This preview shows page 1. Sign up to view the full content.
Unformatted text preview: The Philosophical Quarterly, Vol. , No. ISSN – July A DETERRENCE THEORY OF PUNISHMENT
B A E
I start from the presupposition that the use of force against another is justiﬁed only in self-defence or
in defence of others against aggression. If so, the main work of justifying punishment must rely on its
deterrent eﬀect, since most punishments have no other signiﬁcant self-defensive eﬀect. It has often been
objected to the deterrent justiﬁcation of punishment that it commits us to using oﬀenders unacceptably,
and that it is unable to deliver acceptable limits on punishment. I describe a sort of deterrent theory
which can avoid both of these objections. I
If punishment is to be justiﬁed, it must be predominantly by reference to
deterrence. I hold this for two reasons, neither of which I can argue for here.
First, each of the so-called retributive theories is either internally incoherent or has implausible moral implications.
Secondly, even if there were an otherwise acceptable retributive theory of
punishment, we should have to reject it, for I believe that violence is justiﬁed
only in self-defence against aggression, that is to say, against behaviour
violating a constraint which one has a right to uphold in self-interest (as is
usual, I include in this the defence of others’ self-interest). This claim, of
course, does not itself restrict us to a deterrent theory, for it is consistent with
other possible justiﬁcatory aims: direct prevention, for instance, or reform.
But most punishments have no signiﬁcant preventative eﬀect, and are not
intended to have any; and their beneﬁcial reformative eﬀect is scarcely more
The simplest version of the deterrence theory of punishment, that we
may punish oﬀenders to deter other potential oﬀenders, has usually been
rejected, for two reasons.
First, the theory commits us to accepting that in punishing one person in
order to deter others we are unacceptably using him. There have been many
ways of trying to avoid this. For now, I shall simply register my agreement
© The Editors of The Philosophical Quarterly, . Published by Blackwell Publishing, Garsington Road, Oxford , UK,
and Main Street, Malden, , USA. ANTHONY ELLIS with Tony Honoré that none of them will work.1 Broadly, attempts to avoid
the problem hold that punishing oﬀenders to deter others does not
unacceptably use them (because, for instance, they have implicitly assented
in advance to this treatment, or because they deserve their punishment), or
that they have forfeited the right not to be used. But I think that the
objection is correct, and the only response a proponent of the theory could
make would be to accept this, and simply tough it out.
Secondly, the theory lays down no acceptable limits on punishment.
Presumably if we are concerned simply with deterring oﬀenders, then the
appropriate level of punishment should be whatever is necessary to deter.
And this might be, in particular cases, either too lenient or too severe. Of
course, we can solve the problem by laying down independent moral
constraints upon the acceptable levels of punishment.2 But this may seem ad
hoc ; one might have hoped for a more uniﬁed theory, one in which the point
of the institution itself generates the constraints.3
In this paper I shall sketch a deterrent theory which avoids these
The theory understands punishment as a form of self-defence. I shall assume
that we have the right to use force in legitimate self-defence. I shall also
assume that groups may use force in self-defence, and that a society or a
nation may do so. Do we, in addition, have the right to threaten retaliation
against potential aggressors in order to deter them? It seems plausible to
think that at least sometimes we do; all else being equal, it must surely be
better to try to prevent aggression rather than to have to deal with it
forcefully when it occurs. Of course, all else may not be equal. Greater force
may have to be used to deter aggression than would be required to deﬂect it
once it has started. Or the methods used to deter could be intrinsically
unacceptable. Or the threat may have little chance of succeeding, or may
even constitute a perverse incentive to potential aggressors. But in the
absence of such factors, threats of retaliation are surely acceptable.
However, this raises a familiar question: if it is indeed permissible to threaten
some level of retaliation to deter aggression, is it also permissible to carry out
that retaliation if the threat fails to work? The justiﬁcation for threatening is
that it will prevent aggression (and any harm involved in a self-defensive Tony Honoré, Responsibility and Fault (London: Hart, ), p. .
This is the position of H.L.A. Hart (or one of his positions): see Punishment and Responsibility
(Oxford: Clarendon Press, ), ch. .
3 Cf. P. Montague, Punishment as Societal-Defense (Lanham: Rowman & Littleﬁeld, ), p. .
2 © The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT response). If the threat does not work, however, that justiﬁcation will not
carry over into a justiﬁcation for carrying out the threat, for carrying out the
threat will, by hypothesis, do nothing to prevent these harms. Retaliation,
then, seems unjustiﬁed (at least by the principles of self-defence).
But what if in addition to merely issuing the threat I somehow bound myself
to carry it out?4 I mean by this not just that I pledged myself to carry it out,
but that I somehow made it the case that if the threat were ignored, retaliation would be inevitable. We could suppose, for instance, that I constructed
a booby-trap surrounding my domain which, once it was constructed, I
could not dismantle, and whose operation was automatic as soon as anyone
crossed my border; I might then, with the intention of deterring anyone from crossing the border, announce that I had done this. If the threat
of retaliation failed, the actual retaliation would be automatic.
When retaliation against an oﬀender is carried out by such an automatic
system (I shall just say ‘punished’ from now on), can he complain that he is
being used? Initially, at least, it hardly seems so. That charge was based on
the claim that oﬀenders were being made to suﬀer in order to modify the
behaviour of others, and that seems no longer true. After all, it may be
known that the oﬀender’s punishment will have no such eﬀect; how then
can it be said that he is being punished in order to deter others? His punishment is simply a direct response to his own action, carrying no thought of
how it may aﬀect others.
However, the core of the worry may remain. Any particular oﬀender’s
punishment may be known to have no deterrent eﬀect, but the fact that punishments are actually administered must play some role, otherwise it would
not be justiﬁable to go beyond a system in which the threat is merely a bluﬀ.
The obvious role is making the threat credible to others. It may then seem
that the actual oﬀender is punished because the system requires that
oﬀenders in general must be punished in order to maintain the general
eﬀectiveness of the system. And this may seem essentially no diﬀerent from a
system in which we pick a number of people and punish them as a deterrent
demonstration to others, knowing that, for whatever reason, some of these
punishments would not in fact serve that purpose: here each oﬀender could
correctly say that he was being used to deter others, even though, in some
cases, the particular punishment would not have that eﬀect.
4 Warren Quinn also used this idea in a discussion of punishment: see ‘The Right to
Threaten and the Right to Punish’, Philosophy and Public Aﬀairs, (), pp. –. Daniel
Farrell has shown that Quinn’s argument rests upon a premise which is false: see ‘On Threats
and Punishments’, Social Theory and Practice, (), pp. –. The idea that punishment
can be justiﬁed as a form of self-defence has also been defended by Montague (see Punishment
as Societal-Defense), and by Farrell: see, for instance, ‘Deterrence and the Just Distribution of
Harm’, Social Philosophy and Policy, (), pp. –, and ‘On Threats and Punishments’. © The Editors of The Philosophical Quarterly, ANTHONY ELLIS But maintaining the credibility of the threat is not the only reason we
could have for making it a genuine one. We could imagine that, for various
reasons, a bluﬀ might actually work – suﬃciently so at least to maintain the
system’s eﬀectiveness. In such circumstances, there might still be reasons to
make the threat a real one. It may simply be more convenient. Or again, a
system of bluﬀs would require dishonesty, and we might object to that.
There is, of course, an indeﬁnite number of further possible reasons, having
nothing to do with the deterrent eﬀectiveness of the system. So long as one
of these is our reason, no one is punished because the system requires that
oﬀenders must be punished in order to maintain the deterrent eﬀectiveness
of the system. This should remove the last vestige of the worry that oﬀenders
are being used. It would in no sense be true that they are being punished in
an attempt to modify the behaviour of others. The threat is addressed to
each individually, and each is punished because he, individually, chose to
ignore the threat; the others, and their potential behaviour, are irrelevant.
There are, of course, other possible problems with such retaliatory
systems. One is that such systems may seem to involve the intention to cause
harm at a point in time when this harm may be known to have no beneﬁcial
eﬀects. But that is a tendentious description. Less tendentiously, the system
requires that at time t1, with the intention of preventing an evil, we set in
motion an irreversible process, which, if our intention is not realized,
will cause harm at t 2. Now it is certainly impermissible to intend, tout court, to
cause pointless harm; but it does not follow from this that it is impermissible
to intend to prevent harm by setting in motion a train of events which will, if
one’s intention is not realized, cause harm. That will depend upon the
totality of the circumstances, and in particular whether the agent has
observed certain conditions which I shall outline later.5
A diﬀerent sort of problem we may call the problem of scatter: unless the
system’s method of detection were infallible, the threat would be triggered
even by innocent people whom the system mistakenly took to be aggressors.
(Who is ‘innocent’ is a problem to which I shall return.) But this does not by
itself make such systems impermissible, for there are clearly circumstances in
which it is permissible knowingly to put innocent people at risk as long as
this is not our aim. Indeed, we do this constantly in our criminal justice
systems, requiring only that beneﬁts and risks be appropriately balanced.
And so long as there is the appropriate balance, there is, as yet, no objection
to such deterrent threats.
5 There is, of course, an immense and ancient literature to which I cannot here do justice
on the logic of the intentions embedded in threats. For a relevant discussion, see Farrell, ‘A
New Paradox of Deterrence’, in J.L. Coleman and C.W. Morris (eds), Rational Commitment and
Social Justice: Essays for Gregory Kavka (Cambridge UP, ), pp. –. © The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT One way of making it easier to achieve that balance would be by making
the system slightly less automatic, allowing it to be stopped when there is
reason to think that an apparent aggressor is in fact innocent. I might do this
by dividing the operation into diﬀerent functions and placing each in the
hands of a diﬀerent person. So one agent might be authorized to apprehend
suspected aggressors, another might be authorized to decide whether they
really were aggressors, and another might be authorized to administer
punishment if appropriate. But no one would have authority to deactivate
his part of the system except in special circumstances. From my point of
view, this system would be substantially similar to the simpler model I
sketched a moment ago. Once I have issued my threat, I have nothing
further to do in order for an aggressor to be punished: from my point of
view it is now automatic. It can now properly be said that the aim of the
system is deterrence, but that no one is punished in order to deter others.
It may seem that this model is too remote from reality to tell us anything
about punishment in the real world. However, I shall suggest that it is a
useful model of criminal justice systems in the USA and UK.
A criminal statute, we may say, lays down a threat, which we can think of
as a threat of retaliation against anyone in violation of its prohibition. And it
is, in the important sense, like the threat in the model I have suggested: once
someone has transgressed, the procedure goes forward fairly automatically.
Legal systems diﬀer, of course. In some, such as the German and Austrian
systems, prosecution of all oﬀences is, in principle, mandatory where there is
adequate evidence to sustain a conviction. In others, such as the UK and
US systems, there is broad discretion at most levels. It is enough, of course,
for my present purposes that we can imagine a realistic system in which the
normal course, though not necessarily mandatory, is prosecution. And, I
shall suggest, this is the best understanding of the US and UK systems.
The procedure is automatic in the sense in which the system just
imagined is automatic – a system which, of course, was meant to reﬂect the
separation of powers that characterizes modern democracies. The legislature issues threats, but it has no authority to decide whether those threats
are acted on in particular cases. Prosecutors and police have that authority;
but they in turn have no authority to decide whether a convicted person is
punished. Actually, in some jurisdictions the police can deal with ﬁnes for
minor traﬃc oﬀences. And in some systems prosecutors are empowered to
dismiss cases if the accused is prepared to pay some compensation to the
© The Editors of The Philosophical Quarterly, ANTHONY ELLIS victim, or even make a payment to a charity nominated by the prosecutor.
Judges (and jurors) decide upon sentences, but only in accordance with
previous law. Punishment is then in the hands of the executive, which, in
general, has no authority not to administer due punishment. It could be said
that we, through the legislature, issue a threat; ‘we’, however, cannot revoke
it. From ‘our’ point of view, punishment is automatic.
This may seem to ignore the discretion exercised by each branch of the
penal system. The police are granted some authority not to pursue a
complaint, or merely to issue a warning to an oﬀender; a prosecutor may
choose not to prosecute a case; a judge typically has some discretion in
sentencing. There are also exceptions to the general picture I have sketched,
such as the power of executive clemency. And in so far as this discretion
extended to permit the authority to take account of the deterrent eﬀects of
particular punishments it would, of course, reintroduce the original worry.
That discretion does so extend in our own legal systems seems clear. But
we are not bound to this. Though there is much discretion in the operation
of the law, none of it is arbitrary. When an oﬀender has been apprehended,
the default presumption is that he will be prosecuted and punished, and at
every level, if the law is not to take its normal course, this should be because
allowing it to take its normal course would be against ‘the public interest’. I
need not now give a full account of what is meant by that phrase, but it need
not be, and is not, construed in a narrowly utilitarian way: it already
encompasses notions of what is just or fair, for instance. It would not, then,
be an unrealistically radical step to embody in this notion the value of
individuals, which lays a moral constraint upon merely using them. So we
can plausibly say that if, at any stage in a particular case, the question is
raised of whether the law should take its normal course, that question should
not be answered by reference to whether the normal course will deter
others. (The UK Criminal Justice Act signalled a clear move away
from deterrent considerations in sentencing: individual sentences were to be
based primarily upon the ‘seriousness’ of the oﬀence, with the possibility of
incapacitative sentences for some oﬀenders who posed a great risk to the
We have, then, a system in which the legislature issues a threat whose
execution is, from its point of view, automatic: it need, and can, do nothing
further to ensure that it is carried out. Of course those who are involved in
carrying it out are granted some discretion in this. But so long as that
discretion does not extend to deciding individual cases on the basis of
deterrent eﬀectiveness, it can correctly be said that the point of the penal
system is to deter, but that the oﬀender is not punished in order to deter
© The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT IV
I turn now to the limits on punishment. If deterrent threats are made in selfdefence, then they are subject to the principles of self-defence, a central one
of which is that it is permissible to use in self-defence only such force as is
‘reasonably necessary’. What does that mean?
Imagine that I have only three possible responses to an act of aggression
already under way: doing nothing, rational persuasion, and counter-force.
What I may permissibly do will be a matter of the likely costs and beneﬁts,
For each course of action, the primary beneﬁt aimed at will be that the
threatened harm is avoided or minimized. So we need to know both how
great the threatened harm is, and how likely it is to occur given each course
of action. If the threatened harm is slight and merely speculative, then
normally the correct course of action will be to do nothing, though in such
circumstances there will usually be alternatives besides the three just mentioned. At the other extreme, a high probability of serious harm if I do not
use force will often justify it. But not always: in particular, if it has in any
case no chance of warding oﬀ the threat, then it will be pointless violence,
and cannot be justiﬁed under the principles of self-defence.
As to what count as costs, there will be some dispute. Some will consider the harm to the aggressor a cost, but others will not. Most would think
that the degree of the aggressor’s responsibility would need to be taken
account of, but not necessarily as limiting the victim’s rights; it may be
unkind to use force against an ‘innocent threat’, but it is not necessarily
unjust. Some may consider it a (moral) cost to oneself to indulge in violence,
even if it is justiﬁable to do so, whereas others will not. But there are some
costs that are clear: a forceful response to aggression may simply trigger a
forceful counter-response on the part of my aggressor, for instance; or it may
cause harm to innocent third parties. And on any plausible view, forceful
action will in general be more costly than non-forceful action.
These considerations, which I shall refer to as ‘the restraining considerations’, determine when it is reasonable to use force, and how much force it is
reasonable to use. Dispute about their precise content can be settled only by
appeal to their source. About that too, of course, there will be dispute, and
I can do little more here than state baldly what I believe it is. It stems, I
believe, from the general social necessity to minimize the use of violence. It
is clear that we need disincentives to violence, and recognizing the right to
self-defence is one of the most important of them. But forceful self-defence
© The Editors of The Philosophical Quarterly, ANTHONY ELLIS itself needs restrictions, for it too poses dangers: it may spill over into, or be
confused with, vengeance, with all of the attendant dangers; it may be used
as a cloak for violence desired for other ends; it may provoke further
violence from the aggressor, either in self-defence or in revenge; and it poses
dangers to those who are uninvolved. On the other hand, an absolute
prohibition on the use of force would obviously not be desirable, for this
would remove a major disincentive to aggression. And it would not be
generally adhered to; most would think that, in its encouragement of
aggression, the prohibition made no sense, and they would not respect it. In
any case, self-interest would ensure that people would not routinely forego
the use of force when this entailed a signiﬁcant sacriﬁce to themselves.
So a compromise has to be found between an absolute prohibition and
no restriction at all. It is impossible to say with precision just where it will
be located. But the guiding principle will be that the use of force should be
conﬁned to limits we can sensibly expect people to abide by, and force
which is reasonably necessary is simply force within those limits. There will,
obviously, be some disagreement about what those limits should be; but
there is also considerable agreement. Virtually everyone agrees, for instance,
that the victim of aggression should be required to forego self-defensive force
when there is a signiﬁcant disparity between the harm threatened and the
amount of force required to thwart it: this is a restriction that normal people
will see the point of, and will generally abide by. And when it is reasonably
necessary to use force, the restriction on the amount of force that may be
used has of course the same point. Its eﬀect is to reduce forceful countermeasures to the level that we can sensibly demand. We require, then, that
victims should sometimes be prepared to sustain some loss when responding
forcefully to an aggressor by choosing a less forceful measure than they
might prefer. There is, of course, no way in practice of saying precisely how
great a loss it is reasonable in general to require; it is no surprise, therefore,
that judgements vary over this matter.
To turn now to deterrent threats, again imagine that I have three possible
responses to a threat of aggression: I may do nothing, I may try to persuade
potential aggressors not to aggress, or I may threaten them with retaliation
in order to deter them.
The restraining considerations operate much as before: we need to know
the scale and likelihood of the threatened harm, the likelihood of success of
each course of action, and the costs of each course. But diﬀerent factors will
now become salient. (They, or counterparts, are of course present in the
case of direct self-defence too.) For instance, since we are dealing with
merely potential aggression, we shall need to think of how tempting is
the aggression to potential aggressors: the more tempting it is, the greater
© The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT the harm it will be justiﬁed to threaten (other things being equal). There is
also the possibility that the threat may act as a perverse incentive, or that
retaliation will lead some to seek revenge. And if we are talking of semiautomatic systems of retaliation, the possible harm to innocent third parties
will loom much larger.
It may seem natural to think that it would be permissible to threaten in
deterrence all and only that which it would be permissible to do in direct
self-defence. But this is not, of course, correct. The restraining considerations mentioned above generate crucial divergences. First, we may
take account of the probability that the threat itself may work, and the
retaliation thus may not be triggered. This will often relax the limits. Indeed,
other things being equal, there would perhaps be no limit at all on what
might be threatened if we could be absolutely sure that the threat would be
On the other side, there is the crucial question of immediacy. Potential
aggressors can often be deterred by the threat of much less force than would
be required to prevent their aggression once it has started. (It may be
necessary for you to kill someone who is trying to kill you; but he might have
been deterred if you had let it be known that if someone kills you, your
friends will administer him a beating.) More generally, all sorts of resources
that might have been available in advance will not be available in the heat
of the moment.
The issue of immediacy also raises the question of scatter. If I booby-trap
my house, and post a notice to this eﬀect, the trap may still be triggered by
innocent intruders who have not seen the notice. This problem is not
conﬁned to self-defensive threats, of course: it applies also to direct selfdefence. But it is presumably a more serious problem in the case of
automatic and semi-automatic systems of threatened retaliation than in
most cases of direct self-defence, and requires a greater stringency in the
restrictions governing their use.
Further complications arise when we move from individual self-defensive
deterrence to collective self-defensive deterrence on the part of society against
those who aggress against its members; there will then be special empirical
facts, and, arguably, special normative considerations to take account of.
But the goal of reducing violence will still be appropriate. A threat of retaliation may be a successful disincentive against aggression. But even, or perhaps especially, when wielded by the government, it must be hedged about
with restrictions. For one thing, we are rightly reluctant to give government
more coercive power than is necessary, for its agents are ordinary people, as
likely to misuse power as others. And we are reluctant to allow government
to resort to force too easily, for this is a lapse from the ideal relationship
© The Editors of The Philosophical Quarterly, ANTHONY ELLIS between government and citizens. And even when serious punishment is
completely justiﬁed, it is likely to foster alienation from the government on
the part of at least some citizens (the family and friends of the oﬀender, for
instance). There is also the question of likely harm to innocents. As well as
the wrongly convicted, there are others, such as the innocent families of
properly convicted oﬀenders. And the more serious the punishment we
threaten, the greater the costs we shall feel constrained to pay in the attempt
to avoid miscarriages of justice. The government is required, then, not to
use the threat of punishment to prevent crime if there is a reasonable, less
violent alternative; and it may threaten no more than is reasonably required
So though the principles are the same, what it is permissible for an
individual to do in direct self-defence will often diverge from what it is permissible for society to threaten in self-defensive deterrence. Overlooking this
may make it seem that punishment and self-defence are morally separate
Proportionality in punishment ... is more rigorous than proportionality in self-defence.
Using the death penalty for rape, for example, violates the principle of proportional
punishment.... Yet if a woman is threatened by rape, she may legally resist by killing
the aggressor. Even legal systems that have abolished the death penalty permit the use
of deadly force in the defence of vital interests. While proportionality in punishment
requires that the sentence ﬁt the crime, clearly more is permitted in self-defence.6 Fletcher speaks here of a distinction between what is permissible in selfdefence and what is permissible in punishment. But we could equally well
think of it as a distinction between what is permissible in direct, individual
self-defence and what is permissible in collective, self-defensive deterrence.
A woman faced with imminent rape faces an immediate, serious threat,
which there may be no way of repelling, short of killing her attacker. A
legislature deciding upon the sentence for rape faces a diﬀerent problem. Its
problem is to deter potential rapists, and a threat of death is not reasonably
necessary for this end; for the most part, the threat of lesser punishment
achieves it. So an individual woman may kill in order to resist rape, whereas
the state may not punish rapists by killing them.
But if the threat of imprisonment works only ‘for the most part’, why
should we not say that a threat of death, which may have greater deterrent
force, is reasonably necessary?
First, threats of death would be unlikely to achieve the desired result.
Most of those who are not deterred by the threat of relatively heavy prison
6 George P. Fletcher, A Crime of Self-Defense: Bernhard Goetz and the Law on Trial (New York:
Free Press, ), p. . © The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT sentences would not be signiﬁcantly more eﬀectively deterred by the threat
of capital punishment. In addition, juries would generally be unwilling to
send rapists to their deaths, and this would decrease its deterrent value. This
problem might be alleviated by restricting capital punishment to only a few
types of rape; but this would increase the uncertainty attaching to the
consequences of the oﬀence and would correspondingly reduce its deterrent
value again. Secondly, even if it did decrease the number of rapes, we
should have to set against that the likely costs. They include the destruction
of human life. Many people would think this justiﬁable to prevent an
otherwise certain rape, but not justiﬁed as part of a somewhat speculative
process designed to deter possible rapes. The likelihood of wrongful convictions would also strike most people as an unacceptable cost.7 The point is
illustrated in most legal systems. A woman immediately threatened with
rape may kill to protect herself. But in most civilized jurisdictions she may
not, in order to deter potential rapists (or even potential murderers), display
a deadly weapon with intent to use it; and even those jurisdictions which
allow the open carrying of ﬁrearms would baulk at a woman booby-trapping
her body with a bomb which would explode if she were attacked; these
measures would pose an unreasonable threat to innocent people. The
government, in adopting measures to deter rapists, is in the same position.
In order to explain the phenomenon remarked by Fletcher, then, we do
not need principles of punishment intrinsically more restrictive than the
principles of self-defence. On the other hand, if we derive the principles of
punishment from the principles of self-defence, then anything that is permissible in direct self-defence will indeed be permissible, in some imaginable
circumstances, in deterrent threats. Imagine, for instance, a desert island
inhabited by two people, one of whom is a determined rapist who can be
deterred only by the genuine threat of death. Would it in those circumstances be permissible to set up a deterrent threat of automatic deadly
retaliation? In these unlikely circumstances, it might be. It would depend in
part upon the amount of violence that this threat, and its realization, might
themselves trigger, and in part upon the relative evaluation of sexual
autonomy and human life. About that latter consideration the protectivedeterrent theory has itself, like other theories of punishment, nothing to say.
All that can be said is that if it is permissible to kill to avoid an otherwise
certain rape, then there will be some conceivable circumstance in which it
would be permissible to mount a credible threat of death to deter a potential
rapist. But the legal system in the real world never ﬁnds itself in circumstances like these.
7 Since ‘At least people have been exonerated and freed from death rows in states’ (The New York Times, August th, ). © The Editors of The Philosophical Quarterly, ANTHONY ELLIS V
What would a legal system look like that followed these principles?
The range of oﬀences that could be punished would be those alone which
could plausibly be brought under the umbrella of self-defence. It would be
impermissible to punish behaviour merely on the ground that it was immoral. And it would be impermissible to punish behaviour which harmed
only the agent. The most obviously punishable behaviour would be paradigm acts of aggression, such as rape and murder. But regulatory oﬀences,
such as traﬃc oﬀences, would also be covered, since the relevant
regulations, if legitimate, are constraints that we are justiﬁed in protecting in
It would have a robust and plausible requirement that the punishment
should ﬁt the crime: it would restrict punishment to what it is reasonable to
threaten in order to deter the oﬀence. Deterrent considerations, however,
would play no further part in the treatment of particular cases.
A particular aspect of the idea that the punishment should ﬁt the crime is
the idea that the innocent should not be punished, and the deterrence
theory can explain this. Punishment is retaliation against those who have
ignored a warning. The system is set up so as to be triggered by those
who ignore that warning, not by others; and no acceptable self-defensive
intention would be achieved by setting it up in any other way. ‘Punishing’
those who are innocent, then, will have no justiﬁcation, for they have not
failed to comply with the warning. Unless we were, unacceptably, using
them in a deterrent display, there would thus be no way of bringing their
punishment under the umbrella of self-defence.
We could, of course, set up the system so that innocents might suﬀer the
‘retaliation’. We could issue threats to punish the innocent relatives of
potential lawbreakers, for instance. Or we could threaten to harm entire
groups when only certain members of the group transgress. But it is clear
that punishment in these circumstances would be using people. Those who
are punished would, either directly or indirectly, be made to suﬀer as part of
an attempt to modify the behaviour of others; the deterrence theory can
quite consistently object to that.
Questions of guilt and innocence, of course, refer to more than whether
the defendant committed the prohibited act. First, punishment is hedged
around with the requirements of mens rea, or responsibility. One is not
punished at all unless one acted in an appropriate state of mind. And the
severity of one’s punishment may depend upon one’s state of mind when
© The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT one committed the oﬀence, whether one did it deliberately or negligently,
for instance. Secondly, the law recognizes a number of excuses, such as
necessity or duress, which can lead to an acquittal or mitigation of sentence.
All of this may seem to reﬂect nothing of self-defence, for our right of selfdefence against one who threatens us does not turn upon his level of
responsibility, or whether he is coerced into harming us. Why, then, are
such questions so important in the law?
Taking responsibility ﬁrst, a requirement of responsibility enters into the
limits of self-defence as soon as we move from direct self-defence to deterrent threats. The reason is simple: there is no point in issuing threats to
those who cannot heed them, and any violence occasioned by such threats
would therefore be without justiﬁcation. Anyone who suﬀered such violence would have been done an injustice. The law, then, may not justiﬁably
threaten with punishment, for instance, the severely mentally ill, and so may
not punish them. (It may, of course, detain them for the safety of themselves
and others.) The threat of punishment is not addressed to them.
But some of the mentally ill, kleptomaniacs, for example, are capable to
some degree of being deterred, and threats, though they may eventually be
ignored, are not without eﬀect. Is it then permissible to punish such people?
If it would not be pointless to threaten them with punishment, then punishment, within the restraining considerations, would not be unjust. So they
may be punished without injustice. It is a misfortune for them that they
suﬀer this obsession; but others are not required to bear the burden of their
misfortune, and may use whatever force is reasonably necessary to protect
themselves against its consequences. Whether carrying out punishment in
such circumstances would be stupid, or callous, are, of course, diﬀerent
To turn now to negligence, we seem to have the same right of selfdefence against a negligent threat as against a deliberate one. But, typically,
the law treats deliberate misconduct more severely than negligent misconduct. How can that be, if punishment is a form of self-defence?
We do indeed have the same rights of self-defence against negligent and
deliberate threats, but when we think about self-defensive deterrence, as
opposed to direct self-defence, we should be more concerned about the
prospect of deliberate misconduct than the prospect of negligent misconduct. Deliberate misconduct is in general more likely to cause harm than
negligent misconduct: we have more reason to be concerned about someone
who intends to kill than about someone who simply drives his car recklessly,
for, in general, the former will pose a much more serious threat. Additionally, deliberate misconduct threatens the fabric of the social order in a
way negligence does not. Further, it is generally easier to deter negligent
© The Editors of The Philosophical Quarterly, ANTHONY ELLIS behaviour, which requires only more care, than deliberate misconduct,
which requires abandoning a positively desired course of action. All of this,
other things being equal, justiﬁes a correspondingly less forceful measure.
To turn now to excuses, such as coercion, the actions of one who is
severely coerced are guided by reasons, and so this is not like the case of the
mentally ill. On the other hand, we can be quite sure that those who are
severely coerced will not respond to threats of punishment (those who
coerce them will usually ﬁnd it easy to ensure this); so threats would again be
pointless, and any violence they involved would be unjustiﬁed. So legal
threats are not addressed to those who are severely coerced; coercion is an
excuse. (Typically, only severe coercion is an excuse. It does not usually excuse in murder cases. But if threats of punishment are no more likely to
deter potential murderers than others, coercion should be an excuse here
too, as some legal theorists have argued.)
Does this mean that it is unjustiﬁable to issue threats of punishment
whenever we are sure that this will not deter? And does that in turn mean
that it is unjustiﬁable to threaten punishment to hard-core, recidivist criminals? The answer to the ﬁrst question is ‘Yes’; to the second question ‘No’.
‘Hard-core recidivist criminals’ may oﬀend whatever we do, but they would
commit far more oﬀences if they had carte blanche to do so. Deterrent eﬀectiveness is always a matter of degree, and one who eventually oﬀends may
none the less have been deterred to some degree, in the sense that the range of
circumstances in which he would oﬀend is restricted. The threat is thus not
pointless, and the violence it involves may be justiﬁable.
I shall deal with one last issue, the law of attempts. An unsuccessful
attempt to commit a crime may itself be a separate crime: one who tries to
murder but fails may be convicted of attempted murder. This may seem odd
from the point of view of the deterrence theory:
those who set about crime intend to succeed and the law’s threat has all the deterrent
force it can have if it is attached to the crime; no additional eﬀect is given to it if
unsuccessful attempts are also punished.8 But the deterrence theory can explain why we have separate oﬀences of
attempt. If we were setting up a system of semi-automatic retaliation, we
should not set it up so that the retaliation would be triggered only by the
completed oﬀence. Our ﬁrst thought might be that we should set it up, if
possible, so that it would be triggered before the oﬀence was complete, indeed
as soon as the oﬀender had fully committed himself to his course of action.
But the issue is more complicated. Given the goal of reducing violence, a
rational strategy would not only be to give potential aggressors an incentive
8 Hart, Punishment and Responsibility, p. . (Hart thinks that this is a ‘fallacy’.) © The Editors of The Philosophical Quarterly, A DETERRENCE THEORY OF PUNISHMENT not to aggress in the ﬁrst place; we should also want to give actual aggressors
an incentive to desist from their actions even when they had embarked on
them, and to give them such an incentive for as long as reasonably possible.
However, if we set the trigger point very late, we might encourage potential
oﬀenders to embark on oﬀences, and to continue with them at inconvenience to others, knowing that they could later withdraw if it seemed prudent.
If, on the other hand, we set the trigger point very early – as soon as
they had started planning the act, for instance – then this would deprive
oﬀenders of what might have been an eﬀective incentive to desist once the
act was under way; in addition, enforcing this would require enormous resources, and considerable general deprivation of liberty. The challenge,
then, is to ﬁnd the point at which one’s self-defensive strategy would be
optimized within the restraining considerations mentioned earlier.9
This goal in turn makes intelligible why unsuccessful attempts are usually
punished less severely than completed attempts. The emphasis on moral
desert that characterizes many theories of punishment makes this puzzling.10
But from the point of view of the protective-deterrence theory, moral desert
will of course not be the issue. All that will be relevant are the considerations
mentioned earlier: roughly, what will be the likely costs and beneﬁts of
threatening various levels of retaliation against, say, unsuccessful attempts?
We have reason to want oﬀenders to desist from their oﬀences even when
they have already embarked upon them; given this, it would be perverse to
threaten them with the full punishment as soon as they embark on the
oﬀence, for then they would have no incentive to desist as soon as the likelihood of apprehension is as great as when the oﬀence is accomplished. A
natural thought would be a sliding scale of retaliation: roughly, and other
things being equal, the further along the course of his action the oﬀender had
progressed, the greater would be the retaliation. This would not, of course,
be practical in the criminal law. Our best approximation is to ﬁx a point at
which an attempt really has been made, as opposed to mere preparations,
and a point at which the attempt has been completed; between those two
points we punish, but less severely than for the completed attempt.
Virginia Commonwealth University 9 And we ﬁnd in jurisprudential thought just what we should expect, given this aim. But for
a diﬀerent approach, see Duﬀ, Criminal Attempts (Oxford: Clarendon Press, ).
10 Duﬀ has tried to explain the puzzle in Criminal Attempts. I have criticized Duﬀ’s argument
in my ‘Criminal Attempts’, Journal of Applied Philosophy, (), pp. –. © The Editors of The Philosophical Quarterly, ...
View Full Document