- What Population and What Question?
In 1984, Vernon Quinsey commented that, "by selectively contemplating the various studies, one can conclude anything one wants" about sex offenders' recidivism rates (101). Unfortunately, this observation remains valid two decades later. Ron Langevin, Suzanne Curnoe, Paul Fedoroff, Renee Bennett, Mara Langevin, Cheryl Peever, Rick Pettica, and Shameen Sandhu (2004) draw no general conclusions about specific rates of sex offender recidivism; their discussion is largely given over to methodological issues and the specifics of their own study. It also strikes us that the disapproval of Cheryl Webster, Rosemary Gartner, and Anthony Doob (2006) stems from ideological and pragmatic implications they believe would follow from the general realization that sex offenders exhibit high rates of recidivism, and not from the Langevin et al. study's shortcomings alone.
Some severe criticism by Webster et al. is, therefore, levelled at the failure of Langevin et al. to say enough about the sample, leaving readers unable to assess its representativeness. Representativeness is a thorny issue that depends on the population of interest. Some readers focused on epidemiological concerns might be interested in the reoffense rates for all perpetrators of sexual assault and molestation in the Canadian population as a whole. Others, interested in criminal justice policy around sentencing and conditional release, might be interested in the rates of reoffending among rapists and child molesters released from federal corrections. Still others, professionals conducting risk assessments perhaps, might be interested in subsequent violent behaviour only among male sexual assaulters assessed by forensic clinicians at their own facility. Many possibilities exist here, and we believe that, for many consumers of the literature (though certainly not all), the population of interest is adjudicated adult male rapists and child molesters1 released (or considered for release) from secure custody. This is also the population for which the empirical literature is the largest. Clearly, research designs depend on the populations of interest – population surveys might be appropriate for the first question above (although we doubt there is any methodology [End Page 95] suitable to answer this question), while multi-site, cross-agency follow-ups are required for the second.
The Langevin et al. study provides data about the third question, and Webster et al. are correct to point out that the findings cannot be assumed to inform other questions (although they seem unclear about this themselves: contrast "individuals who committed sex offences" with "all known sex offenders"; emphasis added). Regarding such populations, no one can claim a methodologically adequate follow-up study of representative sex offenders in any jurisdiction. Even the meta-analyses by R. Karl Hanson and Monique Bussière (1998) and by Hanson and Kelly Morton-Bourgon (2004) may not have comprised representative samples of sex offenders. For example, these meta-analyses primarily dealt with child molesters (incest or extra-familial), whereas, as Webster et al. state, most sex offenders charged in Canada have assaulted women.
It is important not to overstate the case, however, especially in concluding that the Langevin et al. results overestimate recidivism. There are empirical findings that do help evaluate the importance of several possible sources of methodological bias. Webster et al. point out that Langevin et al.'s subjects mostly came from Ontario, as if to imply that population rates of sexual aggression are lower in other provinces. There is no reason to think so. Webster et al. also point out that the Langevin et al. sample had a proportion of "incest offenders" higher than that in the sex offender population (whatever that may be). We note that Webster et al. cite the proportion of offenders charged with incest, whereas Langevin et al. define incest (as most researchers and clinicians do) as sexual contact with a child in the same family. Nevertheless, if the Langevin et al. sample did include a higher proportion of incest offenders than the population of interest, this would make their estimated rates of recidivism conservative. Webster et al. note that the Langevin sample had been referred for psychiatric assessment and imply, entirely without evidence, that this means it represents higher-than-average risk. The available evidence indicates that sex offenders referred for psychiatric assessment do not differ...