Correctional Service Canada
Symbol of the Government of Canada

Common menu bar links

Research Reports

Warning This Web page has been archived on the Web.

Circles of Support & Accountability:
A National Replication of Outcome Findings

2008 No R-185

Robin J. Wilson
(Humber Institute of Technology & Advanced Learning)
Franca Cortoni

(Correctional Service of Canada)
Monica Vermani

(Adler School of Psychology)

May, 2007


More than 12 years ago, a group of concerned citizens formed a community-based support group for a recently-released, high-risk sexual offender. This community action was later formalized as Circles of Support & Accountability (COSA), a project stewarded by the Mennonite Central Committee of Ontario and sponsored in part by the Correctional Service of Canada. The original pilot project in South-Central Ontario has since assisted more than 100 such offenders--most of whom were released to media attention and community apprehension. Projects based on this model are now in place in the United Kingdom and several states in the USA. Projects are also underway in all Canadian provinces. Initial research into the South-Central Ontario COSA project [CSC Research Report No R-168] showed that participation by ex-offenders in the pilot project reduced sexual recidivism by 70% or more in comparison to both matched controls and actuarial norms.

The current study consisted of a national examination of the impact of COSA on recidivism. A group of 47 high risk sexual offenders involved in COSA across Canada after having been released at the end of their sentence were matched to a group of 47 high risk sexual offenders who were released at the end of their sentence, but who were not involved in COSA. Offenders were matched on risk, length of time in the community, release date and location, and prior involvement in sexual offender specific treatment. The average follow-up time was 2.8 years (34 months). For the purpose of the study, recidivism was defined as having a new charge or conviction for a new offense or for having breached a condition imposed by the Court. Recidivism information was obtained from CPIC records (Canadian Police Information Centre, a national database of offense histories which indicates whether a charge has been laid or a conviction registered).

Results show that the offenders who participated in COSA had significantly lower rates of any type of reoffending than did the matched comparison offenders who did not participate in COSA. Specifically, offenders who participated in COSA had an 83% reduction in sexual recidivism in contrast to the matched comparison group (2.1% vs. 12.8%), a 73% reduction in all types of violent recidivism (including sexual - 8.5% vs. 31.9%), and an overall reduction of 72% in all types of recidivism (including violent and sexual - 10.6% vs. 38.3%). Overall, COSA participants were responsible for considerably less sexual, violent, and general offending in comparison to the matched comparison group.

These findings suggest that the impact of participation in COSA is not site-specific. In addition, these results provide further evidence for the position that community volunteers, with appropriate training and guidance, can and do assist in markedly improving offenders’ successful reintegration into the community.


The authors would like to thank Andrew McWhinnie, Susan and Pat Love, Peter Huish, Jean-Jacques Goulet, Moira Brownlee, Joan Carolyn, Maureen Donnegan, Dan Laramie, Ed Call, and Rev. David Molzahn for their assistance in facilitating this review. We are also thankful to Colette Cousineau and Renée Meunier for assistance with data extraction and coding. Last, we continue to be in the debt of the many volunteers who contribute to the safety of the community by donating their time to help ex-offenders find new lives and hope with the assistance of a Circle of Support & Accountability.





The Circles of Support and Accountability (COSA) initiative was originally conceived as a means to help high risk sexual offenders who were released at the end of their sentence. Within the Canadian federal correctional system, offenders at highest risk of reoffense tend to remain incarcerated until the end of their sentences (i.e., to Warrant Expiry Date--WED), after which they are released without a formal process of community supervision. As such, COSA projects generally set their sights on men released after having completed their entire sentence, and who have been judged to be at high risk to reoffend. Specifically, COSA target individuals who seem most likely to fail, due to a lack of prosocial support in the community. COSA have also supported those offenders who are likely to attract significant media attention.

The primary aim of Circles of Support and Accountability is "no more victims". The goal of COSA is therefore "to promote successful integration of released men into the community by providing support, advocacy, and a way to be meaningfully accountable in exchange for living safely in the community" (CSC, 2002). In doing so, safety is enhanced for the community, particularly where risk exists for women, children, and other vulnerable persons. Simply put, COSA promotes safety for victims (past or potential) by validating their needs for healing and continued safety while holding ex-offenders accountable for behaving responsibly. In return, their rights as citizens are protected. By supporting ex-offenders and holding them accountable for their choices in the community, harm is reduced.

COSA has grown from an innovative response to a single set of circumstances to becoming a viable community partner in assisting high-risk sexual offenders in their reintegration to society. The original pilot project, centred in Toronto, has now sponsored over 100 Circles, each comprised of a Core Member (the ex-offender) and four to six community volunteers--citizens who have pledged personal time to assist the Core Member in the community. The COSA model has also proliferated across Canada and into the UK and USA, with other countries investigating the model. In the UK, a well-established COSA variant has been jointly managed by the Thames Valley Probation Service and the Religious Society of Friends (Quakers - see Quaker Peace and Social Witness, 2005).

Almost all Canadian COSA projects enjoy the counsel of an advisory committee comprised of professionals from law enforcement, corrections, clinical services, and business. Volunteers are trained to ensure that they understand the roles and responsibilities associated with assisting high-risk sexual offenders in the community (see CSC, 2002). Volunteers act as concerned friends or surrogate family members for the Core Members, with support and accountability being set prominently in their minds. In a model consisting of two concentric circles, the "inner circle" is comprised of the Core Member and his volunteer supports. An equally important component of the model is found in the "outer circle", which is comprised of a pool of volunteer professionals available to provide advice and guidance should the volunteers encounter an issue beyond the scope of their role or expertise.

Empirical validation is an important and ever-present need when offering any risk management service. The initial validation of the Ontario pilot project (Wilson, Picheca, & Prinzo, 2005 [CSC Research Report No R-168]; Wilson, McWhinnie, Picheca, Prinzo, & Cortoni, 2007a; Wilson, Picheca, & Prinzo, 2007b; Wilson, Picheca, & Prinzo, in press) has underscored the efficacy of the COSA model in managing the risk of sexual offenders in the community. Results of that study showed that offenders involved in a Circle demonstrated a 70% lower rate of sexual reoffending than did their counterparts in a comparison to matched offenders who did not participate in a Circle. Results from the aforementioned UK project are equally encouraging. After three years of co-sponsoring a pilot project in the Thames Valley, the Quakers and the UK Home Office reported behavioural outcomes for 22 offenders involved in COSA. Their results showed that none of these offenders incurred a new sexual offense, and that only one offender was convicted of breaching a Sex Offence Prevention Order (Quaker Peace & Social Witness, 2005).

As COSA continues to expand across Canada and in other countries, it is important to determine whether the initial research findings on the efficacy of COSA in reducing recidivism remain valid in other samples. Consequently, the current study consisted of an extension of the initial Ontario validation to a new sample of Canadian offenders involved in COSA. This study examined whether COSA projects across Canada continue to demonstrate efficacy in reducing the recidivism of high risk sexual offenders in the community, regardless of their location.



Two groups of offenders were included in this study. The first group was comprised of 47 offenders who were involved in a COSA after having been released at the end of their sentence. These offenders were drawn from COSA projects in the following Canadian cities: Montreal (N=5), Ottawa (N=8), Kingston (N=9), Winnipeg (N=6), Saskatoon (N=2), Calgary (N=11), and from projects in British Columbia (N=6). The second group consisted of a matched comparison sample of 47 similar offenders who were also released at sentence completion, but did not participate in a COSA. The groups were matched a priori, meaning that there was an intentional process involved in selecting the comparison sample, so that it would be a more comparable group for the COSA participants.

Matching criteria

As mentioned earlier, COSA projects are principally intended to address the post-release needs of detained offenders. To ensure adequate matching, we only included similarly detained sexual offenders in the comparison sample. To further guarantee that the two groups were equivalent in criminality and risk levels, we recorded scores for each subject on the General Statistical Information on Recidivism (GSIR--Nuffield, 1982) scale. Each member of the COSA group was matched with a comparison subject in the same general risk category (e.g., low, low-moderate, moderate, moderate-high, or high).

In matching the two groups of offenders, we also endeavored to make sure that the matched subject was released on or about the same date as the subject in the COSA group and to the same general community. The purpose for doing so was two-fold: First, this process ensured that the matched subjects were released to relatively the same political and community climate; and second, it allowed for an easy comparison of the length of time at risk before failure (in those offenders who committed a new offense).

Last, we ensured that the two groups were matched with regard to prior involvement in sexual offender treatment programming. Given recent results (see Hanson, Gordon, et al., 2002) suggesting that completion of a treatment program adhering to the principles of effective correctional interventions (see Andrews & Bonta, 2003) can have a significant impact on recidivism, it was important to make sure that any differences found between the two groups were not the result of variations in previous treatment experiences.


STATIC-99 (Hanson & Thornton, 1999).

The STATIC-99 is a tool that actuarially assesses risk for sexual and violent recidivism based primarily on static risk variables. This instrument has moderate predictive ability (r = .33, ROC area = .71), and has extensive survival data from which long-term prognosis of risk potential can be established.

Rapid Risk Assessment for Sexual Offense Recidivism (RRASOR - Hanson, 1997).

The RRASOR is a four-item scale designed to actuarially assess risk for sexual reoffending in known sexual offenders. These four items are wholly contained in the STATIC-99 but, on their own, provide a moderately accurate screening of risk potential (r = .27, ROC area = .71).

Psychopathy Checklist Revised (PCL-R--Hare, 2003).

The PCL-R is a 20-item scale designed to measure the presence of particularly severe antisocial personality orientations--known as psychopathy. Although the PCL-R was developed as a diagnostic tool for psychopathy, research has consistently demonstrated a positive correlation between PCL-R scores and propensity for violence (Hare, 2003).

Phallometric testing.

The phallometric test is a psychophysiological procedure in which changes in penile circumference or volume are measured during presentation of audiovisual stimuli. Differential responding to various age, gender, or activity stimulus categories is helpful in diagnosing deviant sexual preferences (or paraphilias). Although conflicting research exists regarding the psychometric properties of the test (see Fernandez, 2002; Freund & Watson, 1991), it is generally accepted as a useful tool for diagnosis and, by extrapolation, risk assessment.


Recidivism was defined as being charged with a new offense or for having breached a condition imposed by the Court. Only official documentation was utilized and, in most cases, this information came in the form of CPIC (Canadian Police Information Centre, a national database of offense histories) records indicating that a charge had been laid or a conviction registered. For this study, sexual recidivism was defined as a new charge or conviction for a sexual offence. Violent recidivism was defined as a new violent charge or offence (including sexual offences). General recidivism was defined as any new charge or offence, all categories confounded. Consequently, the categories are not mutually exclusive.

Statistical Significance.

In this study, statistical significance was observed at the traditional p < .05. However, there are times when it is useful to evaluate results in terms of their social significance (see Gendreau, Little, & Goggin, 1996), that is, the impact the particular finding has on the community.


In order to assess the effects of COSAs on recidivism, we gathered data on 47 men involved in COSAs projects. These men were identified in consultation with COSA project managers across the country. In order to better understand the results we obtained from these men, we also selected 47 similar offenders from CSC’s Offender Management System (OMS) database. The 47 comparison subjects were matched to their COSA counterparts according to the criteria described above. Data from both groups were coded from OMS and from CPIC information.


In the first outcome study, data were presented showing the relative reoffense rates of COSA participants and matched comparison subjects in South-Central Ontario. These data are reprinted here (see Table 1) for comparison purposes.

Equivalency of Groups

As seen in Table 2, there were no differences in age between the two groups in the replication sample. In comparison to the earlier study (Table 1), differences in scores on actuarial measures were reversed, with the comparison group having a significantly higher average score on STATIC-99 (F[1,92] = 8.36, p < .01). The two groups, however, were not different in their average scores on the RRASOR (F [1,92] = .005, n.s.), nor were they different in terms of scores on the PCL-R (X2 [4] = 4.13, n.s.)1. Finally, there were no differences between the COSA participants and the comparison group in terms of the percentage of each group demonstrating sexually deviant profiles on phallometric testing (COSA: 32%; comparison: 30%).


As expected, given the matching procedure to ensure similar release dates in each matched pair of COSA participants and comparison offenders, there was no difference between the groups in terms of mean length of follow-up. There was a trend towards a difference (F[1,21] = 3.36, p < .10) in mean time until first failure (in those subjects who incurred further charges or convictions). Contrary to the results from the initial study [Research Report R-168], the mean time until failure in the comparison sample was longer than that observed in the COSA group.

Table 1. Recidivism Data--Ontario Pilot Sample (Wilson et al., 2005)

  COSA (N=60) Control (N=60)
M(SD) age
47.47 (12.27)
43.62 (10.84)
5.60 (2.22)
5.00 (1.96)
3.18 (1.65)
2.12 (1.31)
M(range-mos) follow-up
54.67 (3-123)
52.47 (3-124)
M(mos) until 1st failure
5.00% (3)
16.67% (10)
15.00% (9)
35.00% (21)
28.33% (17)
43.44% (26)

* p < .05   ** p < .01  ‡ p < .10

Table 2. Recidivism Data--National Replication Sample

  COSA (N=47) Control (N=47)
M(SD) age
43.18 (9.55)
43.52 (8.66)
5.00 (2.14)
6.11 (1.52)
2.72 (1.50)
2.74 (1.36)
Modal PCL-R (recoded 1-5)
5 (high)
5 (high)
Deviant Phallometrics
M Month (range) follow-up
32.53 (6-84)
35.74 (6-95)
M Month (range) until 1st failure‡ a
23.92 (1-39)
50.73 (3-112)
Recidivism (convictions + charges)
2.13% (n=1)
12.77% (n=6)
8.51% (n=4)
31.91% (n=15)
General** b
10.64% (n=5)
38.30% (n=18)
# of charges

* p < .05   ** p < .01  ‡ p < .10

a Because there are so few recidivists in the COSA group, the range is restricted - leading to a lowered average month until failure

b Includes sexual and violent offences

Recidivism Comparison

As seen in Table 2, COSA participants from the national replication sample had 83% less sexual reoffending (X2 [1] = 3.86, p < .05), 73% less violent reoffending (X2 [1] = 7.98, p < .01), and 72% less reoffending of any kind (X2 [1] = 9.73, p < .01) than the matched comparison group. Further, in looking at the actual total number of new charges incurred by the two groups (as opposed to the number of offenders who recidivated), the comparison group garnered 76% more charges (n=68) than the COSA group (n=16).

ROC (area under the curve).

Similar to the earlier study, we believed that, due to the lack of variability in scores, the STATIC-99 would lose its predictive utility when applied to the subgroup of high risk sexual offenders from this study. In fact, as found in the initial study, the ROC for the COSA group regarding STATIC-99 hit-rate was .80 for sexual recidivism. The ROC for the comparison group, however, was only .43. Similar results were also obtained in regard to violent recidivism and any new recidivism.


Prior research looking at the efficacy of the Circles of Support & Accountability model (Wilson et al., 2005) provided very encouraging results suggesting that involvement in COSA contributes to considerably less sexual and other reoffending in high-risk sexual offenders released to the community at sentence completion. The findings reported in this study underscore those earlier findings, in that the levels of reoffending in men who were involved in COSA were markedly (and, for the most part, statistically significantly) lower than for similar high-risk offenders who did not participate. The findings in both the earlier and the present study have also been echoed in the interim results of the COSA project in the Thames Valley, UK (Quaker Peace and Social Witness, 2005), where reductions in reoffending were also observed.

Contrary to the sampling difficulties noted in the earlier study, in which the COSA group exhibited significantly higher actuarial risk than the comparison group (using RRASOR), the opposite was true in this study. These differences are interpreted as being the result of imperfection in the matching process. In the current study, the groups were not different on RRASOR, but the comparison group was significantly higher on STATIC-99 (although, both groups would still be considered to be at high-moderate or above risk for reoffending). This difference in the current study serves to diminish the strength of the findings for this replication study, as some of the difference may be attributable to a somewhat lesser risk profile in the COSA participants, at least according to STATIC-99. Despite these differences, given the sheer size of the differences (i.e., 1 vs. 6, 4 vs. 15, and 5 vs. 18 for the numbers of sexual, violent, and general reoffenders), we are confident that the current findings are robust.

As was found in the earlier study examining subjects in the South-Central Ontario region, the STATIC-99 retained its predictive ability in the current COSA group but, as expected, did not do so for the comparison group. As noted in the earlier study, the higher ROC value for the COSA group suggests that recidivism in this group conforms to logical models of risk prediction, in that recidivism was positively related to higher STATIC-99 scores. The low ROC value found in the comparison group suggests that recidivism is occurring without a link between scores and outcome in this group. Ultimately, this finding suggests that the Andrews and Bonta (2003) risk principle still holds - the highest degrees of support and monitoring should be given to those offenders with the highest risk profiles. In other words, even within COSAs, additional attention should be paid to those offenders who have particularly high STATIC-99 scores.


Putting the current study in context of the greater debate as to whether or not sexual offenders can be managed in community settings, the findings here, in combination with those obtained earlier in Wilson et al. (2005) and the results from the Thames Valley project, strongly suggest that such management is possible. Community engagement into the risk management process, however, is crucial for its success. In contrast to some control practices in other jurisdictions (e.g., 1000-foot laws; coloured license plates; public internet listings) that have received little or no empirical backing in the literature (see Levenson, D’Amora, & Hern, in press), the COSA model appears to provides clear evidence that sexual offenders, particularly high-risk sexual offenders, need not be destined to fail over and over again. Silverman and Wilson (2002) suggested that a viable solution to community violence is found in community engagement with the criminal justice system. COSA is an excellent example of such community engagement helping to increase offender accountability and community safety.


Andrews, D.A. & Bonta, J. (2003). The psychology of criminal conduct (3rd Edition). Cincinnati, OH: Anderson.

Correctional Service of Canada (2002). Circles of Support and Accountability: A guide to training potential volunteers, Training manual 2002. Ottawa, ON: author.

Fernandez, Y. (2002). Phallometric testing with sexual offenders against female victims: Limits to its value. Forum on Corrections Research, 14, 7-12.

Freund, K. & Watson, R. (1991). Assessment of the sensitivity and specificity of a phallometric test: An update of "Phallometric diagnosis of pedophilia". Psychological Assessment, 3, 254-260.

Gendreau, P., Little, T., & Goggin, C. (1996). A meta-analysis of adult offender recidivism: What works! Criminology, 34, 575-607.

Hanson, R.K. (1997). The development of a brief actuarial scale for sexual offense recidivism. User Report 1997-04. Ottawa, ON: Department of the Solicitor General of Canada.

Hanson, R.K., Gordon, A., Harris, A.J.R., Marques, J.K., Murphy, W., Quinsey, V.L., & Seto, M.C. (2002). First report of the Collaborative Outcome Data Project on the effectiveness of psychological treatment for sexual offenders. Sexual Abuse: A Journal of Research and Treatment, 14, 169-194.

Hanson, R.K. & Thornton, D. (1999). STATIC-99: Improving actuarial risk assessments for sexual offenders. User Report 1999-02. Ottawa, ON: Department of the Solicitor General of Canada.

Hare, R. (2003). The Psychopathy Checklist--Revised, 2nd Edition. Toronto, ON: Multi-Health.

Levenson, J.S., D'Amora, D.A., & Hern, A. (in press). Megan's Law and its impact on community re-entry for sex offenders. Behavioral Sciences & the Law.

Nuffield, J. (1982). Parole decision-making in Canada: Research towards decision guidelines. Ottawa: Communication Division.

Quaker Peace and Social Witness (2005). Circles of Support and Accountability in the Thames Valley: The first three years April 2002 to March 2005. London, UK: Author.

Silverman, J. & Wilson, D. (2002). Innocence betrayed: Paedophilia, the media and society. Cambridge, UK: Polity Press.

Wilson, R.J., McWhinnie, A.J., Picheca, J.E., Prinzo, M., & Cortoni, F. (2007a). Circles of Support & Accountability: Engaging community volunteers in the management of high-risk sexual offenders. Howard Journal of Criminal Justice, 46, 1-15.

Wilson, R.J., Picheca, J.E., & Prinzo, M. (2005). Circles of Support & Accountability: An evaluation of the pilot project in South-Central Ontario. Research Report R-168. Ottawa, ON: Correctional Service of Canada.

Wilson, R.J., Picheca, J.E., & Prinzo, M. (2007b). Evaluating the effectiveness of professionally-facilitated volunteerism in the community-based management of high-risk sexual offenders--Effects on participants and stakeholders. Howard Journal of Criminal Justice, 46, 289-302.

Wilson, R.J., Picheca, J.E., & Prinzo, M. (in press). Evaluating the effectiveness of professionally-facilitated volunteerism in the community-based management of high-risk sexual offenders--Recidivism rates. Howard Journal of Criminal Justice.

1 Due to missing data and inconsistent reporting of scores vs. risk ratings, PCL-R scores were recoded to a 5-point scale--low, low-moderate, moderate, moderate-high, high; data were available for only a subset of each group [COSA N=18, comparison N=28].