We are testing a new beta website for gov.scot go to new site

Recidivism amongst Serious Violent and Sexual Offenders


Recidivism amongst Serious Violent and Sexual Offenders


2.1 In Scotland, the MacLean Committee's review of serious violent and sexual offenders noted that no single typology of a 'sex offender' exists. Rather, sex offenders come from all parts of society and do not fit within a single description or schema (Scottish Executive 2001c). In England, West (2000b) notes that sexual offences vary widely in their characteristics and seriousness. More serious sexual offences range from rape and sexual assault to child molestation and incest. Importantly, recent research in England (Hood et al. 2002) found that dividing sexual offenders into subgroups had important implications for assessing the likelihood of reconviction.

2.2 Research elsewhere suggests that offences vary even within offence types. This is evident for example in Knight and Prentky's (1991) classifications of child molesters depending on the degree of fixation with children, level of contact with children, and degree of injury, and of rapists into four different groups (opportunistic, pervasively angry, vindictive, and sexual, with sadistic rape as a subgroup of sexual rapists). Violent offending also covers a range of behaviours and levels of harm. Much of the research suggests, however, that a relatively small group of offenders commit the majority of crimes which involve serious violence against others (see Piper 1985). Self-report interview data from 1,412 adolescents in the United States identified 118 as serious violent offenders (Elliot, Huizinga, and Morse 1986). Despite this, longitudinal data from a 1958 Philadelphia cohort showed that violent offenders accounted for a large proportion of the more serious index offences (Piper 1985).

2.3 Some research into violent and sexual offenders does not draw distinctions between the groups in terms of recidivism or risk of recidivism. The MacLean Committee in Scotland (Scottish Executive 2001c), among others, explain that this can be because the groups are classed as 'serious' offenders and are therefore studied at the same time, and because sexual and violent offending are not always mutually exclusive, such as in the case of rape. Indeed, research often suggests that sexual offending is motivated by a desire for power and control rather than from libido (e.g. Darke 1990).

2.4 With these difficulties in mind, the following information summarises available research into the recidivism of violent and sexual offenders.

Sexual offenders

Main findings:
Studies of sex offending consistently show relatively low rates of recidivism relative to all other types of offending.
Different rates of recidivism are apparent for different types of sex offenders.
Sexual offenders remain at risk of reoffending often long after their discharge from custody

2.5 Studies of sex offenders tend to show relatively low rates of recidivism for this group. Statistics in Scotland show that sexual offenders have amongst the lowest rates of return to custody after two years (Ash and Biggar 2001). Figures in England show a similar pattern: after two years,

Lloyd, Mair, and Hough (1994) found that 25% of sex offenders were reconvicted of an offence (15% of which were sexual offences), compared to 47% of violent offenders (27% of whom committed further violent offences). More recent research in England (Hood et al. 2002) followed 162 sex offenders for four years and 94 for six years after their release from long-term prison sentences. The study found that overall reconviction of this group for another sexual offence after six years was below 10% and did not increase substantially when the rate included serious violent offences.

2.6 The picture internationally reflects the same patterns. After three years, a study in the United States (Beck and Shipley 1989) found that sexual offenders had the lowest rearrest rates for offences of the same type (25%), apart from murderers, only 7% of whom were rearrested for a further murder. They also found rates of 52% rearrest (a higher figure than reconviction) among rapists and 48% among sexual assaulters, though these were among the lowest general rates; only rearrest of people originally convicted of murder and negligent manslaughter were lower, with rearrest rates of 43% and 42% respectively. However, this study followed sex offenders for fairly short periods of time which, as noted above, may not give an accurate picture of reoffending for this group.

2.7 A number of studies have followed up sexual offenders for longer periods after their release from custody. Researchers in Toronto (Wilson et al. 2000) followed 107 sexual offenders on parole supervision over eight years, with an average follow-up of just over three and a half years. Just over a fifth (21.0%) of the subjects reoffended, including 10.3% for offences of violence and 3.7% for further sexual offending. In Australia, records for 560 sex offenders were analysed for up to 12 years after the offenders' release from prison (Broadhurst and Maller 1992). The results showed the probability of reincarceration for any offence to be 35% for non-Aborigines and 80% for Aborigines. For violent crimes, including further sex offences, probabilities were 21% and 62%, respectively. Hanson and Bussière (1996) conducted a meta-analysis of 61 studies from six different countries of recidivism amongst sexual offenders. They found that, after an average follow-up of four to five years, the average rate of recidivism for sexual offenders was 13.4% for sexual offences, 12.2% for non-sexual offences against the person, and 36.3% for any offence.

2.8 A small study of young sex offenders in Sweden who had been subjected to court-ordered forensic psychiatric investigations from 1988-95 (N=46) had a base rate of reoffending of 20% for sexual offences and 65% for offending in general, over an average time at risk of five years (Langström and Grann 2000). A larger long-term study of 541 male sexual offenders in Norway (Grunfeld and Noreik 1986) found that, after 9 - 13 years, 12.8% had committed further offences. Most of the recidivists had committed a single subsequent offence, either of the same type or a less severe offence. Similar research in Canada, following 570 sex offenders over an average of 3.5 years after release from a federal prison, found that about a third were convicted of new criminal offences, about a quarter of violent offences, and just under 10% of new sexual offences (Motiuk and Brown 1996).

Sub-groups of sexual offenders

2.9 While sex offenders tend to be categorised as a single group of people, the nature of their offending and subsequent risk to the public often differs (see below). Little Scottish research has been published on this, but recent work in England by Hood and colleague (2002) emphasised the importance of dividing sexual offenders into varying categories of risk. They found, for example, that all offenders in their study who had offended against children and then reconvicted for a sexual offence had originally been convicted of an extra-familial sexual crime. After six years, 32% of the extra-familial group had been imprisoned again for a sexual or a serious violent crime. Similarly, earlier work published in England (Fisher 1994) noted that non-familial sexual abusers of boys reoffended between 13 and 40% of the time. The comparative figure in Hood's study for men who originally offended against adults was 7.5% reconviction for a sexual offence after six years, and 15% reconviction for a sexual or serious violent crime (see also 'Choice of victims' in Chapter 4 below).

2.10 Similar patterns were evident in other countries. Marshall and Barbaree (1990) noted different rates of recidivism for different types of sex offenders. Exhibitionists, for example, had the highest rate of recidivism, ranging from 41-71%. Non-familial child sexual abusers showed the next highest rate of reoffending (10-40%), followed by rapists (7-35%) and finally familial child sexual abusers (4-10%). Child molesters with male victims showed a higher rate of recidivism than those with female victims (35% v. 18%; Quinsey et al. 1995).

2.11 Another long-term study, this time one which followed 197 child molesters, showed that 42% of the sample were reconvicted of sexual, violent, or both sexual and violent crimes (Hanson, Steffy, and Gauthier 1993). A tenth of this group was reconvicted 10 - 31 years after being released. A similar study in Ontario, Canada (Hanson, Scott, and Steffy 1995) used national police records to compare the recidivism of 191 child molesters and 137 non-sexual criminals over 15 - 30 years. During this time, 61.8% of the child molesters and 83.2% of the other group were reconvicted. Almost all further sexual offences were committed by the child molesters (35% of this group, compared to 1.5% of the non-sexual criminals), while the non-sexual group was responsible for virtually all of the non-sexual violent recidivism. Research into 136 extra-familial child molesters in a maximum security psychiatric institution (Rice, Quinsey, and Harris 1991) found that, after a follow-up of 6.3 years (on average), 31% were convicted of new sexual offences, 43% committed a violent or sexual offence, and 58% were arrested or returned to the institution.

2.12 A study which looked specifically at rape followed 54 rapists released from a maximum security psychiatric hospital in Canada over a period of almost four years. Of this group, 59% committed a further offence, 28% committed a further sexual offence, and 43% committed a further violent offence (Rice, Harris, and Quinsey 1990). A similar study with rapists and child molesters, this time involving 178 patients over a period averaging about five years, found that 28% were convicted of a new sex offence, while 40% were arrested, convicted, or returned to the psychiatric facility for a violent (including sexual) offence (Quinsey, Rice, and Harris 1995). Research into 251 incest offenders at the Royal Ottawa Hospital Sexual Behaviors Clinic (Firestone et al. 1999) found that, roughly six and a half years after their conviction, recidivism was 6.4% for further sexual offences, 12.4% for violent offences, and 26.7% for offences of any kind.

2.13 Two studies by Hagan and Gust-Brey looked specifically at adolescent perpetrators of rape (1999) and of sexual assault against children (2000). The ten-year follow-up of 50 adolescent rapists released from custody in Wisconsin showed that most were convicted of further crimes, but not of a further sexual assault. For those who committed a further sexual offence, the pattern of assault remained stable in years 6 - 10 as it did in years 1 - 5, meaning this group seems to commit such offences at a fairly steady rate for at least 6 - 10 years after release. No decrease in the risk of sexual recidivism was apparent even up to ten years after release. For sexual assaults against children, 12% of the sample reoffended sexually, 66% were convicted of non-sexual offences, and 22% were not reconvicted during a ten-year follow-up. Though the risk of further sexual offences continued throughout the ten-year period, most first sexual offences after release for this group took place in the first five years.

2.14 A methodological analysis of recidivism studies of sex offenders assessed six official sources on 251 rapists and child molesters discharged from a treatment centre in Massachusetts over a 25-year period (Prentky et al. 1997). The research found high variability in reported recidivism depending on the methods and definitions used. However a number of findings were consistent, namely that both rapists and child molesters remain at risk of reoffending often long after their discharge from custody (in some cases 15 - 20 years); simple calculations of recidivism based on those known to have reoffended during a follow-up period consistently underestimated the actual amount; and measures based on conviction or imprisonment also markedly underestimated the actual amount of reoffending.


2.15 The ability to predict the risk of further offending by people who have committed a sexual offence remains contentious (see Wollert and Doren 2001). In England, West (2000b) argues that statistics do not support the stereotype of the 'inveterate' sex offender for whom recidivism is inevitable, even within specific offence types such as paedophilia (2000a). A consistent finding internationally, however, is that recidivism amongst this group is small relative to all other types of offending (see Wormith and Ruhl 1986). A group of adolescents followed into adulthood, for example, showed that while recidivism was higher for sexual offences if the offender committed such offences as a juvenile, the rate of rearrest overall was low (Sipe, Jensen, and Everett 1998). For juveniles who committed non-sexual offences, however, offending in adulthood was higher for all other offence types.

2.16 Some authors, such as Marshall (1994) in England, argue that a person should be considered a potential sexual recidivist if he has ever been convicted of a sexual offence, regardless of the nature of any further offending. A study of the use of preventive detention in Canada (Wormith and Ruhl 1986) found that the 97 people given dangerous sexual offender designations were uniformly high-risk offenders, despite the findings of predictive instruments which suggested dangerous sexual offenders were low-risk compared to the general prisoner population. Interestingly, the authors also found that the designation of 'dangerous sexual offender' was used disproportionately for people convicted of child molestation.

Violent offenders

Main findings:
About a fifth of offenders with an index offence for violence are reconvicted of violence within two years.
Reconviction for violence was highly associated with previous convictions for violence and may be more likely when the index offence of violence is more serious;
People released from prison were reconvicted at a much faster rate than people who had received other sentences;
A more extensive history of offending is associated with a greater likelihood of reconviction within two years; the higher the number of previous convictions, the faster the rate of reconviction.

2.17 Much information on recidivism of violent offenders comes from official statistics. In Scotland, information from the Scottish Offenders' Index (which does not include offences committed under age 16) was analysed for reconvictions over five years (Scottish Executive 2001b). This data showed that a fifth of offenders with an index offence for violence were reconvicted of violence within two years. Equally, a fifth of the 12,526 people with previous convictions for violence were reconvicted for violent crimes (average 1.7 crimes each), and 40% were convicted of any crime. A similar pattern was apparent for serious assault: of 821 people convicted of a serious assault in 1995, 40% were reconvicted for any crime within two years: 2% for the same crime, 20% for a violent crime, and 5% for a serious violent crime. In this case, a serious violent crime was defined as one which was dealt with in the High Court or Sheriff's Solemn Court; how many cases would fit under the definition of a serious violent offence for the purpose of the current review is unclear. The data also said that 9% of these offenders committed a further serious crime, defined as before but including any other conviction for serious assault, robbery, possession of a firearm with intent to endanger life, abduction, attempted rape, or indecent assault.

2.18 The same report found that, of 550 people convicted of robbery in 1995, 13% were reconvicted of serious violence - more than double the next highest proportion from any other category of index offence - and 18% were convicted of a serious crime. Reconviction for violence was highly associated with previous convictions for violence: two-thirds (67%) of the 70 offenders with over ten previous convictions for violence in 1995 were reconvicted for further violence within two years, with an average of just under three further violent offences in total (ibid.). The corresponding rate for the 8,210 people with no previous convictions for violence was 20%.

2.19 The Scottish statistics showed that recidivism for people with an index offence of violence was higher for those discharged from a custodial sentence (35%) or probation (31%) than from other types of disposals. Further, people whose index conviction for violence was in a Sheriff summary court or district court were less likely to be reconvicted for violence within two years (19%) than if the conviction was in a High Court or Sheriff's solemn court (25%). These figures suggest that violent recidivism may be more likely when the index offence of violence is more serious.

2.20 A comparable study in New Zealand showed similar results (Spier 2001). Over half (56%) of people convicted of a violent offence in 1995 were reconvicted within two years (compared, for example, to 68% of property offenders and 46% of traffic offenders). As in Scotland, a fifth (21%) of people convicted of a violent offence were reconvicted for violence within two years, including 7% who were reconvicted of two or more violent offences. As in Scotland, people released from prison were reconvicted at a much faster rate than people who had received other sentences; the vast majority of people with an index offence of violence and released from prison in New Zealand were reconvicted within seven months of release. Again, more extensive history of offending was related to a greater likelihood of reconviction within two years: a quarter of first offenders were reconvicted within this time, compared to 88% of people with more than 50 prior convictions. The higher the number of previous convictions, the faster the rate of reconviction.

2.21 Research in Canada (Johnson and Grant 2000) compared recidivism amongst indeterminate sentenced prisoners (including life), prisoners sentenced to ten years or more, and those serving sentences of less than ten years. In general, people sentenced for first degree murder 2 in Canada serve a minimum of 25 years before they are eligible for parole (or 15 years in exceptional cases), while those sentenced for second degree murder or other offences serve 10 - 24 years. Indeterminate sentences also include people designated as Dangerous Offenders 3 in which case they stay in custody until the National Parole Board determines custody to be no longer necessary for the protection of the public. Over 90% of indeterminate-sentenced prisoners in Canada are there for homicide or related offences Life sentenced prisoners are usually there for homicide (74.3% of all homicide cases), though sometimes for sexual offences (7.6%) or drugs (6.5%), while other prisoners with indeterminate sentences (such as Dangerous Offenders) are usually there for sexual offences (9% of all sexual offenders in custody, compared to 0.4% of all people imprisoned for homicide; Motiuk and Nafekh 2000). Determinant-sentenced prisoners serving ten years or more are more often there for sex offences (12.5% of sex offenders in custody), robbery (14%), or drugs (10.7%; ibid.). Those serving ten years or more (including life) have generally committed most serious violent offences: 54% homicide, as well as robbery, sexual assault, and assault (Johnson and Grant 2000). This group makes up a third of federally sentenced prisoners.

2.22 Johnson and Grant followed these groups for four to seven years after release. At the end of the seven years, they found a rate of 40% recidivism for determinate-sentenced prisoners and 27% for those with indeterminate sentences. The authors note that even this smaller rate for people who had served indeterminate sentences may be inflated by comparison since they remain under supervision for longer periods. The shorter-term determinate group (serving less than ten years) reoffended more quickly - 20% within the first year and 30% within two years. Just over a tenth of longer-term determinate prisoners reoffended in their first years, compared to about 5% of indeterminate prisoners. However, the study found that recidivism of people who served longer-term determinate sentences began to catch up with the shorter-term group after about five years.

2.23 Also after the seven years, 11% of the indeterminate group committed a new violent offence, compared to 20% of both the long- and short-term sentenced prisoners. Only 1% of people who had served indeterminate sentences had more than five new convictions, compared to 5% of the longer-sentenced group and 10% of the shorter-sentenced offenders. The most serious offence for over half of the indeterminate-sentenced group was non-violent, and none of this group were convicted of new homicide-related offences. Important to note, however, is that people released from indeterminate sentences in Canada are older, are more likely to have been convicted of homicide, and are more likely to be released on day parole.

2.24 In contrast to these findings, Bonta and Hanson (1995) compared recidivism for violent offences between detainees 4 in Canada and general federally sentenced prisoners. A sample of over 3,000 federal offenders showed a recidivism rate for violence of 48.6% over a period of 11 years. For detained prisoners, the rate of violent reoffending was 40.4%, but over a time frame of one year or more after release. The authors concluded that violent reoffending amongst the two groups was roughly similar, but that reoffending amongst the detained group happened much more quickly. Prisoners in their study who had been designated as Dangerous Offenders had not yet been released, but they shared significant similarities with the detained offenders in terms of index offence, criminal history, objective measures of risk, and incidence of antisocial personality.

2.25 In Florida, 59 young people sentenced to serve time in an adult prison for murder, attempted murder, and in a few cases, manslaughter, were followed for 15 - 17 years (Heide et al. 2001). More than two-thirds were released prior to November 1999; of these, 60% were returned to prison, usually within the first three years after their release. In Wisconsin, a small-scale study of 44 subjects diagnosed as extremely dangerous during psychological treatment in an institution for adjudicated juvenile delinquents were followed for 2 - 10 years after release (Hagan and King 1998). The results showed that the vast majority had reoffended, and of the recidivists only one had not been involved in a further violent offence. In contrast, another small study of 20 adolescent perpetrators of homicide and attempted homicide released from custody in Wisconsin (Hagan 1997) found that these young people did not commit further homicides up to 15 years following release. The majority were however involved in crime, primarily violent crime. Despite this, a significant number did not become reinvolved in crime, and almost half were not reincarcerated. No significant differences in likelihood of recidivism were evident between this group and other young offenders. The numbers of subjects in all of these studies are, however, likely to be too small to draw wider conclusions.

2.26 A longitudinal study of 322 people arrested for violent offences in Ohio found that, while recidivists were not high-rate offenders, they still tended to engage in above-average levels of index and other violent activity, and many seriously persisted in violent activities (Martinez 1997). In Canada, Serin, Mailloux, and Hucker (2000) tracked 260 violent federal offenders for seven to eight years after release. They found that 55% of the sample committed a further offence, of which 18% committed a violent offence as their first recorded offence after release. Over a third of the sample (37%) committed a violent offence at some point after their release. The authors found that people released on statutory release were more likely to commit a further offence of any kind and to commit a further violent offence at some time than were those released on parole, and that this group reoffended more quickly.

2.27 Research in the United States suggests that only 5 - 7% of young people who offend go on to serious violent careers (Jeffery 1998). An analysis of self-report interview data from 1,412 adolescents in the United States showed that involvement in violent offending was generally less than one year, but also that individual rates of offending increased with time (Elliot, Huizinga, and Morse 1986).

Violent and sexual offenders

2.28 In Scotland, Connelly and Williamson (2001) argued that the preventive detention of violent and sexual offenders is likely to have a limited impact on recidivism since offending declines with age. They based their assessment on a review of the literature elsewhere; very little work on preventive detention has been conducted in Scotland. Research in other countries has examined the effect of preventive detention on further offending, though the numbers are generally very small. In Germany, less than thirty offenders per year are sentenced to preventive detention. They serve an average of four years on preventive detention and have an average age of 40 (Albrecht 1997).

2.29 Some of the most extensive work in this area has been conducted in Canada. There, a follow-up of 424 detainees for at least a year after their eventual release (range up to 7.8 years, with an average follow-up of just over four years; Motiuk, Belcourt, and Bonta 1996) showed that a fifth (83 people) were readmitted into federal custody within a year of release. Fifty were returned for technical violations, but 33 (8% of the total) were readmitted for new offences. This compares to a readmission rate of 46.6% for prisoners released automatically (those not detained under preventive legislation; Nouwens, Motiuk, and Boe 1993). The sample was then divided into three groups, namely those who were initially detained, then released prior to the end of their sentence (Group 1: 148 prisoners/35%); those who were detained for their entire sentence (Group 2: 114 prisoners/27%); and those who were detained until the end of their sentence after a "one-chance" release was revoked (Group 3: 162/38%). Of these groups, only 6.2% of Group 2 and 7.4% of Group 3 returned to custody for any reason in the first year, compared to 43.9% of Group 1.

2.30 The extended follow-up showed that about 60% of detainees were convicted of a new criminal offence, including 40% for a violent offence (also Bonta and Motiuk 1996) and 15% for a sexual offence. No significant difference between the groups was evident for this or for the average number of times they were convicted after release. Of the new convictions, about a third of the most serious convictions were for assault, while about a fifth of the most serious reconvictions were sexual offences. Three offenders, all from Group 3, were convicted of second degree murder after their release, while the remainder of reconvictions were property-related. The authors conclude that:

Perhaps the most striking finding was the relatively low rate of post-detention serious reoffending (such as murder, sexual assault, robbery). Considering this population was seen as one of the highest risk groups of offenders within the system, a much higher percentage was expected. (Motiuk, Belcourt, and Bonta 1996:3, emphasis in original)

Mental disorder and recidivism

Main findings:
Rates of recidivism for people labeled as sexual psychopaths, treated, and released into the community were comparable to those for sexual offenders who had not been committed for treatment.
High risk patients had five times as many violent reconvictions as did low-risk patients.
The level of recidivism amongst patients in a medium secure unit was comparable to that in other forensic psychiatric patients, which was much lower than recidivism amongst ordinary prisoners.
The rate of recidivism for an offender/patient released against expert advice was substantially greater than for those released in accordance with advice.

2.31 A number of studies looked specifically at the role of mental disorder on the recidivism of serious violent and sexual offenders. This section discusses these studies in more detail.

Sexual offenders

2.30 Very little Scottish research into mental disorder and recidivism is available. In England, a study of 49 patients diagnosed with psychopathic disorder from the Young Persons Unit (YPU) at Broadmoor Hospital looked at outcome measures until discharge from statutory supervision (Reiss, Grubin, and Meux 1996). By the end of the follow-up, ten patients (20.4%) had reoffended, including four (8.2%) who committed serious violent or sexual offences. Interestingly, a number of studies found that rates of recidivism for people labelled as sexual psychopaths, treated, and released into the community were comparable to those for sexual offenders who had not been committed for treatment (Horowitz 1995; Lieb 1996; Wilson 1998). This led to a repeal of sexual psychopath statutes, which had originally been put into place in more than half the states by the 1960s in the United States (Connelly and Williamson 2001).

Violent offenders

2.31 Research into impulsivity amongst 44 male patients also from Broadmoor Special Hospital (Howard and Lumsden 1997) found that 5 of 21 high-risk patients (23.8%) and 1 in 23 low-risk patients (4.3%) had a violent outcome (violent recidivism or suicide) after release. High risk patients had five times as many violent reconvictions as did low-risk patients. Admittedly mentally disordered patients hospitalised following more serious offences are likely to be in hospital for longer periods (Lelliot, Wing, and Clifford 1994; Green and Baglioni 1998; Friendship et al. 1999). Perhaps because of this, Australian research into one such group found their reoffending was lower than average, with 20% known to have reoffended and 11% charged with a violent offence (Green and Baglioni 1998). In England and Wales, 234 patients discharged from a medium secure unit showed a rate of recidivism of 24% with at least one conviction, including 12% convicted for a serious offence (Friendship et al. 1999). Again, rates may be lower because of the proportion of patients who remained in an institution for much of the follow-up period. However, the level of recidivism was comparable to other studies of forensic psychiatric patients, which the authors found to be much lower than recidivism amongst ordinary prisoners.

2.32 The TBS (Terbesschikkingstelling) system in The Netherlands is designed for people who have committed very serious crimes (eligible for a sentence of four years or more) but who have some element of reduced responsibility due to a mental disorder. The purpose of the system is the protection of the public, along with the care, treatment, and rehabilitation of the offender with a view to returning him or her to society. TBS patients usually have some form of psychotic illness and personality disorder, 60% have been assessed with addiction problems, and a quarter have been convicted of a sexual offence (5% of all sexual offenders convicted in The Netherlands; Scottish Executive 2001c). In the first five years after release from the TBS system, recidivism is approximately 50 - 60%, with the rate for more serious offences fairly steady at 20% (Kinzig 1997; Leuw 1995). If an offender/patient was released against expert advice, the rate of recidivism has been about 27%, but those released in accordance with advice have shown a rate as low as 8% (Petrunik 1994). Higher recidivism was also evident when patients were released directly into the community with no conditional discharge (Connelly and Williamson 2001). Young people in the TBS with long criminal histories show higher rates of recidivism, though a significant correlation was evident between behaviour while in the TSB and recidivism (ibid.). About 20 - 30 of the 1,100 - 1,200 patients in the TBS (roughly 2%) have been assessed as extremely resistant to treatment and are unlikely ever to be released (Scottish Executive 2001c).

Violent and sexual offenders

2.34 Van Emmerick (1989) conducted a five-year follow-up 138 violent sex offenders discharged from a psychiatric hospital in the Netherlands between 1970 and 1983. During that time 64% were charged with at least one further offence, 49% were convicted, 32% imprisoned, and 25% were convicted of a further sexual offence. Interestingly, these figures did not differ from those for patients who were not sexual offenders. Leuw (1995) studied 91 sexual offenders discharged between 1984 and 1988 and found similar results. These results may well be underestimates of actual reoffending, however, with some authors estimating that sexual recidivism is up to three times higher than official figures (Frenken et al. 1999).

The impact of treatment

Main findings:
Appropriate correctional treatment can reduce rates of reoffending; however different approaches to and duration of treatment appear necessary for different types and personalities of offenders.

2.35 The remit of this literature review did not include an assessment of the effectiveness of the various treatment programmes designed for serious violent or sexual offenders. A proper assessment of the effectiveness of programmes would need to look at recidivism rates of large samples and make comparisons with non-treated samples (SWSI 1997). Matching is not always done in such studies and, depending on the criteria for treatment, may not be possible (Marshall and Pithers 1994); other methodological limitations and problems in evaluation are also common (Finn 1995). However, much of the information of recidivism for this group is reported in the context of treatment programmes, predominantly programmes for sex offenders. A selection of programmes will be discussed briefly here 5.

2.36 The Social Work Services Inspectorate (1997) noted that no systematic evaluation has been conducted of the majority of programmes for sex offenders in Scotland. Early drop-out from programmes has been a problem for some types of treatment (e.g. a 30% drop-out rate amongst sex offenders in the STOP programme at HMP Peterhead) which, due to the structure of the programme, may mean people who drop out are potentially more dangerous than they were previously (Expert Panel on Sex Offending 2001). Indeed, a meta-analysis conducted in Canada of 61 studies of recidivism amongst sexual offenders found that those who fail to complete treatment may be at increased risk of reoffending (Hanson and Bussière 1998).

2.37 In England, Hedderman and Sugg (1996) suggested that some forms of treatment may indeed have positive effects on attitudes and recidivism. Internationally, treatment of sexual offenders and of serious offenders generally has shown a small but robust effect of treatment over comparison conditions (e.g. Lipsey 1999; Plumb 1992; Hall 1995; Walsh 1998). A meta-analysis by Hall (1995) reported an overall rate of 27% recidivism for untreated groups, compared to 19% for treated offenders. A comparison of 200 studies involving serious offenders found consistent treatment effects, with larger effects for juveniles with a range of prior offences and for more serious offenders (Lipsey and Wilson 1998). Some programmes have claimed reductions in both sexual and non-sexual offending (Studer et al. 199). Canadian research into 89 sexual offenders in treatment compared with 89 matched untreated sexual offenders over roughly ten years found a rate of 23.6% sexual recidivism for the treated group, compared to 51.7% for the untreated group (Looman, Abracen, and Nicholaichuk 2000). Non-sexual recidivism also differed significantly between the two groups. Nicholaichuk and colleagues (2000) compared larger matched samples of 296 treated sexual offenders from a Canadian Regional Psychiatric Center with 283 untreated offenders over an average of six years with positive results: convictions for new sexual offences for the treated group were 14.5%, compared to 33.2% of the untreated sample, and 48% remained out of prison compared to 28.3% of the untreated offenders. After ten years, the treated men still reoffended at significantly lower rates. The authors conclude that appropriate correctional treatment can reduce rates of reoffending.

2.38 Relatively short-term follow-ups of treated sexual offenders in Canada also showed reductions in sexual recidivism. A two-year follow-up of 80 higher-risk sex offenders showed a reduction of sexual recidivism of 58.9%, from 14.6% for the control group to 6% in the treatment group over two years (Gordon and Nicholaichuk 1996), while a three-year follow-up of 210 treated sex offenders recorded a reduction of 50%, from 6% for a control group to 3% for the treated group (Motiuk 1998).

2.39 The findings of such research has not however been consistent. An evaluation of the same treatment facility in Canada examined in Looman, Abracen, and Nicholaichuk's work (Quinsey, Khanna, and Malcolm 1998) showed a worse performance of treated offenders compared to untreated groups. In the study by Quinsey and colleagues, 38% of 483 sex offenders were arrested for new sexual or violent offences within an average of 44 months after release. Treated offenders were most frequently arrested for sexual offences; prisoners judged unsuitable for treatment were rearrested less frequently, particularly for sex offences. Those who refused treatment or were judged not to require treatment also had fewer arrests for sexual offences, though were rearrested more often for violent offences. Research into 136 extra-familial child molesters in a maximum security psychiatric institution (Rice, Quinsey, and Harris 1991) found that behavioural treatment, provided to 50 of the 136 subjects, had no bearing on recidivism. Research by Marques and colleagues (1994) failed to show conclusive results as to whether cognitive-behavioural treatment was effective in reducing recidivism between a treatment group and two separate control groups, although the treated offenders had the lowest reoffence rates for both sexual and violent crimes.

2.40 Hagan, King, and Patros (1994b) conducted a short-term study of 50 young people who committed a sexual offence against a child and subsequently completed a Serious Sexual Offenders Program. Their results showed that, after two years, 46% of the participants had committed a further offence, including 8% who had committed a further sexual offence. The same authors conducted an almost identical study of adolescent rapists in the programme and found that 58% were convicted of further charges, 10% were involved in a further sexual assault, and 38% returned to custody. The authors made no comparison of these findings with untreated samples.

2.41 A very small-scale study of Outpatient Behaviour Therapy for sex offenders (Crolley et al. 1998) noted statistically significant differences in entrance and exit scores and clinical improvements (including reduction in deviant sexual arousal and enhancement of appropriate behaviour) amongst the 16 participants, and only one new offence in the 26-month follow-up.

2.42 Craissati and McClurg (1997) also found signs of improvement amongst 25 child sexual abusers compared to a matched group. Both of these studies relied on very small samples and did not compare their findings with untreated groups, so findings should be treated with caution.

2.43 In Germany, special socio-therapeutic prisons have been established for recidivists, though some authors (e.g. Albrecht 1997; Ortmann 2000) have expressed doubt as to whether treatment in such institutions, or indeed in prison generally, has a significant effect on reoffending. Equally, Thornton (1992) notes that psychoanalysis is rarely effective in decreasing sex offending. Indeed, some studies have found greater levels of recidivism following certain types of treatment compared to untreated offenders (e.g. Quinsey et al. 1993).

2.44 Therapeutic communities have shown some success in reducing recidivism generally (Taylor 2000), and for drug-involved offenders (Inciardi et al. 1997), especially with longer periods in therapy (Rawlings 1999; cf Weisz et al. 1990). The Patuxent Institution in Maryland, which houses serious (usually violent) offenders with an 'emotional imbalance', has claimed notable success. According to their figures no one released from Patuxent into the community was rearrested within three years, and only 7% were returned for parole violations, usually drug-related. This compares to a 42% rate of rearrest for mainstream prisons there (Scottish Executive 2001c). The Institution also runs a community-based programme for sex offenders which, over the last 25 years, has shown a rate of 8% recidivism for any offence and 3% for sexual offences. Inmates of Patuxent are not however a representative sample. They must volunteer for placement there, stay in custody for long periods, and are released under a close level of supervision. Research into people in therapeutic communities diagnosed as psychopathic (Harris, Rice, and Cormier 1994) suggested that such communities may be less effective for this group.

2.45 A more controversial method of reducing reoffending amongst sex offenders has been through physical and/or chemical castration. A review by Barker and Morgan (1993) noted that psychosurgical intervention has been unsuccessful in decreasing recidivism. However, some jurisdictions have claimed high rates of success with castration, such as in Germany where the available data on castrated offenders shows a rate of 3% for sexual recidivism (Carpenter 1998). In Denmark from 1935-70 (after which surgical castration was discontinued there), 21 of 43 offenders sentenced to preventive detention after violent rapes were surgically castrated and released on probation within 18 months; those who refused castration were detained for extended periods. After castration, two committed further sexual offences after hormone replacement therapy from their GP, compared to ten who refused castration. A review of research in this area claims that rates of recidivism for physically castrated offenders are "strikingly better" than those for chemical methods (Meyer and Cole 1997).

2.46 Barker and Morgan argue that studies which show low reconviction for castrated offenders have included consensual homosexual acts in their range of index offences; reconviction of castrated rapists and child sex offenders was unclear as of 1997 (SWSI 1997). Sturup (1968) found that reconviction of castrated sexual offenders in Denmark for non-sexual offences was actually higher than for comparable sex offenders who were not castrated. Physical castration has been largely abandoned for several years (Marshall et al. 1991).

2.47 A follow-up of people sentenced to preventive detention at the Herstedvster Institute for Abnormal Offenders in Denmark from 1973-87 showed high rates of recidivism (Hansen and Lykke-Olesen 1997). Three were chemically castrated with the drug Adrocur, two of whom committed a further sexual offence including one sexual murder. These three and the remainder were also subject to individual psychodynamically orientated therapy designed to correct behavioural attitudes. Of the remainder, three committed new sexual offences during their sentence. Twelve were released on probation, half of whom reoffended including four sexual offences, three of which were very serious, and two violent offences. Of the six who had not been reconvicted, one had been charged with raped, with the case dropped due to insufficient evidence. One was later convicted of attempted murder and another of murder.

2.48 In 1989, 33 offenders were referred to Herstedvster for chemical castration. Three refused it, two of whom were later released and committed further serious sexual offences. Of the small number who started and continued treatment, none have reoffended; the numbers here are small and the follow-up short, but the Institute is apparently optimistic about the success of its current regime of chemical castration (ibid.; see also Meisenkothen 1999).

2.49 Treatment of violent offenders has shown some success. A Canadian study of 60 higher risk offenders with a history of violent offences (Motiuk et al. 1996) found that, after two years, general recidivism differed from 40% for the control group to 35% for the treated group (a reduction of 12.5%). Violent recidivism dropped by 16.7%, from 18% for the control group to 15% for the treated group. Research by Borduin and colleagues (1995) suggests that multi-systemic therapy, which addresses wider correlates of antisocial behaviour, is more effective in preventing future serious and violent offending among young people. More novel approaches such as Maharishi's Integrated System of rehabilitation, which includes transcendental meditation, has shown significant decreases in violence and recidivism in some countries (Anklesaria and Lary 1992).

2.50 The available evidence internationally suggests that cognitive-behavioural models may generally be the most promising (Barker and Morgan 1993; Henning and Frueh 1996; Wood, Grossman, and Fichter 2000; also McGrath, Hoke, and Vojtisek 1998), though the results have not been consistent. Some authors claim that cognitive-behavioural methods have been less successful in reducing recidivism for rapists and exhibitionists (Marshall and Barbaree 1990; Marshall et al. 1991), while others have found the opposite effect with some programmes (Marques et al. 1994). With child molesters, the type of treatment found to be effective has varied with the sex of the victim (Marshall and Barbaree 1990). In general, different approaches to treatment appear necessary for different types and personalities of offenders (Saunders 1996; Wood, Grossman, and Fichter 2000), as is the case regarding optimal duration of treatment (Pellerin et al. 1996).

Summary and conclusions

2.51 The available information above shows varying rates of reconviction of sex offenders. Most of the rates are fairly low compared to other types of offending (e.g. Becker and Hunter 1992; SWSI 1997), though the rate increases over longer periods. Important to note is that studies of recidivism which use reconviction or even rearrest as a measure are likely to underestimate the actual rate of reoffending: interviews with people convicted of rape or child molestation found that the majority had been convicted more than once and admitted to committing sexual offences which had not been detected (Groth et al. 1982).

2.52 Recidivism for violent offences is higher than for sexual offences, but again is generally lower than for other types of crime. For both violent and sexual offenders, recidivism is not usually for a further sexual or violent offence. The statistics clearly show that violent and sexual offenders do not necessarily specialise in one type of offence. Rather, a large proportion may commit other types of offences, both before and after their first conviction for a sexual or violent crime (Grunfeld and Noreik 1986). However, the highest risk for further sexual or violent offences was from people who had committed such offences in the past. Greater recidivism was also associated with more extensive offending histories and with more serious offences. The next chapters discuss these patterns in more detail.

2.53 The treatment of sex offenders is problematic, and many approaches have been adopted with limited success. Actual rates of recidivism with this group are equally difficult to assess: the majority who have been detained under provisions for dangerous offenders have not yet been released (Connelly and Williamson 2001) 6. Some types of programmes have shown small but robust improvements in rates of recidivism for both violent and sexual offenders compared to untreated samples.