Risk Assessment for Sexual Offender Recidivism
(i) number of prior sex offenses,
(ii) the offender's current age,
(iii) the gender of the offender's victims, and
(iv) the offender's familial relationship to the victim.
RRASOR yields a score ranging from 0 to 5. The corresponding recidivism rates,
over a 10 year window of opportunity, range from 6.5 to 73.1%. Hanson (1997)
clearly acknowledges there are limitations associated with the RRASOR:
The RRASOR was not intended to provide a comprehensive assessment of all the factors relevant to the prediction of sexual offender recidivism. Instead, the RRASOR should be used only to screen offenders into relative risk levels. These risk levels could then be adjusted by the consideration of other relevant information such as deviant sexual preferences and treatment compliance ( p. 19).
other words, Hanson acknowledges that the RRASOR is still at a preliminary stage
of development; and, as a result, it also amounts to an experimental procedure.
Hanson's recommendations regarding assessment of deviant sexual preferences, and
treatment compliance, create their own problems. In particular, how does a
psychologist undertaking a sexual predator evaluation reliably assess
"deviant sexual preferences"? The number of psychologists trained to
make phallometric assessments is quite limited. Too often, then, the determination
of deviant sexual preference hinges on clinical judgment. There are
psychologists making sexual predator evaluations who assume that all previously
convicted sexual offenders, as a result of their prior conviction(s), exhibit
deviant sexual preferences. Assumptions such as these obviously interfere with
attempts at discriminating between high-risk and low-risk subsets of a
previously convicted offender population.
meta-analysis by Hanson and Bussière (1998) found that the only examined
treatment factor related to recidivism is whether the offender completed a
course of treatment. "Treatment compliance" therefore is a misnomer.
As reported by Hanson and Bussière (1998), considerations such as motivation
for treatment, length of treatment, denial and victim empathy are not re1ated to
recidivism. The only treatment factor to be considered is whether the offender
completed treatment if given the opportunity to do so. Consequently, the
appropriate term for this issue is "treatment completion." The term
"treatment compliance" invites
to again rely on their UC] in assessing this issue. Given the limited resources
of many correctional agencies, convicted offenders often request treatment and
find it unavailable. Classifying an offender in these circumstances as failing
to complete treatment is obviously ill advised.
and Meehl (1997) identified gross disparities between the levels of sensitivity
and specificity realized by the RRASOR. Using the maximum cutoff score of 5,
they reported the specificity of the RRASOR is a near perfect .97. Its
sensitivity, however, is only .15. In other words relying on this cutoff score
would only identify 15% of recidivists. The remaining 85% would be mistakenly
classified as non-recidivists. Janus and Meehl also questioned whether the
RRASOR is even applicable to sexual predator hearings. They reported that the
base rate for Hanson's population of offenders was 41%. Janus and Meehl
indicated that sexual predator evaluations attempt to assess risk for groups in
which the base rate of recidivism may be substantially lower .
The issue of recidivism base rates for previously convicted sexual offenders is controversial to say the least. ]anus and Meehl (1997) estimate the base rate of recidivism for this population at between 20 and 45%. It can be argued that a range of this magnitude ultimately leaves us with the conclusion that the recidivism rate for previously convicted sexual offenders remains unknown.
Lee, Knight, and Cerce (1997) reported data estimating recidivism rates for
rapists as high as 39%, and rates for child molesters as high as 52%. These
data, however, were obtained from high-risk populations. The rapist group had an
average of 3.51 offenses prior to their release, and the child molester group
had an average of 4.59 offenses prior to their release. Citing the data of
Prentky et al. (1997), Doren (1998) nonetheless claimed "... we can
scientifically conclude that a significant set of offenders will likely be
dangerous again" (p.111).
the final analysis, the time has come to cease speculative arguments, premised
on data obtained from atypical populations, regarding the base rate of recidivism
for previously convicted sexual offenders. The six-step validity-reliability
study previously outlined would also clearly define the recidivism base rate for
the population studied. Rather than debate the relevance of currently
available recidivism estimates, based on different populations with varying
windows of opportunity, it would be more appropriate to undertake the necessary
research to resolve this issue with empirical data.
This evaluation of the RRASOR results in rather grim conclusions for those who would rely on it in a sexual predator hearing. Hanson himself acknowledges that the RRASOR is not a comprehensive method for assessing recidivism risk in cases of previously convicted sexual offenders. Except for the estimates of Janus and Meehl, there are no other data identifying the
(i) levels of sensitivity,
(ii) levels of specifity,
(iii) frequencies of false positives and
(iv) frequencies of a false negatives
associated with the RRASOR.
necessary study In accordance with Testing Stanard 1.1 has not been done for the
RRASOR. Despite Testing Standard 5.1, there is no generally available manual for
the RRASOR detailing its proper use.
Regarding Testing Standard 6.1, and Ethical Standard 2.04(a), there are no reliability or validity data reported in a peer-reviewed journal for the RRASOR.
considerations lead to the conclusion that the RRASOR is also an experimental
procedure that cannot support expert testimony in a legal proceeding.
association with the Minnesota Department of Corrections, Epperson, Kaul, and
Huot (1995) have developed the Minnesota Sex Offender Screening Tool (MnSOST).
This is an actuarial instrument with 21 items addressing sexual and nonsexual
offense history, substance abuse history, and treatment compliance. Hanson
(1998) reported that the original MnSOST had undergone revision, further
indicating that many of its items did not significantly correlate with
recidivism. The MnSOST -R is a 17-item instrument that has not yet been
validated on a new sample (Hanson, 1998).
is not surprising that many MnSOST items failed to correlate significantly with
sexual offender recidivism. The data of Hanson and Bussière (1998) demonstrated
that nonsexual offense history, substance abuse history, and treatment
compliance did not correlate significantly with recidivism. Similar to the
RRASOR, Janus and Meehl (1997) report substantial differences in sensitivity
(less than 20%) and specificity (above 90%) for the MnSOST.
is no generally available manual (as previously defined) for the MnSOST or
MnSOST -R. Though this instrument has been discussed in various peer-reviewed
publications, it has not undergone peer review itself. Indeed, the xeroxed
copies of the MnSOST -R seen by this author are designated: "SEX OFFENDER
SCREENING TOOL -- Second Research Edition" (emphasis added).
Sensitivity and specificity estimates (Janus & Meehl, 1997) apply to
the original MnSOST. The levels of sensitivity and specificity, and the
frequencies of false positive and false negative classifications, remain unknown
for the MnSOST -R. Similarly, inter-rater reliabilities have not been identified
for either the MnSOST or MnSOST -R.
obviously, then, the MnSOST -R (and to a similar extent the MnSOST) fail to
satisfy the obligations defined by Ethical Standards 2.02(a) and 2.04(a). The
MnSOST -R also fails to satisfy ethical obligations defined by Testing Standards
1.1, 2.8, 5.1, and 6.1. In view of these glaring shortcomings associated with
the MnSOST and MnSOST -R, they also amount to experimental procedures; and,
therefore, cannot support expert testimony in a legal proceeding.
The SORAG has been developed by Quinsey and his colleagues based on their many years of work in the Ontario correctional system (Quinsey et al., 1998). The SORAG is a 14-item actuarial instrument for assessing the recidivism risk of previously convicted sexual offenders. There is little if any empirical support, however, for at least five of the SORAG items:
history of alcohol problems,
history of nonviolent offenses,
meeting DSM-III criteria for Personality Disorder, and
DSM-III criteria for schizophrenia (coded as reducing recidivism risk).
In their massive meta-analysis, Hanson and Bussière (1998) found no significant correlation between alcohol abuse problems and sexual offense recidivism. Similarly, neither marital status nor nonviolent offenses correlated significantly
recidivism. Additionally, Hanson and Bussière found no significant correlation
between "any personality disorder" and recidivism. In direct contrast
to the SORAG item regarding schizophrenia, Hanson and Bussière (1998) found
that "psychological maladjustment" rated as "severely
disordered" did not correlate significantly with recidivism. Quite clearly,
then, the SORAG relies on some risk factors for which there is little, if any,
SORAG also includes Psychopathy Checklist-Revised (PCL-R) (Hare, 1991) scores in
its overall assessment. Though it is more an objective procedure than not, there
are elements of clinical judgment involved in using the PCL-R. Consequently,
appropriate usage of the PCL-R appears to require specific training (Wakefield
& Underwager, 1998). Psychologists who use the SORAG without specific PCL-R
training are at risk for violating Ethical Standards 1.04(a) and 1.04(b)
addressing "boundaries of competence:"
Psychologists provide services, teach, and conduct research only within the
boundaries of their competence, based on their education, training,
supervised experience, or appropriate professional experience.
Psychologists provide services, teach, or conduct research in new areas or
involving new techniques only after first undertaking appropriate study,
training, supervision, and/or consultation from persons who are competent in
those areas or techniques
Efforts at evaluating the validity and reliability of the PCL-R have been confined almost entirely to Canadian prison populations (Rogers, Duncan, Lynett, & Sewell, 1994). Using the PCL-R with US minority groups, therefore, is especially problematic.
A 1990 study, for example, found substantial differences between African-Americans and Anglo-Americans in:
(i) the distribution of psychopathy scores,
(ii) the relation of psychopathy to measures of impulsiveness, and
the congruence of the underlying factor structure of the PCL-R across these
two ethnic groups
well informed opinion regarding the PCL-R has emphasized: "... we do not
believe the existing validity data justify the use of the PCL-R with minority
groups" (Salekin, Rogers, & Sewell, 1996, p. 208). This is a
particularly sobering consideration in view of the obligations related to
Ethical Standard 2.04(c):
Psychologists attempt to identify situation:, in which particular interventions or assessment techniques or norms may not be applicable or may require adjustment in administration or interpretation because of factors such as individuals' gender, age, race, ethnicity, national origin, religion, sexual orientation, disability, language, or socioeconomic status (American Psychological Association, 1992, p. 1603).
The SORAG also involves potential problems of inter-rater reliability. The item "elementary school maladjustment," for example, is scored in the following manner:
--1, "no problems;"
+2, "slight (minor discipline or attendance) or moderate problems;"
"severe problems (frequent disruptive behavior and/or attendance or
behavior resulting in expulsion or serious suspensions"
between "slight" and "severe" elementary school
maladjustment invites a great deal of subjective opinion. In particular, at what
point do problems escalate from "moderate," requiring "minor
necessitating a "serious suspension"?
More importantly, to what extent would two or more psychologists independently evaluating the same offender reach consistent conclusions regarding this item?
SORAG item related to "history of alcohol problems" further presents
problems of inter-rater reliability. The criteria for this item state:
One point is allotted for each of the following:
teenage alcohol problem,
adult alcohol problem,
alcohol involved in a prior offense,
alcohol involved in the index offense.
0 = -1, 1 or 2 = 0, 3 = +1, 4 or 5 = +2
(Quinsey et al., 1998, p.241).
These criteria obviously raise questions regarding the extent to which two or more psychologists can use them consistently when evaluating the same offender. Ascertaining whether an offender's history involves parental alcoholism, a teenage alcohol problem, or an adult alcohol problem can provoke a wide range of subjective opinions. Similarly, whether or not alcohol was involved in a prior offense, or the index offense, moreover encourages the kind of conjecture that can vary enormously between clinicians.
considering issues of DSM-III criteria for personality
disorder, it is also necessary to remember that the inter-rater reliabilities
for that diagnostic class fell short of the levels recommended by DSM-III
(American Psychiatric Association, 1980, p.470).
In other words, "personality disorders" can be considered an inherently unreliable diagnostic classification per DSM-III criteria (Campbell, 1999).
SORAG relies on phallometric assessment to determine "deviant sexual
preferences." As previously pointed out, however, very few psychologists
have been trained to do phallometric assessments. This consideration then raises
the question of how a psychologist using the SORAG assesses deviant sexual
preference without the availability of phallometric data. Relying on UCJ, or
other indices of deviant sexual preference, is inconsistent with the
standardized procedure for the SORAG. In other words, deviations from
standardized procedure mean that the SORAG no longer qualifies as a standardized
et al. (1998) reported recidivism probabilities corresponding to the
SORAG for 7 and 10 year windows of opportunity. They moreover reported
percentile scores for the entire range of SORAG raw scores. Nevertheless, they
did not clearly identify the sample characteristics of their offender
population. They specifically neglected to report the size of their population,
and other important characteristics such as mean age, mean number of prior
sexual offenses, and number of child molesters versus offenders with adult
victims. Without information regarding the SORAG's normative population,
psychologists using it cannot know whether it is applicable to the offender they
Finally, it is not particularly clear as to what kind of recidivism the SORAG assesses. The table reporting recidivism probabilities for the SORAG states:
"Probability of violent recidivism at two different mean lengths of opportunity as a function of nine equal-sized Sex Offender Risk Appraisal Guide (SORAG) categories" (Quinsey et al., 1998, p.244).
description suggests that the recidivism risk assessed by the SORAG is general
recidivism, or the likelihood of a previously convicted sexual offender
committing any kind of violent offense subsequent to release from incarceration.
Sexual predator statutes, however , typically call for assessing an offender's
future risk for committing violent acts that
specifically sexual. If the SORAG assesses general recidivism risk as it
appears to do, then it likely is not applicable to sexual predator hearings.
Despite its impressive potential, carefully evaluating the SORAG clearly demonstrates it cannot support expert testimony in legal proceedings. The SORAG's levels of sensitivity and specificity, and its frequencies of false positives and false negatives, remain unknown. There are no data available regarding inter-rater reliabilities of the SORAG. Moreover, there is no generally available manual for the SORAG. Additionally, the SORAG has yet to undergo peer review. Not surprisingly, then, Quinsey et al. (1998) indicate: "We are pursuing further testing of the SORAG in our current research" (p. 157).
clearly, there is a compelling need for further testing of the SORAG.
Consequently, it presently amounts to an experimental instrument in need of
further development and revision.