Prior convictions, conduct and disposition: a scientific perspective

ABSTRACT

The use of prior conduct and convictions in criminal proceedings has been a controversial issue for over a hundred years. Rarely does a year go by without another call from lawmakers or prominent citizens for prior convictions and ‘history’ to be made available to tribunals to resolve issues of guilt. Yet the debate seems to have reached an impasse, with the proponents of either side facing off at 20 paces, armed only with the familiar arguments that have failed to prevail in the past. The lack of resolution in the debate suggests that it needs to be reframed. The present article proposes a perspective based on broader scientific methodology, analysing factual decisions in a criminal trial as a form of probabilistic argument. The psychology of probabilistic reasoning is examined with a focus on the effect of attributional information on judgments of probability. The logical structures of factual reasoning about disposition, tendency, similar facts and coincidence are then put under the microscope to expose some new arguments, some weaknesses in old arguments, and some quantitative relations that affect probability judgments in this area.

ORCID

Peter M. Robinson http://orcid.org/0000-0002-5911-4251

Notes

¹‘A justice issue worth debating’, The Australian (online edition), 20 March 2014, http://www.theaustralian.com.au/opinion/editorials/a-justice-issue-worth-debating/story-e6frg71x-1226859513530.

²D’Cruz (2013); ‘Kevin Rudd: Juries should be told about prior convictions', Sunshine Coast Daily (online edition), 25 March 2013, http://www.sunshinecoastdaily.com.au/news/rudd-backs-jury-call/1804208/; ‘A justice issue worth debating’, The Australian (online edition), 20 March 2014, http://www.theaustralian.com.au/opinion/editorials/a-justice-issue-worth-debating/story-e6frg71x-1226859513530; ‘Supreme Court Justice Philip McMurdo backs debate over prior crimes', The Courier-Mail (online edition), 18 March 2014, http://www.couriermail.com.au/news/queensland/supreme-court-justice-philip-mcmurdo-backs-debate-over-prior-crimes/story-fnihsrf2-1226857466225.

³Abbott (2015); ABC News Online, ‘Sydney siege: Tony Abbott announces urgent review, including looking at Man Haron Monis' access to firearms', http://www.abc.net.au/news/2014-12-17/sydney-siege-tony-abbott-joint-commonwealth-nsw-investigation/5973842, 18 December 2014.

⁴Glazebrook (2010).

⁵Edgely (2007).

⁶Cunliffe (2014).

⁷In the UK, sections 101 and 103 of the Criminal Justice Act 2003 (UK) now combine to admit evidence of prior convictions to prove a propensity for the kind of crime charged. For an American perspective, see Eisenberg and Hans (2009), p 1356.

⁸ABC Lateline, ‘Should juries be told about the prior convictions of defendants', http://www.abc.net.au/lateline/content/2013/s3717030.htm, 15 March 2013.

⁹The term ‘hard evidence’ is used to refer to evidence directly attesting to the matters under charge.

¹⁰Wissler and Saks (1985), Greene and Dodge (1995), Lloyd-Bostock (2000, 2006), UK Law Commission (2001), Eisenberg and Hans (2009), Devine (2012), pp 96 et seq.

¹¹Wissler and Saks (1985), Lloyd-Bostock (2000).

¹²Tversky and Kahnemann (1983) described a series of studies that commenced in 1974. Daniel Kahnemann was later awarded the Nobel Prize, partly for this joint work with his late colleague, Amos Tversky.

¹³Tversky and Kahnemann (1983), p 299.

¹⁴Tversky and Kahnemann (1983), p 298.

¹⁵Tversky and Kahnemann (1983), p 301.

¹⁶Tversky and Kahnemann (1983), p 307.

¹⁷Tversky and Kahnemann (1983), p 308.

¹⁸There is much controversy about the precise nature of these conceptual structures that need not concern us here.

¹⁹Tversky and Kahnemann (1983), p 311.

²⁰Tversky and Kahnemann (1983), p 296; Feeney et al (2007).

²¹Beyth-Marom (1981), Fischhoff (1985), Gavanski and Roskos-Ewoldsen (1991).

²²Gavanski and Roskos-Ewoldsen (1991), Hilbert (2012).

²³Pennington and Hastie (1992).

²⁴Pennington and Hastie (1988).

²⁵Epstein (1994), Barbey and Sloman (2007), Wolfe and Reyna (2010), Juslin et al (2011).

²⁶Juslin et al (2011), p 264. A minority of researchers believe that the brain does perform valid probability calculus, though it is disrupted by systematic biases arising from ‘noisy’ informational input – see, for example, Costello and Watts (2014).

²⁷Juslin et al (2011), p 252.

²⁸Juslin et al (2011), p 248.

²⁹See, for example, Simon (1998), Simon et al (2001), Simon (2004), Simon, Snow and Read (2004), and Simon, Strensom and Read (2015).

³⁰Simon (2004), pp 563–564; Simon et al (2004), pp 816, 831–832.

³¹Simon (2004), pp 518, 533–535, 545; Simon et al (2001).

³²Simon (2004), pp 541–542; Simon, Snow and Read (2004), Study 4; Simon et al (2001).

³³Simon (2004), pp 521–522.

³⁴Simon (2004), pp 524, 536–537, 541–542; Simon et al (2004); Simon et al (2015).

³⁵Simon (2004), p 538; Simon et al (2004), pp 821–822.

³⁶Simon et al (2004), Study 2.

³⁷Simon (2004), pp 537–538

³⁸Simon (2002); Simon (2004), p 549.

³⁹The coherence effect could not be used to wholly explain conjunction fallacies, such that there would be only one effect in operation, because conjunction fallacies are found in simplified settings where no complex mental modelling arises to induce coherence.

⁴⁰Laudan and Allen (2011).

⁴¹ Perry v R (1982) 150 CLR 580, at 593–594; [1982] HCA 75, per Murphy J, at para 10.

⁴²Gigerenzer and Goldstein (1996).

⁴³The guarded statement in the text is based on mixed statistical findings. The Sentencing Advisory Council of Victoria (2007) found that in relation to the offence of murder, which one would only expect of individuals with a checkered past, only 42 per cent of those found guilty presented with a prior criminal record of any sort, and far fewer had a record of violence. The figures for manslaughter were even lower. For 2009–10, the NSW Bureau of Crime Statistics and Research reported that 57 per cent of sexual assault offenders and 59 per cent of homicide-related offenders had no prior conviction in the previous five years (Ringland 2011), although figures specifically for murder and attempted murder were lower – 39.3 percent and 33.3 per cent respectively (Taussig 2012). Only 1 per cent of homicide offenders had a record of prior homicide-related offences. Most NSW robbery offenders did have a prior record (81 per cent), but only 15 per cent had a prior robbery-related offence (Weatherburn and Taussig 2012). The figures for England and Wales cited by Redmayne (2015) in his Table 2.4 are broadly consistent with the Australian pattern, with serious crimes normally committed by offenders with no prior record. The American statistics, which lean strongly towards offenders with prior records, also reveal the difficulty of getting any sound statistical measure of this. In the US, the proportion of criminal defendants facing trial with prior convictions rose from 47 per cent in a 1966 study to 79 per cent in a 2000–01 study (see Laudan and Allen 2011, p 504). If prior convictions are over-represented in charging and prosecution decisions, then the conviction rates ensuing from such decisions are not a valid measure of the relative incidence of guilt.

⁴⁴Kahnemann and Tversky (1973).

⁴⁵Bar-Hillel (1980).

⁴⁶Barbey and Sloman (2007).

⁴⁷Pennycook and Thompson (2012).

⁴⁸Welsh and Navarro (2012).

⁴⁹Kahnemann and Frederick (2002), p 54.

⁵⁰See, for example, Redmayne (2015), p 17; Ulliver (1982), p 883; R v Kilborne [1973] AC 729, per Lord Simon of Glaisdale at 757, cited with approval in Sutton v The Queen (1984) 152 CLR 528, at 545–546; [1984] HCA 5, per Brennan J at para 10; Perry v R (1982) 150 CLR 580, at 585; [1982] HCA 75, per Gibbs CJ, at para 7; Harriman v R. (1989) 167 CLR 590, per Dawson J at 597, per Toohey J at 607; [1989] HCA 50, per Dawson J, at para 1, per Toohey J, at para 8.

⁵¹See Cossins (2011b); Redmayne (2015), p 17; R v Kilborne [1973] AC 729, per Lord Simon of Glaisdale at 757, cited with approval in Sutton v The Queen (1984) 152 CLR 528, at 545–546; [1984] HCA 5, per Brennan J at para 10. In Harriman v R ((1989) 167 CLR 590, at 595, 596; [1989] HCA 50, at pars 8 and 11), Brennan J goes close to such an analysis in suggesting that, by reason of the presumption of innocence, the jury might assume the accused to be a legitimate businessman, and past conduct was therefore relevant to rebut such a presumption. However, Brennan J also relied on an argument that the evidence showed a pre-existing ‘illegal business' that the defendant would be motivated to continue.

⁵²Uniform Evidence Acts, s 97.

⁵³ R v Randall [2004] 1 Cr App R 26; [2004] 1 WLR 56; [2003] UKHL 69.

⁵⁴ Lowery v The Queen [1974] AC 85.

⁵⁵If random sampling were used, there would be no guarantee that the sample would include any of the prior convictions.

⁵⁶Gans and Palmer (2014), p 190, specifically note this possibility with respect to tendency evidence under the uniform Evidence Acts.

⁵⁷The absolute value for that rate would not be needed, so long as the commission rate of past offenders relative to the general population was known. The relative rates of commission by offenders and non-offenders are recoverable from recidivism and general population statistics, but one potential problem is that recidivism statistics usually deal with rates of reoffending within particular time spans, such as two or three years. Estimates for offences older than that may therefore be troublesome.

⁵⁸‘Kevin Rudd: Juries should be told about prior convictions', Sunshine Coast Daily (online edition), 25 March 2013, http://www.sunshinecoastdaily.com.au/news/rudd-backs-jury-call/1804208/.

⁵⁹Finkelstein and Fairley (1970), Finkelstein and Fairley (1971), Tribe (1971a), Tribe (1971b), Fairley (1973), Fairley and Mosteller (1974), Lempert (1977).

⁶⁰Tiller (2011).

⁶¹The uniform Evidence Acts use the terms ‘character’, ‘reputation’, ‘conduct’ and ‘tendency’ – e.g. Evidence Act 1995 (Cth.), Evidence Act 1995 (NSW), Evidence Act 2008 (Vic.), s 97. The common law cases may use those terms, or alternatively ‘disposition’ or ‘propensity’.

⁶² Pfennig v R. (1995) 182 CLR 461, at 497; 127 ALR 99, at 125; 69 ALJR 147, at 166; [1995] HCA 7, per Toohey J at para 25.

⁶³Cossins (2011a, 2011b).

⁶⁴ R. v O’Keefe [2000] 1 Qd R 564, at 568; [1999] QCA 50, per Thomas JA at para 4.

⁶⁵Anderson 2012, pp 1933–1934.

⁶⁶Australian Law Reform Commission (1985), pars 794–800.

⁶⁷Australian Law Reform Commission (2006), pars 3.9–3.14.

⁶⁸Saks and Thompson (2003).

⁶⁹Anderson (2012), pp 1932–1933.

⁷⁰Kurland (2013).

⁷¹The term ‘extrinsic’ is used to refer to evidence that does not directly attest to the facts under charge (the opposite of ‘hard evidence’). It may be prior conduct, but may also post-date the offence, as in Pfennig.

⁷²For a useful summary of the legal perspective on these patterns, see Gans and Palmer (2014), pp 185–191.

⁷³These will usually be in the past, but need not be.

⁷⁴Han and Kamber (2006), p 228; Giudici and Figini (2009), p 175.

⁷⁵Bellazzi and Zupan (2008).

⁷⁶Here, the problem is formulated in terms appropriate to court proceedings. In scientific fields, the false negative rate (e.g. failure to predict a cancer risk) may be more important than the false positive rate.

⁷⁷ R. v Butler (1986) 84 Cr App R 12.

⁷⁸In market basket analysis, the item frequency is generally more important than entropy. Entropy is more often used in cluster analysis to remove redundant dimensions or attributes from a predictive set.

⁷⁹ Phillips v R (2006) 225 CLR 303; 224 ALR 216; [2006] HCA 4.

⁸⁰ Phillips v R (2006) 225 CLR 303, at 312; 224 ALR 216, at 222; [2006] HCA 4, at para 28; and the Court of Appeal decision, R v PS [2004] QCA 347, at para 60.

⁸¹ Phillips v R (2006) 225 CLR 303, at 317; 224 ALR 216, at 227; [2006] HCA 4, at para 42, and R v PS [2004] QCA 347, per Williams JA at para 63.

⁸²Hamer (2007), p 627.

⁸³Other critics of the High Court decision have steered clear of these errors in formulating arguments based on similarity – see Gans (2006); Cossins (2011a).

⁸⁴In the cross-admissibility cases, the evidence may appear to present as a coincidence pattern, but it has the potential to convert to the tendency pattern if the court finds guilt in one or more of the cases (which is the assumption on which admissibility is assessed).

⁸⁵ Makin v Attorney-General (NSW) [1894] AC 57; [1891–4] All ER 24 (PC).

⁸⁶ Smith v R. (1916) 11 Cr App R 229.

⁸⁷Cossins (2013b), p 740.

⁸⁸Cossins would argue that secret baby burials were also commonplace in those days, based on contemporary evidence described in Cossins (2013a).

⁸⁹The ‘relevant period’ raises another issue about properly framing the base classes which is too complicated to detail here. In Makin, the relevant period was arguably the period during which the defendants were allegedly engaged in ‘baby farming’.

⁹⁰See Cossins (2013b), at 734. More recently, this issue was noted in Velkowski v The Queen [2014] VSCA 121, at p 33, para 173(f).

⁹¹Cossins (2011b).

⁹² Perry v R. (1982) 150 CLR 580, at p 594; [1982] HCA 75, at para 11.

⁹³ Pfennig v R. (1995) 182 CLR 461, at 497, 506, 506; 127 ALR 99, at 125, 131; 69 ALJR 147, at 166, 171; [1995] HCA 461, per Toohey J, at paras 26, 47, 50.

⁹⁴ABC News Online, ‘Louise Bell murder: Man faces court on cold case’, http://www.abc.net.au/news/2013-11-19/louise-bell-murder-court-dieter-pfennig-forensic-breakthrough/5101812, 20 November 2013.

⁹⁵ ‘New DNA forensic technology used in reopened Louise Bell murder case’, The Advertiser (online edition), 9 October 2011, http://www.adelaidenow.com.au/news/south-australia/new-dna-forensic-technology-used-in-reopened-louise-bell-murder-case/story-e6frea83-1226162292688; ‘Raymond John Bolte jailed for seven years after abusing sisters', Daily Telegraph (online edition), 29 January 2010, http://www.dailytelegraph.com.au/raymond-john-bolte-jailed-for-seven-years-after-abusing-sisters/story-e6freuyi-1225824642260. It should be noted that the conviction was overturned in 1985 on the basis of doubtful witness testimony, so the miscarriage was discovered before the exculpating DNA evidence became available.

⁹⁶Cossins (2013b), p 741.

⁹⁷Plutinski (2011).

⁹⁸Morris (1998).

⁹⁹Imwinkelried (2005).

¹⁰⁰Wissler and Saks (1985), Lloyd-Bostock and Thomas (1996), Lloyd-Bostock (2000), Laudan and Allen (2011), Stanchi and Bowen (2014).

¹⁰¹Laudan and Allen (2011); UK Law Commission (2001), paras 6.18–6.22.

¹⁰²Cossins (2011b).

¹⁰³Barbey and Sloman (2007).

¹⁰⁴Simon (2004), p 544.

¹⁰⁵This would be identified in step 3 if a large sample were available, but it may not be evident from the limited data available to a court.

¹⁰⁶Of course, in a legal case, such data is partly the subject of surmise.

¹⁰⁷Redmayne (2008). Sadly, Mike Redmayne died at an early age while this article was in the process of being written. His substantial contributions to research on probability and the law over an extended period of time should be acknowledged here.

¹⁰⁸The example given by Gans (2006), p 228, of our reliance on customer reports on eBay sellers, is similar. We avoid sellers with adverse reports not because we have any confidence that they will offend on an isolated transaction, but because there is no point in taking the risk. However, in the situation of multiple adverse reports which Gans describes, it could well be that offending conduct was predominant in the seller's transactions, making it quite different from evidence of prior convictions.

¹⁰⁹ Boardman v DPP [1975] AC 421; [1974] 3 All ER 887.

¹¹⁰ DPP v P [1991] 2 AC 447; [1991] 3 All ER 337.

¹¹¹Imwinkelried (2005).

¹¹²To get a taste of the rich flavour of these issues, see Tribe (1971a) and Redmayne (2008, 2015).