Cossins, Annie --- "The Behaviour of Serial Child Sex Offenders: Implications for the Prosecution of Child Sex Offences in Joint Trials" [2011] MelbULawRw 29; (2011) 35(3) Melbourne University Law Review 821
THE BEHAVIOUR OF SERIAL CHILD SEX OFFENDERS: IMPLICATIONS FOR THE PROSECUTION OF CHILD SEX OFFENCES IN JOINT TRIALS
ANNIE COSSINS[*]
[This article summarises the literature on child sex offender (‘CSO’) recidivism and sexual behaviour, in order to consider the implications for the rules used at common law and under the uniform Evidence Acts (‘UEAs’) for excluding propensity/tendency and coincidence/similar fact evidence in trials involving multiple complainants. By analysing two distinct lines of authority from courts of appeal in New South Wales and Victoria, this article demonstrates that the use of the ‘striking similarities’ test and similar formulations under the UEA shows a lack of understanding of the wide variety in the behaviour of CSOs who do not demonstrate a propensity for strikingly similar behaviours with different victims. In recommending reforms to the UEA to reflect the reality of CSO sexual behaviour, the article concludes that the striking similarities test and similar formulations should be abandoned when assessing the cross-admissibility of the evidence of multiple complainants in a joint trial involving child sex offences.]
I INTRODUCTION
Serial and repeat sex offending is a phenomenon that is well documented in the psychological literature.[1] Although the observed recidivism rate for child sex offenders (‘CSOs’) is considered to be between 10 and 20 per cent after five years,[2] this is likely to be a substantial underestimation since a majority of children only disclose sexual abuse one or more years after it has occurred, or not at all.[3] In fact, ‘[s]tudies using data from police and child protection files have estimated that actual recidivism rates are more than twice those indicated by reconviction data’.[4]
There are many factors that affect a child’s ability to report sexual abuse.[5] Children are less likely to disclose if they are abused by a relative, while stranger abuse is the single best predictor of rapid disclosure.[6] Less intrusive forms of abuse may be more likely to be reported, while children who experience multiple abuse and younger children are less likely to disclose.[7] Under-reporting means that the true recidivism rate for serial offenders may be impossible to estimate, although the rate is likely to be much higher than the estimation of 10–20 per cent.[8]
While the recidivism of CSOs presents key issues in relation to child safety and detection by authorities, it also poses problems for the prosecution of sex offences involving multiple offences and/or multiple complainants — problems that have not been previously discussed in the legal literature.[9]
This article summarises the literature on CSO recidivism, choice of victim and types of sexual behaviours. It presents a 16 year literature review which shows that estimates of reoffending are highly dependent on methodology, including the length of the follow-up of the offenders under study.
Since this information has not informed the development of trial processes, procedure and evidence law in Australia, this article will consider its implications for the prosecution of serial CSOs in trials with multiple complainants. In particular, the article will consider the implications of serial CSO behaviour for the rules used at common law and under the uniform Evidence Acts (‘UEAs’)[10] for excluding evidence known as tendency/propensity and coincidence/similar fact evidence.
In analysing two distinct lines of authority from the courts of appeal in New South Wales (‘NSW’) and Victoria and their use of ‘striking similarities’ and similar formulations, the article identifies the existence of a belief that CSO behaviour is distinctive in nature and CSOs are highly selective in their choice of victim. These beliefs demonstrate a lack of understanding of CSO behaviours, including: the fact that a core group of offenders will abuse more than one child; the wide variety in offenders’ sexual and grooming behaviours; and the fact that recidivism is higher for offenders who exhibit crossover behaviours and commit a diverse range of sex offences.
The article argues that the striking similarities test (and similar formulations) should be abandoned when assessing the admissibility of multiple complainants’ evidence in a joint trial and makes recommendations for reform.
II THE LEGAL SIGNIFICANCE
OF CHILD SEX
OFFENDER BEHAVIOUR
In order for the evidence of two or more complainants to be cross-admissible as propensity or similar fact evidence in a joint trial for child sex offences, the common law requires the evidence to be relevant and to have ‘striking similarities’. As discussed below in Part IV, the application of this formulation or test in child sexual assault trials has resulted in detailed comparisons of the types of alleged sexual behaviour committed against multiple complainants by an accused. Where the sexual behaviour is considered to have sufficient similarities, the striking similarities test will be satisfied and the evidence will be admissible if, in addition, the ‘no rational view of the evidence’ test (also know as the Pfennig test) is satisfied.[11]
The striking similarities test has been adopted in relation to the admissibility of tendency and coincidence evidence under the UEAs,[12] suggesting that when presented with new statutory tests for admissibility judges will still be guided by the common law in the absence of clear guidance from legislatures on how to apply these tests.
If an accused’s alleged sexual conduct with one victim differs in type from his alleged sexual conduct with another, it is generally considered that there will be no striking similarities between the evidence of the victims. For example, in a Queensland case involving four victims who ranged in age from 6 months to 12 years,[13] Holmes JA held that ‘[t]he fact that the appellant had interfered with A [aged 12 years] in certain ways did not go very far to prove that he had molested B [aged 8 years] in quite different ways.’[14] Holmes JA also stated that
none of the evidence concerning indecent dealing with A and B themselves was admissible in respect of [the charges involving the defendant’s infant children]; by no stretch of the imagination could one say that the dealings with them had any similar fact relationship with what happened to the infants.[15]
These and similar views expressed in other cases indicate the existence of a belief that CSO behaviour is distinctive in nature and CSOs are highly selective in their choice of victim. These views raise a number of questions which are the subject of this article: to what extent are CSOs likely to engage in repeat and serial offending? Do they always choose victims of the same age and sex? To what extent does their sexual behaviour with children vary?
III THE CHARACTERISTICS OF CHILD SEX OFFENDER BEHAVIOUR
A The Specialisation and Persistence of Child Sex Offenders
There are many features concerning the behaviour of CSOs that can be discerned from the literature, including reoffending rates and whether all types of CSOs reoffend at the same rate. For example, in studies of the specialisation of rapists compared to CSOs, researchers have found that all types of sex offenders commit a range of non-sex offences during their criminal careers, although CSOs are much more likely than rapists to ‘specialise’ in sex offences.[16] This degree of specialisation means CSOs are more likely to be repeat sex offenders and to have multiple victims.
Nonetheless, there is considerable debate about the recidivism rates of CSOs. Table 1 sets out recidivism rates reported in the literature from 1995 to 2011. Where possible, CSOs are categorised according to their relationship with a child victim;[17] however, most studies in Table 1 did not make a distinction between different offender types. While it is beyond the scope of this article to engage with the literature on actuarial predictions of an offender’s risk of recidivism, measures of risk are dependent on the measure of recidivism used.[18]
Table 1: Review of Recidivism Rates of CSOs 1995–2011[19]
Study
|
Sample Size
|
Offender Type*
|
Follow-Up Period
|
Measure of Recidivism
|
Recidivism Rate (%)
|
---|---|---|---|---|---|
Hanson, Scott and Steffy
(1995)[20]
|
191
|
intra- and extra-familial
|
15 to 30 years
|
re-conviction
|
35.1
|
Prentky et al
(1997)[21]
|
115
|
CSOs
|
25 years
|
charge
re-conviction
imprisonment
|
32
(52 FR**)
25
(41 FR**)
23
(37 FR**)
|
Firestone et al
(1999)[22]
|
251
|
intra-familial
|
average 6.7 years
|
rearrest or re-conviction
|
6.4
|
Firestone et al
(2000)[23]
|
192
|
extra-familial
|
average 7.8 years
|
rearrest or re-conviction
|
15.1
|
Soothill et al
(2000)[24]
|
763
|
heterosexual CSOs
|
average 15 years
|
re-conviction
|
19.3
|
Beech et al
(2002)[25]
|
53
|
CSOs
|
6 years
|
re-conviction
|
15
|
Greenberg et al
(2000)[26]
|
400
|
extra-familial
|
average 7 years
|
charge or
re-conviction
|
16.2 (acquaintances)
9.7 (strangers)
|
Serin, Mailloux and Malcolm
(2001)[27]
|
3
8
|
intra-familial
extra-familial
|
7 years
|
re-conviction or violation of conditions of release
|
3.0
1.5
|
Hood et al
(2002)[28]
|
35
19
|
intra-familial
extra-familial
|
6 years
|
re-conviction
|
0
26.3
|
Scalora and Garbin
(2003)[29]
|
194
|
treated and untreated CSOs
|
average 4.5 years
|
charge
|
24.7
|
Bartosh et al
(2003)[30]
|
37
59
|
intra-familial
extra-familial
|
5 to 5.5 years
|
rearrest
|
11
14
|
Langevin et al
(2004)[31]
|
46
142
|
intra-familial
extra-familial
|
25 years
|
re-conviction
charge
self-report and official data***
re-conviction
charge
self-report and official data***
|
51.1
60.0
84.2
73.7
79.0
94.1
|
Soothill et al
(2005)[32]
|
103
116
|
convicted CSOs
suspected CSOs
|
average 15 years
|
re-conviction
|
20
9
|
Patrick and Marsh
(2009)[33]
|
447
|
CSOs
|
2 to 15 years
|
re-conviction
|
9.2
|
Goodman-Delahunty
(2009)[34]
|
214
|
intra-familial
|
3 to 18 years survival rate
|
charge or
re-conviction
|
10.7
|
Parent, Guay and Knight
(2011)[35]
|
275
|
CSOs
|
10 years
|
charge
|
17.4
|
Notes to Table 1:
* This table only includes studies that reported the recidivism rate for CSOs and does not include studies that did not differentiate between rapists and CSOs.
** FR means ‘failure rate’, which is ‘an estimate that takes into account the amount of time each offender has been on the street and [is] thus able to reoffend.’[36] It is considered to be a more accurate estimate of recidivism.
*** Official data included police databases, hospital records and legal databases.
Before discussing the data in Table 1, three meta-analytic studies and one large-scale study are also reviewed here. Such studies are statistically more powerful because they comprise a larger cohort of offenders than the individual studies in Table 1. They also contain analyses which enable the predictors of recidivism to be identified.
In 1998, Hanson and Bussière’s large-scale study used a meta-analytic technique for integrating data from 61 studies involving 28 972 sex offenders.[37] For the 9603 CSOs in the study, who were not differentiated into types, they reported a recidivism rate of 12.7 per cent for an average follow-up period of four to five years.[38] They also identified the factors that significantly correlated with recidivism, including sexual arousal to children and prior arrests or convictions for a sex offence.[39] Recidivism was higher for those CSOs who had committed a diverse range of sex offences,[40] suggesting that serial offenders may be more likely than other offenders to engage in different types of sex offences.
In 2004, the meta-analytic study by Harris and Hanson provided recidivism rates for 2595 CSOs at 5, 10 and 15 years.[41] Most had been released from prison and came from 10 different samples, a majority of which used re-conviction as the recidivism measure. After 15 years, the recidivism rates were 13 per cent (intra-familial offenders), 16 per cent (heterosexual extra-familial offenders) and 35 per cent (homosexual extra-familial offenders).[42]
Generally, these rates are similar to the rates from the studies listed in Table 1 that used re-conviction as the recidivism measure, but are lower than those from the studies that used charges and rearrest. Notably, ‘[o]ffenders with a prior sexual offence conviction had recidivism rates about double the rate observed for first-time sexual offenders (19% versus 37% after 15 years)’,[43] while those under the age of 50 years were more likely to reoffend after 15 years compared to older offenders.[44]
In a review of 16 British studies, Craig et al reported the mean re-conviction rates of eight incarcerated (n = 5195) and eight non-incarcerated (n = 1274) samples of CSOs at 2 years, 4 years and 6 years or more.[45] Incarcerated offenders showed a slightly higher recidivism rate after 6 years or more (19.5 per cent) compared to non-incarcerated offenders (15.5 per cent).
Craig et al’s comparison of a group of relatively high-risk CSOs with a relatively low-risk group can be contrasted with Freeman’s study that included all registered male CSOs on probation (a relatively low-risk group) as at August 2005 in New York State (n = 4700).[46] Freeman found that the reoffending patterns of probationers mirrored those of incarcerated CSOs, an unexpected result given that incarcerated offenders are generally considered to be more likely to reoffend. Using rearrest as the measure of recidivism and a follow-up period of three years, Freeman reported a recidivism rate of 5.7 per cent. This was identical to the rate reported by Craig et al for non-incarcerated offenders after two years and almost the same as the rate for incarcerated offenders (6 per cent) after two years.[47]
However, the CSOs in Freeman’s study were more likely to be rearrested for a non-sex offence than a sex offence (30.1 per cent compared to 5.7 per cent), which may be due to the visibility of the particular non-sex offences (assault, drug offences, vehicle/traffic violations and failure to register with the New York sex offender register) compared to child sexual assault.
Freeman found that
the best predictor of rearrest for a sexual offense for both rapists and child molesters was number of prior sexual offense arrests. Each additional prior sexual offense arrest increased the probability of a rapist being rearrested for a sexual offense by 0.1%, whereas it increased the likelihood of a child molester being rearrested by 15.7%.[48]
There were three additional variables that were significant predictors for the rearrest of CSOs in Freeman’s study: the victim’s sex, the number of victims in the ‘instant offence’ and the CSO’s age at the time of the instant offence.[49] Those CSOs who had more than one victim in the instant offence and who had a male victim were more likely to be rearrested — each additional victim increased the likelihood of a CSO being rearrested by 2.7 per cent, and an offender was also 2.7 per cent more likely to be rearrested if the alleged victim was male.[50] Similarly, younger CSOs were more likely to be rearrested for a sexual offence — every one year increase in age reduced the likelihood of rearrest by 0.3 per cent.[51]
1 Recidivism Studies: Their Strengths and Limitations — Methodological Factors
Generally speaking, the broader the definition of recidivism and the longer the follow-up period, the higher the recidivism rate. In analysing the data reported in Table 1, the most important caveat is that ‘recidivism rates … are gross underrepresentations of the real number of offenses committed, no matter what criteria are used’.[52] This can be seen in the wide variation in the rates in Table 1 (from 0 per cent to 94.1 per cent). This variation arises from a number of methodological factors, such as the sample size of the studies (from 11 to 763) the proportions of different types of offenders included in the sample, the follow-up period (from 2 to 31 years), the measure of recidivism, and whether the sample included high-risk offenders.
While there is a good argument that the recidivism rates reported in meta-analytic studies should be given greater weight than the rates from individual studies, Hanson and Bussière warned that their recidivism rate should be treated with caution because it was based on 61 studies with a diverse range of measures of recidivism, offender types and follow-up periods.[53] Meta-analyses can also ‘easily neglect potentially important differences between studies’ such as offender types and offender settings.[54]
CSOs are known to commit a wide range of both sexual and non-sexual offences,[55] although estimates of the recidivism rate of American CSOs are limited by the frequency of plea bargaining in the United States: ‘sex offences are often reclassified through plea bargaining as violent non-sexual charges … and even as property offences (such as break and enter) if, for example, an offender is foiled in an attempted rape’.[56] The use of new charges as the measure of recidivism ‘avoids certain limitations associated with the use of conviction as [the] measure … especially those related to plea bargaining’, where sexual charges are modified into non-sexual charges in order to obtain a guilty plea,[57] thus turning high-risk offenders into apparently low-risk offenders.
From the data listed in Table 1 and reported in the meta-analytic studies, several trends emerge. Recidivism rates will vary widely based on the measure of recidivism that is used. Where re-conviction is used as the measure of recidivism, lower rates of recidivism are reported than when measures such as rearrest, charges or self-report data are used. For example, in the study by Prentky et al ‘[t]he rate of recidivism using conviction or incarceration was two thirds that of the rate using charge throughout the study period.’[58] Similarly, Doren recalculated the recidivism rate reported by Hanson, Scott and Steffy by reference to the likely charge rate.[59] As a result, Doren found ‘a more accurate’[60] recidivism range of 44.6 per cent to 51.6 per cent compared to the recidivism rate of 35.1 per cent based on reconviction data that had been reported by Hanson, Scott and Steffy.[61] Similarly, recidivism rates are higher where lifetime failure rates are calculated.
Langevin et al reported the highest recidivism rates out of any study in Table 1. Using both official and self-report data as their recidivism measures, 94.1 per cent of 142 extra-familial offenders and 84.2 per cent of 46 intra-familial offenders had reoffended after 25 years, one of the longest follow-up periods in Table 1.[62] Furthermore, Langevin et al reported the percentage of undetected offences in the criminal histories of 142 extra-familial CSOs as 76.5 per cent, and of 46 intra-familial offenders as 71.8 per cent, indicating that even after a 25 year period around three quarters of these CSOs’ crimes remained undetected. Langevin et al’s findings stand in stark contrast to those from other studies which have concluded that CSOs reoffend at relatively low levels and intra-familial offenders are less dangerous than extra-familial offenders because they are less likely to reoffend.[63]
The high rates reported by authors such as Doren and Langevin et al suggest that the rates from other studies listed in Table 1, which range from 0 to 35.4 per cent, could be significant underestimates. However, there are key limitations to Langevin et al’s study. The first is the relatively small size of the study samples, particularly in relation to intra-familial offenders (n = 46). The second is that other studies which used long follow-up periods did not produce similarly high rates. This suggests that the particular sample of CSOs studied by Langevin et al — those referred for psychiatric treatment — was another limitation because these offenders may reoffend at higher rates than other CSOs. Even so this limitation is not clear-cut since the period of referral (1966–74) for the CSOs in Langevin et al’s study was a time when psychiatry was sometimes used as an alternative to a criminal charge and there was a belief that deviant sexual behaviour could be cured by standard psychiatric treatment.[64]
Another trend that emerges from Table 1 is the longer the follow-up period, the higher the recidivism rate. This is graphically illustrated in the study by Harris and Hanson, who calculated recidivism rates at 5, 10 and 15 years.[65] For ‘boy victim’ extra-familial offenders the rate of recidivism increased from 23 per cent at 5 years to 35 per cent at 15 years, and for ‘girl victim’ extra-familial offenders the rate of recidivism increased from 9 per cent at 5 years to 16 per cent at 15 years. For intra-familial offenders, the rate of recidivism doubled (from 6 per cent to 13 per cent) within the same period.[66]
Similarly, Prentky et al, who followed CSOs over a 25 year period, noted that ‘[i]f [they] had restricted [their] follow-up to the conventional exposure period of 12 or 24 months, [they] would have erred by approximately 45% or 40%, respectively.’[67] The results of Langevin et al’s study also support long follow-up periods since the average time span of the criminal histories of the CSOs in their study was almost two decades.[68] In fact, long follow-up periods are essential to estimating recidivism rates, since between 47.9 per cent[69] and 69 per cent[70] of children will not have reported within five years of the abuse.
Recidivism rates will also vary depending on sample size, sample composition and the source from which offenders are obtained. Where a study sample has a larger proportion of serial offenders, this will increase but also skew the recidivism rate, as will the inclusion of a larger proportion of offenders older than 50 years who are either less likely to reoffend compared to younger age groups[71] or are more adept at avoiding detection. Similarly, ‘offenders committed to a state hospital … [and offenders from] a maximum security psychiatric hospital … are likely to differ in ways that would affect their recidivism rates and make cross-sample comparisons difficult.’[72] This argument also applies to samples which include CSOs who have completed a treatment program and those who have not. For example, Scalora and Garbin reported that treated offenders were much less likely to reoffend compared to those who had dropped out of treatment or for whom treatment was unsuccessful (2.1 per cent compared to 25 per cent).[73] The proportion of different types of offenders (intra-familial versus extra-familial and mixed type) in a study sample is also considered to account for variations in recidivism rates in different studies, because intra-familial offenders are thought to reoffend at a much lower rate than extra-familial offenders.[74]
However, just over half of the studies in Table 1 did not differentiate between types of CSOs. Of the ones that did, extra-familial offenders reoffended at a considerably higher rate than intra-familial offenders. But even this is a generalisation with exceptions, since intra-familial offenders under the age of 25 years when released from prison show higher recidivism rates than extra-familial offenders.[75] As well, few studies have considered the rates of disclosure by children and the particular factors that predict disclosure, such as stranger abuse, and that predict non-disclosure, such as long-term intra-familial abuse.
In fact, a possibility not recognised in most of the studies listed in Table 1 is that intra-familial offenders may sexually abuse their own children for many years, engaging in repeat abuse of one or two victims, while extra-familial offenders may abuse several victims. A single charge of child sexual assault ‘may represent years of abuse of a single victim’,[76] thus resulting in an underestimation of an intra-familial offender’s criminal behaviour, particularly where follow-up periods are relatively short.
In a recent study of low-risk intra-familial offenders, Goodman-Delahunty reported that the vast majority (90 per cent) had offended on more than one occasion[77] and 86 per cent had committed a penetrative offence.[78] For a majority of the 214 participants in the study, ‘the index offence was not the most extensive abusive behaviour by the offender, demonstrating that intrafamilial offenders are more deviant … than has often been presumed.’[79] In fact, the wide range of abusive acts committed by these offenders ‘indicated that the sexually deviant practices exhibited by this group of intrafamilial [CSOs] was similar to those of extrafamilial [CSOs].’[80] It is, therefore, possible that, given the relatively low reporting rate of children victimised by a family member,[81] intra-familial offenders are less likely to be detected and ‘merit longer-term follow-up and a re-evaluation of their sexual preferences and general criminality’.[82]
2 Recidivism Studies: Their Strengths and Limitations — Criminal Justice and Victim Factors
An individual’s recidivism risk will differ from his group’s rate ‘to the extent that the offender differs from “typical” members of the group’ by committing more or less sex offences.[83] However, rates of recidivism are dependent on a child or parent making a report to the police. If recidivism measures such as charges and re-conviction are used, then high attrition rates at the investigation and prosecutorial stages must be recognised as having a significant effect on recidivism rates. However, the discussion in many recidivism studies assumes that each CSO who reoffends has exactly the same chance of being arrested, charged and/or convicted.
In fact, there is no level playing field when it comes to the detection and prosecution of child sexual abuse because of diverse factors such as a child’s age, relationship with the offender, relationship with their parents, access to the police and emotional stability to give evidence; an offender’s silencing methods; parental denial; quality of the police investigation; presence or absence of corroborating evidence; low guilty plea rates; and low conviction rates at trial.[84]
Although Harris and Hanson concluded that 73 per cent of all sex offenders in their meta-analytic study had not reoffended after 15 years, and that ‘strong contradictory evidence is necessary to substantially change these recidivism estimates’,[85] the above factors mean such confidence may be misplaced, at least in relation to CSOs. In fact, two reviews of studies on the reporting rates of child sexual abuse confirm that long delays are a typical feature of this particular crime.[86] This means that child sexual abuse is unlike many other types of offences which are more visible because of high-level policing (eg assault and public order offences) or higher rates of reporting (eg motor vehicle theft and serious violent crime).[87] While CSOs are more likely to be re-convicted of violent or property offences,[88] what cannot be discounted is the visibility of these crimes compared to child sexual abuse, and the impact of plea bargaining.
Another issue not discussed in the literature is that some recidivism rates are remarkably similar to the estimated likelihood of an offender being convicted once a report is made to the police which, in Australia, is 8 to 10 per cent.[89] Given the low reporting rates by children, combined with high attrition rates of child sex offences at the policing and prosecutorial stages, recidivism rates using re-conviction as the measure might be measuring nothing more than the (low) probability of an offender proceeding to trial and conviction.[90] Statistically speaking, the likelihood of a child reporting the abuse and that report resulting in a conviction is very low indeed.
3 Summary
This literature review has shown that, depending on the offender sample, the follow-up period and the measure of recidivism, reoffending will be an issue for a significant proportion of CSOs — from around 5 to 30 per cent for low-risk offenders to around 90 per cent for high-risk offenders. Not only are there several methodological issues that make it difficult to estimate precise rates, there are also many victim and criminal justice factors that militate against the detection and prosecution of child sex offences which are not taken into account in most recidivism studies.
B Crossover and Variation in Child Sex Offender Behaviour
CSO behaviour is characterised by certain elements that challenge many of the beliefs in the case law about CSO choice of victim and sexual misconduct. These elements include ‘crossover’ behaviours (ie the extent to which CSOs choose victims of different gender, age and relationship) and variation in sexual behaviour with different victims. This information has been reported by a number of researchers who have conducted self-report studies with offenders. The key findings from these studies are presented below.
1 Features of Self-Report Studies
Self-report studies of CSOs are difficult and time-consuming to conduct since they require surveys of a known population of offenders (such as clinical or incarcerated groups) who must be given guarantees of confidentiality for the data to be considered reliable. However, ‘the ability of this approach to measure undetected offending makes it a valuable complement to research based on official sources of data.’[91]
While it is possible that offenders under-report their offences in self-report studies, generally these studies reveal ‘a greater proportion of the “true” prevalence of offending than do official reports’,[92] indicating that ‘the gap between actual and officially recorded sex offending … is enormous’.[93] This gap suggests that ‘a certain proportion of [the] child molester population are actively perpetrating acts of sexual violence against children for some two decades before they are detected by an authority agency’.[94]
Indeed, if the recidivism rates for serial CSOs are reflected in the cases that come to court (and no study has as yet followed the progress of CSOs who reoffend through the criminal justice system), then it can be expected that police and prosecutors will be dealing with a significant proportion of cases involving an offender with multiple victims.
2 The Crossover Behaviour of Child Sex Offenders
More than 20 years ago, Abel et al conducted the first study to determine whether or not sex offenders sexually abused only one type of victim or whether they chose multiple victim types.[95] They reported that 20 per cent of 561 sex offenders crossed over between male and female victims, 23.3 per cent crossed over between children related to them and unrelated victims, and just over 42 per cent crossed over between age groups (children, adolescents and adults).[96]
Other studies have reported similar results. In 1991, Weinrott and Saylor found that 34 per cent of extra-familial CSOs sexually abused children related to them while 50 per cent of intra-familial CSOs reported abusing an unrelated child.[97] Twelve years later, in a study of 255 CSOs, Heil, Ahlmeyer and Simons found that a much larger proportion of intra-familial offenders said they had sexually abused unrelated children (64 per cent) while 53 per cent of extra-familial offenders reported sexually abusing children who were related to them.[98] In 2010, an Australian study reported that 48 per cent of 128 CSOs selected children from different age groups, while 25.8 per cent selected both related and unrelated victims.[99]
In fact, in various studies, significant minorities of CSOs have admitted to selecting victims from both sexes, with proportions including 21 per cent,[100] 22 per cent,[101] 28 per cent[102] and 63 per cent.[103]
These crossover behaviours have implications for recidivism since CSOs who abused both boys and girls were three times more likely to have preschool aged victims than offenders who had older victims,[104] while offenders who engaged in crossover behaviour were more likely to have more victims and to be at risk of recidivism[105] than offenders who did not. Similarly, offenders who selected victims from different age groups also had more victims.[106]
These studies suggest that between one fifth and two thirds of CSOs will have multiple victims and that their sexual behaviour will vary according to the age and sex of the child. This hypothesis is supported by a detailed Australian self-report study carried out by Smallbone and Wortley in relation to 182 incarcerated male CSOs who completed a self-report questionnaire about their sex offending history.[107] The sample comprised 79 intra-familial offenders, 60 extra-familial offenders, 30 mixed-type offenders (those with related and unrelated victims) and 13 deniers.[108] Most (73.3 per cent) reported they had female victims.[109] Most extra-familial offenders (86.7 per cent) and mixed-type offenders (89.6 per cent) reported sexual contact with children known to them[110] while only 6.5 per cent of offenders reported having their first sexual contact with a child who was a stranger.[111]
The 169 offenders who admitted having at least one victim disclosed offences involving 1010 children (748 boys and 262 girls).[112] 34.4 per cent of offenders reported sexual contact with two to five children while almost 17 per cent reported contact with six or more.[113] Thus, while just over half of the CSOs reported they had more than one victim, only 38.9 per cent were associated with official convictions. Intra-familial offenders disclosed an average of 1.5 victims, compared to extra-familial offenders (average 6.1 victims) and mixed-type offenders (average 20 victims).[114] Most offenders (72 per cent) ‘had more than one sexual encounter with their first victim and 28 per cent had more than 10 sexual encounters.’[115] In fact, even though intra-familial offenders are considered to be less dangerous than other offender types, they were more likely to have multiple sexual contacts with their first victim and more likely to offend over a longer period of time with their first victim.[116]
Although a majority of all offenders admitted to a previous sex offence, this was not reflected in previous convictions: only 21.3 per cent had a previous conviction for a sexual offence.[117] Offenders admitted to having their first sexual contact with a child many years before being convicted for their current
offence — on average, 8.6 years for intra-familial offenders, 10.6 years for extra-familial offenders and 11.5 years for mixed-type offenders[118] — supporting the view that convictions are not a reliable method for assessing the frequency of sex offending behaviour. Extra-familial offenders (30.5 per cent) and mixed-type offenders (41.4 per cent) were more likely to have a previous sexual conviction than intra-familial offenders (10.8 per cent),[119] most of whom committed their self-reported sexual offences in their own homes,[120] making detection less likely.
3 The Variation in Child Sex Offender Behaviour
The results from Smallbone and Wortley’s study confirm the findings from previous self-report studies that CSOs display high levels of sexual misconduct that is never detected by authorities. In other words, CSOs’ sexual behaviour is more varied in terms of number of victims and types of sexual misconduct than is recognised by judges and lawyers dealing with child sex offence cases.
Rather than being confined to one particular type, CSO sexual behaviour is best described as involving a continuum, beginning with non-sexual touching, moving to less serious sexual touching to test out a child’s reactions and sometimes progressing to serious sexual assault involving oral, digital or penile penetration, depending on access, opportunity, the age of the child, the resistance of the child and the successful use of emotional manipulation, coercion or threats.
Smallbone and Wortley’s study revealed the strategies used by 169 CSOs for targeting children, as well as their grooming and sexual behaviours.[121] The study did not support the judicial assumption that offenders specialise in particular types of sexual behaviour.[122] While this study was not designed to measure the particular sexual behaviours committed by individual offenders, as a group, the 169 CSOs committed a diverse range of sexual behaviours. Most admitted to touching a child’s genitals (89 per cent), in addition to 17 other sexual acts, including oral sex on the victim (59.8 per cent), penetration of the vagina/anus with a finger (69.6 per cent) or a penis (47 per cent), and masturbation of the victim (46.9 per cent);[123] many also admitted to having a child perform sexual acts on them.[124]
Intra-familial offenders admitted to an average of 5.6 different sexual behaviours, while extra-familial offenders admitted to an average of 5.5 and mixed-type offenders admitted to an average of 6.4. As a combined group, the 169 offenders admitted to an average of 6.1 different sexual behaviours.[125] In other words, CSOs do not appear to ‘specialise’ in one particular type of sexual contact to the exclusion of other types.
Similar findings were reported by Goodman-Delahunty, who carried out an evaluation of a NSW sex offender treatment program which only accepts intra-familial CSOs who admit their offences.[126] The 214 male offenders who were the subject of the study (of whom 93 had been accepted into the program and 121 had been declined)[127] had an average of 1.2 victims.[128] The mean duration of offending was 3.2 years and the mean number of reported incidents of child sexual abuse was 128.7.[129] Although the intra-familial offenders had extensive offending histories,[130] the mean number of charges was 5.7,[131] indicating that a majority of incidents of child sexual abuse do not result in charges being laid.
Goodman-Delahunty also documented the types of abuse that intra-familial offenders had committed. While the most common abusive act was sexual touching or fondling (89.7 per cent of all offenders), offenders admitted to 11 other different types of sexual behaviours, including oral abuse (45.8 per cent), exposure (46.7 per cent), digital penetration (37.4 per cent) and vaginal intercourse (34.8 per cent).[132] The fact that the percentages for each of the 12 different types of offences add up to more than 100 per cent demonstrates that some offenders engaged in two or more sexual behaviours with their victims.
In fact, Goodman-Delahunty undertook analyses of the types of sexual acts committed by offenders and found that 90 per cent committed more than one act of sexual abuse[133] while only 5.1 per cent committed one type of sexual abuse.[134] Together, the offenders committed an average of 4.5 different types of sexual acts with the victim of the subject offence, with only 5.6 per cent of offenders committing one abusive act during the subject offence.[135] Offenders who completed treatment admitted to abusive acts that were wider in range and more intrusive than the acts reported by victims,[136] suggesting that victims under-report the types of sexual abuse perpetrated against them.
4 Summary
Overall, the literature shows that a significant proportion of offenders (between one fifth and two thirds) will have victims of different ages, sexes or relationship (related/unrelated) and that offenders who engage in crossover behaviours are more likely to be at risk of recidivism. In fact, the recidivism studies reviewed in this Part suggest that serial child sex offending is not a rare or unusual event and that offenders who have had more than one victim are more likely to reoffend. However, when multiple victims report abuse by the same person, there are major hurdles to proving that such serial abuse occurred, as discussed in the next Part.
Of those offenders who do have multiple victims, most will have engaged in a range of sexual behaviours with their victims. These findings have serious implications for the prosecution of serial CSOs, and ought to inform the decision-making of judges in such cases.
IV IMPLICATIONS FOR THE PROSECUTION OF CHILD SEX OFFENCES
The CSO literature shows that ‘one of the best predictors of sexual recidivism is a previous sex offense’ conviction[137] along with prior charges.[138] In fact, ‘any previous accusations of sexual abuse may signal a pattern of sexually abusive behaviour’[139] — that is, (in legal terminology) a tendency to act in a particular way. The case law discussed in the next two sections reveals the assumption that serial CSOs specialise in relation to the age, sex and relationship of their victims and in relation to particular types of sexual behaviours. Since this assumption is not supported by the literature on CSOs, it appears to have arisen out of the demands of the striking similarities test, rather than being based on empirical data.
Cases involving alleged serial offenders pose the greatest difficulties for prosecutors because of the multiplicity of victims and the existence of tendency/propensity and coincidence/similar fact evidence. While a judicial decision to accept or reject a defendant’s application for separate trials ‘is a decision on a matter of procedure’,[140] that decision depends on whether the complainants’ evidence is cross-admissible in a joint trial as tendency/propensity or coincidence/similar fact evidence. This means the evidence of one complainant must be relevant to and sufficiently probative in relation to the counts involving the other complainant(s) for it to be admitted. The same applies to the evidence of the other complainant(s) in relation to the counts involving the first complainant. A decision that the complainants’ evidence is cross-admissible enables all of them to give evidence in a joint trial.
As discussed in Part II, at common law, cross-admissibility is dependent upon satisfaction of the striking similarities test, a test that has had a considerable degree of application under the UEA. The question is whether the striking similarities test (and similar formulations) is the appropriate measure for deciding cross-admissibility and whether or not a joint trial should be held in relation to charges against an alleged serial CSO.
A Striking Similarities at Common Law
The common law has a long history of excluding evidence of a defendant’s past criminal conduct in criminal trials. In Australia, exceptions to this general rule were developed mostly in relation to a series of poisoning cases in the late 1800s.[141] These cases reflected both the ease with which poisons, such as arsenic and antimony, were available at the time and judges’ frustration in relation to the difficulty of proving the act of poisoning by defendants.[142] Previous evidence of a defendant’s involvement in a person’s death by poisoning became an exception to the rule against admitting evidence of past criminal conduct.
Although the term was not in use at the time when these early cases were decided, evidence of past criminal conduct eventually became known as propensity or similar fact evidence and was only admissible if it met the striking similarities test.[143]
The admissibility of tendency/propensity and similar fact/coincidence evidence is governed either by the common law or statute depending on the jurisdiction.[144] While the terminology of ‘tendency’ and ‘coincidence’ evidence is used under the UEA,[145] it is clear that these words were intended to deal with the type of evidence known as propensity and similar fact evidence at common law.[146] The terms ‘propensity’ and ‘tendency’ evidence refer to ‘[e]vidence of the character, reputation or conduct of a person, or a tendency that a person has or had … to act in a particular way, or to have a particular state of mind’.[147]
The common law term ‘similar fact evidence’ refers to evidence that may, because of the degree of similarity in the evidence of different witnesses, show the identity of the offender or ‘the improbability of coincidence if a number of similar accounts are all true.’[148] The admission of such evidence does not rely on propensity reasoning ‘but on explaining the coincidences between events by [reference to] the perpetrator’s tendency.’[149] Similarly, coincidence evidence is defined under s 98 of the UEA with reference to the similarities in the evidence in question — it refers to two or more events being used to prove the improbability of the events occurring coincidently, having regard to the similarities of the events and/or the circumstances in which they occurred.
In Perry v The Queen,[150] Sutton v The Queen[151] and Harriman v The Queen,[152] the High Court recognised that relevant similar fact and propensity evidence could be admissible but only if its probative value outweighed its prejudicial effect. Similar fact evidence was considered to derive its probative force from
the fact that the evidence reveals ‘striking similarities’, ‘unusual features’, ‘underlying unity’, ‘system’ or ‘pattern’ such that it raises, as a matter of common sense and experience, the objective improbability of some event having occurred other than as alleged by the prosecution.[153]
In Pfennig v The Queen (‘Pfennig’), Mason CJ, Deane and Dawson JJ emphasised that ‘usually the evidence will lack the requisite probative force if the evidence does not possess such characteristics’,[154] although their Honours held that these characteristics were not essential prerequisites to admissibility in every case. In R v Phillips the High Court, while apparently disclaiming a need for striking similarities,[155] nonetheless held that a strong degree of probative force, in the form of sufficient similarities, was needed to admit the evidence of multiple complainants in a joint sexual assault trial.[156]
Both similar fact and propensity evidence can be highly prejudicial to the defendant because of the human tendency to infer guilt from past conduct.[157] In order to prevent the admission of this type of prejudicial evidence, the common law developed the ‘no rational view of the evidence’ test (also known as the ‘Pfennig test’), which ensures this type of evidence will only be admitted if there is no rational view of the evidence that is consistent with the defendant’s innocence.[158] However, the focus of this article is on whether or not there is an empirical basis for the application of the striking similarities test for determining the admissibility of multiple complainants’ evidence in a joint trial involving child sex offences. This has not been previously discussed in the legal literature,[159] and is analysed in light of the findings from the psychological literature, discussed above, and recent moves in Western Australia (‘WA’) and the United Kingdom to abolish the test.
The degree of similarity or unity required to give similar fact evidence sufficient probative force is, essentially, a subjective evaluation. As Lord Mackay observed in Director of Public Prosecutions v P, the issues of relevance, sufficient probative force and whether it is just to admit the evidence are all ‘matters of degree’.[160] This question of degree in a sexual assault trial is made that much easier where the complainants all report very similar offences, a similar modus operandi and similar circumstances in which the offences occurred, because it can then be argued it is improbable that so many complainants would make similar allegations unless they were true.[161] Admissibility problems will arise where the circumstances, the offences and the defendant’s modus operandi appear to differ in key ways.
Where an alleged offender’s sexual behaviour differs in relation to each complainant, lawyers and judges commonly focus on those differences, thus reducing the apparent probative value of the evidence.[162] However, if the striking similarities test is applied without reference to the variety and lack of similarity that is typical of CSOs’ behaviour, it amounts to no more than judicial guesswork which unwittingly benefits defendants whose behaviour does not meet the test and disadvantages victims whose experiences of sexual abuse are too varied.
From time to time, judges have argued that the striking similarities test should be relaxed in cases where identity is not a fact in issue in a sexual assault trial. In Director of Public Prosecutions v P, Lord Mackay observed that
[w]here the identity of the perpetrator is in issue, and evidence of this kind is important in that connection, obviously something in the nature of … a signature or other special feature will be necessary. To transpose this requirement to other situations where the question is whether a crime has been committed, rather than who did commit it, is to impose an unnecessary and improper restriction upon the application of the principle.[163]
In England, Lord Mackay’s observations constituted a ‘reformulation of the “similar fact” evidence test … [which] was subsequently described by Lord Lloyd … in R v H … as having released the law from the bondage of “striking similarity”.’[164]
The only Australian jurisdiction which has followed the United Kingdom’s lead is WA. Along with a package of other reforms, s 31A was inserted into the Evidence Act 1906 (WA) to deal with the problems associated with admitting propensity evidence in a joint trial of sex offences.[165] Under s 31A(2), where the evidence of multiple complainants is considered to have significant probative value, a balancing test weighs its probative value against the degree of risk of an unfair trial — that is, the test asks whether ‘fair-minded people would think that the public interest in adducing all relevant evidence of guilt must have priority over the risk of an unfair trial.’[166]
It has been accepted that s 31A, together with a provision which allows for joint trials to be held even if the evidence of multiple complainants is not cross-admissible,[167] has ‘substantially altered the common law’ in WA.[168] Because the WA legislature articulated a particular threshold test of admissibility — that the evidence must have significant probative value — the common law test of ‘striking similarities’ is no longer a necessary criterion for admissibility.[169]
While the WA Court of Appeal has held that the word ‘significant’ has the same meaning as that given to the word under the UEA,[170] the Court has accepted, in several child sexual assault cases, that propensity evidence about a defendant can have significant probative value without there being a need to show striking similarities or sexual interference by the defendant in a particular way.[171] If the evidence in question reveals an underlying unity, system or pattern, that will be sufficient to establish that the evidence has significant probative value under s 31A(2), ‘irrespective of what physical acts [were] individually involved’[172] and even if the sexual conduct of the defendant was alleged to have occurred ‘in a variety of ways and at various times and places.’[173]
B Striking Similarities under the UEA
Despite the fact that the UEA contains its own formulations under pt 3.6 for the admission of tendency and coincidence evidence,[174] the striking similarities test is still a test of common currency under the UEA in contrast to the situation in WA. Once the relevance test under s 55 is satisfied, tendency or coincidence evidence adduced by the prosecution is subject to a two-stage scheme for its admission under the exclusionary rules contained in pt 3.6:
1 a ‘significant probative value’ test under ss 97(1)(b) and 98(1)(b); and
2 a balancing test under s 101(2) which requires the probative value of the tendency or coincidence evidence to substantially outweigh its prejudicial effect.
The words ‘significant probative value’ create a second test of relevance[175] since ‘[t]he question of probative value is a question of relevance’.[176] The assessment involves asking whether ‘the evidence is capable, to a significant degree, of rationally affecting the assessment … of the probability of the existence of a fact in issue.’[177] This is a subjective judgment or evaluation,[178] as can be seen from the wording in ss 97(1)(b) and 98(1)(b), which refer to whether ‘the court thinks that the evidence will … have significant probative value’ (emphasis added). As the former Chief Justice of NSW observed, the presence of the words ‘the court thinks’ ‘introduce[s] an element of subjectivity’.[179] In fact, a majority of the NSW Court of Criminal Appeal (‘NSWCCA’) has held that the approach under ss 97(1)(b) and 98(1)(b) is involves ‘a degree and value judgment’[180] which will vary from case to case because of the requirement that a trial judge must have regard to any other evidence adduced by the prosecution.[181] For example, where there is no forensic or eyewitness evidence, tendency evidence that shows a defendant has engaged in sexual conduct with another child is likely to be ‘significant’ (ie ‘important’ or ‘of consequence’)[182] in establishing whether or not the accused committed the sexual conduct complained of.[183] But because a judge’s interpretation of the word ‘significant’ will be subjective, the word’s application will depend on a trial judge’s knowledge and assumptions about the behaviour of CSOs.
In relation to the second stage of admissibility, the balancing test under s 101(2) is a less stringent rule of exclusion than the common law Pfennig test.[184] Section 101(2) does not assume that tendency or coincidence evidence will automatically have a prejudicial effect, because the section refers to ‘any prejudicial effect it may have on the defendant’ (emphasis added). The balancing test also involves a degree and value judgment since, as a test directed towards the exclusion of evidence, it is ‘quintessentially a judgment … as to the probative value of the evidence relative to any potential prejudicial effect’[185] — in particular, a value judgment as to whether the jury ‘may make some improper use’ of the evidence.[186] Thus, it is appropriate at this stage for a judge to consider ‘the utility of, and protection afforded by, directions that may be necessary or available as the evidence is given or in summing up’[187] as a way of ameliorating the likely prejudicial effect.[188]
1 Tendency Evidence under the UEA: The Role of the Striking Similarities Test
Section 97 does not contain any references to the need for similarities in particular evidence for that evidence to be admissible to prove a relevant tendency. Nonetheless, the case law shows that the significant probative value test under s 97(1)(b) will be satisfied if there are striking similarities in the evidence of two or more complainants,[189] or if that evidence reveals a similar pattern of behaviour on the part of the accused.[190] This is despite the fact that in R v Ellis, Spigelman CJ observed there were at least eight indications in pt 3.6 of the UEA that Parliament ‘intended to lay down a set of principles to cover the relevant field to the exclusion of the common law principles previously applicable.’[191]
Tendency evidence ‘may be strong or weak, depending upon the nature of the evidence’, but the probative value ‘must not be so weak as to be bereft of “significance”’.[192] While ‘[t]he level of generality of the evidence may affect’ its significance,[193] concepts such as generality and significance inevitably involve subjective judgments. It is not the fault of trial judges that they may not understand the significance of the totality of an alleged serial offender’s behaviour or his different behaviour on different occasions with different victims. Yet engagement with the psychological literature is necessary if a trial judge’s assessment of offender behaviour is to bear any relationship to reality.
The striking similarities test has been applied in several cases under the UEA when assessing whether the evidence of multiple complainants was sufficiently similar to have significant probative value. Appeal cases show that this test is regularly argued in submissions by prosecutors and defence counsel and often applied by trial judges and appeal judges to admit or exclude evidence under ss 97 and 98.[194] In some cases, the Crown has expressly disclaimed reliance on striking similarities, instead basing its argument for admissibility on the fact that the tendency or coincidence evidence establishes a pattern of behaviour or modus operandi on the part of the accused.[195] However, in order to show such a pattern or a modus operandi it is necessary to identify similarities in the accused’s behaviour[196] — meaning that an approach that is in substance the same as striking similarities is being applied.
Because the striking similarities test has been the focus for considering the admissibility of evidence under s 97, it is necessary to analyse the validity of this approach when considering the sexual conduct of an alleged serial CSO. In CGL v Director of Public Prosecutions (Vic) (‘CGL’), the Victorian Court of Appeal (‘VSCA’) remarked that
[w]hen s 97(1) speaks of a tendency ‘to act in a particular way’, we hardly think that Parliament had in mind a tendency which would be expressed as generally as ‘a tendency to act upon sexual attraction to young girls aged between eight and 13 years’.[197]
Therefore, the question is: how general or specific should the sexual conduct of an alleged CSO be before it is accepted that the CSO’s behaviour amounts to ‘acting in a particular way’?
In R v Barton (‘Barton’),[198] the NSWCCA upheld the accused’s appeal against his conviction in a case where the trial judge had found the evidence of seven adolescent inmates of a juvenile detention centre cross-admissible in a joint trial. This was despite the apparent lack of striking similarities in the accused’s sexual behaviour, which ranged from minor sexual behaviour such as voyeurism and sexual touching to masturbation, fellatio and intercourse. Grove J (Dunford J and Kirby J agreeing) criticised the trial judge’s decision because of his failure to discriminate between the less serious and more serious sexual behaviour, since ‘[b]etween the two assemblages of … conduct, there was no discernible pattern or “signature”.’[199] For this reason, the less serious behaviour was not considered probative in relation to the occurrence of the more serious behaviour. In light of the literature on the extent and variety of CSOs’ behaviour, it is clear that Grove J was making a subjective judgment that may have been logical from a legal standpoint focused on similarities but not from a scientific perspective focused on whether an offender is a repeat offender.
In R v Fletcher (‘Fletcher’),[200] a case concerning a priest who had been charged with nine counts of child sexual assault, the NSWCCA condoned the trial judge’s decision to exclude the evidence of all but one of eight witnesses because of lack of similarities with the complainant’s evidence. Where there are no striking similarities, Simpson J cautioned against concluding that ‘evidence of a tendency to sexual misconduct with adolescent boys could rationally affect the assessment of the probability that the [the defendant] sexually misconducted himself with the complainant as an adolescent.’[201] Yet in studies going back at least 15 years, the psychological literature has found that evidence of prior sexual behaviour with a child is the best predictor as to whether a person will commit another child sex offence.
Simpson J also appeared to have had little understanding of the purpose of an offender’s grooming behaviour in testing a child’s reaction to non-sexual touching, and the way this behaviour can lead to more serious sexual behaviour involving genital contact.[202] There was also a failure to acknowledge the evidentiary significance of the defendant’s sexual games and interactions with male adolescents, as indicated by Simpson J’s warning:
While it may be tempting to think … that evidence of a sexual attraction to male adolescents has probative value … an examination must be made of the nature of the sexual misconduct alleged and the degree to which it has similarities with the tendency evidence proffered. … [T]here will be cases where the similarities are of so little moment as to render the evidence probative of nothing.[203]
In both Barton and Fletcher, the NSWCCA adopted the common law approach[204] by focusing on whether or not the sexual practices of the defendant with the complainant were sufficiently similar to his sexual practices with another complainant/witness to allow the evidence of those witnesses to be considered admissible.
More recently, however, there has been a discernible shift in the NSWCCA’s approach to tendency evidence in child sexual assault trials. In Dao,[205] the defendant, a priest, had been accused of sexually abusing six adolescent boys. The trial judge ordered that the counts in respect of three complainants be separated, but permitted a joint trial for the counts in respect of the remaining three complainants (SM, JC and MB). The trial judge held that the evidence of the former three complainants was not cross-admissible in the joint trial, on the grounds that the defendant’s ‘opportunistic’[206] sexual conduct in relation to the first three boys lacked the planning, grooming and special attention that characterised the alleged sexual conduct involving SM, JC and MB.
On appeal, Simpson J recognised that ‘[e]vidence of a tendency may cast light on the conduct or state of mind of a person without being evidence of conduct of the same kind’,[207] although her Honour still held that similarity in conduct was relevant to the assessment of significant probative value under s 97(1)(b).[208] Contrary to her views in Fletcher, however, Simpson J observed that ‘[i]t could scarcely be contested that evidence that showed that the applicant had a sexual interest in, and attraction to, adolescent boys had probative value in respect of an allegation that he had sexually abused another adolescent boy.’[209]
This more realistic approach to the admissibility of tendency evidence was also evident in the NSWCCA’s decision in R v PWD (‘PWD’).[210] In that case, four male students at a private school had accused the principal of various forms of sexual misconduct. This behaviour did not exhibit the high degree of similarity that the NSWCCA said was required in Barton and Fletcher since it ranged from minor sexual touching and fondling to masturbation and oral sex.
The trial judge refused to admit the evidence of the four complainants and two other witnesses as tendency evidence because of the lack of similarities in the alleged sexual conduct. With apparently little knowledge about the grooming and sexual behaviours of CSOs, it was easy for the trial judge to argue that tendency evidence given by three complainants ‘did not make it “more likely to a significant extent” that the charged acts occurred’.[211] In an appeal by the Crown, the NSWCCA held that the trial judge had erred in making the lack of similarity between the accused’s sexual acts with different boys the determining factor in assessing the probative value of the evidence.[212]
Simpson J observed in Dao that ‘Barton and PWD are illustrative of the divergence of facts and circumstances that will inevitably result in different outcomes without any misapplication of principle.’[213] This analysis is not terribly satisfactory where the principle is based on subjective assessments about the similarities or non-similarities in the conduct of alleged CSOs. Both Barton and PWD involved a range of sexual misconduct, as well as differences in the gravity of the offences alleged; yet in Barton the complainants’ evidence was not cross-admissible. While the sexual misconduct in PWD was very similar to that in Barton in terms of its variety, the evidence of the three complainants in PWD was held to be cross-admissible on appeal.
More recently, in Dao’s case, Simpson J (Spigelman CJ, Allsop P, Kirby and Schmidt JJ agreeing) concluded that ‘[e]vidence of more serious [sexual] conduct may support allegations of less serious [sexual] conduct just as evidence of less serious [sexual] conduct may support allegations of more serious [sexual] conduct’, despite the wide variety in Dao’s sexual behaviour.[214] These views from a bench of five of the NSWCCA is a welcome development, although the recognition that each case will depend on its own facts[215] means that subjectivity rather than informed knowledge of CSO behaviour is likely to prevail at both trial and appeal levels.
In fact, no consistency of approach can be identified in the many appeal cases heard in NSW; nor is there any guarantee that in future cases the striking similarities test will not be applied, given the frequency with which counsel rely on it and the frequency with which trial judges still apply it.
In Victoria, a similar lack of consistency is evident in trial and appeal judges’ approaches to the need, or otherwise, for striking similarities in the evidence adduced in a joint trial of child sex offences. What is evident in the appeals being heard under the Evidence Act 2008 (Vic) is frequent reliance on the common law striking similarities test.
In MR v The Queen, the trial judge held that any requirement of similarity in the sexual activity of the accused was met by the fact that each count against the accused ‘involved sexual activity involving his young daughter.’[216] Thus, the trial judge focused on whether or not there was evidence of sexual conduct generally speaking, rather than evidence of specific types of sexual activity by the accused. While the trial judge’s approach was upheld on appeal, the VSCA was careful to note that a degree of similarity may still be required in order to identify the ‘relevant tendency’ of an accused and ‘to sustain admissibility under s 97’ where there are several counts involving multiple complainants and where cross-admissibility is at issue.[217]
In Director of Public Prosecutions (Vic) v BCR (‘BCR’), the trial judge had distinguished between three types of the accused’s alleged conduct with different complainants: ‘indecent touching while comforting students who had been injured playing sport, … alleged touching after punishment of students and touching after comforting students who had not been injured’.[218] Because not all the evidence contained ‘the distinctive feature of “comforting after injury”’,[219] the trial judge concluded that there was ‘an insufficient identifiable pattern of similarities in the circumstances of the offending “to make out a tendency[”]’.[220]
On appeal, the Crown argued that the trial judge had incorrectly applied the principles from three previous cases — NAM v The Queen (‘NAM’),[221] PNJ v Director of Public Prosecutions (Vic) (‘PNJ’)[222] and CGL[223] — in deciding that evidence from two or more complainants had to have distinctive or unusual features to be cross-admissible, and that these Victorian authorities were inconsistent with recent NSW authority.[224] The Crown argued that nothing in s 97 of the UEA ‘warrants the importation of a requirement that there be something “distinctive” about the evidence’ before it can have significant probative value.[225] While the Crown accepted there had to be some degree of similarity in the accused’s alleged conduct to enable a tendency to be identified, that need not be a ‘striking’, ‘distinctive’, ‘remarkable’ or ‘unusual’ similarity.[226]
The VSCA refused to be drawn on the Crown’s submission that the decisions in NAM, PNJ and CGL were ‘plainly wrong’, because an interlocutory appeal was not ‘an appropriate vehicle for challenging an existing line of authority [except] in exceptional circumstances.’[227] This means that on current authority in Victoria, if the evidence of multiple complainants is to be cross-admissible in a child sexual assault trial, it must exhibit some distinctive behaviour on the part of the accused in order to link him with each complainant. Yet Smallbone and Wortley’s study showed that extra-familial offenders, engage in a variety of strategies to gain access to their victims,[228] so that differentiating between sexual abuse while comforting and not comforting students — as the trial judge did in BCR — is largely irrelevant to whether or not a particular offender has a tendency to engage in the sexual abuse of teenage boys.
Indeed, NAM, PNJ and CGL are inconsistent with the recent progressive approach taken by the NSWCCA in relation to the cross-admissibility of the evidence of two or more complainants. But does it matter that the Victorian line of authority represented by NAM, PNJ and CGL concerned s 98, rather than s 97, of the UEA?
2 Coincidence Evidence under the UEA: The Role of the Striking Similarities Test
While the NSWCCA recently decided that ‘for evidence to be admissible under s 97 there does not have to be striking similarities, or even closely similar behaviour’,[229] the admissibility of coincidence evidence under s 98 is based on finding similarities in two or more events, in the circumstances in which they occurred, or in both the events and the circumstances. If there are such similarities, the next question is whether they are ‘such that it is improbable that the events occurred coincidentally’.[230] If the answer is yes, only then can the events be used to prove that the accused did a certain act (such as sexual abuse) or had a certain state of mind (such as sexual attraction). Like the approach at common law, coincidence or improbability reasoning is used where there are features of the accused’s sexual behaviour with multiple complainants that are strikingly similar or where the circumstances in which the behaviour occurred are such that coincidence can be eliminated as a reasonable explanation. But what degree of similarity in the sexual conduct of the accused (the events) or in the circumstances in which the alleged sexual abuse occurred is required?
As a result of the inquiry into the operation of the UEA in 2005,[231] s 98 was amended to eliminate a drafting anomaly which had meant that dissimilar evidence was admissible to prove a lack of coincidence.[232] What degree of similarity will be required under the new s 98, in the events or the circumstances or both, will differ from case to case, although case law on the former s 98 stated that, for the events to be related, they had to be ‘substantially and relevantly similar, and that the circumstances in which they occurred were [required to be] substantially similar’.[233] The requirement that the events be ‘related’ reflected the requirement of ‘striking similarities’ for the admission of similar fact evidence at common law. The definition of ‘related events’ in the former s 98
ha[d] the effect that the intended controls on admissibility [would] only apply if the events and circumstances in which the events occurred [were] substantially similar. Paradoxically, therefore, there [was] a test of admissibility for ‘related events’ … but not a test for unrelated events. As a result, [the former] s 98 [did] not apply to exclude evidence where the events [were] not substantially and relevantly similar … [and] the other intended control, s 101, [had] no application.[234]
Presumably, therefore, a wider interpretation of the amended s 98 will be required. This will ensure that the provision applies to the type of evidence that the former s 98 did not by capturing evidence that does not have a high degree of similarity as required by the common law ‘striking similarities’ test. According to the Australian Law Reform Commission (‘ALRC’), the NSW Law Reform Commission (‘NSWLRC’) and the Victorian Law Reform Commission (‘VLRC’), although the amended s 98 requires
consideration of similarities in events and circumstances, it does not require … that there be similarities in both the events and circumstances and that the similarities be striking. … The Commissions’ view is that to require both a striking similarity of events and a striking similarity of circumstances would be to raise the threshold too high and would be likely to exclude highly probative evidence. For example, highly probative evidence of unusual, similar acts occurring in different circumstances would be excluded.[235]
In recent VSCA cases which have applied s 98, however, a high degree of similarity has been required, with no reference being made to the reasons why s 98 was amended or the fact that the greater the degree of similarity required in the events and circumstances, the less likely it will be that s 98 will capture dissimilar evidence.[236] According to the VSCA’s approach, the more there are apparent dissimilarities in the evidence of multiple complainants, the less likely their evidence, as a whole, will be considered to have significant probative value on the basis that it disproves coincidence.
The interpretation of s 98 by the VSCA has been based on the approach taken under the former s 398A of the Crimes Act 1958 (Vic)[237] and in common law cases.[238] While the VSCA has said it is unnecessary to demonstrate striking similarities, since what is required is an underlying unity, a common modus operandi or a pattern of conduct,[239] the Court has also expressed the contradictory view that ‘it is the distinctiveness of the act which provides the requisite degree of similarity.’[240] The VSCA has also noted that ‘factors which [are] to be considered in determining whether … tendency evidence [has] significant probative value [are] also relevant in determining whether the evidence [has] significant probative value’ as coincidence evidence.[241]
This requirement of distinctiveness meant that in CGL the alleged conduct of the accused in rubbing the genitals of two young girls was ‘so non-specific … as to reveal nothing distinctive about any particular alleged act’[242] and occurred in ‘unremarkable circumstances that are common to sexual offences against children’.[243] The apparent lack of relevant similarity which rendered the complainants’ evidence inadmissible as coincidence evidence also meant their evidence was inadmissible as tendency evidence.[244] In fact, the VSCA proposed the general proposition that
the greater the degree of specificity with which the similarities can be identified, the more likely it is that the evidence will be probative of a tendency to act in a distinctive way or to do acts of a distinctive kind. Conversely, the greater the degree of generality, the more difficult it will be to demonstrate that the evidence in question has ‘significant’ probative value and … to demonstrate that its probative value ‘substantially outweighs’ the very real prejudicial effect of evidence of this kind.[245]
A similar approach was taken by the VSCA in PNJ. This case involved a youth officer in a youth training centre who had been charged with 14 counts of sexual assault, ranging from indecent assault to penetration, against resident teenage boys. The accused had argued ‘that the events relied on by the prosecution lacked “any meaningful similarity”’.[246] The VSCA in PNJ went one step further than the Court in CGL by stating it was ‘a mistake to treat as relevant similarities … features of the alleged offending which reflect circumstances outside the accused’s control’ since these reflected the setting in which the offending occurred.[247] The Court held that ‘[t]o qualify as a relevant similarity in [these] circumstances … there must be something distinctive about the way in which the accused allegedly took advantage of the setting’.[248] Thus, no similarities could be drawn from the complainants’ ages since that was the age range for detention at the youth centre, nor could similarities be drawn from the location of the alleged offending (the complainants’ or the defendant’s bedrooms) since that merely reflected the custodial setting;[249] the defendant had no choice in the setting or locations that were common to each complainant. Although each complainant complained of similar sexual acts (masturbation and oral sex), the VSCA considered these to be ‘unremarkable’ and nothing more than common types of sexual behaviour which could not ‘distinguish the applicant’s offending from that of any other such offender.’[250]
However, the purpose of s 98 is not to distinguish one sex offender from another. Where a joint trial is being sought by the prosecution, the role of s 98 is to decide on the cross-admissibility of the complainants’ evidence. To take an approach which requires distinctive patterns of sexual behaviour on the part of the accused narrows the scope of the amendments to s 98, contrary to the intentions of the ALRC, NSWLRC and VLRC when they recommended those amendments. The narrow approach of the VSCA also constitutes a major impediment to the successful prosecution of ‘unremarkable’ sexual abuse and for victims who through no fault of their own happen to have been sexually abused in ‘commonplace’ or ‘unremarkable’ ways. The approach also fails to recognise the type of offender who seeks victims through his paid or unpaid work[251] or who uses his employment to gain access to children. In their self-report study, Smallbone and Wortley reported that while only 1.2 per cent of offenders joined an organisation ‘for the purpose of locating victims’, 19.7 per cent of offenders took advantage of the employment situation they found themselves in.[252] For these types of offenders, control over the setting of the alleged abuse is irrelevant to whether or not sexual abuse has occurred since the pattern of behaviour is the selection of children in an employment situation to whom the offender is sexually attracted.
The VSCA’s approach in CGL and PNJ also contradicts the view of a majority of the NSWCCA that in some cases
the probative value of the coincidence evidence arises not merely from the … events, but also, and especially, from the circumstance that two or more persons independently give evidence of the … events, where it is improbable that they would have given accounts with such similarity unless both or all accounts had foundation in fact. In such cases, … the … events need not necessarily have such striking similarity as is required where it is only the events themselves that are said to have probative value.[253]
This suggests that in a situation where the setting is outside the control of the victim and the apparent control of the offender, the fact that several children give accounts of sexual abuse by the same offender, whether the acts of sexual abuse are commonplace or distinctive, means that their evidence can be used to prove the improbability of the abuse occurring coincidentally.
Nonetheless, the narrow approach in PNJ has been adopted in subsequent VSCA cases,[254] including NAM[255] where most of the sexual acts of the accused were described as ‘commonplace’ except those where two complainants gave almost identical accounts about the accused’s use of a banana. What these cases show is that many acts of child sexual abuse are now regarded as ‘commonplace’ and ‘unremarkable’ by the VSCA, illustrating the ‘straightjacket’ represented by the striking similarities test which has led the VSCA on a search for distinctive sexual acts and distinctive circumstances. Worryingly, it seems that strange and unusual sexual acts will be required for multiple complainants’ evidence to be cross-admissible as coincidence evidence. This demonstrates a lack of understanding of the phenomenon of child sexual abuse and the huge variety of sexual and grooming behaviours committed by CSOs. With empirical evidence showing that recidivism is higher for offenders who commit crossover behaviours, most serial CSOs will not demonstrate the type of distinctive behaviour now required by the VSCA.
V UEA REFORM OPTIONS AND RECOMMENDATIONS
In light of the above case law analysis, this Part considers the reform options in relation to the cross-admissibility of multiple complainants’ evidence in child sexual assault trials. By adopting the view of Spigelman CJ in R v Ellis that the UEA was intended to operate on its own terms to the exclusion of common law principles,[256] and recognising that the significant probative value test is a statutory formulation not found at common law, it is evident that tendency evidence about an accused does not need to demonstrate striking or indeed any similarities before it can have significant probative value.[257] In fact, striking similarities is the standard required when identity is in issue, which is rarely the case in sexual assault trials.[258]
The admissibility of tendency evidence under the UEA should be guided by the actual wording of s 97. For example, it has been observed that when deciding whether evidence is, in fact, tendency evidence, the prosecution ‘need not demonstrate a tendency to commit a particular crime’, but merely that the evidence shows a tendency to act in a particular way.[259] This wording in s 97 suggests the provision was not drafted to focus on the striking similarities in, or the distinctiveness of, a defendant’s conduct. In fact, tendency evidence does not need to show a tendency ‘to commit acts that are closely similar to those that constitute the crime with which [the] accused is charged.’[260] Contrary to the defendant’s arguments in R v Ford, Campbell JA held there was no need for the evidence from three complainants about the accused’s sexual conduct to show a ‘striking pattern of similarity’.[261] Nor did the tendency behaviour have to be ‘“compellingly rare or exceptional” before it [could] have significant probative value’.[262]
To engage in sexual misconduct with a child is to act in a particular way, sexually speaking, although it may involve a wide range of behaviours from sexual touching to intercourse. It is also to act in an unusual way since only a minority of men will engage in child sexual abuse. This means that any previous sexual conduct with a child shows that a defendant has a tendency to act in this unusual sexual way. If there is evidence that the accused has sexually abused one child, that evidence would have probative value in relation to allegations by other children of sexual misconduct towards them.
Such an approach, which was endorsed by the NSWCCA in Dao,[263] would lower the threshold test for the admissibility of tendency evidence in a child sexual assault trial. It is also an approach consistent with the literature on CSOs since there is considerable evidence to show that any sexual misconduct with a child is the best predictor of future and past behaviour with children.[264]
The striking similarities formulation implies that the evidence needs to show a tendency to commit a particular act or a particular crime, rather than ‘to act in a particular way’, which, as argued above, connotes a broader range of behaviours. Smallbone and Wortley’s study showed that 169 CSOs employed 28 different strategies for gaining access to children,[265] 38 strategies for developing a child’s trust[266] and 67 strategies for getting a child to engage in sexual activity.[267] Some of these strategies were evident in Fletcher,[268] Barton,[269] PWD[270] and Dao,[271] where the defendants had been accused of engaging in a variety of sexual and grooming behaviours with different adolescent boys. All of that behaviour amounted to acting in an unusual sexual way, that is, sexual misconduct with children.
Section 98 should also be interpreted in light of the CSO literature, as well as the rationale for its amendment. To ensure that the amended s 98 captures dissimilar evidence about a CSO’s sexual behaviour, the striking similarities or distinctiveness approach must be abandoned as a criterion of admissibility.
As discussed previously, the use of striking similarities or distinctiveness under the UEA for the admission of tendency and coincidence evidence in child sexual assault trials demonstrates a lack of understanding of the way CSOs behave, in particular that core group of high-risk offenders who will abuse more than one child. The literature shows that this group commits a wide variety of sexual and grooming behaviours. Because of this, the serial offender is unlikely to demonstrate the distinctive behaviour required by some judges.
It is a difficult task to formulate ‘appropriate rules to deal with probative but prejudicial evidence … [because] [i]n this area there is a stark conflict between the policy objectives of receiving all probative evidence and minimising the risk of wrongful conviction.’[272] In many ways the public policy conflict is even starker since what is also at stake is the conflict between protecting children from serial CSOs and the public policy of protecting defendants from prejudicial evidence. Some will argue that it is not the role of the criminal justice system to consider the welfare of children but if prosecuting CSOs is, at present, the only means by which offenders can be deterred or rehabilitated then child protection becomes, even in a de facto sense, an essential public policy objective of child sexual assault trials.
Under current approaches to tendency and coincidence evidence, where there is a lack of striking similarities or distinctiveness between the evidence of two or more complainants, a joint trial is less likely to be held. This, in turn, will result in more separate trials, and thus a decreased likelihood of serial CSOs being convicted, since corroborative evidence of child sexual abuse is uncommon[273] and lack of corroborative evidence is likely to be one of the reasons why low conviction rates are found in child sexual assault trials.[274]
There have been a number of joint child and adult sexual assault trials in which juries have returned a mixture of guilty and not guilty verdicts,[275] which suggests that properly instructed juries are capable of looking beyond the prejudicial effect of tendency and coincidence evidence and carefully considering the evidence pertaining to each count separately. This seems contrary to the view that ‘the dangers inherent in propensity evidence generally are greatly enhanced in sexual offence cases’.[276] Indeed, there is considerable empirical evidence that the particular prejudices of jurors are in fact associated with misconceptions about the effects of child sexual abuse and the reliability of children’s evidence, along with beliefs in rape myths and stereotyped views of women and children who complain of sexual assault.[277]
Although restricting the admissibility of tendency and coincidence evidence is based on preventing the desirable goal of impermissible reasoning by juries, arguments that justify excluding these types of evidence do not withstand scrutiny. For example, it is assumed that juries will not engage in impermissible propensity reasoning[278] when the evidence has sufficient similarity to justify its admission, but will do so where the dissimilarities are more marked. Logically, juries would be more likely to engage in impermissible propensity reasoning where there are greater similarities in the evidence of multiple complainants about a defendant’s sexual conduct.
Perhaps the real difficulty with ss 97 and 98 is that, without clear legislative guidance, they are unlikely to be correctly applied by judges versed in the common law. In contrast, in WA, where a significant probative value test was enacted to deal with the admissibility of propensity evidence, the then WA Attorney-General stated that this and other reforms were designed to address the limitations of the criminal justice system when dealing with sexual assault.[279] It was envisaged the reforms would change the way that sexual assault trials were conducted in WA by increasing the number of joint trials where there was an alleged serial offender in order to ensure that the jury does not operate in a ‘vacuum’[280] by being unaware that the accused has been charged with multiple offences involving more than one complainant. This legislative intention has been subsequently adopted by the WA Court of Appeal in its interpretation of s 31A (as discussed above in Part IV(A)).[281]
In UEA jurisdictions, similar legislative guidance will be necessary given the fact that a large number of appeals in relation to ss 97 and 98 arise from child sexual assault trials. Yet these provisions represent a significant hurdle for the successful prosecution of serial CSOs, since they provide no guidance to judges about what previous sexual misconduct on the part of a defendant amounts to evidence of significant probative value, with the result that some judges continue to import the common law striking similarities approach into the statutory scheme. It is recommended that there is a need for greater uniformity in the UEA jurisdictions at least in relation to the admissibility of relevant tendency and coincidence evidence in joint child sexual assault trials. Because of the arbitrary nature of the striking similarities approach, it is recommended that ss 97 and 98 be amended to reflect (i) the literature on CSO behaviour reviewed in this article; (ii) the interpretation given to s 97 by Campbell JA in R v Ford;[282] and (iii) the rationale underlying the amendments made to s 98, as expressed by the ALRC, NSWLRC and VLRC.
Thus, the following provisions are recommended as amendments to ss 97, 98 and 101:
Insert the following sub-sections into s 97 of the UEA:
(3) Sub-sections (4) and (5) apply to proceedings in respect of a prescribed sexual offence and despite any other rule of law to the contrary if 2 or more counts charging sexual offences involving different complainants are joined in the same indictment [information, presentment].
(4) In considering whether evidence about the sexual conduct of the defendant has significant probative value for the purposes of sub-section (1), the court must not have regard to whether that evidence has striking similarities or distinctive features with other evidence about the sexual conduct of the defendant.
(5) It will be sufficient for evidence about the sexual conduct of the defendant to have significant probative value if that evidence shows the defendant has committed a charged or uncharged act of a sexual nature of any kind against a child.
Insert the following sub-sections into s 98 of the UEA:
(3) [As above].
(4) In considering whether evidence of two or more events which describe the sexual conduct of the defendant have significant probative value for the purposes of sub-section (1), the court must not exclude the evidence solely on the basis that the evidence does not reveal striking similarities, distinctiveness, unusual features, an underlying unity, system, pattern or common modus operandi in the defendant’s alleged sexual misconduct or in the circumstances in which the alleged sexual misconduct occurred.
(5) It will be sufficient for the evidence of two or more events to have significant probative value if the evidence shows that the defendant has committed a charged or uncharged act of a sexual nature against a child, irrespective of whether the act(s) involve the same or different sexual conduct on the part of the defendant.
Note: For example, if event 1 describes sexual touching by the defendant of a witness and event 2 describes sexual penetration by the defendant of another witness, these events will have the required degree of similarity to satisfy the significant probative value test in sub-section (1).
Insert the following sub-sections into s 101 of the UEA:
(5) Sub-section (6) applies to proceedings in respect of a prescribed sexual offence and despite any other rule of law to the contrary if 2 or more counts charging sexual offences involving different complainants are joined in the same indictment [information, presentment].
(6) In considering, for the purposes of sub-section (2), whether the probative value of evidence about the sexual conduct of the defendant substantially outweighs its prejudicial effect, the court must not have regard to whether the evidence has striking similarities or distinctive features with other evidence about the sexual conduct of the defendant.
It is also recommended that other Australian jurisdictions should follow the example of WA and Victoria[283] in enacting a provision that would allow joint trials to be held even in situations where the evidence of multiple complainants is not cross-admissible. However, the following recommendation contains an extra caveat that there is a presumption in favour of a joint trial where the evidence of multiple complainants reveals any type of sexual misconduct against a child on the part of the defendant. It is therefore recommended that the following provision, based on s 194 of the Criminal Procedure Act 2009 (Vic), be enacted in all UEA jurisdictions:
(1) If 2 or more counts for child sexual offences are joined in the same indictment there is a presumption in favour of those counts being tried together.
(2) The presumption created by sub-section (1) arises where the evidence of two or more complainants reveals any type of sexual misconduct against a child on the part of the defendant.
(3) The presumption created by sub-section (1) is not rebutted merely because the evidence on one count has been held to be inadmissible in relation to another count on the grounds that the evidence does not reveal similarities, striking similarities, distinctiveness, unusual features, an underlying unity, system, pattern or common modus operandi in the defendant’s alleged sexual misconduct or in the circumstances in which the alleged sexual misconduct occurred.
VI CONCLUSION
For the first time in the legal literature, this article has presented a review of the studies on the rates of recidivism of CSOs, the extent of their crossover behaviours, and the range and variety of their behaviour with children. While recidivism rates vary from study to study depending on methodological issues such as the follow-up period, the measure of recidivism and the offender type, studies which employed lengthy follow-up periods along with measures of recidivism other than re-conviction suggest that 5 to 30 per cent of low-risk offenders and approximately 90 per cent of high-risk offenders will reoffend.
Offenders who engaged in crossover behaviours, in terms of age, sex and relationship with a child, were more likely to have more victims and to be at greater risk of recidivism. Crossover studies indicated that approximately one fifth to two thirds of CSOs will have multiple victims and will engage in a variety of sexual behaviours. Self-report studies supported this data, showing that just over 50 per cent of a mixed group of offenders admitted to having more than one victim, while a study of intra-familial offenders showed they had extensive offending histories with fewer numbers of victims. Importantly, these studies demonstrated that CSOs’ sexual behaviour is not highly specialised, similar or distinctive and that they engage in many different types of sexual behaviours.
In fact, this literature review indicates that the criminal justice system will be regularly confronted with serial CSO cases involving a wide variety of sexual behaviours over relatively long periods of time. Unfortunately, case law under the UEA shows that an approach based on ‘striking similarities’ or ‘distinctiveness’ in sexual behaviour has been regularly used to create artificial distinctions in relation to CSO behaviour which are out of step with the psychological literature. While recent case law from the NSWCCA reveals a much more realistic approach to the cross-admissibility of the evidence of multiple complainants, no consistency of approach can be identified in the many appeal cases heard in NSW, with virtually the same range of sexual behaviours by a defendant resulting in separate trials in some cases and a joint trial in others.
In Victoria, a line of authority has recently emerged that has decided that unless there is something distinctive about the defendant’s sexual behaviour or the circumstances in which the allegations occurred, the evidence of multiple complainants will not be cross-admissible. This approach, which has no empirical basis to support it, appears to reflect a subjective belief that CSOs are highly specialised in the type of sexual behaviour they engage in.
Irrespective of whether judges are considering cross-admissibility under s 97 or s 98 of the UEA, the fact is that when dealing with sex offences against children, a distinction cannot be made between offences that are ‘unremarkable’ and offences that are unusual, distinctive or bear striking similarities. The sexual behaviour of CSOs encompasses a broad continuum of sexual and grooming behaviours, including a range of crossover behaviours, and will vary with the age and sex of the child, opportunity, access, and the degree of emotional entanglement with, and the passivity or resistance of, the child. Not only is any previous sexual misconduct with a child predictive of sexual recidivism but there is evidence that serial CSOs are more likely to engage in a range of sexual behaviours — exactly what many cases in this article and the self-report studies have revealed.
The application of the striking similarities or distinctiveness approach under the UEA amounts to a false measure for assessing the probative value of evidence about a defendant’s sexual behaviour. The approach will result in more separate trials with a decreased likelihood of serial CSOs being convicted, thus placing more children at risk of sexual abuse. Arguably, a lower threshold for determining the probative value of the evidence of multiple complainants is appropriate in child sexual assault cases where the identity of the offender is not in issue, in order to capture the extensive range of sexual and grooming behaviours of serial CSOs reported in the literature.
This article also discussed the rationale behind the reforms to s 98 of the UEA. That rationale shows that the provision needs to be interpreted widely in order to give effect to the stated objectives of the amendments to s 98: this would mean abandonment of the striking similarities test and recognition of the fact that any two incidents of sexual behaviour with children will be sufficiently similar, given that sexual behaviour with children is an unusual practice.
The importation of the striking similarities and distinctiveness formulations has occurred despite the fact that Parliament intended pt 3.6 of the UEA to cover the field to the exclusion of the common law.[284] Indeed, when faced with exactly the same significant probative value test under s 31A of the Evidence Act 1906 (WA), along with a clearly stated legislative intention, the WA Court of Appeal accepted that propensity evidence need not show ‘striking similarities’ or sexual interference by the defendant in a particular way in sexual assault trials. Clearly, greater legislative guidance is needed for judges in UEA jurisdictions.
In order to provide that guidance, this article made recommendations to amend ss 97, 98 and 101 of the UEA. It is acknowledged that some people will argue these reforms constitute too radical a change to an area of the law that has long sought to protect defendants in criminal trials from the effects of prejudicial evidence. Yet there is sufficient evidence to show that properly instructed juries are capable of returning verdicts that confound the belief that jurors routinely engage in impermissible propensity reasoning when faced with evidence of an accused’s past criminal conduct. Indeed, empirical evidence shows that jurors and jury-eligible citizens hold prejudicial views that favour the defence case, rather than the other way around. While propensity evidence may create the risk of an unfair trial, that risk needs to be weighed against competing public interests: those of the community, the victim and child safety. The reforms suggested in this article will allow these often unheard public interests to be considered in child sexual assault trials.
[*] BSc (Hons), LLB, PhD (UNSW); Associate Professor, Faculty of Law, The University of New South Wales.
[1] Unless specified otherwise, all references to offences committed by child sex offenders (‘CSOs’) refer to sex offending and sex offences. A CSO is a person who commits a sex offence against a child under the age of 18 years. Serial offending is defined as a CSO having two or more victims while repeat offending refers to more than one offence with the same victim.
[2] See Terance D Miethe, Jodi Olson and Ojmarrh Mitchell, ‘Specialization and Persistence in the Arrest Histories of Sex Offenders’ (2006) 43 Journal of Research in Crime and Delinquency 204, 207; R Karl Hanson and Kelly E Morton-Bourgon, ‘The Characteristics of Persistent Sexual Offenders: A Meta-Analysis of Recidivism Studies’ (2005) 73 Journal of Consulting and Clinical Psychology 1154, 1154; David Greenberg et al, ‘Recidivism of Child Molesters: A Study of Victim Relationship with the Perpetrator’ (2000) 24 Child Abuse & Neglect 1485, 1486.
[3] Kamala London et al, ‘Disclosure of Child Sexual Abuse: What Does the Research Tell Us about the Ways That Children Tell?’ (2005) 11 Psychology, Public Policy, and Law 194, 204. See the review of the literature in Annie Cossins, ‘Time Out for Longman: Myths, Science and the Common Law’ [2010] MelbULawRw 3; (2010) 34 Melbourne University Law Review 69.
[4] Karen Gelb, ‘Recidivism of Sex Offenders’ (Research Paper, Sentencing Advisory Council, January 2007) 21, citing Nancy Loucks, ‘Recidivism amongst Serious Violent and Sex Offenders’ (Report, Scottish Executive Social Research, 2002) 2.
[5] See Cossins, ‘Time Out for Longman’, above n 3, 79–82.
[6] Ibid 80.
[7] Ibid 81.
[8] See Grant T Harris et al, ‘A Multisite Comparison of Actuarial Risk Instruments for Sex Offenders’ (2003) 15 Psychological Assessment 413, 416.
[9] For example, the following articles on propensity evidence do not discuss recidivism rates: Andrew Palmer, ‘Propensity, Coincidence and Context: The Use and Admissibility of Extraneous Misconduct Evidence in Child Sexual Abuse Cases’ [1999] NewcLawRw 4; (1999) 4 Newcastle Law Review 46; Mike Redmayne, ‘The Relevance of Bad Character’ (2002) 61 Cambridge Law Journal 684; Jeremy Gans, ‘Similar Facts after Phillips’ (2006) 30 Criminal Law Journal 224; David Hamer, ‘Similar Fact Reasoning in Phillips: Artificial, Disjointed and Pernicious’ [2007] UNSWLawJl 36; (2007) 30 University of New South Wales Law Journal 609; C R Williams and Sandra Draganich, ‘Admissibility of Propensity Evidence in Paedophilia Cases’ [2006] DeakinLawRw 9; (2006) 11 Deakin Law Review 1.
[10] See Evidence Act 1995 (Cth); Evidence Act 2011 (ACT); Evidence Act 1995 (NSW); Evidence Act 2001 (Tas); Evidence Act 2008 (Vic).
[11] Pfennig v The Queen [1995] HCA 7; (1995) 182 CLR 461 (‘Pfennig’). See text accompanying below n 158 for an explanation of the Pfennig test.
[12] See below Part IV(B).
[13] R v KP; Ex parte A-G (Qld) [2006] QCA 301 (22 August 2006).
[14] Ibid [60].
[15] Ibid [61].
[16] Harris et al, above n 8, 416–17; Naomi J Freeman, ‘Predictors of Rearrest for Rapists and Child Molesters on Probation’ (2007) 34 Criminal Justice and Behavior 752.
[17] These categories include intra-familial offenders (those who have sexually abused related children), extra-familial offenders (those who have sexually abused unrelated children) and mixed-type offenders (those who have sexually abused related and unrelated children). Although there are good reasons for arguing that this categorisation is artificial because of the degree of overlap between intra-familial and extra-familial offenders, the distinction remains: Anne Cossins, ‘A Reply to the NSW Royal Commission Inquiry into Paedophilia: Victim Report Studies and Child Sex Offender Profiles — A Bad Match?’ (1999) 32 Australian and New Zealand Journal of Criminology 42.
[18] Leam A Craig et al, ‘Sexual Reconviction Rates in the United Kingdom and Actuarial Risk Estimates’ (2008) 32 Child Abuse & Neglect 121, 130.
[19] This 16 year period was chosen because 1995 represents the publication of the first landmark recidivism study (R Karl Hanson, Heather Scott and Richard A Steffy, ‘A Comparison of Child Molesters and Nonsexual Criminals: Risk Predictors and Long-Term Recidivism’ (1995) 32 Journal of Research in Crime and Delinquency 325), a study that, at the time, used the longest follow-up period (15 to 30 years) of any study conducted and followed both intra- and extra-familial CSOs.