1 Evidence of Bad Character in Criminal Proceedings Law Com. No. 273, Cm. 5257 (London 2001).

2 Review of the Criminal Courts of England and Wales: Report (London 2001), pp. 563–568Google Scholar.

3 Justice For All Cm. 5563 (London 2002), pp. 79-82. The proposals are rather opaque, but the examples given do suggest that an attempt will be made to relax the current rule, perhaps using a test of whether admission will have a “disproportionate” effect on the trial.

4 Op. cit. n. 1, para. 2.2.

5 [1975] A.C. 421, 451.

6 See, e.g., L.H. Hoffman, “Similar Facts After Boardman” (1975) 91 L.Q.R. 193.

7 Op. cit. n. 5, p. 456.

8 A useful exploration of these concepts is A.L.-T. Choo, “The Notion of Relevance and Defence Evidence” [1993] Crim. L.R. 114. See also R. Pattenden, “The Discretionary Exclusion of Relevant Evidence in English Civil Proceedings” (1997) 1 E. & P. 361.

9 (1858) 27 L.J.C.P. 241.

10 Ibid. p. 242.

11 Ibid.

12 For elaboration, see, e.g., R.D. Friedman, The Elements of Evidence (2nd edn., St Paul, Minn. 1998), pp. 42-65; Lempert, R.O., “Modeling Relevance” (1977) 75 Mich. L. Rev. 1021Google Scholar.

13 See Spector, R.G., “Rule 609—A Last Plea for its Withdrawal” (1979) 32 Oklahoma L. Rev. 334, 351Google Scholar.

14 See, generally, Caprara, G.V. and Cervone, D., Personality: Determinants, Dynamics, and Potentials (Cambridge 2000), pp. 38–50Google Scholar. Though now slightly dated, a useful review of research, with application to the rules of evidence in mind, is S.M. Davies, “Evidence of Character to Prove Conduct: A Reassessment of Relevancy” (1991) 27 Criminal Law Bulletin 504. In this vein, see also M.A. Mendez, “The Law of Evidence and the Search for a Stable Personality” (1996) 45 Emory L.J. 221. Scepticism about character has recently emerged in the philosophical literature. For discussion, see G. Sreenivasan, “Errors About Errors: Virtue Theory and Trait Attribution” (2002) 111 Mind 47.

15 Associated, in particular, with the work of Allport. See, e.g., Allport, G.W., Personality: A Psychological Interpretation (London 1937)Google Scholar.

16 Mischel, W., Personality and Assessment (New York 1968)Google Scholar, esp. ch. 2.

17 See, e.g., Kenrick, D.T. and Funder, D.C., “Profiting from Controversy: Lessons From the Person-Situation Debate” (1988) 43 American Psychologist 23Google Scholar.

18 Caprara and Cervone, op. cit. n. 14, p. 64. They add at p. 110: “there simply are no longer any situationists”.

19 The most prominent approach is sometimes called the “five factor model”, which analyses character in terms of five broad traits (the “big five”): extraversion, agreeableness, conscientiousness, emotional stability and intellect. See Wiggins, J.S. (ed.), The Five-Factor Model of Personality: Theoretical Perspectives (New York 1996)Google Scholar. For a more general review of the current status of trait theory, see Matthews, G. and Denny, I.J., Personality Traits (Cambridge 1998)Google Scholar, esp. chs. 1, 2, 11.

20 Not because the debate is unimportant, but because we are interested in the reliability of prediction, and can therefore afford to treat personality as a black box. All the same, we should bear in mind that without an understanding of exactly what personality is and how it works, we should be somewhat cautious about predictions based on it.

21 Mischel, W. and Shoda, Y., “Reconciling Processing Dynamics and Personality Dispositions” (1998) 49 Annual Review of Psychology 229, 244Google Scholar.

22 Mischel, W. and Shoda, Y., “A Cognitive-Affective System Theory of Personality: Reconceptualizing Situations, Dispositions, Dynamics, and Invariance in Personality Structure” (1995) 102 Psychological Review 246, 257Google Scholar.

23 Funder, D.C., Personality Judgment: A Realistic Approach to Person Perception (San Diego 1999), p. 43Google Scholar.

24 See Mischel and Shoda, op. cit. n. 21, pp. 233-235.

25 Of course, armed robbers can have less choice over their motivations, and here we can expect situational variation to be significant. The armed robber who wins the lottery is unlikely to rob again.

26 This point is often overlooked in the psychological literature: see Funder, op. cit. n. 23, p. 9. An implication of Mischel and Shoda's research, however, is that such probabilistic inference might run into problems in some situations. We might, from one instance of behaviour, generalise that a child is verbally aggressive, but this could mislead if used to ground an inference about verbal aggression in a different situation, where the child is actually less verbally aggressive than his peers. While this problem might arise in stark form in characterbased reasoning, we should note that it is a potential problem in all inductive inference. Any induction might be undermined were we to gather evidence at a more microscopic level than that on which we are relying to ground it. A match between two DNA profiles might, given the information we have, be very powerful evidence of identity. But our inference would be untenable were we to test one more locus and so discover a difference between the two profiles. Obviously, the possibility of confutation at a more microscopic level of analysis does not always paralyse induction.

27 Evidence in Criminal Proceedings: Previous Misconduct of a Defendant Law Com. Con. Paper No. 141 (London 1996), paras. 6.7-6.38.

28 Op. cit. n. 1, paras. 6.11-6.13. It is worth noting, in passing, that the psychological literature reviewed in the Consultation Paper is rather dated.

29 The present law allows, in some circumstances, the introduction of evidence that the defendant has engaged in misconduct which he has not been convicted of (as when more than one count is joined in an indictment, e.g. Boardman, op. cit. n. 5; or when uncharged but suspicious conduct is used to bolster a case, e.g. R. v. Smith (1915) 11 Cr. App. R. 229), and even evidence of misconduct of which he has been acquitted: R. v. Z. [2000] 3 W.L.R. 117. Evidence of behaviour which it is difficult to describe as misconduct may also be admissible, e.g. R. v. Butler (1987) 84 Cr. App. R. 12.

30 See Ashworth, A., Sentencing and Criminal Justice (3rd edn., London 2000), pp. 180–181Google Scholar; Walker, N. and Padfield, N., Sentencing: Theory, Law and Practice (2nd edn., London 1996), pp. 357–358Google Scholar. While emphasising the difficulty of predicting offending, the literature does acknowledge that prior offending behaviour is the most significant predictor of future offending: see Monahan, J. et al., Rethinking Risk Assessment: The MacArthur Study of Mental Disorder and Violence (New York 2001), pp. 44–45Google Scholar.

31 For an example of incautious use of statistics, see Law Commission, op. cit. n. 27, p. 124, n. 10.

32 Park, R.C., “Character at the Crossroads” (1998) 49 Hastings L.J. 717Google Scholar.

33 Kershaw, C., Goodman, J., and White, S., Reconvictions of Offenders Sentenced or Discharged from Prison in 1995, England and Wales, HOSB 19/99 (London 1999)Google Scholar. The following discussion uses data from tables 2 and 3.

34 Statistics are taken from Criminal Statistics England and Wales 1996 Cm. 3764 (London 1997), table 5.10. This provides statistics on offenders found guilty or cautioned. Ignoring cautions would give slightly larger comparative propensity figures (see table 5.6) and might be thought a better basis for the calculation, given that cautions were not counted in the recidivism study. However, because factors likely to exaggerate comparative propensity are the most significant for our purposes, and since recidivists are less likely to be cautioned, cautions have been included in the comparative propensity denominator. A second problem with the data in the Criminal Statistics is that they provide the rate of offending, not the number of offenders. Within the two-year period, some offenders will commit more than one offence; all but the first offence should be ignored when calculating the denominator. There is no easy way of achieving this adjustment and, in any case, the result of ignoring this factor is again an underestimation of comparative propensity. Further problems with the comparative propensity calculation are discussed below.

35 See Redmayne, M., Expert Evidence and Criminal Justice (Oxford 2001), p. 194Google Scholar.

36 See Park, op. cit. n. 32, p. 763 and n. 27. Park provides the following comparative propensity figures: homicide: 224; rape: 163; burglary: 56; larceny: 19; drug offence: 24. Differences in the statistics mean that these are not directly comparable with the UK figures.

37 In terms of legal category, anything from recklessly inflicting actual bodily harm (Offences Against the Person Act 1861, s. 47) to intentionally causing grievous bodily harm (OAPA s. 18). The figures only include standard list offences, so common assault is not included.

38 Cf. M.E. Rice and G.T. Harris, “The Scientific Status of Research on Sexual Aggressors” in D.L. Faigman, D.H. Kaye, M.J. Saks, and J.L. Sanders (ed.), Modern Scientific Evidence: The Law and Science of Expert Testimony Vol. 3 (St Paul, Minn. 1999) pp. 91-97.

39 See C. Friendship and D. Thornton, “Sexual Reconviction for Sexual Offenders Discharged from Prison in England and Wales: Implications for Evaluating Treatment” (2001) 41 Brit. J. Criminol. 285; cf. R. Hood, S. Shute, M. Feilzer and A. Wilcox, “Sex Offenders Emerging from Long-Term Imprisonment” (2002) 42 Brit. J. Criminol. 371, 392. A further possibility is that the small size of the samples makes these figures less reliable than others.

40 See Kershaw et al., op. cit. n. 33, pp. 12-14.

41 Ibid. pp. 14-15. On the significance of multiple convictions, see further Prime, J., White, S., Liriano, S. and Patel, K., Criminal Careers of Those Born Between 1953 and 1978 HOSB 4/01 (London 2001)Google Scholar, paras. 3.13-6.

42 Farrington, D.P., “Human Development and Criminal Careers” in Maguire, M., Morgan, R., and Reiner, R. (ed.), The Oxford Handbook of Criminology (2nd edn., Oxford 1997), pp. 365–369, 373Google Scholar.

43 The following different-crime comparative propensities were calculated from the custody data: violence: 14; burglary: 46; robbery: 30; theft/handling: 22; fraud/forgery: 19; drug offence: 10. These figures measure the tendency of a person with a previous conviction for any offence to commit the specified offence within two years of release from custody, compared with offending in the general population in a two-year period. The number of sexual offences was too small to appear in the summary statistics in the study. A larger sample might show a reasonably large comparative propensity. Ignoring sexual offences, all these figures are smaller than the same-crime comparative propensities, suggesting a tendency for offenders to both specialise and generalise. Note that the figures are aggregated over “previous conviction for any offence”. A more detailed analysis might reveal quirks at a finer level: for instance, forgers might just turn out to be less likely to commit sexual offences than members of the general population.

44 See Farrington, op. cit. n. 42, p. 380; Britt, C.L., “Versatility” in Hirschi, T. and Gottfredson, M.R. (ed.), The Generality of Deviance (New Brunswick 1994)Google Scholar; Soothill, K., Francis, B. and Fligelstone, R., Patterns of Offending Behaviour: A New Approach Home Office Research Findings no. 171 (London 2002)Google Scholar.

45 For more detailed discussion of the Law Commission's character evidence project, see J. McEwan, “Previous Misconduct at the Crossroads: Which ‘Way Ahead’?” [2002] Crim. L.R. 180; M. Redmayne, “The Law Commission's Character Convictions” (2002) 6 E. & P. 71; P. Mirfield, “Bad Character and the Law Commission” (2002) 6 E. & P. 141.

46 S. Lloyd-Bostock, “The Effects on Juries of Hearing About the Defendant's Previous Criminal Record: A Simulation Study” [2000] Crim. L.R. 734. The research is also summarised in the Consultation Paper.

47 S. Lloyd-Bostock, The Effects on Magistrates of Hearing that the Defendant has a Previous Conviction LCD Research Series No. 3/00 (London 2000). The research is summarised in the report.

48 Lloyd-Bostock herself notes that “the effects … are not large”, ibid., p. 54. Magistrates increased the probability of guilt from 0.4 to 0.6; juries from 0.5 to 0.66. In interpreting the studies, it is assumed that subjects were expressing probabilities; Lloyd-Bostock tends to interpret the results in this way, but, to be precise, subjects were asked to express “likelihood ratings” on a scale of 1-100.

49 There is some equivocation on this point: cf. paras 6.37 and 6.56 of the report, op. cit. n. 1.

50 Ibid. para. 6.55.

51 Lloyd-Bostock, op. cit. n. 47, p. 24.

52 Op. cit. n. 1, para. 6.44.

53 See n. 43 above. Note also the finding of Hood el al., op. cit. n. 39, that sexual offenders against children are not infrequently reconvicted for other, non-sexual, offences.

54 The studies do explore the effect of previous convictions on verdicts, which is where moral prejudice would exert any influence it has. But the results appear to be explained by the subjects’ probabilities of guilt; they do not look to have been influenced by any other factor. One particular clue is that guilty verdicts did not become more common in the “recent dissimilar” condition, where one might expect to find moral but not reasoning prejudice.

55 See, generally, Coleman, C. and Moynihan, J., Understanding Crime Data: Haunted by the Dark Figure (Milton Keynes 1996)Google Scholar, ch. 2.

56 See McConville, M., Sanders, A., and Leng, R., The Case for the Prosecution: Police Suspects and the Construction of Criminality (London 1991)Google Scholar; Hillyard, P., Suspect Community: People's Experiences of the Prevention of Terrorism Acts in Britain (London 1993)Google Scholar.

57 Recently, a complex argument has appeared which might be taken to suggest that rounding up the usual suspects results in previous convictions being evidence of innocence. See N. Fenton and M. Neil, “The ‘Jury Observation Fallacy’ and the Use of Bayesian Networks to Present Probabilistic Legal Arguments” (2000) 36 Mathematics Today 180; C. Jowett, “Sittin’ in the Dock with the Bayes” (2001) 151 N.L.J. 201. This argument is not considered in detail because it applies, not to juries, but to external observers of their verdicts. It is also based on the unrealistic assumption that cases can be brought to trial on the basis of a previous conviction alone.

58 On the presumption of guilt in defence lawyering, see McConville, M., Hodgson, J., Bridges, L. and Pavlovic, A., Standing Accused: The Organisation and Practices of Criminal Defence Lawyers in Britain (Oxford 1994)Google Scholar, esp. ch. 6.

59 Defendants who testify sometimes face a choice of restricting their defence, so as not to cast aspersions on the character of prosecution witnesses or give evidence against a co-defendant, or else having their previous convictions revealed to the jury. Now that inferences can be drawn from a failure to testify, the third horn of this trilemma—avoiding the revelation by not testifying—has been sharpened. Those with recent previous convictions will also be prevented from making a claim of good character, and previous convictions may sometimes be revealed under the similar fact rule. On these various rules, see, e.g., Dennis, I.H., The Law of Evidence (London 1999), pp. 406–413Google Scholar, 584-614, 632-653.

60 See P. Darbyshire, “Previous Misconduct and the Magistrates’ Courts—Some Tales from the Real World” [1997] Crim. L.R. 105; Lloyd-Bostock, op. cit. n. 47, pp. 43-53.

61 For a somewhat similar thesis, see Box, S., “The Criminal Justice System and ‘Problem Populations’“ in Lacey, N. (ed.), A Reader on Criminal Justice (Oxford 1994)Google Scholar.

62 See, e.g., M. McConville and J. Baldwin, “The Role of Interrogation in Crime Discovery and Conviction” (1982) 22 Brit. J. Criminol. 165, 168. The 2000 British Crime Survey found that in two thirds of violent incidents the offender was known to the victim: Kershaw, C. et al., The 2000 British Crime Survey England and Wales HOSB 18/00 (London 2000), p. 38Google Scholar. Obviously, this still leaves the police with some choice over whether—and against whom—to pursue charges, decisions which may be influenced by previous convictions. On this, see Cretney, A. and Davis, G., Punishing Violence (London 1995)Google Scholar, ch. 4.

63 Although this happens rarely, it is significant because it provides recidivism statistics of a sort. There is a 40 per cent chance that DNA taken from a crime scene will match a sample already on the DNA database (Forensic Science Service, Press Release, 7 April 2000). Given that it is early days for the database, this implies that the police are not massively ignoring sections of the offending population in the sort of crime where DNA is used.

64 M. Maguire, “Crime Statistics, Patterns and Trends: Changing Perceptions and their Implications” in Maguire et al., op. cit. n. 42, p. 175.

65 Ibid. See further Coleman and Moynihan, op. cit. n. 55, ch. 5. At p. III the authors comment: “[a]s far as ordinary offending goes, the characteristics of the dark figure are probably fairly well represented by known offenders”.

66 Op. cit. n. 1, para. 6.56.

67 See Ayer, A.J., “Two Notes on Probability” in The Concept of a Person and Other Essays (London 1964), pp. 198–208Google Scholar.

68 A brief account is provided in Popper, K.R., A World of Propensities (Bristol 1990), pp. 3–26Google Scholar.

69 Gillies, D., “Varieties of Propensity” (2000) 51 British Journal for the Philosophy of Science 807Google Scholar.

70 On Peirce's propensity theory, see Burks, A.W., “Peirce's Two Theories of Probability” in Moore, E.S. and Robin, R.S. (ed.), Studies in the Philosophy of Charles Sanders Peirce (Cambridge, Mass. 1964)Google Scholar. Peirce also used the suggestive term “habit”.

71 Gillies, D., Philosophical Theories of Probability (London 2000), pp. 120–136Google Scholar.

72 Ibid. pp. 119-125. See also Ayer, op. cit. n. 67.

73 Ibid. p. 123.

74 Ibid. p. 120.

75 For an elaboration and defence of the legal application of subjective probability, see D. H. Kaye, “The Laws of Probability and the Law of the Land” (1979) 40 U. Chi. L. Rev. 40.

76 See, e.g., Bonjour, L., “The Dialectic of Foundationalism and Coherentism” in Greco, J. and Sosa, E. (ed.), The Blackwell Guide to Epistemology (Oxford 1999)Google Scholar.

77 A vivid example is provided by the decisions of the Court of Appeals in US v. Shonubi, 998 F. 2d. 84 (1993) and 103 F. 3d. 1085 (1997). For discussion, see J.L. Gastwirth, B. Freidlin and W. Miao, “The Shonubi Case as an Example of the Legal System's Failure to Appreciate Statistical Evidence” and A.J. Izenman, “Assessing the Statistical Evidence in the Shonubi Case” in J.L. Gastwirth (ed.), Statistical Science in the Courtroom (New York 2000). There is a philosophical literature which explores epistemological and moral objections to the use of statistics to prove individual instances of behaviour. The moral issues fall beyond our remit; the epistemological ones are none too convincing. See Thompson, J.J., “Liability and Individualized Evidence” in Rights, Restitution and Risk: Essays in Moral Theory (Cambridge, Mass. 1986)Google Scholar; F. Schoeman, “Statistical vs. Direct Evidence” (1987) 21 Nous 179; D.T. Wasserman, “The Morality of Statistical Proof and the Risk of Mistaken Liability” (1991) 13 Cardozo L. Rev. 934.

78 Op. cit. n. 71, p. 203.

79 Op. cit. n. 46, pp. 752-753. A similar finding was made in the magistrates study: op. oil. n. 47, p. 34. Research on real juries in New Zealand is equivocal on this point. It was common for jurors to speculate about previous convictions—this occurred in all trials where the jury was not told whether the accused had a criminal record; “[i]n a number of cases, their speculations were based upon factors which were indicative of a prior record or previous involvement with the police”. However, other jurors would often remind such speculators that this issue was not really relevant. See Young, W., Cameron, N. and Tinsley, Y., Juries in Criminal Trials Part Two: A Summary of the Research Findings NZ Law Com. Prelim. Paper 37, Vol. 2 (Wellington 1999), pp. 43–44Google Scholar.

80 Op. cit. n. 46, p. 751; op. cit. n. 47, p. 37.

81 Searle, J., Mind, Language and Society: Doing Philosophy in the Real World (London 1999), pp. 107–108Google Scholar.

82 Op. cit. n. 5, p. 451.

83 Ibid. p. 439.

84 See Dennis, op. cit. n. 59, pp. 47-48. Dennis here follows Hoffman, op. cit. n. 6.

85 Makin v. AG for New South Wales [1894] A.C. 57.

86 For a study which suggests people do sometimes think in this way, at least when confronted with statistics, see Wells, G.L., “Naked Statistical Evidence of Liability: Is Subjective Probability Enough?” (1992) 62 Journal of Personality & Social Psychology 739Google Scholar.

87 At least in situations involving historical fact-finding. People are probably far more prepared to base decisions about risk on information about past behaviour. Someone with a previous conviction for assaulting his children is going to finding it difficult to gain employment as a baby-sitter.

88 See Boardman, op. cit. n. 5. See now DPP v. P. [1991] 2 A.C. 447. In practice, the move to a different test of admissibility in P. has probably made little difference.

89 See P.R. D’Agostino, “The Encoding and Transfer of Stereotype-Driven Inferences” (2000) 18 Social Cognition 281; Z. Kunda and P. Thagard, “Forming Impressions from Stereotypes, Traits and Behaviours: A Parallel-Constraint-Satisfaction Theory” (1996) 103 Psychological Review 284; G.V. Bodenhausen, “Stereotypic Biases in Social Decision Making and Memory: Testing Models of Stereotype Use” (1988) 55 Journal of Personality & Social Psychology 726. One study of character evidence supports this view: V.P. Hans and A.N. Doob, “Section 12 of the Canada Evidence Act and the Deliberations of Simulated Juries” (1975) 18 Crim. L.Q. 235, 246. The authors found that “[w]hen the record of defendants is made known to the jurors, the evidence in the case changes meaning and appears more damaging”.

90 An example of how the case law has taken this sort of line is R. v. Lewis (1982) 76 Cr. App. R. 33.

91 See Pennington, N. and Hastie, R., “The Story Model for Juror Decision Making” in Hastie, R. (ed.), Inside the Juror: The Psychology of Juror Decision Making (Cambridge 1993)Google Scholar.

92 This was generally the case in the New Zealand study (see Young et al., op. cit. n. 79, p. 44) as well as in Lloyd-Bostock's research.

93 For discussion of some of these issues, see P. Tillers, “What is Wrong with Character Evidence?” (1998) 49 Hastings L.J. 781.