Current Legal Problems Redmayne

The Ethics of Character Evidence
Mike Redmayne*
In everyday life, we often judge people on the basis of their past behaviour.
We avoid contact with those who have previously been rude to us, do not
revisit the restaurant where we got bad service, and appoint to jobs those
with good CVs and references. None of this is usually seen as problematic.
But are things dierent if we hold a defendants previous misbehaviour
against him in a criminal trial? To give an example (one which we will
keep returning to): suppose that D is on trial for burglary, and he has
a previous conviction for burglary. Should a court admit the previous
conviction as evidence of his guilt? This deceptively simple question raises
a number of complex issues.
In Australia, Canada, New Zealand, and the United States, the burglary conviction would be inadmissible, unless there was some factual
similarity to link it to the present charge. England and Wales, however,
has recently made a dramatic break with that common law position. The
* Law Department, London School of Economics. A version of this paper was given
as a lecture at UCL Faculty of Laws in November 2007: I am grateful to the audience for
their interest and their questions. Amit Pundik and Tom Poole gave helpful comments on
a written draft.
On Australia, and the English position prior to the Criminal Justice Act 2003,
see C. R. Williams, Approaches to Similar Fact Evidence: England and Australia, in
P. Mireld and R. Smith (eds.), Essays for Colin Tapper (London: LexisNexis, 2003),
2142; note also the High Courts conrmation of the Australian approach in Phillips v
R [2006] HCA 4, discussed in D Hamer, Similar Fact Reasoning in Phillips: Articial,
Disjointed and Pernicious (2007) 30 University of New South Wales Law Journal 60938.
For the Canadian position, see R v Handy [2002] 2 SCR 908, discussed in M. Redmayne,
Similar Facts, Familiar Obfuscation (2002) 6 International Journal of Evidence and Proof
243250. On New Zealand, see the New Zealand Law Commissions paper, Disclosure
to Court of Defendants Previous Convictions, Similar Oending and Bad Character, Issues
Paper 4 (Wellington, 2007), Ch 3. The US approach is covered in R. O. Lempert, S. R.
Gross, and J. S. Liebman, A Modern Approach to Evidence: Text, Problems, Transcripts and
Cases (3rd edn, St Paul, Minn: West Publishing, 2000), Ch. 5.
For example, if in the previous burglaries D had gained entry by using a blowtorch to
crack the glass, and that method was used in the burglary now charged, that might have
been sucient similarity to gain admissibility: see R v Mullen [1992] Crim LR 735.
Book 1.indb 371
12/11/2008 5:01:19 PM
Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015
Introduction
372
Mike Redmayne
The leading case is R v Hanson and Others [2005] 2 Cr App R 21. Various other cases
give a reasonable idea of where the line is now being drawn: R v M [2006] EWCA Crim
3408; R v Brima [2006] EWCA Crim 408; R v Atkinson [2006] EWCA Crim 1424; R v
Wood [2006] EWCA Crim 2270.
This paper generally refers to, and focuses on, the admissibility of previous
convictions. Many of the arguments surveyed here could be adapted to cover other
types of bad character evidence, such as non-criminal misconduct (e.g. R v Dolan
[2003] 1 Cr App R 18) or patterns of behaviour which have not (yet) led to conviction
(e.g. R v Smith [191415] All ER 262). However, occasionally the arguments apply
peculiarly to previous convictions: see further note 51 below.
See M. Redmayne, The Relevance of Bad Character (2002) 61 CLJ 684714.
See S. Lloyd Bostock, The Eects on Juries of Hearing about a Defendants Previous
Criminal Record: A Simulation Study [2000] Crim LR 73455; T. Eisenberg and
V. P. Hans, Taking a Stand on Taking the Stand: The Eect of a Prior Criminal Record on
the Decision to Testify and on Trial Outcomes Cornell Legal Studies Research Paper, 2007,
No 07012.
Book 1.indb 372
12/11/2008 5:01:19 PM
Criminal Justice Act 2003 has made it much easier than it was previously
to introduce previous convictions to prove guilt. The burglary example
described above is probably now a borderline case, but if D had two or
three previous convictions for burglary, they would in all likelihood be
admitted as evidence of his propensity to burgle, even if there were no particular factual similarities between the burglaries. This is controversial.
The interesting question, of course, is why this should be so, especially in
the light of our reliance on character in everyday decision making. This is
the question taken up in this paper.
The usual reason given for excluding previous convictionsand bad
character evidence more generallyfrom trials is that it might prejudice
the fact nder. While that claim should not be dismissed outright, there
are reasons to be sceptical of it. Crimes such as burglary are relatively
rare among the general population, but they have high recidivism rates.
It seems to follow from this that a previous conviction for burglary has
considerable probative value in proving the commission of a new burglary. While empirical studies do suggest that juries give some weight
to previous convictions, the eects are relatively modest. The studies,
therefore, do not obviously support the argument that previous convictions are more prejudicial than probative. For the purposes of this
paper, then, it will be assumed that juries do not give too much weight
to character evidence. Whether or not it is right, this assumption will
allow us to focus on the question whether there might be other reasons for objecting to the use of bad character evidence in criminal trials.
Might there be something morally problematic in judging people on
the basis of their previous bad acts? The remainder of this paper explores
this question.
373
Moral Accounts of the Criminal Trial

and their Limits
For a paradigmatic expression of this view, see L. Laudan, Truth, Error, and Criminal
Law: An Essay in Legal Epistemology (Cambridge: Cambridge University Press, 2006), especially Chs. 7 and 9.
See, for example, I. H. Dennis, The Law of Evidence (3rd edn, London: Sweet &
Maxwell, 2007), Ch. 2; P. Roberts and A. Zuckerman, Criminal Evidence (Oxford: Oxford
University Press, 2004), Ch. 1. The most inuential statement is probably R. A. Du,
Trials and Punishments (Cambridge: Cambridge University Press, 1986), Ch. 4.
R. A. Du, Punishment, Communication and Community (New York: Oxford
University Press, 2001). The view is applied to the trial in Dus Trials and Punishments,
ibid, and at greater length in R. A. Du, L. Farmer, S. Marshall, and V. Tadros, The Trial on
Trial: Volume 3: Towards a Normative Theory of the Criminal Trial (Oxford: Hart Publishing,
2007). Du also expands on the theme, with particular reference to the criminal law, in
Answering for Crime (Oxford: Hart Publishing, 2008).
Book 1.indb 373
12/11/2008 5:01:20 PM
It might at rst seem odd to claim that the rule against admitting a
defendants previous convictions can be justied by moral concerns. The
criminal trial, it might be thought, is about deciding who is guilty and
who is not. On this view, so long as defendants are treated humanely and
evidence is given appropriate weight, the idea that there can be a moral
objection to evidential practices makes little sense. There is, however,
a dierent, and inuential, view of the criminal trial, which holds that
there are moral constraints on trial practices. One way of grounding this
view is to claim that the criminal trial involves the communication of
censure to those who are found guilty or, in the slightly more ambitious
terms of Antony Du, that criminal trials call defendants to account for
their wrongdoing. From this, it can be argued that defendants cannot
be expected to understand the courts censure, and that the court cannot
call defendants to account, unless the trial respects their dignity and treats
them as moral agents. This moral view of the trial might explain why, for
example, courts exclude improperly obtained evidence: the message of
censure, or the courts moral standing, would be undermined if the court
were to condone ocial wrongdoing in the process leading up to the
nding of guilt. Although this account of the criminal trial raises various
questions, we will take it as a given for the purposes of the current enquiry.
The question then becomes: does this sort of view of the criminal trial create a problem for bad character evidence? To return to the initial example:
would using Ds previous conviction for burglary to prove his guilt be in
any way incompatible with treating him as a moral agent, or be in tension
with the censuring message conveyed by the verdict?
374
Mike Redmayne
R v Straen [1952] 2 QB 911.

In fact, Straen had been found unt to plead, but this detail is not important for the
purposes of the present discussion. Given the repeated use of Straen as an example, and as
one where guilt is obvious, it should be noted that not everyone is convinced of Straens
guilt: see B. Wonden, Insane, Guilty, or Neither?, The Guardian, 26 May 2001.
See e.g. Dennis, note 8 above, 1016; Roberts and Zuckerman, note 8 above, 15760;
A. L.-T. Choo, Evidence (Oxford: Oxford University Press, 2006), 12937. The phrase
is taken from Y. Kamisar, Comparative Reprehensibility and the Fourth Amendment
Exclusionary Rule (1987) 86 Michigan Law Review 150.
Book 1.indb 374
12/11/2008 5:01:20 PM
Suppose that we decide that there is a moral argument against using

previous convictions to prove guilt. We will then be faced with the
question whether this leads to a total prohibition on the use of previous
convictions. Those who take a moral view of the criminal trial presumably
think that criminal trials are justied; what justies them, they would
probably argue, is the good of retribution. But if retribution is valuable, a
complete ban on the use of previous convictions to prove guilt will be costly.
Straen provides a good example. Straen had previous convictions for
murdering young girls. The attacks were non-sexual and Straen made
no attempt to hide the bodies. When the unhidden, unmolested body
of a young girl was found close to where Straen had just escaped from
prison, his previous convictions were admitted as the primary evidence
against him, and led to his conviction. We would have to view the moral
structure of the criminal trial as being astonishingly rigorous if it were to
lead to the conclusion that previous convictions could not be used against
Straen, because this would almost certainly result in his acquittal; a high
price to pay in terms of lost retribution. Were the previous convictions
less probative, or the crime less serious, exclusion of the evidence would
be more acceptable. For this reason, moral accounts of the character rule
are likely to have some sort of exception built into them.
Given the foregoing, the simplest way to think about an exception to
a moralized character evidence prohibition would be in terms of balancing the probative force of the evidencewhich speaks to the question
of whether D deserves to be found guiltyand the seriousness of the
oence, against the moral cost of using the evidence. This assessment of
comparative reprehensibility is close to what many advocate as the appropriate test for the admissibility of improperly obtained evidence, such as a
confession gained after D was wrongfully denied access to legal advice.
Conceptually sound though this may be, when it comes to character
evidence it might not be very satisfactory in practice. There is no obvious
way to weigh the relevant factors against each other. It might be that even
minimal probative value would always outweigh the moral cost of admitting previous convictions, at least for mala in se, or it might be that the
375
For strong judicial reactions to torture evidence, see A (FC) and others v SOS for Home
Dept [2005] UKHL 71. In R v Nathaniel [1995] 2 Cr App R 565, the Court of Appeal did
exclude improperly retained DNA evidence, but this decision is controversial: cf Dennis,
note 8 above , 3145 and Attorney-Generals Reference (No 3 of 1999) [2001] 1 All ER 577,
where the House of Lords hints that it would have admitted the evidence in Nathaniel.
Some commentators are prepared to endorse the decision in Nathaniel, but they approach
the issue from a perspective dierent to the legitimacy/balance one discussed here: see
K. Grevling, Fairness and the Exclusion of Evidence Under Section 78(1) of the Police and
Criminal Evidence Act (1997) 113 LQR 66785; A. Ashworth and M. Redmayne, The
Criminal Process (3rd edn, Oxford: Oxford University Press, 2005), 3301.
See Redmayne, Similar Facts, note 1 above.
Book 1.indb 375
12/11/2008 5:01:20 PM
moral cost is so high that we should conclude that even Straens previous
convictions are inadmissible. Perhaps things are not quite as indeterminate as this, though. When it comes to improperly obtained evidence,
gut reactions do seem to lead to a reasonable degree of consensus, at least
in extreme cases: evidence obtained by torture will always be illegitimate,
but it is appropriate to use improperly retained DNA to prove rape.
And the balancing process may well become less hazy as regards previous
convictions once we have said more about the moral concerns that they
do raise. With improperly obtained evidence, though, the comparison
of the wrong done by the state with the wrong done by the defendant
does look to involve the balancing of rough commensurables: it seems as
though we will just have to accept that things will be much vaguer if we
take a balancing approach to previous convictions.
There may be a way for advocates of a moralized character evidence
prohibition to avoid this problem. It might be argued that the use of
character evidence in cases like Straen is permissible because it does not
engage the moral prohibition at all. Straen and cases like it might be
seen as dierent in kind, rather than just in degree, from our burglary
example. In fact, the common law provides some resources for thinking about the character evidence prohibition in this way. The common
law admitted Straens previous convictions, along with those of various other defendants. It is not easy to generalize about the common law
approach to previous convictions, because the shape of the exception to
the character rule varies across time and between jurisdictions. But at least
one inuential formulation of the rule suggests that the exception reects
a dierence in kind: the rule against bad character evidence is sometimes
said to exclude forbidden reasoning, a concept often explicated in terms
of reasoning from a propensity to commit crime (or reasoning from
general propensity, contrasted with permissible reasoning from a specic
propensity to commit crime in a particular way, such as by killing young
girls and not hiding their bodies). This theme in the common law of
376
Mike Redmayne
The Signicance of Desistance

The moral accounts we will examine all seem to revolve around similar themes, though the themes get developed in various dierent ways.
A central theme involves the signicance of desistance from crime, so
we should begin by saying something about desistance, and noting one
relatively straightforward way in which it complicates the inference that
might be drawn from character evidence. The argument for the admissibility of previous convictions assumes that a previous conviction shows a
propensity to commit crime. Support for the argument can be found in
the fact that recidivism is one of the brute facts of criminology. However,
desistance is also a signicant fact, and one that has recently drawn the
attention of criminologists. At some stage, nearly every oender stops
committing crime. Does the possibility of desistance undermine the use
of previous convictions to demonstrate propensity to commit crime?
At rst sight, the concept of desistance might not seem to add anything
to the concept of propensity. Desistance is the ip-side of propensity; a
strong propensity to desist is simply a weak propensity to commit crime.
And the statistics which form the basis of the argument that previous
convictions show propensity already take desistance into account; desistance is thus part of the measure of propensity. For example, if a group of
subjects with previous burglary convictions is followed up over a two-year
Previous oending has been said to be the most signicant predictor of future
oending: see J. Monahan et al, Rethinking Risk Assessment: The MacArthur Study of Mental
Disorder and Violence (New York: Oxford University Press, 2001), 445.
Book 1.indb 376
12/11/2008 5:01:20 PM
character evidence might even be taken to support the argument for a

moralized version of the rule, for it hints that the common law has long
been aware of the ethical objections to character reasoning, even if it has
not attempted to theorize them nor make them explicit.
For now, it suces to point out that there would be a high cost if the
character rule was taken to be an outright ban on using previous convictions
to prove guilt, that there are two possible approaches to justifying the
admission of bad character evidence, and that the forbidden reasoning
approachwhich sees a sharp distinction between Straen and cases such
as our burglary exampleis initially the more attractive one. We will say
more about the viability of that approach below. With that by way of
introduction, we can now turn to explore various ways in which a moral
objection to character evidence might be formulated.
377
S. J. Gould, The Median Isnt the Message, in The Richness of Life: The Essential
Stephen Jay Gould (London: Vintage, 2007), 29.
See J. H. Laub and R. J. Sampson, Shared Beginnings, Divergent Lives: Delinquent
Boys to Age 70 (Cambridge, Mass: Harvard University Press, 2003), especially Ch. 6.
R. J. Sampson and J. H. Laub, A Life Course View of the Development of Crime
(2005) 602 Annals, American Academy of Political and Social Science, 1245, 41.
Book 1.indb 377
12/11/2008 5:01:20 PM
period, and 60 per cent of them commit a burglary during this time but
the other 40 per cent desist, then the 60 per cent conviction rate provides
the basis for a measure of propensity. However, this very example hints
at the problems lurking here. What if the defendant in our example is
one of the 40 per cent who desist? One response to this concern is that the
strategy of averaging over the group is perfectly proper. Indeed, it appears
to be an inevitable part of evidential inference. This eyewitness might be
one of the 20 per cent who make false identications, but we take this
into account, not by refusing to draw an inference, but by tempering the
weight we put on the inference through noting that the probability of
true identication is only 0.8. We should, however, bear in mind Stephen
Jay Goulds warning about the use of averages: variation is the hard reality,
not a set of imperfect measures for a central tendency. Means and medians
are the abstractions. If there are identiable dierences between those
who reoend and those who do not, then basing an inference on the
average tendency could be said to over-emphasize the abstraction and
to ignore the hard reality. Criminal career research does indicate some
factors that bear on the likelihood of desistance: certain life events, such
as marriage and employment, can lead to sharp breaks in oending
careers. But Robert Sampson and John Laub caution against thinking
that we can easily place subjects in distinct groups in terms of predicting
risk of re-oending: Even if all risk factors were measured without error,
our framework posits the continuous inuence of human agency and
randomizing events, leading to heterogeneity in outcomes . . . and a lack
of causal prediction. The reference to agency highlights the fact that the
defendant himself may have tried to change, may have tried to distance
himself from the grim mean of recidivism.
Various issues are emerging here. As stated at the outset, the aim of
this paper is to assess moral objections to the use of character evidence
to prove propensity to commit crime. Some of the issues relating to the
signicance of desistance, however, seem to involve epistemic objections
to the use of previous convictions, for they highlight the inferential danger of relying on average tendencies. It appears, though, that in this area
epistemic and inferential objections are often dicult to separate: treating
D as if he were like the average might be both morally objectionable and
378
Mike Redmayne
For further discussion of this approach, and the contrast with Alex Steins theory of
evidence law, see M. Redmayne, The Structure of Evidence Law (2006) 26 OJLS 80522.
For the detail of Steins account, see A. Stein, Foundations of Evidence Law (Oxford: Oxford
University Press, 2005).
This is the concern ultimately raised by Peter Tillers in What is Wrong with Character
Evidence? (1998) 49 Hastings Law Journal 781834.
A. Camus, Ltranger (Paris: Gallimard, 1942).
(1692) 12 How St Tr 833, 864.
Book 1.indb 378
12/11/2008 5:01:20 PM
inferentially suspect. For now, we will concentrate on the more epistemic

end of the spectrum. The rest of this paper will examine the signicance
of desistance from more moral perspectives.
If we have no information about the defendant that might place him in
a group with a lower than average risk of re-oending, then the analogy
with eyewitness evidence, above, suggests that treating him as the average
is not problematic on inferential grounds. After all, for all we know,
D might have characteristics that put him at a higher than average risk
of re-oending; by treating D as the average we may not be presuming
the best about him, but nor are we presuming the worst. There is, however, some reason to be suspicious of the eyewitness analogy. With the
eyewitness, we will want to enquire into the evidence, to see if there is
any particular reason to doubt itor, indeed, not to doubt it. This will
typically be done by cross-examination of the witness about factors such
as identication conditions, possible bias, and the like. But what is the
parallel with previous convictions and the risk of recidivism? If D wants
to show that he is less of a risk for re-oending than the average person,
just how is he supposed to do that? It might be very dicult for him to
distance himself from the bad acts of others that give the evidence its probative force. It is not easy for someone to show that they have changed,
but even in the situations mentioned above, where D can point to factors
such as marriage and employment which lower his risk of re-oending,
there is something quite disturbing about the idea of these being explored
in evidence: should D be open to cross-examination about the state of his
marriage? Could his probation ocer be called to give evidence about his
risk status? The concern here is that the trial might become a general moral
enquiry into Ds life, something we would nd distasteful. Our fears
about this sort of trial are captured in Camus LEtranger, where the central character has his lack of emotion at his mothers funeral used against
him in court. They also seem to be captured in Holt LCJs response to
the prosecutions attempt to introduce previous convictions in Harrisons
Trial: Hold, what are you doing now? Are you going to arraign his whole
life? Away, away! That ought not to be; that is nothing to the matter.
379
S. Field, State, Citizen and Character in French Criminal Process (2006) 33 Journal
of Legal Studies 52246.
Book 1.indb 379
12/11/2008 5:01:20 PM
While this is an interesting line of objection to character evidence,

there are a number of reasons to be sceptical of its ability to provide a
foundation for the exclusion of bad character evidence. While our initial
reaction might be that the criminal trial should not be a general enquiry
into a persons life, it turns out that something like this occurs in France.
Stewart Fields fascinating research into the French criminal trial shows
that in more serious cases the trial starts by presenting the accuseds life
story. Previous convictions are freely admitted. One way of interpreting French practice, in fact, is to assume the two have to go together: if
we want to reveal previous convictions, we need to set them in context,
to show how they might be connected to circumstances in a persons life
that no longer hold. This might give the defendant some protection from
having a propensity inference drawn against him too automatically.
A second problem with the sort of objection outlined above is that it
is potentially too strong. Straen might argue that if his previous convictions are introduced it will be dicult for him to rebut the inference
drawn from them, hard for him to show that he has changed. As was
argued earlier, excluding the evidence in Straen is an uncomfortably
high price to pay for the moral values that underlie the exclusionary urge.
There are various responses to the Straen problem. One is to argue that
we might want to make an exception for cases like Straen because the
crime is serious and the evidence strong. That is a reasonable response
but, as noted above, it provides us with no obvious threshold at which
to draw the line. It would not tell us, for example, whether the common
law or the Criminal Justice Act approach is preferable. A second response
is to argue that cases like Straen are dierent in kind to our burglary
example. In Straen it might be said that we are relying on a specic propensity rather than a general one. That is a popular move in attempts to
defend the common law threshold, and more will be said about it below.
For now, though, we should note that the distinction between general
and specic propensity does not appear to be helpful here. The objection
to character evidence that we are considering is based on the diculty
a defendant might face in challenging bad character evidence, or in the
concerns about privacy and the moralization of the trial that might arise
if we attempted to place the signicance of the previous conviction in
context. These concerns just do not seem to drop out of the picture in
cases of specic propensity. A third way of dealing with Straen is to argue
that the evidence proves that Straen has not changed, so the concern
380
Mike Redmayne
But it may be that something like this is driving the Australian approach to admissibility which, in theory, does require a very high threshold for admissibility. Pfennig, the
leading authority, seems to condition admissibility on there being no reasonable view of
the evidence consistent with the innocence of the accused (Pfennig v R (1995) 182 CLR
461, 484, 490). See further Hamer, note 1 above.
See S. Maruna, Making Good: How Ex-Oenders Reform and Rebuild Their Lives
(Washington, DC: American Psychological Association, 2001), Ch. 1.
M. Dhami, D. Mandel, G. Loewenstein, and P. Ayton, Prisoners Positive Illusions
of Their Post-Release Success (2006) 30 Law & Human Behavior 63147; R. Moore,
Adult Oenders Perceptions of Their Underlying Problems: Findings from the OASys
Book 1.indb 380
12/11/2008 5:01:20 PM
that it might be dicult for him to show otherwise has no traction. This
manoeuvre, too, has problems, for it seems to assume Straens guilt; and
even if that were thought to be acceptable, it appears to establish a high
threshold for admissibility.
There is another reason why the objection we are considering
which involves the diculty D faces in distancing himself from the
character inferencemay be too strong. It might apply to types of
evidence other than bad character. Motive is an obvious example. If the
defendant had a motive to kill his rich uncle, that can be used as evidence against him at a trial for murdering the uncle. But this defendant
seems to face similar diculties to the ones just outlined: it is going to
be hard for him to show that he is not the sort of person to be swayed by
nancial motives, at least without opening up a wide-ranging inquiry
into his life.
Finally, there is a more complex point, which touches on the nature
of propensity. The line of thought we are pursuing presumes that D
may have a previous conviction, but has desisted from re-oending
and thus lost the propensity to commit crime. But the realist metaphysics that seem to be relied on here might be questioned. Propensity
is not necessarily some almost tangible characteristic of oenders that
comes and goes; propensity may be a construct that wethe fact nders,
as it wereuse to express our uncertainty about human behaviour.
Criminologists agree that desistance is dicult to identifyone can
never be sure that a person has desisted, at least until they dieand it
is surely something that is dicult to achieve, just as is any long-term
change in behaviour or character. Even outside the context of crime,
if we want to claim that we have changedgot rid of some bad habit,
saywe should be honest enough to be circumspect about the claim.
It also turns out that oenders are over-optimistic about the possibility
of change; prisoners predictions of their chances of re-oending fall
some way below the grim reconviction rates, so an oenders argument
that he has changed probably has to be viewed sceptically. Propensity
381
Autonomy
We do not have to look far to nd a further objection to using bad
character evidence to prove guilt. In the previous section we noted the
diculty that defendants may face in trying to show that they are not
average members of the group of previous oenders; there is also a more
purely moral side to this problem. By drawing an inference based on the
behaviour of other people, we might be treating defendants in a way that
conicts with the moral context of the criminal trial. According to David
Wasserman, courts are reluctant to base verdicts on the frequency of
misconduct by others or by the defendant himself , because this is inconsistent with laws commitment to treat the defendant as an autonomous
individual, free to determine and alter his conduct at each moment. An
objection to using previous convictions to prove guilt that draws on the
value of autonomy turns out to be quite common; it occurs, in various
Self-Assessment Questionnaire, Home Oce Research Findings 284 (London: Home

Oce, 2007).
There is some controversy on this point. Gottfredson and Hirschi have argued
that propensity does not decline with age, but the better view seems to be that it does:
cf M. R. Gottfredson and T. Hirschi, A General Theory of Crime (Stanford, CA: Stanford
University Press, 1990); Sampson and Laub, note 18 above. But no one denies that
oending declines with age: see, generally, A. R. Piquero, D. P. Farrington, and A. Blumstein,
Key Issues in Criminal Career Research: New Analyses of the Cambridge Study in Delinquent
Development (Cambridge: Cambridge University Press, 2007), Ch. 4.
D. T. Wasserman, The Morality of Statistical Proof and the Risk of Mistaken
Liability (1991) 13 Cardozo Law Review 93576, 94353.
Book 1.indb 381
12/11/2008 5:01:20 PM
may weakenindeed, this seems to hold for most oenders as part

of the aging processbut claiming that it vanishes, or, at least, that
we would ever have reason to think that it has vanished, is more problematic. This is not to deny the possibility of change, nor to deny the
signicance of factors such as marriage and employment, identied
above. But the points just made do take some of the force out of the
objection that it is unfair to use previous convictions against a defendant
who claims to have changed.
This rst attempt to identify moral reasons for excluding previous convictions has come to a rather indeterminate conclusion. But it by no means
exhausts the moral arguments that can be made. As we proceed to examine
other arguments, we will see that the signicance of desistanceof moral
changecontinues to play a signicant role.
382
Mike Redmayne
Tillers, note 20 above. Though raising the autonomy issue, Tillers ends up attaching
little signicance to it.
H. L. Ho, A Philosophy of Evidence Law: Justice in the Pursuit of Truth (Oxford: Oxford
University Press, 2008), Ch. 6, which builds on H. L. Ho, Justice in the Pursuit of Truth:
A Moral Defence of the Similar Facts Rule (2006) 35 Common Law World Review 5179.
A. E. Acorn, Similar Fact Evidence and the Principle of Inductive Reasoning: Makin
Sense (1991) 11 OJLS 6391, 68.
Note 9 above. A very brief sketch of the arguments developed in The Trial on Trial
appears in Du, note 8 above, 131.
See N. Shover, Great Pretenders: Pursuits and Careers of Persistent Thieves (Boulder,
Colo: Westview, 1996), Ch. 4; Maruna, note 25 above, passim.
Book 1.indb 382
12/11/2008 5:01:20 PM
dierent ways, in work by Peter Tillers, Hock Lai Ho, and Annalise
Acorn; it also appears in The Trial on Trial, the recent account of criminal trials developed by Antony Du, Lindsay Farmer, Sandra Marshall,
and Victor Tadros.
Autonomy-based objections to bad character evidence face a
number of initial problems. First, the fact that behaviour is predictable does not necessarily imply a lack of autonomy: I am regularly to be
found in a particular pub on a Tuesday night, but the fact that my presence there can be predicted with a fair degree of reliability does not say
much about my autonomy. Second, there is something rather paradoxical
in the claim that oending behaviour, in particular, is non-autonomous.
Autonomous means, literally, making ones own law, and there is an obvious sense in which criminals are better at doing that than the law-abiding.
Indeed, the criminological literature identies a certain rebellious, antiauthoritarian streak in many persistent oenders; the defendant with
previous convictions would thus seem to display his autonomy more by
continuing to oend than by ceasing to oend. Third, there is a marked
asymmetry in the worries about autonomy here. Motive, again, is a useful counter-example. When we say that the fact that the defendant has a
motive for murdering his rich uncle increases the probability that he did
so, the autonomy concern does not seem to be engaged, even though,
in Wassermans terms, we are basing a verdict on the frequency of misconduct by others, for motive is only evidence of guilt because we know
that there is a general tendency to act on motive. Similarly, when the
defendant oers good character evidence in his favour, we do not worry
about the possibility that he has suddenly demonstrated his autonomy by
becoming a bad person. Fourth, it is in just those cases where autonomy
might appear to be most in question that we are most ready to admit character evidence. We might wonder whether Straen the child murderer is
really autonomous: from Straens point of view, at least, he may have
little choice but to murder the young girls he comes across, but Straen
383
Motives and Reasons

We have twice used the example of motive as a way of suggesting that
certain objections to character evidence are over-broad. This strategy,
however, might be undermined if there was a signicant dierence
between evidence based on the existence of a motive and evidence based
on previous convictions. In fact, both Ho and Du et al suggest that there
might be a dierence between the two types of evidence, a dierence
that has moral resonance. Hos account is especially signicant because,
he argues, it can be used to explain why some previous convictions are
dierent in kind to others; in other words, it could buttress the moral
account of the character evidence prohibition against the sort of criticism suggested earlier: the criticism that moral accounts are vulnerable if
they cannot explain why we should draw the admissibility threshold at a
particular point.
Something like the following argument can be found in both Ho and
Du et al: when we say that D has a motive for committing crime
he killed his wife because she was having an aair, or his rich uncle in
order to inherit under the willwe are oering an explanation for his
behaviour. While motive evidence may have a similar inferential structure to bad character evidencewe look at the behaviour of others to
See e.g. N. Arpaly, Which Autonomy?, in J. K. Campbell, M. ORourke, and
D. Shier (eds.), Freedom and Determinism (Cambridge, Mass: MIT Press, 2004), 17388.
Du et al note the problem of the vagueness of the term (note 9 above, 130 fn 5), and
suggest that it is better to talk in terms of responsibility. Nevertheless, autonomy is used
in their discussion of character evidence, perhaps because in that context replacing it with
responsibility would make the arguments sound less intuitively persuasive.
Du et al, Trial on Trial, note 9 above, 2546; Ho, Philosophy, note 30 above,
298306.
Book 1.indb 383
12/11/2008 5:01:20 PM
is a case where there is a strong argument for admissibility of previous

convictions.
We should resist jumping from these objections to the conclusion that
there is nothing to the autonomy objection. There is plenty to niggle with
in the points above; further, autonomy is a vague concept, so perhaps the
clearest lesson to draw from the foregoing is that the autonomy objection
needs to be rened and expressed more carefully. We will shortly move
on to look at the particular ways in which the autonomy objection is
developed in the literature. But rst, we need to side-track slightly to deal
with one somewhat separate point.
384
Mike Redmayne
Makin v AG NSW [1894] AC 57.

R v Ball [1911] AC 47. This account also seems to face diculties in cases where
what links D to the crime is a particular modus operandi which throws no light on Ds
motivations (e.g. Mullen, note 2 above), or where D is connected to dierent crimes partly
through geographical and temporal proximity rather than through particular similarities
between the crimes (e.g. R v John W [1998] 2 Cr App R 289).
Against this, it might be argued that oenders motivations are not this transparent.
Katz, for example, makes much of the complex layers of meaning which surround criminal
activity (J. Katz, Seductions of Crime: Moral and Sensual Attractions of Doing Evil (New
York: Basic Books, 1988)). But if this is right, it surely applies to Ball and the Makins as
much as to the burglar.
Book 1.indb 384
12/11/2008 5:01:20 PM
infer the likely behaviour of Dthe explanatory structure in motive

cases, in particular the fact that it taps into Ds reasons for action, might
be thought to be signicant. This sort of inferential structure does seem
to respect D as a responsible agent, in the sense of someone who can
respond to reasons. It does not treat D as simply having some opaque
and ungovernable urge to commit crime. This argument could then be
extended to mark the dierence between general and specic propensity, and so to justify the common law admissibility threshold. Cases
of specic propensity, it could be argued, illuminate Ds reasons for
committing crime.
This is an intriguing argument, but it relies on distinctions which
do not really hold up. An initial question is: why should the fact that
D responds to reasons be signicant in the context of the moral structure
of the trial? The answer might seem obvious: only those who respond
to reason are appropriate subjects of criminal responsibility. But this
response will collapse the distinction between character and motive. If
we use Ds previous conviction for burglary as evidence that he has committed burglary, there is no obvious sense in which we are treating him as
unresponsive to reason (that is, unless we interpret reason in the Kantian
sense of right reason, but surely the D who murders his unfaithful wife
also does not respond to reason in this sense). Indeed, it is not hard to
read a motive into Ds actions: he burgles houses to make money, just as
we read into the Makins actions (Makin being a case where the common
law had no doubt about admitting the evidence) that they kill children
to make money, and just as we read into Balls actionsto pick an
example used by Hothat he is sexually attracted to his sister.
There is nothing here, then, to suggest why motive should be treated
dierently to character evidence, or how we might distinguish general
from specic propensity. But a dierent way of framing the objection to
bad character might be developed from the preceding discussion. If Ds
responsiveness to reasons is important to the criminal trial, this might
385
be because such people are able to respond appropriately to the censuring

message in the criminal verdict. If we use Ds previous conviction as
evidence that he has a propensity to burgle, might we be implying that D
has not responded to censure, and hence reason, in the appropriate way?
And might that be problematic? In the following sections, we will examine
arguments that take up these points.
In our initial assessment of the autonomy argument against previous

convictions, we noted that autonomy concerns do not seem to surface
when it comes to good character evidence, and that there is a sense in which
criminals are more autonomous than the law-abiding. It was suggested
that, rather than taking these to be knock-down objections to the autonomy argument, we might instead respond to them by trying to rene
the concept of autonomy. Perhaps, then, what is really important is not
autonomy per se, but the freedom to change our ways, or, slightly dierent, to become better people than we already are. Some writerswho are
not addressing character evidence issuesprovide resources for thinking
that this particular sort of freedom may be morally signicant. There is a
bizarre, but intriguing, philosophical debate on whether pre-punishment
is permissible: that is, whether it could be appropriate to punish someone
for a crime before they commit it. To make sense of this question we
have to suppose that we have strong evidence that a person will commit a crime, and that after the event it will not be practicable to punish
him, perhaps because he will leave the jurisdiction. We also need to be
clear that we are talking about retributive punishment, not preventive
detention. Saul Smilansky suggests that there are moral reasons for not
allowing pre-punishment in this situation. For him, the unacceptability
of pre-punishment ows from deep Kantian intuitions about respecting
autonomous moral personality and choice. Pre-punishment would treat
people as objects, as if they had no choice, as if they could not change
See C. New, Time and Punishment (1992) 52 Analysis 3540; S. Smilansky, The
Time to Punish (1994) 54 Analysis 5053; C. New, Time and Punishment (1995) 55
Analysis 6062; D. Statman, The Time to Punish and the Problem of Moral Luck (1997)
14 Journal of Applied Philosophy 12935; R. Sorensen, Future Law: Prepunishment and
the Causal Theory of Verdicts (2006) 40 Nos 16683. The problem of prepunishment is
taken in a slightly dierent direction in S. Smilansky, Determinism and Prepunishment:
The Radical Nature of Compatibilism (2007) 67 Analysis 34749; I. Haji, Libertarian
Openness, Blameworthiness, and Time, in Campbell et al, note 34 above, 13550.
Smilansky, The Time to Punish, ibid.
Book 1.indb 385
12/11/2008 5:01:20 PM
Moral Improvement
386
Mike Redmayne
A. Margalit, The Decent Society (trans. N. Goldblum; Cambridge, Mass: Harvard

University Press, 1996), 7075.
Du et al, note 9 above, 1134.
Ho, Philosophy, note 30 above, 300.
For a similar point in a dierent context, see R. Reiner, Law and Order: An Honest
Citizens Guide to Crime and Control (Cambridge: Polity Press, 2007), 1820.
This example intentionally reects a debate in punishment theory about incapacitative punishment. Various views are usefully collected in A. von Hirsch and A. Ashworth
(eds.), Principled Sentencing Readings on Theory and Policy (Oxford: Hart Publishing, 1998),
88140. The bomb example is used by Norval Morris at 108. For Antony Dus contribution to this debate, see R. A. Du, Dangerousness and Citizenship in A. Ashworth and
M. Wasik (eds.), Fundamentals of Sentencing Theory: Essays in Honour of Andrew von Hirsch
(Oxford: Oxford University Press, 1998), 14163.
Book 1.indb 386
12/11/2008 5:01:20 PM
their minds. In a similar vein, though on a dierent topic, that of decency,

Avishai Margalit argues that there is a deep link between respect and the
possibility of change: respecting people preserves the idea that the future
is open; even hardened criminals, Margalit suggests, deserve respect
because of the possibility that they might change.
When Du et al and Ho talk of autonomy in relation to character evidence, it seems that this is what they are really getting at. Du et al argue
that [t]o take the fact of prior wrongdoing as evidence of . . . guilt of a new
oence is inconsistent with . . . respect for the defendant as a responsible
agent: as a responsible agent she could have put her past crime behind
her, and come to guide her actions by the appropriate reasons that the law
provides or expresses; but we treat her as if her past conduct determines
her present conduct. For Ho, [r]espect for the accused requires that
the court must not be dismissive of his capacity to revise, or act against,
his bad character.
There is much to be said for the idea that we should respect peoples
capacity for moral improvement, and that such respect should be embedded in our criminal justice institutions. But it really is not clear that using
bad character evidence against defendants is inconsistent with respect
for their ability to change. How exactly does the inference from previous
conviction to present propensity involve treating the defendant as if past
conduct determines present conduct? We are not treating the defendants
actions as pre-determined, we are just saying that there is a certain probability that he has done something. An analogy may be useful. Suppose,
as sometimes happens, that during building work an unexploded Second
World War bomb is found in London. It may be that, so long after the
bomb was dropped, there is very little chance of it now exploding. But
when bomb disposal experts clear the area, they are not dismissing the
possibility of the bomb being safe; they might even take it into account
in deciding what degree of precautions to take. What they are doing,
387
See C. Kershaw, J. Goodman, and S. White, Reconvictions of Oenders Sentenced or

Discharged from Prison in 1995, England and Wales, Home Oce Statistical Bulletin 19/99
(London: Home Oce, 1999), 1214.
If we treat Ds present conduct as being inuenced by her past, we are not denying
that she could have put her past crime behind her, and come to guide her actions by
the appropriate reasons. Nor does there seem to be any conict with treating D as a
responsible agent.
See J. Raz, The Morality of Freedom (Oxford: Clarendon Press, 1986), 1556.
Book 1.indb 387
12/11/2008 5:01:21 PM
though, is allowing the possibility that the bomb remains dangerous. In

the same way, when using the previous conviction against the burglar,
we should take into account both the possibility that he no longer has a
propensity to burgle as well as the possibility that he remains dangerous.
But that is exactly how the concept of comparative propensity works,
because it averages over recidivists and desisters. So when we presume that
the defendant remains somewhat more likely to burgle than other people,
we are taking into account his capacity to change, not dismissing it. We
can also, and should, take the possibility of change into account by noting the length of time between the previous conviction and the present
charge; reconviction rates drop o sharply during the rst three years
after release from prison. It is true that the bomb analogy is in some
ways inapt; people obviously are not inanimate objects. But introducing
agency makes no dierence to the argument, which is simply a conceptual
one about probability.
Might it be argued, though, that we demean the defendant by treating
his conduct as partially determined? One problem here is that determination is usually taken to be a bivalent, all or nothing concept. Something
is either determined or it is not. It is true that we are treating the defendants choice as being inuenced by his character; inuence is something
that can come in degrees and may be what the arguments under consideration have in mind. But it seems that those arguments need the strong,
bivalent concept of determinism if they are to work: the argument of
Du et al, which objects to treating the defendant as if his past determines his future, loses its pull if expressed in terms of inuence. As the
unexploded bomb analogy shows, the inference that there is a possibility that the defendant has burgled again does not involve assuming that
he is unable to change. Moreover, there does not seem anything wrong,
or disrespectful, in treating the defendant as less than fully free. Radical
autonomy, where we choose arbitrarily, is not an attractive vision of the
world, nor does it seem respectful to think of a defendant as autonomous
in this way. Incomplete autonomy has got to be the realistic starting point
for any theory which values autonomy.
388
Mike Redmayne
Though we should recall the points made earlier about the diculty of complete
reform.
Book 1.indb 388
12/11/2008 5:01:21 PM
It is, however, possible to reframe the argument about the signicance

of change in a way which avoids the problems just described. When we
take precautions in dealing with the unexploded bomb, even though
we are not dismissing the possibility that the bomb is harmless, we are
presuming that the bomb has not completely changed. Similarly, when we
use previous convictions against the burglar, we are presuming that he has
not completely reformed. So what has just been said might seem to be
splitting hairs. Du et al and Ho could argue that it is the presumption
that the defendant has not completely reformed that is incompatible with
the censuring function of the criminal verdict, and that this presumption
fails to treat the defendant as a responsible agent. To assess that claim, we
need to look in more detail at the concept of censure, to see if criminal
verdicts do express a message that is incompatible with the use of character evidence to prove guilt. Before doing that, however, it is worth taking
another slight side-track, to note a further diculty with the account of
Du et al.
Suppose we accept that using previous convictions to prove guilt is
in conict with respect for the capacity for moral change, and thus with
the purposes of the criminal trial. If we conclude that we can therefore
never use previous convictions to prove guilt, we face a by-now familiar
problem. There is a high cost to complete exclusion, a cost we have been
symbolizing in terms of the acquittal of Straen. It seems that Du et al
do not accept that the exclusionary rule is absolute: there are also, at least
sometimes, reasons to allow evidence of prior convictions . . . which
might indeed sometimes outweigh the reasons against. If we take from
this that we should balance probative value and crime seriousness against
the moral cost of admissibility, there is a diculty, and it is not just the
problem of the opaque nature of the balancing exercise that we noted
earlier. If we presume that D has put her past behind her, it is hard to
see how the balancing process could ever get going; we simply have no
reason to suspect D at all, and there is no probative value to be placed
on the admissibility side of the scales. There is a general lesson here: if
we take a forbidden reasoning approach to previous convictions, then,
unless we combine it with an argument that admissible character evidence is dierent in kind from inadmissible evidence, it will be dicult
ever to justify admissibility, because the reasoning that establishes probative force is always forbidden. It might be argued that the rule against bad
389
character applies, in jury trials, to the fact nder and not to the judge. But
if the presumption of moral change does not apply to judges, this would
at least weaken the moral integrity of proceedings.
Censure and Reform
The arguments analysed in this section put considerable emphasis on the process of
censure embodied in a guilty verdict; they thus apply very much to previous convictions
rather than to other types of bad character evidence (see note 4 above). It might seem odd
to have a theory that justies the exclusion of just one type of bad character evidence; but
then again, if the arguments are thought convincing, the feeling of oddness will doubtless
wear o.
Book 1.indb 389
12/11/2008 5:01:21 PM
Should the criminal trial operate on a presumption that a defendant

has responded to the censure of a previous conviction by completely
reforming? As we noted right at the start of this paper, in everyday
morality we do not appear to be under any obligation to presume that
people who have behaved badly on some occasion in the past have completely changed. If you invite an acquaintance to dinner and he spends
the evening telling racist jokes, you do not seem to do anything wrong
by not inviting him again. In acting in that way, you do not disrespect
him in the way that concerns Smilansky and Margalit: they talk in terms
of possibilities, not presumptions. Of course, the criminal trial is very
dierent to a dinner invitation, and once again it may be in a theory of
the trial, and in particular in the censuring functions of the verdict, that
we nd support for the presumption we are interested in. The obvious
way to nd some support for the presumption is to look to sentencing theory to deepen our understanding of the way in which verdicts
communicate the censuring message to defendants. In fact, Du et al
take this route, in a proposal which supplements the analysis explored
above. If the criminal verdict is seen as part of a process of moral communication with the defendant, then we might suppose that, when the
defendant was previously convicted of burglary, he should have taken
the message of moral censure seriously. He should have responded to
the moral message by taking on board the wrongness of burglary and
desisting from future oending. This might provide the foundation
for various objections to the use of previous convictions. By using the
previous conviction we might, as Du et al suggest, be failing to treat
the defendant as a responsible agent, as someone who responds properly to censure. The criminal process might also be thought to be acting
390
Mike Redmayne
While the focus here is on retributive theories of punishment, a similar argument

might be made under a deterrence framework: if punishment is justied on grounds of
individual deterrence, then the use of previous convictions might be thought to conict with that justication. The response would be that, given that deterrence does not
usually claim to prevent crimeonly to deter it to an appropriate degreethere is again
no incompatibility.
Du, Punishment, note 9 above, 122.
A. von Hirsch and A. Ashworth, Proportionate Sentencing: Exploring the Principles
(Oxford: Oxford University Press, 2005), 23. See also A. von Hirsch, Censure and Sanctions
(Oxford: Clarendon Press, 1993), 13.
Book 1.indb 390
12/11/2008 5:01:21 PM
cynically, or hypocritically, if it punished people in the expectation that

they would reform, but then presumed that they had not done so by
using their previous convictions against them once they were suspected
of a later crime.
The diculty this analysis faces is that it does not easily mesh with
sentencing theory. If we look at two inuential theories of sentencing,
which both emphasize the communicative nature of punishment, we
do not nd anything as strong as a normative expectation that the
defendant will respond to censure by changing. Du himself argues
that punishment does aim to modify conduct by persuading oenders
to recognize and repent their wrongdoing, but a commitment to the
autonomy of defendants means that punishment must leave them free
to reject its message. Perhaps it could be argued that ideal defendants would take the moral message on board and lose any propensity
to commit crime. But to treat someone as a responsible agentwhich
is what is motivating us heredoes not seem to mean that we should
treat them as an ideal agent. Indeed, there would be something rather
odd about a theory of the trial and punishment that treated defendants
as ideal moral agentswhy would such angelic people commit crime
in the rst place? And why would they need hard treatment, rather than
simple symbolic censure?
We noted earlier that, like any change in behaviour, desistance from
crime is something that is not easy to achieve. We need to take this on
board if our theory of the trial is to be psychologically realistic, and
doing so underlines the point just made. Andrew von Hirschs account
of punishment actually embraces this psychological realism as one of
its central features. For von Hirsch, punishment treats us as moral but
fallible agents. Punishment operates as censure because the defendant, as a moral agent, can respond to moral criticism, but because the
defendant is not an ideal agent the censure is supplemented by hard
treatment. For our purposes, it is worth noting that this conception of
the subject of criminal law is prominent in von Hirschs account of why
391
A. von Hirsch, Desert and Previous Convictions, in Principled Sentencing, note

45 above, 1917. For helpful discussion, see J. V. Roberts, Punishing Persistent Oenders:
Exploring Community and Oender Perspectives (Oxford: Oxford University Press, 2008),
Chs. 34.
Ibid, 1956. It should be noted that von Hirsch tentatively goes further: perhaps
punishment asserts that the oender has a duty to attend to the censure and make extra
eorts at self-restraint (196). But even recognizing such a duty does not seem incompatible
with using previous convictions to prove guilt.
Book 1.indb 391
12/11/2008 5:01:21 PM
rst-time oenders should receive lighter sentences than recidivists.

Desert theory, which tends to focus on crime seriousness as the key variable in determining sentence severity, generally rejects the claim that
repeat oenders deserve longer sentences; repetition does not increase
seriousness. But von Hirsch does argue that rst-time oenders should
have their sentences reduced below the level indicated by the seriousness of the oence. Initial sentences should be mitigated because, as a
fallible moral agent, the defendant can be presumed to take censure
seriously. His rst crime is regarded as a lapse rather than as a serious
commitment to wrongdoing. To quote von Hirschin a passage that
resonates with the themes we are exploring[t]he rst-oender discount reects . . . an ethical judgement: it is a way of showing respect
for any persons capacity, as a moral agent, for attending to the censure
in punishment. . . . [P]eoples capacity to take condemnation of their
acts seriously is something that has a moral dimension and should be
acknowledged in the criminal law. But, as the defendant returns to
court, von Hirsch suggests, we have reason to increase the emphasis on
hard treatment because he has, to some extent, proved deaf to the moral
message.
We should certainly take from this that moral opportunitypeoples
ability to change, and to improve themselves morallyis something that
deserves particular respect in criminal justice. But that merely emphasizes
the points drawn from the wider philosophical literature via Margalit
and Smilansky, and it takes the argument for a moral objection to bad
character evidence no further. To spell it out: in using previous convictions against a defendant, the court seems to be acknowledging that he
might not have taken the censuring message to heart when previously
sentenced. That does not really conict with the new message of censure
when he is tried and convicted, a message which calls on him to recognize the moral wrong he has done and to mend his waysbut does not
demand that he does nor presume that he will. It is true that, when the
defendant is convicted of burglary with a previous conviction being part
of the evidence against him, the court appears to be acknowledging that
he did not mend his ways. But it would look like this even if the previous
392
Mike Redmayne
Labelling
The discussion so far has suggested that it is not easy to develop a
coherent moral objection to the use of previous convictions to prove
guilt. Accounts which suggest that such a practice is somehow incompatible with the moral framework of the trial have failed to pay close
attention to the reasoning underlying the use of previous convictions to
prove a propensity to commit crime. The reasoning is not dismissive of
Ibid, 196. It also does not apply so easily to carefully planned oences: see Roberts,
note 56 above, Ch. 3.
Book 1.indb 392
12/11/2008 5:01:21 PM
convictions had not been used in evidence against him, and, in any case,
the acknowledgment that he did not change is not to suggest that he
could not have done so, or cannot do so in the future. The only way to
generate a conict between the censuring message and the use of previous
convictions is to read a much more authoritarian message into censure:
a demand that the defendant reforms, along with a rm expectation that
he will do so. But that view of censure is, as we have seen, unattractive
to both von Hirsch and Du. And it is interesting that the reason why
censure is not viewed in this forceful way is a concern for autonomy. So
a respect for the autonomy of defendants, which was the starting point
for the ideas we have pursued to here, turns out to be one of the things
that makes the use of character evidence permissible, rather than, as the
arguments surveyed suppose, problematic.
There is, however, a dierent argument against the use of previous
convictions that might be drawn from von Hirschs discussion of progressive loss of mitigation. As we have seen, von Hirsch suggests that an initial
oenceand perhaps some subsequent onesshould be seen as lapses
rather than as displaying a serious commitment to oending. If that is
how we should view a defendant, then after an initial crime we should not
consider him more likely to oend than other people. He remains like the
rest of us, a basically good though awed moral agent who lapsed on one
occasion. The initial censure does not ask him to change his moral personality so much as to be more careful, to exert better self-control. As von
Hirsch notes, however, the theory of progressive loss of mitigation may
not apply to more serious crimes, and this seems right. It would be odd
to view Straens rst child killing as a lapse from an otherwise moral life.
It will be dicult, however, to say how serious a crime needs to be before
we should see it as something more than just a lapse.
393
S. Shute, R. Hood, and F. Seemungal, A Fair Hearing? Ethnic Minorities in the

Criminal Courts (Cullompton: Willan Publishing, 2005), 456.
Ibid, 46.
Ibid, 49.
Roberts, note 56 above Ch. 7. All the oenders interviewed did seem to accept that
weight should be put on previous convictions at sentencing; the objection was to doing
this too automatically.
Book 1.indb 393
12/11/2008 5:01:21 PM
the defendants capacity for moral improvement, it merely assumes that

he has not completely changed. Nevertheless, the very fact that it is easy
to misread the reasoning involved may be signicant. We could modify
the arguments slightly, so that what is emphasized is not the precise inferential structure of reasoning from bad character, but the way the use of
previous convictions is likely to be interpreted by defendants.
Accounts of the criminal trial which emphasize the communicative
role of the criminal verdict have tended to be rather armchair-based
aairs, with scholars speculating about how various practicessuch as
the admission of improperly obtained evidencemight be interpreted
by defendants, or by the public. Of course, proponents of the communicative view can easily move to the normative plane, and argue that what
matters is how certain trial practices should be interpreted, not how they
are interpreted. There are dicult questions here about the relationship
between the normative and the empirical, but surely any account which
stresses communication must have one eye on how trial practices are
actually understood. That, however, is something we know very little
about.
There is, though, a little empirical evidence on how, in some situations,
the use of previous convictions is understood by defendants. In research
on the experiences of ethnic minority defendants in the courts, Stephen
Shute, Roger Hood, and Florence Seemungal noted that when it came to
sentencing some felt that their past records when they had been young
and wild were being given too much weight in sentencing policy. Not
enough attention had been paid to them as they were now, due to a
failure properly to take into account the social and cultural circumstances
in which they had grown up. As one defendant put it: The judge didnt
look at my past to my present, he looked at the oence. He said it was my
way of life, but Ive changed from a drug user, Ive found a job. Another
defendant said [t]hey think that if you have been in trouble before you
are still the same person. Ive been out of trouble for six years . . . I am a
family man. Ive changed. They dont care. . . . They think Im the same
person. Similar comments about the importance of interpreting previous convictions in the light of the potential to change can be found in
Julian Roberts research on oenders reactions to sentencing. In the
394
Mike Redmayne
K. D. Vohs and J. W. Schooler, The Value of Believing in Free Will: Encouraging a

Belief in Determinism Increases Cheating (2006) 19 Psychological Science 4954.
See e.g. D. Downes and P. Rock, Understanding Deviance (5th edn, Oxford: Oxford
University Press, 2007), Ch. 7.
M. E. Ezell and L. E. Cohen, Desisting from Crime: Continuity and Change in
Long-Term Crime Patterns of Serious Chronic Oenders (Oxford: Oxford University Press,
2005).
Pfennig, note 24 above, 513.
Evidence in Criminal Proceedings: Previous Misconduct of a Defendant (London:
Stationery Oce, 1996), 124, fn 10.
Book 1.indb 394
12/11/2008 5:01:21 PM
context of our discussion, these reactions are striking. Some defendants

do seem to feel that not enough recognition has been given to their eorts
to change, and one can imagine similar reactions being made to the use
of previous convictions to prove guilt. A cynical response would be that,
given that these defendants have been convicted, they have not in fact
changed; we could refer back to the point made earlier that people tend to
be over-optimistic about their ability to change, and interpret these reactions as self-serving delusions. Even taking this view, however, we might
still feel that there is something signicant here. If what matters is how
defendants interpret trial practices, then these negative reactions may be
noteworthy in themselves. If the courts are too readily dismissive of the
possibility of moral change, might there be a danger that defendants
come to adopt this bleak attitude too?
In a recent psychological study, Kathleen Vohs and Jonathan Schooler
found evidence that undermining subjects belief in free will increased
propensity to cheat on a simple test. It is a big leap, but the thought
would be that if, through their use of previous convictions, courts give
defendants the impression that they are prisoners of their past lives, then
defendants will come to think this way about themselves, and this will
make them more likely to commit crime. Criminologists in fact have a
somewhat similar concern: labelling or interactionist theory suggests that
the process of being labelled a criminal may reinforce deviance, a concern which a recent large-scale study lends support to. The High Court
of Australia may have been thinking along these lines when it worried
that rehabilitation schemes might be undermined if the accuseds criminal record could be used in evidence against him or her, and the Law
Commission, in its Consultation Paper on character evidence, echoed
this line of thought: [w]e should be wary of exaggerating the likelihood
that those convicted will reoend, or of assuming that individuals cannot
change. High conviction rates for those with previous convictions do not
make reconviction inevitable . . . Such change is clearly something the law
should encourage, rather than presuming it does not occur.
395
Slippery Slopes
An underlying worry about previous convictions is that, if we admit
them, we will be pushed down a slippery slope. The argument used to
justify the admissibility of previous convictions is that, because those
with convictions commit more crime than those without, possession of a
previous conviction is an indicator of guilt. But this form of argument can
be used to justify the admissibility of all sorts of evidence that we would be
uncomfortable with. Should race or socio-economic status be admissible
evidence, if people from certain racial or socio-economic groups commit
more crime than others?
It is worth underlining the signicance of this for sentencing. If one is attracted by
von Hirschs account of the rst-oence discount, then it seems that it is not only the rst
oence that should receive a discount, but also ones committed after a signicant break in
oending. A lapse can occur during, as well as at the beginning of, an oending career.
See, e.g. Piquero et al, note 27 above, Ch. 4.
See, generally, T. R. Tyler and Y. J. Huo, Trust in the Law: Encouraging Public
Cooperation with the Police and the Courts (New York: Russell Sage Foundation, 2002);
Roberts, note 56 above, Ch. 6.
Book 1.indb 395
12/11/2008 5:01:21 PM
The diculty here is that we have so little to go on. While the

quotations from the Hood study should make us think, the defendants might have been expressing a legitimate concern that their previous
convictions were too stale to warrant any inferences about whether the
latest oence was part of a pattern of oending as opposedto return
to von Hirschs termsto a lapse from an otherwise successful attempt
to go straight. Nothing that is said here should be taken to support the
introduction of previous convictions that are simply not relevant. Rates
of criminal oending tend to peak in the late teens, and then drop o
in early adulthood. If a defendant has a series of convictions in his late
teens, but then a clean record until the mid-20s, there might be a good
argument simply in terms of probative value for putting no weight on the
previous convictions as evidence of guilt; it might also be wrong to use
such a previous record as an aggravating factor in sentencing. Defendants
do seem to be sensitive to the fairness of the criminal justice system, so we
should be aware that admitting previous convictions when they have little
or no probative value may be counterproductive. Beyond that, though,
the concerns about labelling and feedback loops just seem too speculative
to provide the foundation for our policy on the admissibility of previous
convictions.
396
Mike Redmayne
This is not to deny that dierences of degree can be importantthey play a

crucial role in what is sometimes referred to as threshold deontology (see L. Alexander,
Deontology at the Threshold (2000) 37 San Diego Law Review 893912). And just as
piles can become heaps, and hair loss baldness, there may be some sense in talking of
specic propensity, even though propensity is a continuum. But it is very hard to see
how anything of moral signicance can come out of the distinction between general and
specic propensitybetween Straens and burglarsexcept in relation to the balance
of probative value and prejudicial eect. And, to repeat a point made before, we have no
reason to think that only specic propensity outbalances prejudice.
See, generally, J. Chalmers and F. Leverick, Fair Labelling in Criminal Law (2008)
71(2) MLR 21746. As an example, in 1998 England and Wales created specic oences
of racially aggravated assault; would racial assault have demonstrated a specic propensity
in 1997, but only a general one in 1999?
Book 1.indb 396
12/11/2008 5:01:21 PM
One response is that we are all on a slippery slope. If we admit previous

convictions against Straen, or against the Makins, or whoeveras the
common law doesthen we are relying on the sort of reasoning just
described. We are presuming that some peoplethose who kill children
in a particular wayare more likely to kill children in that way than others.
We may be a long way up the slope, but unless we can draw distinctions
between the cases we are still vulnerable to being pushed down towards
the racial inference. This raises, once again, the question whether the
burglary example we started with is dierent in kind, rather than just in
degree, to cases such as Straen and Makin. That is a complex question
which could lead us far into the thickets of similar-fact case law and scholarship, so nothing said here can be denitive. Briey, though, it is at least
dicult to point to a sound distinction between the cases. Earlier it was
suggested that an account based on reasons or motivation cannot do the
work. And it is no good relying on the simple distinction between general
and specic propensity, by saying that our D has a general propensity to
commit burglary whereas Straen has a specic propensity to kill young
girls in a particular way, unless some convincing account can be given of
the general/specic distinction. The distinction cannot simply rest on the
fact that burglary is a crime label whereas murdering young girls and not
hiding their bodies is not, because the labels are, at least to some extent,
conventional. If we ignore the labels, there is no obvious place to draw
the line: why should we not see burglary as a specic instance of a general
propensity to commit property crime? Or Straen as having a general
propensity to kill young girls, as opposed to, say, blonde girls?
It may be possible, though, to distinguish between inferring propensity
from previous convictions and inferring propensity from race and socioeconomic statusto draw the line further down the slope than the burglary example, rather than above it, as it were. In the burglary example,
we are inferring propensity from the fact that D has already made a bad
397
R v Watters, CACD 19 October 2000.

The parallels with racial proling are obvious. It has proved dicult to specify what
is wrong with statistically sound racial proling, and some accounts conclude that racial
sensitivity is the main problem, i.e. that there is nothing wrong with the reasoning involved:
see F. Schauer, Proles, Probabilities, and Stereotypes (Cambridge, Mass: Harvard University
Press, 2003), Ch. 7; R. Kennedy, Suspect Policy New Republic, Sept 13/20 1999, 30.
For more general debate, see M. Levin, Responses to Race Dierences in Crime (1992)
23 Journal of Social Philosophy 529; L. Thomas, Statistical Badness (1992) 23 Journal
of Social Philosophy 3041; K. Lippert-Rasmussen, Racial Proling Versus Community
(2006) 23 Journal of Applied Philosophy 191205; K. Lippert-Rasmussen, Nothing
Personal: On Statistical Discrimination (2007) 15 Journal of Political Philosophy 385403;
M. Risse and R. Zeckhauser, Racial Proling (2004) 32 Philosophy & Public Aairs 131;
A. Lever, Why Racial Proling is Hard to Justify: A Response to Risse and Zeckhauser
(2005) 33 Philosophy & Public Aairs 94110; B. Harcourt, Against Prediction: Proling,
Book 1.indb 397
12/11/2008 5:01:21 PM
moral choice by committing the previous burglary. The other attributes,

however, are unchosen; they are beyond Ds control. If we are concerned,
as Du et al and Ho are, to respect D as a responsible agent, then an
inference from one of these demographic attributes could be especially
worrying. While, if we use the demographic evidence, we are not assuming that D has made a bad moral choice, perhaps this is an arena where
we should not even allow that D is more likely to have made a bad moral
choice than those who do not share the relevant characteristic. Treating D
as responsible could be said to involve assuming that he is no more likely
to be tempted into crime than other people.
While it is very comfortable to agree with this line of thought, we
should not ignore the hard questions here. In Watters, the prosecution, in
a rather desperate attempt to make a weak DNA case look strong enough
to support a burglary conviction, argued that the fact that D was male
was evidence against him, because most safe-crackers are male. If that
argument strikes us as odd, it may be because Ds sex would have been
obvious to the jury, who might already have taken it into account, albeit
at some unconscious level (they might well have been more reluctant to
convict had Watters been a woman). But do we really want to say that the
prosecutors argument is morally impermissible in court? There is also,
once again, the problem of motive. Evidence that Ds wife is having an
aair could be said to place him in an unchosen demographic group the
members of which are more likely to commit murder than others. Does
that make motive evidence ethically objectionable? It is true that, as noted
earlier, motive treats D as having a reason to commit crime, but that
cannot be the key to these puzzles, or else we would allow evidence of
poverty to prove theft. It may be that we can say little more than that
factors such as race and class are ones we are especially sensitive about,
whereas having an unfaithful spouse is not. That is not a very satisfying
398
Mike Redmayne
explanation of our moral landscape, but it might be the best we can do.
While it is tempting to use our reactions to various cases as the basis for
some ambitious theory about the moral structure of the criminal trial, we
should at least be circumspect about the possibility of developing any such
theory. Our intuitions may simply not t into any coherent theory.
Conclusion
Policing, and Punishing in an Actuarial Age (Chicago: Chicago University Press, 2007),
Chs. 4, 5, 7.
Alan Wertheimers comments on the problem of discrimination are relevant here
(A. Wertheimer, Reections on Discrimination (2006) 43 San Diego Law Review
945980, 952):
I am reasonably condent that our present intuitions about discrimination are
of little help, because those intuitions may well be closely tied to the actual social
consequences of discrimination, or to the history of those consequences. If whites
and blacks had relatively equal wealth, income, longevity, education, and so forth,
it is entirely possible that we would not in fact regard isolated acts of racial discrimination as seriously wrong. It is possible, of course, that we would be wrong
not to regard such acts as seriously wrong, but then we need an argument that does
not appeal to intuitions. My general point remains that it is extremely dicult
to know what is driving our intuitions about acts of discrimination when such
discrimination has been part of a pattern that has generated massive social and
economic inequalities.
Book 1.indb 398
12/11/2008 5:01:21 PM
This paper has surveyed several dierent arguments under which the
use of previous convictions as evidence of propensity to commit crime
is morally problematic. The basic conclusion is that there are no very
strong arguments against this use of bad character evidence, but there
are various caveats to underline. The argument that the inference from a
previous conviction may be dicult for a defendant to rebut has something to be said for it, though, given that this problem may also apply
to other types of evidence, it is not obvious what to make of it. There
is also the argument, drawn from von Hirsch, that if we see an initial
conviction for a crime of modest seriousness as no more than a lapse,
then such convictions should not be used as evidence of propensity to
commit crime. Whether that convinces obviously depends on whether
one agrees with von Hirsch about the relevance of previous convictions
to sentence.
If these arguments give us some reason to exclude previous convictions, they hardly challenge the admissibility regime under the Criminal
Justice Act 2003. Though it is early days, the courts do seem to be taking
the line that a single previous conviction for a minor crime does not
399
See note 3 above.
Book 1.indb 399
12/11/2008 5:01:21 PM
show propensity. It is not easy, though, to argue that this shows some
innate moral awareness on the part of the courts, because those previous
convictions will often have very little probative value in any case.
Beyond this, it is worth emphasizing a vaguer, but more widespread
caveat. The intuition that there is something troubling about the use
of the previous convictions in the burglary example that has been our
leitmotiv is hard to shake o. While the closer we look at attempts to
justify this intuition, the weaker they seem to be, the feeling of unease
lingers on. In the penultimate section of the paper, however, we allowed
the possibility that our intuitions simply might not be amenable to
systematization. If so, then it may be that we should be reluctant to
dismiss our lingering unease. But if that is the best we can do, we do not
have much with which to respond to those who argue for the widespread
admissibility of previous convictions.

Current Legal Problems Redmayne

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Current Legal Problems Redmayne

Transféré par

Droits d'auteur :

Formats disponibles

The Ethics of Character Evidence

Book 1.indb 371

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

Book 1.indb 372

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

The Ethics of Character Evidence

Moral Accounts of the Criminal Trial

Book 1.indb 373

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

R v Straen [1952] 2 QB 911.

Book 1.indb 374

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

Suppose that we decide that there is a moral argument against using

The Ethics of Character Evidence

Book 1.indb 375

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

The Signicance of Desistance

Book 1.indb 376

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

character evidence might even be taken to support the argument for a

The Ethics of Character Evidence

Book 1.indb 377

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

Book 1.indb 378

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

inferentially suspect. For now, we will concentrate on the more epistemic

The Ethics of Character Evidence

Book 1.indb 379

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

While this is an interesting line of objection to character evidence,

Book 1.indb 380

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

The Ethics of Character Evidence

Self-Assessment Questionnaire, Home Oce Research Findings 284 (London: Home

Book 1.indb 381

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

may weakenindeed, this seems to hold for most oenders as part

Book 1.indb 382

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

The Ethics of Character Evidence

Motives and Reasons

Book 1.indb 383

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

is a case where there is a strong argument for admissibility of previous

Makin v AG NSW [1894] AC 57.

Book 1.indb 384

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

infer the likely behaviour of Dthe explanatory structure in motive

The Ethics of Character Evidence

be because such people are able to respond appropriately to the censuring

In our initial assessment of the autonomy argument against previous

Book 1.indb 385

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

A. Margalit, The Decent Society (trans. N. Goldblum; Cambridge, Mass: Harvard

Book 1.indb 386

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

their minds. In a similar vein, though on a dierent topic, that of decency,

The Ethics of Character Evidence

See C. Kershaw, J. Goodman, and S. White, Reconvictions of Oenders Sentenced or

Book 1.indb 387

Downloaded from http://clp.oxfordjournals.org/ at University of Warwick on March 1, 2015

though, is allowing the possibility that the bomb remains dangerous. In

Book 1.indb 388