
of causal connection. It is another matter whether it is part of the
meaning of “causally relevant condition” or “cause”.
Philosophers debate whether causal connection is a relation
between events or between facts.22 An event can be described in
many different ways but a fact is tied to a specific description. That
Brutus stabbed Caesar and that he betrayed Caesar describe the same
historical event but the fact that Brutus stabbed the dictator is not the
same fact as the fact that he betrayed him. This difference has a bear-
ing on what a plaintiff must prove in a lawsuit. In a wrongful death
action by Caesar’s widow, she would have to prove, first, the fact that
Brutus stabbed her husband and, second, the further fact that
Caesar’s death was caused by the first fact of Brutus’ stabbing; it
would not be enough to show that Brutus betrayed Caesar, even
though what Brutus did could be truthfully described as betrayal,
because not all forms of betrayal will found tort actions.
21. But see Mackie, above n.2, at pp.54–5, expressing the contrary view.
99
In my view both events and facts can be causally connected; and in

law, including tort law, both events and facts can be relevant. A plain-
tiff must (1) identify the event or events that give rise to the claim and
(2) prove that the fact that the defendant acted in a certain way
caused the harm of which he complains. He must point to the time,
place and persons involved, which can be described in many different
ways. But he must go on to show that some specific aspect of the
events he identifies (such as the fact that the defendant drove at an
excessive speed) brings the defendant within the relevant legal cate-
gory, and supports the conclusion that his wrongful or risk-
creating conduct caused the harm of which the plaintiff complains.
He must show that the element that makes the conduct wrongful or
creates the undue risk was relevant to the harmful outcome for which
the law provides a remedy.
Thus, if the plaintiff alleges that the defendant drove at excessive
speed, he must show that the fact that he drove at that such-and-such
a speed rather than at the proper speed was relevant to the fact that
the plaintiff suffered such-and-such injuries. Though the incident in
which the defendant is alleged to have driven at an excessive speed
can be described in different ways – as Dan Dawes driving down
Main Street, or as the engineer in the yellow Jersey hurrying to get
home – these descriptions serve merely to identify the incident in
issue. Again, if the plaintiff injured in an explosion relies on strict lia-
bility for the defendant’s use of explosives, he must show that the fact
that the defendant used explosives was relevant to the damage for
which he seeks redress, though what the defendant did can be
described in various other ways – quarrying for stone, or trying out a
new fuse. Law is so structured that a plaintiff must prove a connec-
tion between the facts that specify those features of the events that are
both causally relevant and legally relevant in other ways.
It follows that the law determines the way in which the causal
inquiry is framed. Causal connection (in German terminology
Kausalzusammenhang) cannot be separated in cases of fault liability
from unlawful connection (Rechtswidrigkeitszusammenhang); their
separation in the writings of some German theorists is in my view a
source of error. The inquiry is into whether certain faulty conduct (or
risk-creating conduct entailing strict liability) caused certain harm. In
a country where liability for driving an automobile depends on neg-
ligence, the legal issue will be whether the fact that the defendant
100
drove negligently (e.g. without keeping a proper lookout) caused the

injuries. But in a jurisdiction where automobile liability is strict, the
question will be whether the fact that the defendant drove caused
the harm. The answer to these two questions may be different. The
defendant’s driving may have caused the injury, yet his negligent dri-
ving may not have done so, because his negligence made no differ-
ence to the outcome; the same collision would have occurred had he
driven carefully. In tort law, the issue is not the open-ended one
“What caused this harm?” but “Did the fact that the defendant
behaved in a certain unlawful or undue risk-creating way cause it?”
The first question calls for an explanation of how the harm came
about, the second for an assessment of the defendant’s responsibility
for conduct described in categories fixed by the law.
A good illustration is the case involving the navigation of a ship by
an officer who was competent and experienced but who did not pos-
sess the certificate required by law.23 The officer navigated negligently
and the ship was involved in a collision. The defendants admitted
that the negligent navigation caused the collision, but denied that the
lack of a certificate was causally relevant. Had it been relevant the
amount of damages would have been greater. Given that the basis of
liability was fault, it was rightly held that the lack of a certificate was
irrelevant, since the officer would have navigated no better had he
possessed one.
The lack of a certificate would, it seems to me, have been causally rel-
evant had it been the case that the officer was incompetent and that, in
order to obtain a certificate, he would have had to acquire the neces-
sary competence. On the other hand, had there been strict liability for
navigating the ship without a certificate, so that if the ship was navi-
gated by a pilot without a certificate the navigation was at the defen-
dants’ risk, the navigation would have been held to have caused the
collision. Since strict liability is liability not for wrongful conduct, but
for engaging in risk-creating activity, there would have been no need in
this case to show that the lack of a certificate was causally relevant. In
such a strict liability case, it would have been enough that, had the ship
not been navigated by X (who did not in fact possess a certificate),
no collision would have occurred. In the actual case, however,
what explained the collision was the fact that the ship was negligently
23. The Empire Jamaica [1955] P. 52, 57–8.

101
navigated, not that it was navigated. There is a distinction between

conduct that counts as a cause in an explanatory inquiry, on the one
hand, and conduct that is made by the law a basis of liability provided
it is causally connected with the harm suffered, on the other.
This point is misunderstood by Wright,24 who thinks that the
Hart and Honoré criteria of voluntary human action and abnormal-
ity (in the sense of actions and events that are unusual in the context),
used to distinguish causes from mere causally relevant conditions, are
meant to identify tortious conduct.25 We are said to have difficulty in
accommodating strict liability, where the conduct on which liability
is based (e.g. using explosives) may not be either abnormal or volun-
tary in the sense of intended to do harm. But the role of our criteria
is to mark a backward limit in explanatory causal inquiries, and only
a forward limit to whatever responsibility is appropriate under other,
normative, concepts of tort law. Our causal criteria are not a substi-
tute for the other principles of tort law that define what conduct is
wrongful or entails strict liability. That depends on positive law, and
varies from system to system, some systems imposing strict liability
or no liability in circumstances where others insist on fault (e.g. for
auto accidents). It remains true, however, that often the law for good
reason designates conduct as wrongful when it is intended to do harm
or is a departure from a recognized norm. In this way wrongful con-
duct that causes harm often coincides with what would count as a
causal explanation of the harm that ensues. But in cases of strict lia-
bility, where the wrongfulness of the defendant’s conduct is not in
issue, there is no occasion to trace a causal path from wrongfulness to
the plaintiff ’s harm.
A third point is that, whichever theory we adopt, a hypothetical
question must be put and answered. This involves a counterfactual
proposition, namely that if a condition that in fact occurred had not
occurred, the outcome would have been so-and-so.26 On the but-for
theory we must ask whether in the circumstances the consequence
would have occurred had the condition not occurred. On the NESS
theory we must ask whether in the circumstances the condition was
a necessary member of a set of conditions that was together sufficient

25. They are, we argue, meant to identify the limits of responsibility in most cases: Hart and
Honoré (2nd ed.), above n.6, at pp.68–81.
26. Wright, Pruning, above n.2, at pp.1039–42.
102
to produce the consequence, i.e. would the remaining set of condi-

tions have produced the consequence in the absence of the condition
in issue?
But to speak of the absence of a condition, or the hypothesis that
it did not occur, can mislead. It suggests that the condition we are
interested in can be notionally eliminated and that we can then cal-
culate what would have happened if only the remaining circum-
stances had been present. Wright says that we hypothetically
eliminate the condition being tested without adding or subtracting
any other conditions.27 This view of elimination also has adherents
in German legal theory, partly because theorists there tend to separate
causal connection from unlawful connection. If this were the right
way to test a causal hypothesis, we should indeed be comparing the
real world with an imaginary world. For in the real world conduct is
often a choice between alternatives. If John had not done so-and-so,
he would have done something else. If he had not intervened, some-
one else would have, perhaps in a different way. Sometimes these
alternatives would have had a bearing on the course of events we are
investigating. Mackie rightly says that both necessity in the strong
sense and sufficiency in the strong sense (viz. both the but-for and
NESS theories) “involve assertions about how the world would have
run on if something different had been done”.28
But this is rather loosely expressed, for it leaves open the question
“what is the something different that we must suppose to have been
done”?29In many contexts the answer is “whatever would in the cir-
cumstances have occurred had the condition in question not
occurred”. Sometimes there is no reason to suppose that anything
causally relevant would have occurred in that event, and then it looks
as if the condition in issue is simply being eliminated as causally irrel-
evant. John dropped a lighted match on some brushwood. Did he
cause the destruction of the forest? To test this, ask whether if John
had not dropped the lighted match, the forest fire would have
occurred. We begin by supposing the contradictory of the condition
viz. “John did not drop the lighted match”. Often it is clear that, had
he not dropped the match, whatever else would have happened
27. Id. pp.1041–1042.

29. For the reasons given in the text Wright’s view and that of David K. Lewis, Counterfactuals
(1973) pp.72–7 are unsatisfactory.
103
would not have resulted in a fire. But the position would be different
if we knew that, had John not dropped a match, David would have
done so anyway. In that case, though John will indeed have destroyed
the forest, he destroyed a forest that was anyhow doomed, and that
may affect the forest owner’s claim. However, there is usually no rea-
son to suppose any such alternative event and therefore no need
expressly to substitute anything for the condition eliminated.
But the notion that we do not substitute anything for the condi-
tion whose causal relevance we are testing is an illusion. Suppose we
want to test the assertion that Churchill kept Britain in the war in
1940. Was the fact that he was Prime Minister the cause of Britain
remaining in the war? Here we cannot eliminate Churchill without
substituting another Prime Minister whose hypothetical conduct of
affairs would have had a bearing on whether Britain remained in the
war. The contradictory of “Churchill was Prime Minister” viz.
“Churchill was not Prime Minister” would have implied in the world
of 1940 that someone else was Prime Minister (say Halifax); and we
must then ask whether that person would have wanted and been able
to convince Parliament and the British people to continue the war.
The causal statement about Churchill implicitly contrasts him with
this hypothetical substitute. It is true that logically “Churchill was
not Prime Minister in 1940” is compatible with there being no Prime
Minister at all. But it is the real world, not an imaginary world or a
logical calculus, that determines what is implied in the context by the
contradictory of the condition in question, and what should there-
fore be substituted for it in testing the counterfactual proposition. In
that real world, had Churchill not been Prime Minister, someone else
would have held that office. And that someone else might or might
not have kept Britain in the war.
In a legal context the same approach may be applied, except that
when the inquiry concerns the causal relevance of wrongful conduct, as
is usual in tort claims, we must substitute for the wrongful conduct of
the defendant rightful conduct on his part.30 When liability is based on
fault the comparison is not with what would have happened had the
defendant done nothing, but with what would have happened had he
acted properly. The reason for making this substitution is similar to that
30. Hart and Honoré (2nd ed.), above n.7, at pp.lviii–lx, 411–14; Toepel, above n.20, at
p.106.
104
in Churchill’s case. The contradictory of “the defendant acted wrong-

fully” is “the defendant did not act wrongfully”, and this is logically
quite unspecific; for instance it is compatible with his doing nothing.
Once more we must turn to the features of the real world – in this case,
the world of positive tort law – which tells us that what is implied by
his not acting wrongfully is his acting rightfully. The reason why this is
the appropriate substitution is that the aim of the legal inquiry is to dis-
cover not whether the defendant’s conduct as such made a difference to
the outcome, but whether the fact that it was wrongful did so.
Again, tort law normally requires the plaintiff to show that the
wrongful aspect of the defendant’s conduct was causally connected
with the harm.31 To eliminate driving at an excessive speed therefore
requires us to substitute for the excessive speed driving at a proper
speed, and then to ask whether that would have led to the injuries
that in fact occurred. Moreover, the hypothesis of rightful conduct
involves supposing not merely that the defendant drove at a proper
speed, but also that the likely consequences of his doing so be taken
into account. For example, if the pedestrian who was run over would
have seen the defendant’s automobile coming in time to avoid it, had
he been driving at a reasonable pace, this must be taken into account.
In the same way if someone other than Churchill had been in office,
his colleagues would have been less afraid of contradicting him, and
this must be taken into account in answering the historical question
about 1940. In most legal contexts, the hypothetical inquiry requires
that all conditions be kept steady, except that we substitute rightful
conduct and its likely consequences for the wrongful conduct of
which the defendant is alleged to have been guilty.
In the context of strict liability, however, the inquiry is more like
that in the dropped match case.32 Substitution of lawful conduct is
not possible and is not required in cases of strict liability, since the
defendant’s conduct, though it creates a risk, is not unlawful unless it
causes harm. In such a case the hypothetical inquiry thus must be
31. This is not the case if the law provides that, if the defendant acts unlawfully, he acts at his
risk. Then, as in cases of strict liability, it has only to be shown that his action was causally
connected with the harm, not that the wrongful aspect of it was so connected. Hart and
Honoré, Causation (2nd ed.), above n.7, at lx–lxi. And of course there are torts, such as
trespass, in which harm need not be proved.
32. The dropped match incident might crop up in the context of negligence in tort law. In that
case the comparison would be between what the defendant did and his exercising due care
in regard to the match e.g. by putting it out before throwing it away.
105
whether the plaintiff would have suffered injury had the defendant
not engaged in the activity (e.g. using explosives) that entails strict lia-
bility. There is normally no reason to suppose that, if the defendant
had not used explosives for blasting, he or someone else would then
and there have injured the plaintiff by some other means.
Both inside and outside the law, therefore, the purpose of the
inquiry determines how we should frame the hypothesis to be tested.
Tort law lays down what counts as wrongful conduct or conduct
entailing strict liability and so what the plaintiff has to prove. It aims
to protect people against wrongful infringement of their rights and
exposure to undue danger. So, to ascertain whether an infringement
has occurred, the wrongful conduct of the defendant must normally
be compared with the notional rightful conduct that the plaintiff was
entitled to expect in the circumstances. The question is, whether the
difference between the wrongful conduct that occurred in the real
world and the rightful conduct (together with its likely conse-
quences) that we imagine as occurring, would have led to a different
result in the hypothetical world that resembles the real world in all
other respects. In the strict liability context, where the law protects
against risks created by lawful conduct on the part of others, the
causal inquiry is adapted to that aim.
Does the fact that in tort law we normally compare wrongful with
rightful conduct show that the causal question is a normative ques-
tion in disguise? It does not, for the inquiry into what would have
occurred had the defendant obeyed a legal norm is no more norma-
tive than the question what would have occurred had the Prime
Minister been someone other than Churchill. The answer to these
causal inquires has no normative component. It remains the same,
even if we suppose that it would have been better for Britain to make
peace in 1940, or that it would be better, in order to combat over-
population, for those who drive automobiles to drive blindfold.
There are those who think that counterfactual propositions cannot
be true or false.33 The tort process however assumes, in my view
rightly, that they can be. We can often know what on a certain
hypothesis would happen in the real world. We know that if we strike
a match under certain conditions a flame will appear. The statement
“if a match is struck under conditions XYZ a flame will appear” can
33. Mackie, above n.2, at p.54; Toepel, above n.20, at p.55.

106
be true or false and its truth value can be tested. Why not, also, “if a
match had been struck . . . a flame would have appeared”? The only
difference is that the proposition about the past can be verified only
indirectly, on the assumption that in similar conditions similar results
recur. This assumption cannot itself be verified, but unless it is well-
founded we cannot rely on causal regularities or scientific laws. Of
course we are often uncertain what would have happened on a given
occasion had the defendant acted lawfully. Here, though the coun-
terfactual statement “had X not done what he did Y would not have
happened” is in principle true or false, we cannot tell for certain
which it is. We must then have recourse to rules of law that instruct
the judge or jury to go by the probabilities and, if necessary, to fall
back on rules about the burden of proof.
3. NESS VERSUS B UT -F OR : D IFFERENCES IN THE T HEORIES
I now turn to some of the examples that supporters of the NESS

and but-for theories interpret differently. Some concern over-
determination, others indeterminacy. Over-determination cases, it
will be recalled, involve two or more independent acts each of which
alone would have sufficed to bring about the harm. In such cases, the
separate causal processes may be similar or different, and they may
culminate in harm either at the same time or successively. When (1)
similar causal processes culminate in harm at the same time the NESS
and but-for theories point to different causal conclusions. But when
(2) the causal processes are different or (3) one culminates in harm
before the other, they point to the same causal conclusion. In cases of
(4) causing or inducing people to act, neither theory is adequate, but
the notion of sufficiency, in a modified form, explains why these are
thought of as causal or quasi-causal relations.
A. Over-determination: Similar Causal Processes Culminating at the

Same Time
Two similar causal processes may culminate in the same harm at the
same time. Two shots, negligently but independently fired by A and
B when out hunting, simultaneously hit C and destroy his left eye.
Each shot was sufficient to do this without the other, and the effects
107
of the two shots cannot be distinguished. Neither shot was necessary

in the strong but-for sense, but each was necessary in the weak sense
that it was a necessary member of a set of conditions together suffi-
cient to destroy the sight of the victim’s eye.34 It would be counter-
intuitive to hold that neither A nor B is responsible for the loss of C’s
eye. But is that because legal policy, or intuitions about responsibil-
ity, lead us to disregard the lack of causal connection between each
shot and the loss of the eye?
Mackie, supporting the but-for theory, holds that when this type
of over-determination is present the cause of the harm consists in the
acts of A and B taken as a cluster.35 In the absence of both acts C’s
eyesight would not have been lost. But it makes no sense, he thinks,
to ask which was the cause. Perhaps not, but why need we ask which
was the cause? Cannot each shot have been a cause of the loss of eye-
sight in a sense adequate both to explain what happened causally and
to ascribe responsibility to A and B? It seems that, in Mackie’s view,
if A and B are held liable in tort or criminal law this must be based
on a legal policy that overrides the lack of causal connection, or must
rest on a special legal concept of cause. Indeed he says in one passage,
“whatever our actual concept may be, it is obvious that we can con-
struct causal concepts as we wish, which may be useful for particular
purposes”.36 It is worth asking, therefore, whether in such cases of
simultaneous over-determination lawyers resort to a special legal
notion of cause.
Our concepts are tools shaped by the purposes for which we use
them. Causation is used mainly for three purposes: to provide recipes,
to explain events and to allot responsibility. Its roots probably lie in
the first of these. Even non-human primates find recipes for what
they want. To crack a nut, you need a stone of a certain size and
weight and you must bring it down on the nut with a certain force.
You learn to construct what your sophisticated cousins of a later age
will call a set of conditions together sufficient to produce a given
result, nut-cracking. Having assembled what seems to be a set of suf-
ficient conditions, you whittle them down by eliminating those that
trial and error show to be unnecessary. It turns out to be necessary
that the stone be of a certain weight, that a certain degree of force be
34. Necessity in this weak sense is described above. See, above, text accompanying nn.5–8.
35. Mackie, above n.2, at p.47. See also Toepel, above n.20, at pp.71–2, 84.
36. Mackie, above n.2, at p.58.
108
employed and that the nutshell be not too thick, but unnecessary that
the wind lie in a northerly direction or that the agent utter a blood-
curdling whoop. The recipe is general, built up over time as instances
accumulate and conditions are tested.
The discovery of conditions that are necessary members of a set
sufficient to produce some result can be used to explain events as well
as to bring them about. It can extend to events over which we have
no control. We seek to discover the conditions of these events, per-
haps in order to control similar events in future, or simply to under-
stand how the universe works. The knowledge thus acquired may
come in handy, but it may be pursued for its own sake. It can be used
to trace the effects of natural events and inanimate forces. It can also
be used to assign responsibility – social responsibility – to human
beings for the outcome of their conduct.
For all these purposes – making recipes, explaining events and
attributing responsibility – we can use the concept of the necessary
members of a set of conditions together sufficient to produce a result
of a given type, though we may not be able to specify or quantify all
the conditions. This limited ability to specify conditions does not
prevent our using causal ideas to explain events backwards in terms
of previous events and conditions, or to allot responsibility for con-
duct forwards in terms of outcomes. These are not normative func-
tions. It makes no difference whether the event to be explained, or the
conduct to be assessed, is judged good or bad. It is tort law, not the
concept of cause, that invites us to focus on the outcome of wrongful
conduct.
What bearing has this on the problem of over-determination? The
lesson is surely that the NESS test is not confined to the context of
responsibility in general or tort law in particular. To revert to the two
independent shots by A and B that put out C’s eye, the acts of A and
B each exemplify how to put C’s left eye out, if that was, perhaps, the
objective of snipers in a military campaign. Each also provides an
explanation of C’s loss of his left eye. Why, then, should not each pro-
vide a basis for assigning legal and moral responsibility to A and B?
No recourse to legal policy, or to normative considerations, is needed
to reach the conclusion that both A and B independently caused the
harm to C. None of the purposes for which causal concepts are used
requires us to adopt the but-for theory. There can be different recipes
for attaining a given type of result, more than one of which is available
109
on a given occasion. There can be different explanations of an event

of a given type in terms of prior or concurrent conditions, more than
one of which is present on a given occasion. To prevent the con-
struction of a brick wall, one recipe is to withhold bricks, another to
withhold mortar. If A withholds bricks and B independently with-
holds mortar and the wall is not built, or built on time, have not both
A and B prevented its construction? We do not need a special “legal”
concept of cause to deal with simultaneous over-determination by
similar causal processes.
Toepel in the context of criminal law denies that either A or B is
responsible for putting out C’s left eye in a case like the shooting
case.37 Apart from pointing to the presumption of innocence, he
argues that it is arbitrary to disregard B’s shot when deciding whether
A’s was a necessary member of a set of conditions sufficient to pro-
duce the harm, and vice versa. The answer to this objection is that B’s
shot can be provisionally disregarded when we are considering the set
of conditions of which A’s shot is a member because that set may be
sufficient without B’s shot. The proper procedure in investigating the
cause of a physical sequence of events is to assemble a set of condi-
tions apparently sufficient to produce and hence explain the result.
This we do in the light of what we know to be INUS conditions of
results of that type.38 But the set we provisionally assemble may con-
tain some conditions that turn out to have been unnecessary for the
result.
For example, the provisional set of conditions might consist of A’s
shot, the distance between A and C, the strength of the wind, the
colour of A’s jacket, and B’s shot. We then reduce the set by elimi-
nating those conditions that were not necessary to it. This disposes of
the colour of A’s jacket. B’s shot we provisionally eliminate not
because it was clearly not necessary, but because our knowledge of
causal regularities and scientific laws leads us to think that A’s shot
may have been sufficient to put C’s eye out without B’s. Having
reduced the set so far, we then test the relevance of A’s shot (more pre-
cisely the fact that A shot negligently) by hypothetically substituting,
for A’s negligent shot, his taking proper care in shooting. We ask
whether in that event C would have lost his eye from that set of con-
37. Toepel, above n.20, at pp.66, 72–5.

38. See, above, nn.8–15 and accompanying text.
110
ditions. If not, A’s negligent shooting was a cause of C’s loss of eye-
sight. But suppose that careful investigation shows that A’s shot
missed. In that case A’s shot was not causally relevant to the outcome
and we must construct another set of conditions, including B’s shot,
and test whether B’s shot was a necessary member of that set of con-
ditions which together produced the loss of eyesight. Toepel objects
that this procedure involves deciding that A’s shot was causally rele-
vant before we test it for causal relevance.39 That is not so. All we have
decided before testing A’s shot for relevance while provisionally omit-
ting B’s is that, as we know some of the INUS conditions for loss of
eyesight, A’s shot may on this occasion have been a causally relevant
condition even in the absence of B’s shot. It is true that the testing
procedure is possible only if we know some causal regularities before
we start. But we need not assume from the start that the outcome of
the test will be positive.
A similar causal over-determination analysis may be applied to
other cases in which two similar processes appear responsible for an
event, for example, where two fires merge and jointly culminate in
harm40 or where the pollution created by a number of persons inde-
pendently pollutes a stream or lake.41
B. Over-determination: Different Causal Processes
In other over-determination cases the two sets of conditions involve

qualitatively different causal processes, such as poisoning and dehy-
dration, the various stages of which have been studied and can be dis-
tinguished. In that case, if we know enough about the stages by which
the events came about we can generally tell, even in cases of over-
determination, which causal process ran its course and which was
frustrated. So in the case of the desert traveller42 – where A poisons
C’s water keg, B empties the keg before C drinks the poisoned water,
and C dies of thirst – it now seems to me that B causes C’s death. I
39. Toepel, above n.20, at p.69.

40. Wright, Pruning, above n.2, at pp.1018, 1022.
41. See generally id. p.1035–7.
42. James A. McLaughlin, “Proximate Cause”, 39 Harv. Law Rev. 149, 155 n.25
(1925–1926); Hart and Honoré, Causation (2nd ed.), above n.7, at pp.239–41; Mackie,
above n.2, at pp.44–6.
111
regret that, despite criticism by Mackie,43 since endorsed by

Wright,44 the second edition of Hart and Honoré resisted this con-
clusion.45 My current reasoning is that B’s conduct introduces a con-
dition, lack of water, that in the circumstances, including the absence
of an alternative water supply, is sufficient to bring about and does
bring about C’s death from dehydration.
It is also true that B saved C from dying of poison. But this is not
relevant in an explanatory inquiry, since B’s act adequately explains
how C’s death came about (through dehydration), nor should it be
relevant in a wrongful death action or prosecution for homicide. If B,
not realizing that the water keg was poisoned, emptied it intending
to harm C or negligently failing in his duty to C, the deceased’s
widow ought therefore to succeed in a wrongful death action46 –
though since C was doomed in any event the damages will be mini-
mal. On the other hand if B realized that the water was poisoned and
emptied the keg to prevent C drinking the poisoned water, and sup-
posing that an alternative supply could be found in time to save C, B
would still unintentionally have “caused” C’s death; but he would not
be legally responsible for doing so. There is no need to postulate any
special legal concept of cause in order to understand this type of case.
In cases of over-determination, where different causal processes are
associated with the two conditions that are causal candidates, there
should be no need to show that one of them was a NESS condition
of death as and how the death occurred (e.g. by dehydration on
Tuesday rather than by poisoning on Monday).47 That would be to
require establishing the very point in issue, namely how the victim’s
death came about. What has to be shown in a tort action is that the
defendant’s wrongful act caused the harm, in this case the victim’s
death. We know from the way in which the law structures actions for
wrongful death that what is legally relevant is death, not death at this
or that time or place or by this or that process.48 However, the plain-
tiff (the widow) has to identify the death for which she claims redress.
43. Mackie, above n.2, at pp.44, 46–7.

45. Hart and Honoré, Causation, above n.5, at pp.219–20; (2nd ed.), above n.7, at
pp.239–41. See Mackie, above n.2, at pp.45–6.
46. And, if the harm was intentional, B will be guilty of homicide.
47. As suggested by Mackie, above n.2, at p.46; Toepel, above n.20, at pp.70, 78–9.
48. Wright, Causation, above n.17, at pp.1777–8; Wright, Pruning, above n.2, at pp.1025–6;
Puppe, Der Erfolg, above n.19, at pp.863, 867–8.
112
This requires that the victim, place and time be specified. She has also
to show that the defendant’s wrongful act caused the death, which
involves showing that some causal process initiated by the defendant
brought it about. In the water keg example, that B’s conduct (in emp-
tying the water keg) was a condition of C’s death by dehydration is
relevant because the process of dehydration connects B’s act by a
series of stages with C’s death. To establish that this process occurred
is relevant because, given the plurality of possible causes of death, it
is necessary for C’s widow to point to a specific causal process that
links the two, and dehydration is such a process.
C. Over-determination: Harm that Has Already Occurred
Although philosophers may debate the possibility of backward cau-

sation, it seems clear that it is impossible to cause an event that has
already occurred. One can flog a dead horse but not kill it. On the
other hand one can kill someone who has only a moment to live, as
in the case where defendants were alleged to have negligently electro-
cuted the deceased boy as he fell to certain death.49 The fact that our
concept of cause does not allow for causing harm that has already
occurred can give rise to a problem of over-determination when the
harm consists of depriving someone of future opportunities.
Suppose that A negligently injures C, a wrestler, so that he is per-
manently disabled from wrestling. Six months later B independently
and negligently runs C over so that, even had he not been injured the
first time, he would not have been able to wrestle in future. That B’s
act was independent implies that the first injury did not contribute
to the second, for example by preventing C from avoiding the second
accident.50
What is the status of C’s claim to loss of future earnings as a
wrestler? A, and only A, is clearly responsible for what C would have
earned in the six months following the first injury. As regards C’s lost
wrestling earnings following the second injury, however, responsibil-
ity is much less clear. B will plead that he did not cause C’s disability
as a wrestler even in part since C was permanently disabled already.
49. Dillon v. Twin State Gas & Elec. Co., 163 Atl. 111 (1932).
50. The facts are suggested by Baker v. Willoughby [1970] AC 467. Cf. Jobling v. Associated
Dairies, [1982] AC 794.
113
Yet A will plead that his liability is limited to the six months’ earnings
alone, since, as it turns out, C would not have been able to wrestle for
more than that period in any event. If both of these arguments are
accepted, C, having been wrongfully injured on two occasions,
obtains less than had he been injured only once, which is unsatisfac-
tory. Yet this is not like the case of simultaneous over-determination
where each agent can properly be said to have caused the same injury.
Someone who claims to have lost future opportunities of earning
money must show that he would have been able to take advantage of
those opportunities in the absence of the wrongful act of which he
complains. For example, if after six months C had died of a heart
attack, not brought on by the injury inflicted by A, his estate could
not claim loss of earnings for the period after he was dead. A dead
man cannot wrestle, and there is no reason to impose on A the risk
that C could die of natural causes unconnected with A’s wrongful act.
There are two possible ways round the difficulty, one of which is
consistent with causal principles while the other rests on a basis of
risk-allocation. The first solution is to hold that there is a cause of
action in tort law for depriving someone by a wrongful act of a tort
remedy that would otherwise have been available to him. In the
injured wrestler case, B has deprived C of a tort remedy against A for
his lost wrestling earnings following the second injury. This cause of
action, if it existed, would be like the suggested remedy, now much
discussed, against a defendant who wrongfully deprives a plaintiff of
a chance.51 The alternative way of solving the injured wrestler conun-
drum is to treat is to treat A as guaranteeing C that he will not be
deprived of his prospective remedy for loss of earnings by a later
wrongful act of someone else. To construct such a guarantee is to pro-
vide a non-causal basis of liability, an example of the familiar princi-
ple by which the risk of harm to C is sometimes placed on A. Here
the risk would be imposed on A as a result of A’s wrongdoing, just as
in the law of sale risk is sometimes transferred to the seller because of
his failure to deliver goods on time. The justification for putting the
risk on A would be that it is the function of the law to protect people
against wrongful invasions of their rights. Rather than let a person
51. Wright, Pruning, above n.2, at pp.1067–72. A “lost chance” case may arise, for example,
if a doctor negligently fails to diagnose a progressive disease such as cancer in a timely man-
ner, and so deprives the patient of, say, a twenty per cent chance of recovery. Causes of
action along these lines, whatever their intrinsic merits, do no violence to causal principles.
114
who has suffered two wrongs go without remedy, it is better to make

the first wrongdoer guarantee him against the economic conse-
quences of a later tort by some third person. To do this would not be
to introduce a special legal sense of cause. It would be to substitute,
for liability based on causing harm, liability based on risk-allocation.
D. Indeterminacy
Indeterminacy presents a difficulty for the NESS theory. If the world

is indeterminate, we cannot assemble a set of conditions invariably
sufficient to produce a given outcome. The outcome may usually
follow, but there will be occasions when it does not. At most there
may be a certain probability that on a given occasion the outcome
will follow. An indeterminate world also presents a difficulty for the
but-for theory, since in an indeterminate world we cannot accurately
calculate what would have happened in the absence of a particular
wrongful act. But, if it is permissible with Mackie to fall back on our
intuitive grasp of but-for relationships,52 the difficulty is less serious
for the but-for than for the NESS theory.
Mackie, though leaving open whether the world really is indeter-
minate, argues that what we would say about causation in an inde-
terminate world supports the but-for theory.53 He imagines two
candy machines, both “indeterministic”.54 One never produces
candy unless the proper coin is inserted, but does not always produce
candy when it is. The other always produces candy when a coin is
inserted, but sometimes does so without a coin. Applying the but-for
test, Mackie thinks that we would say of the first machine that it
caused the candy to appear when it did appear, but deny this for the
second machine.
Mackie’s argument is not convincing.55 If we thought that the
candy machine world was indeterminate we would have no reason to
suppose that the first machine would not in future behave like the
52. See, above, text accompanying n.15.

53. Toepel, above n.20, at pp.61, 90–5.
54. Mackie, above n.2, at pp.40–3, 47.
55. Hart and Honoré (2nd ed.), above n.7, at pp.xl–xli, 235 n.56; Wright, Pruning, above n.2,
at p.1029; Wright, Causation, above n.17, at pp.1777–8; cf. Toepel, above n.20, at
pp.92–3.
115
second machine and produce candy without a coin being inserted,

and vice versa for the second machine. We would have no reason to
hold that the insertion of the coin caused the candy to appear in the
one machine but not in the other. But, though their operation may
be erratic, we believe that candy machines are deterministic.56 We
explain their deviant behaviour on the basis that, if the machine does
not produce candy when a coin is inserted or does produce it when a
coin is not inserted, the mechanism is faulty. In the first case, a coun-
teracting factor is present; something in the machine has jammed and
prevents the candy from emerging. Recall that in most sets of jointly
sufficient conditions, one condition is the absence of counteracting
conditions.57 In the second case, given the plurality of causes, we
conclude that an alternative set of INUS conditions for the produc-
tion of candy is present, which might consist for example not in
inserting a coin but in giving the machine a hefty kick when the
mechanism is jammed.
The argument from an imagined indeterminate world does not
support the but-for theory. What of the argument that the world
really is indeterminate? The micro-level indeterminacy associated
with quantum mechanics is not in point. As Wright neatly observes,
it might be an obstacle to actions between fundamental particles, but
it does not affect the macro-phenomena which are the law’s concern.
More baffling is the difficulty of tracing causal connection, or some-
thing like it, with human actions done for reasons – when one per-
son provides another with a reason for doing so-and-so.58 Although
we have up to now discussed physical sequences of events as if they
took place untroubled by human intervention, there are in practice
few situations in tort law where a physical sequence of events is not
partly determined by human decisions taken in the light of a situa-
tion as it develops. To take an example, adapted from Mackie,59 sup-
pose the question is whether Alec’s statement caused Bill to hit
Charles. If Alec said, for instance, that Charles was having an affair

57. See above, nn.11–12 and accompanying text.
58. Hart and Honoré, Causation (2nd ed.), above n.7, at pp.51–61. For want of space I do not
deal in this paper with the provision of opportunities which another person exploits or the
furnishing of help of which another takes advantage. These are important legally and in
ordinary life but are still more remote from physical sequences than is acting on a reason
provided by another.
59. Mackie, above n.2, at p.43.
116
with Bill’s girl-friend, it would be intelligible that what he said caused

Bill to hit Charles. If one objects to the use of “caused” in a context
where Bill should have controlled himself, other expressions such as
“led”, “induced”, and “provoked” could replace it. The situation is of
a type that often arises in tort law, for example in connection with
false statements that induce others to invest money or persuasions
that lead contracting parties to break their contracts. How should
sequences of this sort, which involve providing reasons on which
others act, be analysed?
It seems impossible to suppose a set of NESS conditions that
would together be sufficient to produce Bill’s assault on Charles.
Many people believe that human conduct is not strictly determined.
If so, one cannot even in principle assemble a set of conditions suffi-
cient to induce a person to act in a particular way. Even if human
action is determined, it will still not be possible to assemble a set of
conditions of this sort. This is not merely because the reactions of
human beings vary. A greater difficulty is that a person who has been
confronted with a similar set of conditions in the past may, in the
light of that experience, react differently on this occasion. It does not
follow from Bill’s reaction on this occasion that he would react simi-
larly if told the same thing again. It does not follow that David would
react in the same way if Alec told him the same thing about his
(David’s) girl-friend. Even with a great deal of information about
people’s characters and background, their reactions simply are not
totally predictable. Moreover, any prediction, if known by the person
about whom it is made, may provide a reason for reacting differently.
Thus, determinism in regard to human conduct, even if true, can-
not be tested. Yet Bill’s reaction is explained by what Alec said to him,
and to that extent functions as a causal (or quasi-causal) explanation.
What Alec said was in some sense sufficient in the circumstances to
lead Bill to react as he did. Yet if we are to speak either of “causing”
people to act or, more modestly, of influencing them in their deci-
sions, a different account of the connection is called for.
Mackie thinks that our attitude to interpersonal transactions
favours the but-for theory.60 Had Alec not said what he did, we may
be sure that Bill would not have hit Charles. But how can we tell this?
Presumably because we have assembled, from our experience and that

117
of others, not sets of INUS conditions relating to human behaviour,

but some loose generalizations that tell us what sorts of reason people
act on. In the present case the only reason that could explain Bill’s
attack was what Alec said to him. If Bill admits that this was why he
struck Charles, the diagnosis is confirmed.
But this is not enough to show that the but-for theory explains the
sort of connection involved in acting for a reason. For a person often
has two or more (typically many more) reasons for reaching a deci-
sion and acting on it. A potential investor may be influenced by a
false statement in a prospectus and by advice from his stockbroker.
Suppose that for these two reasons he makes an ill-fated investment
in Eldorado Mines. Nothing turns, it seems to me, on whether each
reason was necessary or sufficient to persuade him to invest, or to dis-
suade him from changing his mind.61 Would we conclude that the
false statement did not induce him to make the investment if the
advice from the stockbroker was also sufficient to persuade him to do
so? If the investor is prepared to say truthfully that each of the reasons
influenced him, it seems immaterial whether he is prepared to say
that, but for one of them, he would not have reached the decision he
did. The reasons on which he acts are precisely that – the reasons that,
either individually or jointly, appeared to him sufficient and moti-
vated him to decide and act.
Sam is offered a post in Middletown at an increased salary.
Middletown is his wife’s home town and she wishes to return there.
Both of these factors – the increased salary and his wife’s wishes – may
be reasons sufficient to induce Sam to accept the offer, in which case
neither is a but-for reason. Or both may be necessary elements in a
set of reasons together sufficient to persuade Sam to accept the offer.
In either case, Sam acts on those two reasons, which are either singly
or jointly sufficient to persuade him. But along with these reasons that
induce him to go to Middletown, there will be other conditions such
that, had they not existed, he would have refused the offer. There are
good schools in Middletown for the children; Middletown has a
lively choral society. Had there not been good schools and a lively
choral society Sam would have refused the offer; but they were not his
61. See Australian Steel & Mining Corp. v. Corben 1974 2 N.S.W.L.R. 202, 209; Hart and
Honoré, Causation (2nd ed.), above n.7, at p.193. Compare, for criminal law, judgement
of Feb. 24, 1959, BGH Gr. Sen. St., 13 Entscheidungen des Bundesgerichtshofes in
Strafsachen [BGHSt] 15 (F.R.G.); contra Toepel, above n.20, at p.93.
118
reasons for accepting. So far as these reasons are concerned he might

just as well have stayed where he was. But-for reasons are often not
reasons for taking a decision or acting on it, but reasons against not
taking it. They do not cause, lead, induce or prompt Sam to move,
any more than the existence of the capitalist system induces the
investor to put his money in Eldorado Mines.
It seems to me that we reach our conclusions about what caused,
induced, persuaded, or provoked someone to act as he did in a way
somewhat like but not identical with the way in which we reach con-
clusions about the causes of physical events. We cannot construct
INUS conditions and NESS conditions for situations in which
human beings make and act on decisions. But we know enough
about the sorts of reason that motivate us to be able often to conclude
that certain factors were singly or together sufficient for the decision.
All that is meant by “sufficient” in this context is that they provide an
adequate explanation of the decision, and that the person who takes
the decision would acknowledge, if truthful, that these were his
reasons. They were sufficient for this individual in this situation,
though there is no implication that they would be sufficient for him
or anyone else in a similar situation on another occasion. What we
mean when we speak of “inducing”, “persuading” etc. is to refer to
reasons that seemed adequate to this person on this occasion and on
which he therefore acted. Some or all of them may also have been
necessary to the decision; and it is often a good procedure heuristi-
cally to ask whether the same decision would have been reached in
the absence of a given reason. But the answer to this question is not
decisive of the existence of a causal or quasi-causal relation between
reason and action. It is no part of the meaning of “cause”, “induce”
etc. in interpersonal transactions that the person concerned would
not have acted as he did, apart from the conduct of the person who
is said to have induced him to act. The meaning of causal or quasi-
causal connection is here closer to sufficiency that to necessity, but
sufficiency not in the sense in which that term is used in regard to
physical sequences, but rather in the sense of what someone regards
as an adequate ground on which to decide and act in the circum-
stances.
119
4. C ONCLUSION
Why is causation important in tort law? One reason is that to insist

on causal connection between conduct and harm ensures that in gen-
eral we impose liability only on those who, by intervening in the
world, have changed the course of events for the worse. But what is it
to change the course of events for the worse? Is it to do (1) something
such that in the conditions prevailing harm comes about, or (2)
something in the absence of which harm would not have come about?
The first interpretation is that of the NESS theory (for physical
sequences); and a variant of that theory can be used to explain what
it is to influence human decisions and actions. The second inter-
pretation is that of the but-for theory.
The first interpretation is to be preferred. Causing a given harmful
result means, in relation to physical events, completing a set of con-
ditions sufficient to bring it about, i.e. providing a NESS condition
of that harm. In relation to influencing human action, it means pro-
viding a reason that the human agent accepts, often along with other
reasons, as sufficient to induce him to decide and act as he does (to
his or someone else’s detriment).
On the other hand the but-for theory, though a useful heuristic
device, provides the intuitively wrong answer in certain areas of over-
determination and indeterminacy. The explanation for this seems to
be that our concept of cause derives ultimately from our search for
recipes to bring about the outcomes we want. We look for sets of con-
ditions that are sufficient (or, in the case of human decisions, that may
be regarded by the agent as sufficient) to bring about a certain out-
come. Knowledge of these sets of conditions, though nearly always
imperfect, has many applications. In particular, it often enables us to
explain puzzling events and, in many contexts of law, including tort
law, to fix the outer limits of social responsibility for conduct.
That does not make the concept of cause a normative one. The
same concept of cause is used for discovering recipes, for explaining
events, and for assigning responsibility for outcomes. The normative
elements are supplied by substantive tort law, which define the con-
duct that entails or may entail tort liability, and prescribes when
causal connection between conduct and harm must be proved.
120

Tony Honore-Responsibility and Fault (1999) - Ch2,5

Uploaded by

Copyright:

Available Formats

Tony Honore-Responsibility and Fault (1999) - Ch2,5

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Tony Honore-Responsibility and Fault (1999) - Ch2,5

Uploaded by

Copyright:

Available Formats

!

1. N EGLIGENCE : T HE O BJECTIVE S TANDARD

or character, or a deficiency in intelligence, learning or experience.

incapable, in some the insane,9 in some the elderly.10 The incapable

may be possible to correct some faults of character, or to avoid acquir-

What justifications of this more ambitious sort are on offer?

One possibility is promising. At times the person whose conduct is

21. W. Blackstone, Commentaries (1st ed. 1769) vol. 3 p.122.

and to philosophers and many lawyers strict liability is a stumbling

2. R ESPONSIBILITY FOR O UTCOMES

On the contrary, given certain conditions, strict liability can be

22. T. Nagel, “Moral Luck”: Mortal Questions (1979) pp.24–38, at p.31.

26. Justinian, Institutes 1.17; Digest 17.2.55; 50.17.10.

quentialist argument for avoiding serious harm reinforces the non-

to ascribe them to others. Others in turn not only hold us responsible

3. C APACITY AND F REEDOM

It seems possible, then, to justify both outcome responsibility and the

outcome only when we could have done something different that

avoided the accident if “could have avoided” refers to a capacity exer-

ble interpretation. The capacity to act otherwise has hitherto been

37. H. and L. Mazeaud and A. Tunc, above n.15, vol. 1 s.431.

the time and place in question – it is implausible to suppose that on

40. Above n.2.

incapable of meeting the objective standard of care on the road. The

That is not an acceptable conclusion. Moreover, we are not driven

holing a six-foot putt. Whether their choices and performances are

1. T HE C AUSATION S TORY : M ODERN E FFORTS TO

In 1959 Herbert Hart and I proposed a thesis that we thought would

(contrary to our view) that every causally relevant condition is a

result – an INUS (Insufficient but Necessary part of an Unnecessary

12. They could also, of course, belong to the same set.

statement need not imply even the vaguest generalization”.16 If in a

2. C OMMON E LEMENTS IN THE NESS AND B UT -F OR T HEORIES

In my view both events and facts can be causally connected; and in

drove negligently (e.g. without keeping a proper lookout) caused the

23. The Empire Jamaica [1955] P. 52, 57–8.

navigated, not that it was navigated. There is a distinction between

24. Wright, Pruning, above n.2, at p.1014.

to produce the consequence, i.e. would the remaining set of condi-

27. Id. pp.1041–1042.

in Churchill’s case. The contradictory of “the defendant acted wrong-

33. Mackie, above n.2, at p.54; Toepel, above n.20, at p.55.

3. NESS VERSUS B UT -F OR : D IFFERENCES IN THE T HEORIES

I now turn to some of the examples that supporters of the NESS

A. Over-determination: Similar Causal Processes Culminating at the

of the two shots cannot be distinguished. Neither shot was necessary

on a given occasion. There can be different explanations of an event

37. Toepel, above n.20, at pp.66, 72–5.

B. Over-determination: Different Causal Processes

In other over-determination cases the two sets of conditions involve

39. Toepel, above n.20, at p.69.

regret that, despite criticism by Mackie,43 since endorsed by

43. Mackie, above n.2, at pp.44, 46–7.

C. Over-determination: Harm that Has Already Occurred

Although philosophers may debate the possibility of backward cau-

who has suffered two wrongs go without remedy, it is better to make

Indeterminacy presents a difficulty for the NESS theory. If the world

52. See, above, text accompanying n.15.