Predictors of Treatment Completion in a Correctional Sex Offender Treatment Program

AuthorSteven R. Gray,Judith V. Becker,Daniel Krauss,Tracey M. Geer
Date01 June 2001
Published date01 June 2001
DOI10.1177/0306624X01453003
Subject MatterJournal Article
International Journal of Offender Therapy and Comparative Criminology
Predictors of Completion
Predictors of Treatment Completion
in a Correctional Sex Offender
Treatment Program
Tracey M. Geer
Judith V. Becker
Steven R. Gray
Daniel Krauss
Abstract: The purpose of this study was to examinewhat factors increase the likelihood that a
sex offender will complete a correctional sex offendertreatment program. Participants were
179 incarcerated adult male sex offenders who volunteered to participate in a sex offender
treatment program.The two groups evaluated were treatment completersand noncompleters.
By examining archival recordsthe following data were collected: scores on the Multiphasic
Sex InventoryII (MSI II), the Minnesota Multiphasic Personality Inventory (MMPI), the Abel
and Becker Adult Sexual InterestCard Sort, and plethysmograph. Demographic information
was obtained from presentence investigationreports. Results from a logistic regression indi-
catethat more years of education, not having a history of sexual victimization, and lower levels
of minimizing or excusing sexual crimes predicttreatment completion (p
suggest that it is possible to predict which incarceratedsex offenders are likely to complete a
prison-based sex offender treatment program.
Few would denythe importance of eliminating sexual crimes. It appears society is
willing to do almost anything to reduce recidivismrates including civilly commit-
ting convicted sex offenders for an indefinite period of time. Despite society’s
eagerness to reduce recidivism there is a prevalent attitude that treatment is too
costly. This attitude is based more on myth than fact. Marshall, Eccles, and
Barbaree (1993) examined the cost benefitin Canada and suggest that a mere 25%
reduction in recidivism would savesociety 4.3 million Canadian dollars a year for
treating 100 sex offenders.This is the predicted savings after the cost of treatment.
It is difficult not to favortreatment when it could have such economic benefits for
society.This economic savings does not even include the societal benefits of hav-
ing fewer sexual assault victims and fewer problems associated with such
victimizations.
Much research has been done examining treatment outcome of sexual offend-
ers, and recent reviews of such studies are providing evidence that many sex
offenders are amenable to treatment (Hall, 1995; Marshall & Pithers, 1994). An
NOTE:Address all correspondence to Tracey M. Geer, Department of Psychology, Universityof Ari-
zona, Tucson, AZ 85721; phone: 520-621-7447.
International Journal of Offender Therapy and Comparative Criminology, 45(3), 2001 302-313
2001 Sage Publications
302
important precursor to this research, however, has been largely ignored—
treatment completion. It is impossible to evaluate a particular treatment’s out-
come unless offenders complete the treatment program.
Unfortunately, sex offendertreatment dropout is common. Studies examining
voluntary outpatient sex offendertreatment programs have reported dropout rates
between 35% and 54% (Abel, Mittelman, Becker, Rathner, & Rouleau, 1988;
Miner & Dwyer, 1995). At least one study has reported that noncompletion of
treatment may be even more pervasive for incarcerated sex offenders (Shaw,
Herkov, & Greer, 1995). Shaw et al. (1995) found that about 86% of the sex
offenders admitted to a correctional treatment program were either terminated
during the evaluation stage for their unwillingness to cooperate during treatment
or discharged for incompletion of treatment modules or inappropriate behavior,
leaving only 14% who successfully completed the program.
There are few existing studies that examine treatment dropouts. These studies
have been limited to outpatient populations (e.g., Abel et al., 1988; Miner &
Dwyer, 1995). Abel et al. (1988) found three factors that differentiatedthose who
completed treatment from those who did not in a population of noninstitu-
tionalized child molesters. These factors were “(1) the amount of pressure the
subject was under to participate in treatment, (2) the diagnosis of antisocial per-
sonality,and (3) lack of discrimination in the choice of sexual victim or paraphilic
act” (p. 226). That is, those who felt they were very pressured to participate in
treatment, those who were diagnosed with antisocial personality, and the partici-
pants who were nondiscriminating in their choice of victims were more likely to
drop out.
Miner and Dwyer (1995) found both demographic factors and scores on psy-
chological assessments were effective predictors of treatment completion. Treat-
ment dropout was highly associated with the demographic factors of being single
or divorced. Those who dropped out also tended to have lowerscores on the Ten-
nessee Self-Concept Scale, lower scores on the K scale of the Minnesota Multi-
phasic Personality Inventory (MMPI), and higher scores on the L scale of the
MMPI.
Shaw et al. (1995) completed one of the few studies focusing on incarcerated
sexual offenders.Although this study examined variables that predicted treatment
outcome, not dropout, it provides valuable information about this population.
Offenders with better reading abilities and stable marital relationships had a more
favorable response to treatment. Offensetype and presence of Antisocial Person-
ality Disorder did not predict treatment outcome.
The purpose of this study was to directly explore predictors of treatment com-
pletion in an incarcerated population. Wehad the opportunity to use data collected
in a prison from 1988 to 1996. Such archival data, although not possessing rigor-
ous control, allow us to explore the likelihood that certain clinical and demo-
graphic factors affect treatment completion in a prison population. The goal of
this study was to identify factors that may predict who will successfully complete
the treatment program. Once predictors of completion status are identified it may
Predictors of Completion 303
be possible to determine who would be a good candidate for treatment. Exam-
ining treatment completers’ and noncompleters’ assessment scores prior to treat-
ment can identify these predictors. Any variables the two groups differon signifi-
cantly may be predictive of completion status.
It is important to investigatewhat factors predict treatment completion for sev-
eral reasons. First, treatment dropouts may be at higher risk for reoffense than
those who did not receive any treatment (Marques, Day, Nelson, & West, 1994).
Thus, it is important to determine what is different between those who complete
treatment and those who drop out. Decreasing treatment dropout rates would
likely lead to more offenders’ completing treatment and could subsequently
reduce recidivism rates. Effective treatment programs are useless if the partici-
pants are not completing them. A second reason for investigating who will drop
out of treatment deals with resources. As a result of financial constraints in many
correctional treatment programs, only a small percentage of those desiring treat-
ment will actually receive it (Shaw et al., 1995). Because treatment resources are
so limited, it is imperative to know who is more likely to complete the program.
Economic limitations dictate who receives treatment; therefore, any information
on how to keep offenders in effective treatment is worth investigating because of
the great costs of their crime to sexual assault victims and to society.
METHOD
PARTICIPANTS
The participants were 179 male inmates serving sentences for sexual offenses
at a correctional facility in the western United States. They voluntarily partici-
pated in assessments prior to entering the sex offender treatment program. These
participants will be referred to as completers and noncompleters. Completers are
defined as those who attended the sessions and completed the posttests (described
below) at satisfactory levels (n= 95). Noncompleters are defined as those who
were terminated from the treatment program (n= 84). Reasons for termination
included voluntary withdrawal, lack of attendance or participation, parole, and
institutional misconduct. Attendance was critical in this treatment program.
Absences due to medical or security reasons (lockdowns, riots, etc.) were
excused. Otherwise, absences were not tolerated. Institutional misconduct refers
to a number of behaviors that violate prison rules and regulations,including use of
drugs, use of alcohol, physical assault, intimidation, and sexual contact with other
inmates or visitors. In addition, the participants in the treatment program were not
allowed to possess pornography.
The two groups were comparable in age and race distribution. The mean age
for the completers was 42.15 (SD = 8.56) and 42 (SD = 9.57) for the noncomplet-
ers. There was little variability in the groups’ race composition. Approximately
304 International Journal of Offender Therapy and Comparative Criminology
77% of the completers and 72% of the noncompleters were Caucasian, 19% of the
completers and 20% of the noncompleters were Hispanic, approximately 5% of
the completers and 6% of the noncompleters were African American, and no
Native Americans completed treatment and about 3% of noncompleters were
Native American.
PROCEDURE
All participants signed consent forms prior to undergoing assessment and
entering treatment. The participants had also signed release forms stating the
information in their files could be used for research and educational purposes. The
information from the pretreatment assessments was obtained by examining archi-
val records. The following is a brief description of the treatment program.
Prior to entering treatment the participants were assessed with the following
standardized assessment instruments: the Multiphasic Sex Inventory II (MSI II),
the MMPI, the Abel and Becker Adult Sexual Interest Card Sort, the plethysmo-
graph, and clinical interviews. Following the assessments, the participants entered
the education component that required the completion of a community college
general human sexuality course and a sexualityand paraphilia class. The sexuality
and paraphilia class covered the following topics: overviewof the offense cycle,
introduction to communication skills, introduction to types of paraphilias, the
impact of early childhood experiences on adult abusive behavior, cognitive
behavioral strategies, and an introduction to general sexuality. After completion
of the education component, the participants began the relapse preventioncompo-
nent of the sex offender treatment program. This section involvedweekly individ-
ual and group therapy sessions. Individual therapy focused on decreasing arousal
to deviant stimuli. In contrast to individual therapy, group therapy focused on
examining the cycle of abuse through the use of homework assignments and
group presentations. Each member of the group developed a relapse prevention
plan that required the participant to identify high-risk situations and develop cop-
ing strategies for these situations.
The social skills training component of the treatment program involved8 hours
of intensive training centering on communication, values clarification, assertive-
ness training, vocational readiness, and anger management. Written materials and
modeling by instructors and therapists were used in this section of training.
Decreasing deviant arousal was the focus of the behavioral reconditioning com-
ponent and consisted of ammonia aversion or verbal satiation. Behavioral recon-
ditioning continued until the participant’s deviant response was decreased to less
than 10% of full arousal, and appropriate response was increased to more than
20% as measured by penile plethysmograph. The treatment program lasted
approximately 18 months. On completion of the treatment program the partici-
pants were administered a full battery of posttests that were identical to the pre-
treatment assessments described above.
Predictors of Completion 305
ASSESSMENT INSTRUMENTS
During the pretreatment assessment phase the inmates were administered a
full battery of assessments. They included paper/pencil assessments as well as
psychological and psychophysiological assessments.
MSI II. The MSI II is a standardized self-report measure of psychosexual char-
acteristics of sexual offenders. It is composed of 560 true/false items and uses 40
scales or indices. Seven scales were used in this study.According to Nichols and
Molinder (1996), these scales have high Cronbach coefficientalpha values, rang-
ing from .73 to .94, which suggests the scales are reliable. In addition, validity of
the items was demonstrated by the low intercorrelations among the items and
other demographic variables.
For the purposes of this study, the Social-Sexual Desirability, Lie, Child
Molest, Rape, CognitiveDistortions, Denier, and Justifications scales were exam-
ined. Depending on the inmate’s offense type, either the Child Molest or Rape
score was recorded. The Obsession scale did not include information relevant to
this study, and there was virtually no variability (ceiling effect) on the inmates’
scores on the Sex Knowledge and Beliefs scale; consequently these data were not
utilized.
The MSI II also contains three items regarding sexual abuse history. Partici-
pants were coded as a victim if they answered true to at least one of the following
items: “I have been raped,” “I have been molested by a male,” or “I have been
molested by a female.”
MMPI. This personality inventory contains 566 statements about oneself with
which the participant either agrees or disagrees by responding in a true/false for-
mat. Over the years, scores on the MMPI have been used to classify sexoffenders
and to predict treatment completion (Hall, Graham, & Sheperd, 1991; Miner &
Dwyer, 1995). Because Miner and Dwyer (1995) found the L, F,K, and PD scales
to be predictors of treatment dropout in outpatient sex offender populations, the
Tscores on these scales were examined in this study.
The Abel and Becker Sexual Interest Card Sort. The Abel and Becker Sexual
Interest Card Sort is a 260-item self-report measure of sexual interest. The Card
Sort includes sexual vignettes the participant reads and rates on a 7-point scale.
For the purposes of this study,only pertinent categories were examined. Excluded
categories were Voyeurism, Exhibitionism, Sadism, Masochism, Frottage, and
Transvestitism.To compare all the participants, regardless of offense type, a max-
imum deviant score and a maximum appropriate score were recorded. Spe-
cifically, the maximum deviant score was determined by choosing the highest
card sort average from the vignettes labeled as Pedophilia Female, Pedophilia
Male, Incest Female, Incest Male, and Rape. The maximum appropriate score was
306 International Journal of Offender Therapy and Comparative Criminology
determined by choosing the highest average out of either the Adult Heterosexual
or Adult Homosexual category.
Plethysmograph. Penile plethysmograph was utilized to objectively assess
sexual arousal in participants. The instruments and procedures used are described
in full detail in Gray (1995). The inmates were seated alone in a room during the
assessment. The inmates viewed slides that contained images of males and
females at varying ages. Each inmate viewed two slides from each of the follow-
ing categories: adult, adolescent, beginning pubescent, and prepubescent. The
participants were also administered plethysmographs while listening to auditory
stimuli.
Presentence investigationreport. Each inmate had a presentence investigation
report, an official report containing information about his criminal history includ-
ing the number of times he had been incarcerated and convicted as well as his
instant offense. The following demographic data were also collected from this
report: marital status, years of education, and race. Marital status was broken
down into two groups, never married and married at least once.
ANALYSES
Due to financial constraints and changes in the program over time, some of the
inmates were not administered all of the assessments. This resulted in a small
amount of missing data for some of the participants. As a result of the types of
analyses used, all the inmates needed complete data. Mean substitution was
employed after demonstrating that the data were missing at random and that the
cause for the missing data was uncorrelated with the other variables(Noll, 1996).
After mean substitution was completed, univariatetests were performed on the
22 variables. It was necessary to exclude at least 4 of the variables to ensure an
appropriate ratio between the number of predictors and the number of partici-
pants. It is recommended in regression analyses that there is only 1 predictor for
every 10 participants. This ratio results in 18 possible predictors for this sample.
Independent samples ttests were performed on all the continuous variables, and
ANOVAs were performed on the 2 categorical variables. Six variables were sig-
nificant at the .05 level and 1 approached significance (see Table 1). Because the
dependent variable was completion status (a dichotomous variable), it was not
appropriate to use regular regression techniques. Logistic regression is designed
for data with a dichotomous dependent variable (Mender, 1995).
The 6 significant variablesand the 1 approaching significance were included in
the logistic regression as well as 11 others that, despite their lack of significance
on the ttest, were theoretically important. Four variables were not included in the
regression. These variables did not produce significant tvaluesand had less theo-
retical importance than the others. Age was the first variable excluded, as the
Predictors of Completion 307
means for the two groups were virtually identical. Previousnumber of convictions
was also nonsignificant, and because it was highly correlated with previous num-
ber of incarcerations (.61), which was significant, it was not included. Due to low
variability in the race variable (nearly all were Caucasian) and the fact that the ttest
produced no significance, the race variable was excluded. As mentioned before,
there was too much missing data on the DSM diagnosis of Antisocial Personality
Disorder to examine this variable.
A hierarchical logistic regression was performed on the variables that demon-
strated statistical significance and on the other variables with theoretical impor-
tance. The variables were entered in blocks that represented a certain category of
data. The order in which the blocks were entered was not guided by a strong the-
ory because all the assessments were administered at the same point in time. The
demographic variables were entered in first with the assessment variablesfollow-
308 International Journal of Offender Therapy and Comparative Criminology
TABLE 1
MEAN DIFFERENCES BY COMPLETION STATUS:
tTEST RESULTS (N= 179)
Completer Noncompleter
Variable MMtTest
Age 42.15 8.55 42 9.57 0.11
Education 12.33 1.78 11.47 1.9 3.15**
Marital 0.78 0.42 0.71 0.45 0.99*
Victim 0.67 0.47 0.85 0.36 –2.70***
Prior incarcerations 0.54 0.81 1.08 1.79 –2.60**
Prior convictions 1.57 2.07 2.31 2.84 –2.02
Child molest/rape 19.73 9.31 18.16 8.3 1.18
Cognitive distortions 7.58 3.88 8.33 3.86 –1.29
Denier 3.4 2.84 4.9 3.35 –3.26
Lie 4.31 3.98 4.89 3.54 –1.03*
Justifications 3.55 3.01 4.58 4.03 –1.96
Social sexual desirability 26.55 4.58 24.55 5.26 2.72
L scale 52.28 11.80 54.84 8.85 –1.62*
F scale 54.73 9.51 59.05 12.40 –2.63
K scale 51.53 9.96 49.35 9.12 1.51
PD scale 66.75 10 67.23 8.52 –0.35*
Card Sort max app 6.45 0.62 6.22 0.85 2.09*
Card Sort max dev 4.02 1.89 3.82 1.84 0.70
Plethysmograph app 0.40 0.30 0.48 0.30 –1.80
Plethysmograph dev 0.50 0.31 0.53 0.29 –0.82
*ppp
ing. These demographic variables were marital status, education, and whether the
offender had been the victim of abuse. These variables were followed by the
inmate’s criminal history as defined by number of previous incarcerations. The
assessment variables were entered after the demographic variables. The type of
assessment categorized scores on various assessments. The MSI II variableswere
all grouped together, as were the MMPI scales. The Card Sort and plethys-
mograph data were combined in a block referred to as Arousal. The assessment
blocks were entered in random order. The MSI II block was the first assessment
block entered, followed by the MMPI block, and the Arousal block was entered
last. The results of the logistic regression are displayed in Table 2.
Predictors of Completion 309
TABLE 2
HIERARCHICAL LOGISTIC REGRESSION: PREDICTING COMPLETION
STATUS WITH DEMOGRAPHICS, CRIMINAL HISTORY, MSI II,
MMPI, AND AROUSAL VARIABLES (N= 179)
Step/Variable χ2βOdds Ratio
I. Demographic variables 17.34***
Education .21* 1.23
Marital .18 1.20
Victim –.86* 0.13
II. Criminal history 5.58*
Incarcerations –.34* 0.71
III. MSI II variables 11.50
Child molest/rape .01 1.01
Cognitive distortions .01 1.01
Denier –.16* 0.85
Lie .05 1.06
Justifications .04 1.04
Social sexual desirability .05 1.05
IV. MMPI variables 3.72
L scale –.02 0.98
F scale –.03 0.97
K scale .01 1.01
PD scale .02 1.02
V. Arousal variables 6.63
Card Sort max app .24 1.27
Card Sort max dev .12 1.13
Plethysmograph app –1.19 0.31
Plethysmograph dev –.22 0.80
NOTE:MSI II = Multiphasic Sex Inventory II; MMPI = Minnesota Multiphasic Personality Inventory.
*pp
RESULTS
The demographic variables were significant in predicting completion status,
χ2(3, N= 179) = 17.34, pχ2(1, N=
179) = 5.58, pχ2(6, N= 179) =
11.50, p
(see Table 2).
In the overall model, there were four significant variables: (a) years of educa-
tion (greater number of years completed predicting completion; p
(b) whether the participant was a victim of sexual abuse, with those without a his-
tory of victimization being more likely to complete (p= .05); (c) previous number
of incarcerations, with those with fewer incarcerations more likely to complete (p
excuse their behavior more likely to complete (p
DISCUSSION
The men in the two groups differed on their pretreatment assessments. This
suggests that it may be possible to predict which individuals will be more likely to
complete a correctional sex offender treatment program. Analyses on both demo-
graphic information and scores on assessments resulted in significant differences
between the two groups (see Table2). These differences can provide information
as to why some of the participants did not complete the treatment program.
Inmates who completed the treatment program tended to have more years of
education than the noncompleters. This effect may be a result of the cognitive
nature of the treatment program. This treatment program was verycognitively ori-
ented, involved reading a great deal of educational materials, and utilized work-
books. The education component, in particular, with its community college class
may have been too challenging, frustrating, or confusing for the individuals with
less education. These feelings may have led the participants to drop out. These
findings are consistent with the literature on treatment outcome of incarcerated
sex offenders (Shaw et al., 1995).
Those who claimed to be victims of sexual abuse were less likely to complete
the treatment program. There are several possible hypotheses as to whythis effect
occurred. It could be due to the detrimental effects childhood sexual abuse has on
an individual’s psychosexual development.Experiencing sexual abuse can some-
times make it more difficult for the individual to developappropriate sexual rela-
tionships. This hypothesis is supported by the data indicating that those who did
not complete treatment rated as less arousing on the Abel and Becker Adult Sex-
ual Interest Card Sort the vignettes that described peer consensual sexual interac-
tions. The noncompleters also showed less arousal to sexual stimuli that assessed
peer, consensual sexual interactions as measured by the plethysmograph (see
Table 1 for means).
310 International Journal of Offender Therapy and Comparative Criminology
Another potential explanation for this phenomenon is that those who had been
victimized may be more likely to justify their own inappropriate behavior. Vic-
tims sometimes believe that “if it wasokay for someone to do this to me, it is okay
for me to do it to someone else.” This attribution would make it more difficultfor
the individual to stay in a treatment program that holds them responsible for their
actions.
The final hypothesis for the reported victimization effect may be a direct result
of the treatment program. This treatment program focuses on victimization
issues. Those men with a history of being victimized during childhood or adoles-
cence may havefound it upsetting to be reminded of their own victimization expe-
riences. Their termination of treatment may have been a way to avoid the painful
memories of their own abuse.
The criminal history of the two groups also differed with completers’ having
fewer previous incarcerations than noncompleters. This could be related to
self-control issues. One theory of criminology suggests that people who commit
criminal acts suffer from low self-control (Gottfredson & Hirschi, 1990). Those
with a greater number of incarcerations may have lower self-control and conse-
quently may have a more difficult time completing a program that requires com-
mitment and responsibility.
The two groups also differ in their minimization of their offense. The Denier
score on the MSI II measures the degree to which an individual acknowledges a
sexual offense has occurred, but minimizes and excuses the offense. A higher
score represents a higher proclivity to minimize and excuse.On average, complet-
ers had lower scores on this scale whereas noncompleters had higher scores.
Because this treatment program focused on eliminating minimizations and
excuses, it may have caused cognitive dissonance in some of the deniers. This
cognitive dissonance probably leads to a lower completion rate for the deniers.
There was little support for the use of MMPI scales in predicting treatment
dropout. Contrary to the findings of Miner and Dwyer (1995), the L and PD scales
achieved significance on the ttest, but they did not withstand the logistic regres-
sion.
Although these findings are intriguing, these results should only be general-
ized to comparable populations. This was an incarcerated population, and their
participation was voluntary. In addition, there were some incentives to stay in
treatment (e.g., lowering of sex offender score). In other programs in which par-
ticipation is required, different predictors may be discovered.
It is also important to keep in mind the heterogeneity of the noncompleter
group. The reasons for termination were various, and it is possible that reasons for
dropout were different for each subgroup. It would be worthwhile to examine a
similar treatment population where the reasons for treatment dropout were not so
diverse.
To determine the valueof these predictors, it is important to discover the effi-
cacy of this treatment program. Predictors of treatment dropout are important
only if the treatment benefits the participants by lowering recidivism rates.
Predictors of Completion 311
It is too early to consider the recidivism rates of these inmates. Some of the men
are still incarcerated, and those who have been released have not been out of
prison long. A substantial follow-up period must pass before recidivism data are
collected to answer this question. A recent meta-analysis by Hall (1995) suggests
that it is optimal to have at least a 5-year follow-up period in sex offender treat-
ment outcome studies. Marshall and Barbaree (1988) have also written about the
importance of a sufficient follow-upperiod for any sexual offender population. A
long follow-up period is essential with this population because evenafter they are
released from prison, many participants are on a strict parole and are closely mon-
itored. This close supervision could reduce the recidivism rates of both treatment
completers and noncompleters. Future research will investigate recidivism rates
once an appropriate follow-up period has passed.
These findings have at least two specific implications for treatment providers
who are attempting to reduce the dropout rate in these types of treatment pro-
grams. First, consideration should be given to encouraging men who are thinking
about entering a cognitive behavioral treatment program to obtain their General
Equivalency Diploma (GED) or further their education in some other way while
in prison, because this may decrease their likelihood of dropping out given that
those with more education were more likely to complete the program. Second,
because men with a history of sexual abusewere more likely to drop out, consider-
ation should be given to adding a component at the beginning of treatment that
deals with the offenders’ own trauma issues. Treatmentproviders can implement
these changes and determine whether it does improve completion rates.
REFERENCES
Abel, G. G., Mittelman, M., BeckerJ. V., Rathner,J., & Rouleau, J. (1988). Predicting child molesters’
response to treatment. In R. A. Prentky,& V. L. Quinsey (Eds.), Human sexual aggression: Cur-
rent perspectives (pp. 223-234). New York: New York Academy of Science.
Gottfredson,M. R., & Hirschi, T. (1990). A general theory of crime. Stanford, CA: Stanford University
Press.
Gray,S. R. (1995). A comparison of verbal satiation and minimal arousal conditioning to reduce devi-
ant arousal in the laboratory. Sexual Abuse: Journal of Research and Treatment,7(2), 143-153.
Hall, G.C.N. (1995). Sexual offender recidivism revisited:A meta-analysis of recent treatment stud-
ies. Journal of Consulting and Clinical Psychology,63, 802-809.
Hall, G.C.N., Graham, J. R., & Shepherd, J. B. (1991). Three methods of developing MMPI taxono-
mies of sex offenders. Journal of Personality Assessment,56(10), 2-13.
Marques, J., Day,D., Nelson, C., & West, M. (1994). Effects of cognitive-behavioraltreatment on sex
offender recidivism. Criminal Justice and Behavior,21(1), 28-54.
Marshall, W. L., & Barbaree, H. E. (1988). The long-term evaluationof a behavioral treatment pro-
gram for child molesters. Behaviour Research and Therapy,25, 499-511.
Marshall, W.L., Eccles, A., & Barbaree, H. E. (1993). A three-tiered approach to the rehabilitation of
incarcerated sex offenders. Behavioral Sciences and the Law,11, 441-455.
Marshall, W.L., & Pithers, W. D. (1994). A reconsideration of treatment outcome with sex offenders.
Criminal Justice and Behavior,2(1), 10-27.
Mender, S. (1995). Applied logistic regression analysis. Thousand Oaks, CA: Sage.
312 International Journal of Offender Therapy and Comparative Criminology
Miner,M. H., & Dwyer, M. (1995). Analysis of dropouts from outpatient sex offender treatment. Jour-
nal of Psychology & Human Sexuality,7(3), 77-93.
Nichols, H. R., & Molinder, I. (1996). Multiphasic sexual inventory II handbook. Tacoma, WA:
Nichols & Molinder Associates.
Noll, J. G. (1996). Missing data by design: The good news about missing data. The American Profes-
sional Society on the Abuse of Children,9(3), 21-24.
Shaw,T. A., Herkov,M. J., & Greer, R. A. (1995). Examination of treatment completion and predicted
outcome among incarcerated sexoffenders. Bulletin of the American Academy of Psychiatry & the
Law,23(1), 35-41.
Tracey M. Geer
Department of Psychology
University of Arizona
Tucson, Arizona 85721
USA
Judith V. Becker
Department of Psychology
University of Arizona
Tucson, Arizona 85721
USA
Steven R. Gray
Psychological and Consulting Services
Mesa, AZ 85210
USA
Daniel Krauss
Claremont McKenna College
Claremont, CA 91711
USA
Predictors of Completion 313

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT