Académique Documents
Professionnel Documents
Culture Documents
This paper is based on the doctoral dissertation of the first author. The research was supported,
in part, by funding from Correctional Service Canada. The authors would like to thank the women
who agreed to participate in this study. The authors would also like to thank Dr. Sharon Kennedy
for her helpful comments on an earlier draft of this paper.
Eastern & Northern Parole District Office, City Place II, 473 Counter Street. Kingston, Ontario,
Canada K7M 8Z6.
Department of Psychology, Kingston, Ontario, Canada K7L 3N6.
SUMMARY
This study involved 100 adult female federal offenders housed within the
multilevel security Prison for Women in Kingston, Ontario. Data were collected
through structured interviews, file reviews, and self-report tests on a wide variety
of variables, classified into four categories: social, personal, and criminal history;
history of maladaptive behavior (including drug and alcohol use); history of
abuse (physical, sexual, and psychological); and current personality, ability, and
emotional functioning.
Descriptive results indicated that the study group was representative of the
population at Prison for Women, and of the population of serious female
offenders in Canadian prisons. Also, there were a number of interesting
similarities, and some differences, on many of these variables when the study
group was compared to a sample of serious male offenders.
We examined the relationship of all variables with previous offending, previous
violent offending, prison misconduct, and violent prison misconduct during the
current sentence, using multiple regression analysis. This method allows one to
determine the independent contribution of each variable to the prediction of the
outcome variables, taking the contribution of all other variables into account.
This portion of the study was postdictive. Psychopathy was the preeminent
variable in the prediction of all outcome variables. Other variables that entered
into the regression functions differed across the outcome measures. These
included measures from each of our categories, although measures of
personality and current functioning were generally the strongest, and indices of
maladaptation were the weakest.
Releases and re-admissions from the original sample were examined
approximately five years after initial data collection, with an average follow-up
time of 38 months. This provided a test of how well our measures performed in
a truly predictive fashion. Forty-seven percent of those released had been
returned to prison, prior to their warrant expiry date, for conviction on a new
2
charge or for a major violation of their release conditions. For this portion of the
study, previous convictions was used as a predictor variable. In examining the
predictive contribution of all variables together, psychopathy, previous criminal
convictions, and substance abuse by the father made significant contributions.
Psychopathy remained the most important predictive variable.
A survival analysis was completed using these three variables as predictors.
The resultant predictor function had a total correlation of .32 with our recidivism
variable. If the released group was split by taking the median value of this
survival function, one would estimate (by visual inspection of the survival curve)
that approximately 5 of 6 subjects above the median would still be in the
community three years after release, while about 4 of 6 subjects below the
median would have returned to prison. The proportion of correct predictions was
found to be in the high end of the range characteristically found for conventional
actuarial scales.
Our analyses support the position that there are considerable similarities in the
factors that help to predict recidivism in serious offenders regardless of gender,
and that, by and large, the differences between genders are not predictive of
continued criminal behavior.
restricted and arbitrarily chosen sets of variables. The majority of studies have
assessed the effects of measures taken in isolation, and there has been little
attempt at integration or comparison of different classes of measures. Moreover,
the applicability of much of the data to serious repeat offenders is limited,
because most of the published research sampled from a population that included
minor or casual offenders.
The purpose of this study was to look at a comprehensive set of predictive
factors in a population of the most serious female offenders. Several outcome
measures were used, including disciplinary infractions in prison, criminal records,
and recidivism after release. A variety of predictor measures taken from different
theoretical approaches was assessed against these various outcomes, to allow
comparisons of relative strength and to show relationships among them. In this
way, it may be possible to make some general conclusions about the most
fruitful approach to explaining serious female criminality, and to address the
similarities and differences between male and female offenders.
METHOD
Subject Recruitment
Subjects were recruited over an 18-month period during 1989 and 1990, within
the sole (at that time) Canadian federal facility for adult female offenders, the
multilevel security Prison for Women in Kingston, Ontario. This facility held the
majority of female offenders in the country who had received sentences of at
least two years. In a minority of cases, women who were a federal responsibility
were kept in provincial institutions under cost-sharing arrangements, but almost
all offenders classified as medium or maximum-security were housed in the
Prison for Women.
The aim was to recruit as large a proportion of the population as possible, so that
the results would be generally representative of the women in the institution.
Subjects were paid a sum of ten dollars for participation, deposited into their
institutional account. Participants were asked to sign an informed consent
document stating that participation would in no way affect decisions about them
by correctional authorities.
Because the study included a number of written inventories, one potential
participant was excluded because of limited literacy. Over the recruitment
period, 120 women consented to participate. During that period, approximately
270 women were at some time housed at Prison for Women, at a satellite
minimum-security unit, or at a satellite behavioral treatment unit. However,
minimum time periods of three months in the institution before and after testing
were required for collection of institutional outcome data, and this disqualified
most women who were scheduled to leave near the beginning of the study or
who arrived near the end. Although an exact figure on participation rate is
difficult to ascertain, we estimate that 70% of the eligible population did agree to
participate. (The first author had worked in the institution as a psychologist for
several years, and by all indications he had earned the trust of the inmate
10
11
13
RESULTS
Representativeness
The first question to be addressed is the extent to which the present sample is
representative of Canadian federally-sentenced women in general, and of the
Prison for Women population in particular. Descriptive data for the study sample
were compared to those of both the institutional population and the population of
federal female offenders in Canada (Shaw, 1991). Data summarized in the
upper part of Table 1 indicate that this sample provides a good match of offence
types to those for the population at Prison for Women and the general population
of federal female offenders. The distribution of sentence lengths is also a
reasonable approximation, although there is an under-representation of longterm offenders compared to the entire institution.
On other measures where authoritative figures were not available but informal
estimates could be made from unpublished figures, e.g., age, the fit appears to
be quite good with values for the institutional population from which the sample
was taken. In general, from the data on offences and criminal history one can
see that this sample had a substantial number of serious and repeat offenders,
consistent with the reputation and stated role of the Prison for Women. Thus,
we conclude that the women in this study were representative of the institutional
population, and also generally representative of serious female offenders in
Canadian prisons.
14
Table 1:
Variable
Current
Sample
Prison for
Women
Population
Male
Samplea
Federally Sentenced
Femalesb
54.0
31.0
15.0
50.0
20.0
30.0
----
----
37.0
12.0
13.0
4.0
14.0
19.0
1.0
36.0
12.0
13.0
1.0
15.0
21.0
2.0
36.0
12.0
13.0
1.0
15.0
21.0
2.0
40.4
11.8
16.8
-14.8
11.8
4.4
71.0
--
72.0
--
22.0
37.0
---
13.9
54.7
---
4.7
--
--
--
10.9
--
12.6
--
82.0
18.0
---
90.0
10.0
---
74.0
26.0
---
81.0
19.0
---
71.0
29.0
---
---
---
15
16.1
-15.6
Mean age left school
Substance abuse problem in
family
Father
48.0
--Mother
27.0
--Siblings
39.0
--Spouse
50.0
--Criminal record in family
Father
18.0
--Mother
9.0
--Siblings
38.0
--Spouse
38.0
--Note. Entries are percentages, unless otherwise noted. Dashes indicate
data were not available
a
male data from Zamble & Porporino (1988)
b
federally-sentenced female data from Shaw (1991)
Exposure to Abuse
During their years of living at home, 54% of the women reported being exposed
to at least one type of "extreme" physical abuse (viz., being beaten, burned,
scalded, or threatened or attacked with a knife or gun) on at least one occasion,
and 49% were exposed more than once. For comparison, Bonta et al. (1995)
reported 14% of their sample of federally sentenced women had experienced
16
--
---------
physical abuse prior to age 12, and 61% had experienced physical abuse during
their lifetime. In Canadian samples of male offenders, Dutton & Hart (1992)
reported 31% had experienced childhood physical abuse, and Robinson & Taylor
(1994) reported a rate of 35%.
Eighty-one percent of the study sample reported experiencing at least one
incident of some level of physical abuse in their lifetime. Similarly, Lightfoot &
Lambert (1992) reported a physical abuse exposure rate of 81% in their Prison
for Women sample. Gal et al., (1992) refer to the work of Das (1990), who
reviewed available studies and estimated a 25% rate for physical abuse, for both
males and females in the general population; clearly, levels in the present
sample were much higher. Exposure to psychological abuse (viz., use of insults,
swearing, spiteful behavior) from a parent or guardian, as defined by five Conflict
Tactics Scale items, was universal.
Similarly, 80% of subjects reported having experienced at least one incident of
sexual abuse during their lifetime, consistent with a rate of 76% reported by
Lightfoot & Lambert (1992) in a sample from the same institution. In another
sample of federally sentenced women, Bonta et al. (1995) reported 54% had
experienced sexual abuse during their lifetime, and 13% had experienced
childhood sexual abuse. Robinson & Taylor (1994) reported a rate of 13% in
their male offender sample.
In the literature concerning the general population, estimates of sexual abuse
among females have ranged from 6% to 62% (Saunders et al., 1992), depending
on how sexual assault was defined, age limitations placed on respondents, the
population examined, and the screening methods used to detect sexual assault.
Sexual abuse exposure rates in the present sample therefore exceed even the
highest estimates reported in the general population. Thirty-eight percent of the
sample had intercourse forced upon them prior to age 13, and sixty-two percent
of the sample had experienced some form of sexual abuse pre-adolescence.
Sixty percent of the sample had intercourse forced upon them after age 13, and
17
seventy-two percent of the sample had experienced some form of sexual abuse
post-adolescence. Sexual abuse reported by subjects generally occurred on
more than one occasion, and over half the sample had experienced sexual
abuse in both pre- and post-adolescence. For both pre- and post-adolescence,
over 20% of the sample reported that sexual abuse occurred at a frequency of
"too many times to count".
Maladaptation History
There was a high rate of involvement in mental health treatment, even when
treatment during previous incarcerations was not considered. Sixty-eight percent
of the women had been involved in mental health treatment during the current
incarceration. This finding is consistent with that of Shaw (1991) in her study of
the Prison for Women population. In contrast, Zamble & Porporino (1988)
reported that only 19% of their male sample admitted to previous treatment for
psychological problems.
Substance use patterns in the study sample were similar to those found by
Lightfoot and Lambert (1992). Twenty-six percent reported at least moderate
alcohol abuse, while 54% reported at least moderate drug abuse. The study
sample displayed a lower rate of alcohol abuse when compared to larger male
samples (Lightfoot & Hodgins 1988; Zamble & Porporino, 1988), but proportions
with at least moderate drug abuse were equivalent.
18
Wuerscher, 1984), and with a more recent American sample of female offenders
(Salekin et al., 1998). The mean total score for the current sample is
approximately 3 points lower than that found by Neary (1990) for a sample of
120 female Black and Caucasian inmates, and 5 points lower than a sample of
40 provincially-sentenced female inmates studied by Strachan et. al. (1990).
Table 2:
Measure
W.A.I.S. Equivalent
Multidimensional Anger
Inventorya
Anger arousal
Range of anger-eliciting
situations
Hostile outlook
Anger-in
Anger-out
PRF-E - Desirability
Psychopathy Checklist
(Revised)b
Total score
Factor 1
Factor 2
Socialization Scale
Social Withdrawl
Sociability
Time Framing Questionnairea
Beck Depression Scale
Beck Hopelessness Scale
Notes.
Study
Sample
Mean
101.9
Standard
Deviation
Standard
Deviation
15.1
Male
Sample
Mean
--
17.8
23.2
7.7
6.6
17.5a
22.5a
7.3
6.6
12.4
13.2
6.9
9.7
3.7
4.9
2.3
3.4
11.6a
13.2a
6.1a
--
3.8
5.4
1.6
--
18.0
8.1
8.1
9.0
4.1
5.7
23.6b
8.9b
11.7b
7.9
3.9
3.9
5.3
5.5
17.8
15.3
3.3
2.8
3.8
5.3
9.5
3.3
--14.7c
10.4c
3.1c
--6.1
---
19
--
With a cutoff score of 30, commonly used for male prison inmates, 11% of the
current sample would be defined as psychopaths. Salekin et al., (1997) found a
rate of 16% in an American sample of female offenders. The median incidence
of psychopathy for a number of samples of male prison inmates (covering all
security levels), based on data described by Hare (1990), is approximately 20%.
The mean coping efficacy score for the female sample was 12.4, close to the
mean of 11.6 reported for male offenders beginning their prison terms by Zamble
& Porporino (1988), and substantially lower than the mean of approximately 16
found for a non-offender population (Hughes & Zamble, 1993).
Criterion Variables
Seventy-six percent of the study sample had a current or past criminal conviction
involving violent behavior. In contrast, detected incidents of violent behavior in
the institution resulting in a conviction in institutional court were relatively
infrequent, with only 16 incidents involving 10% of subjects. Thirty-five percent
of the women received at least one conviction for any type of offence during the
12-month measurement period. The rate of recidivism of the released sample,
over an average of 3 years, was 47%. This is consistent with the findings of
Bonta et al. (1995), whose sample was "at risk" over a similar time period.
Correlation Analyses
Distributions for all variables were plotted and examined for significant departure
from normality. Criminal convictions and violent convictions represented a fair
approximation of a normal distribution. Institutional convictions and institutional
violence were bimodal and therefore they were dichotomized for analyses.
Correlation matrices for all independent variables were examined, and one of
any pair correlating beyond .80 was removed from further consideration. Missing
data were excluded pairwise. For the correlation analyses, this resulted in ns
ranging from 43 to 100. N's below 75 were found exclusively in the personal,
20
social, and criminal history class of variables, where there were frequently
missing data on family history.
As stated earlier, we were interested in studying the relative predictive ability, not
only of individual variables, but also of different variable types. Variables were
classified into 4 main groups: personal, social and criminal history; abuse history;
maladaptation history; and current personality, ability, and emotional functioning.
Table 3 provides a summary of the significant correlations of variables from all
classes with the outcome criteria. Because of the binary nature of the two
institutional criteria, Spearman correlations are preferable to the Pearson
statistic. However, values for the two show an almost identical pattern, so all
values shown are Pearson correlations, to maintain comparability.
Within the class of personal, social, and criminal history variables, age at first
arrest showed the most consistent relationship, being negatively related to all
criteria. Family-related variables, specifically those related to criminal record,
substance abuse, and family cohesiveness, revealed a wider range of criterion
relationships, although these were not as substantial as those for age at first
arrest.
Data within the abuse variable set suggest that early sexual abuse is the primary
associate of violent behavior. Only psychological abuse was associated with
criminal convictions, while several abuse variables were associated with
institutional convictions. In contrast, with the exception of drug usage, which
showed a positive relationship, measures of maladaptation history did not show
significant relationships with any of the criteria.
21
Table 3:
Variables
Age
-.19
Age first arrest
-.42***
Age left school
-.14
Criminal record:
father
.21*
siblings
.12
Family cohesiveness:
ages 6 to 11 yrs.
-.13
age 12 yrs. or more
-.24*
Racial/ethnic origin:
Aboriginal
.07
Black
-.09
Sentence length
-.24*
Substance abuse:
mother
.02
siblings
.11
Pre-adolescent Abuse
Total sexual abuse
.11
Physical abuse
.17
Psychological abuse
.33***
Post-adolescent Abuse
Total sexual abuse
.06
Physical abuse
.16
Drug Use
Social desirability
-.04
Anger-out
-.14
Psychopathy (total
.49***
score)
Psychopathy - Factor 2
.42***
Psychopathy - Factor 1
.43***
Social outgoingness
.03
Time-framing
-.03
a
Notes
- continuous variable
b
- binary variable
*p .05. **p .01. ***p .001.
Total
Institutional
b
Convictions
-.34**
-.48***
-.42***
.15
.26*
22
Total Violent
Criminal
a
Convictions
-.17
-.27**
-.25*
.28**
.16
Total Violent
Institutional
b
Convictions
-.14
-.27*
-.21
.09
.36***
-.25*
-.33**
-.01
-.13
-.13
-.19
.15
-.12
-.18
.21*
-.20*
-.04
.15
-.05
.29**
.37***
.13
.31**
.32**
.36***
.37***
.29**
.29**
.28**
.14
.19
.23*
.08
.18
.20
.40***
.34**
-.30**
.06
.63***
.20*
.22*
.23*
-.13
.30**
.46***
.13
-.00
.54***
.42***
.14
-.36***
.36***
.29**
.21*
-.23*
.38***
.23*
-.08
-.25*
-.24*
.10
.38***
23
24
25
Table 4:
Predictor Variables
Social, Personal,
Criminal History
Age
Age first arrest
Age left school
Criminal record
siblings
Family cohesiveness
age 12 yrs. or more
Racial/Ethnic origin
Black
Substance abuse
mother
siblings
spouse
Sentence length
Total R
Total R squared
(adjusted)
Maladaptation History
Alcohol abuse
Drug abuse
Community mental health
treatment
Total R
Total R squared
(adjusted)
Abuse History
Pre-adolescent
psychological abuse
Post-adolescent physical
abuse
Pre-adolescent sexual
abuse
Post-adolescent sexual
abuse
Total R
Total
Criminal
Convictions
a
b
-.42
Total
Institutional
Convictions
a
b
-.44
Total Violent
Criminal
Convictions
a
b
Total Violent
Institutional
Convictions
a
b
-.23
-.23
.42
.17***
.32
.57
.31***
.28
.39
.13***
.33
.42
.26
.16*
*
.34
.34
.10**
.33
.29
.23
.23
.04*
.44
.23
.39
.28
-.32
.33
.51
.28
.23
26
.22
.23
Total R squared
(adjusted)
Personality, Ability,
Emotional
Coping Efficacy
Anger-out
Time framing
Social desirability
Psychopathy
Total R
Total R squared
(adjusted)
Combined Variable
Classes
Total R
Total R squared
(adjusted)
10***
23***
.07**
.04*
-.31
-.29
.24
.34
.60
.54
.58
.31***
.61
.61
.36***
.25
.30
.42
.34
.49
.22***
.38
.35
.56
.29
.11*
*
.62
.36***
.67
.43***
.55
.28**
*
.45
.18***
adolescent sexual abuse is not predictive of violence, and that neither sexual nor
physical abuse variables significantly predicted the total number of criminal
convictions. Evidently, early sexual abuse specifically increases tendencies
toward violence rather than criminality generally.
No variable within the maladaptation history set reached significance in the final
set of equations. Even in the preliminary analyses, drug abuse was the only
variable from the group to enter any of the equations, and then only weakly, for
two of the four criteria. Although drug abuse appears to be a widespread
problem in this population, it seems not very predictive of criminal misbehavior.
Comparing across outcome criteria, one can see that the best prediction was for
institutional convictions, although a useful proportion of the variance for criminal
convictions was also accounted for. Violence was not predicted as well, and the
weakest prediction was for institutional violence, not surprisingly given the low
incidence rate recorded.
In general, the R2 values for the final equations were not substantially higher
than those where measures of personality, ability, and emotional functioning
alone were the predictors. From this we conclude that measures of personality
and current functioning are the most important in explaining, and probably in
predicting, the measures of criminal behavior and prison misconduct chosen as
outcome criteria for this study.
Prediction of Recidivism
Approximately five years after the completion of the initial study, records in the
Offender Management System were searched, to determine releases, parole
revocations and reconvictions from the original sample. Eighty-one subjects had
been released at some time subsequent to original testing, an average of 37.8
months before the follow-up. Of these, 38 (47%) had been returned to prison for
conviction on a new charge or for a major violation of their release terms. (There
28
were a few minor violations resulting in short temporary suspensions, but these
were not counted.)
The earlier data were then analyzed again, with recidivism as the target. This
provides a test of how well the results seen in the previous section will generalize
across a new dependent measure, and, more important, whether the measures
in this study can be used in a truly predictive fashion.
The analyses paralleled those with the measures of prison misconduct and
criminal history, except that the number of previous criminal convictions was
used here as a predictor. As before, simple correlations were calculated
between the target measure (revocation of release) and the set of independent
variables in the study. Measures that showed significant correlations with
revocation were then entered into one of four logistic regression analyses, for
each division of variables, as in the postdictive analyses in the preceding section.
The results were similar to those shown before, probably closest to those for
criminal history, as one might expect. For each set of predictors, only a single
variable entered the equation. In the class of background and historical
measures, the variable to enter the regression was substance abuse by the
subjects father, although, if this variable was eliminated from the analysis, the
closely related measures of substance abuse by the mother or siblings took its
place. Among measures of abuse, only psychological abuse in childhood met
the criteria for entry into the equation. The number of previous criminal
convictions was the sole significant predictor among measures of previous
maladaptation, while among current psychological measures the PCL-R score
alone entered the equation.
These four measures were then tested together in a single logistic regression. In
this case, three variables entered the prediction before termination: the PCL-R
score, previous criminal convictions, and substance abuse by the father.
29
Finally, these three variables were entered into a Cox regression survival
analysis (Norusis, 1994). There are several advantages of this type of analysis
for recidivism data. Survival analysis generally has the ability to provide more
than the simple binary measure of success or failure. In particular, the Cox
procedure allows one to test the predictive utility of a specific set of variables. In
addition, linear predictor scores can be generated and saved as summary
measures.
The resultant linear predictor scores (X' Beta) had a correlation of .32 with
revocation. Values across subjects were split at the median, and the results
plotted as shown in Figure 1. A substantial difference is clearly visible: using
these curves, one would estimate that approximately 5 of 6 subjects above the
median would still be in the community three years after release, while about 4 of
6 subjects below the median would have returned to prison. The model is clearly
able to differentiate among subjects. Without prior knowledge of the actual
recidivism rate, one can assume that scores above the median (the upper curve)
predict success, while those below the median (the lower curve) predict failure.
To test the accuracy of these predictions, a 2x2 contingency table was
constructed comparing these predictions to actual results, and is given in Table
5. Although 81 participants had been released, complete data was only
available on 80. The proportion of correct predictions was 76%, (1) = 22.17,
2
p<.001, with roughly equal numbers of false positives and false negatives. This
proportion of correct predictions is at the high end of the range of accuracy with
conventional actuarial scales.
30
DISCUSSION
The results of this study indicate that serious female offenders are in many ways
quite similar to populations of serious male offenders, although there are also
some important differences. Our analyses support the position that there are
considerable similarities in serious offenders regardless of gender, and that by
and large the differences between genders are not predictive of criminal
behavior.
For example, psychopathy is as important in predicting general offending in
female serious offenders as it is in male serious offenders, and for females it
plays an important role in the prediction of violent behavior and prison
maladjustment, as it does for males.
On the other hand, although we find evidence of high levels of previous sexual
and physical abuse in our respondents, these factors do not appear to be very
much related to criminal or violent behavior, and, at best, they are secondary to
some endogenous personality or behavioral factors. Among the abuse
measures, pre-adolescent sexual abuse is the only one that correlates
significantly with more than one of the outcome measures. Certainly, there is
ample evidence in the literature that various forms of childhood abuse can have
a profound effect on behavioral and emotional adjustment in both childhood and
adulthood, but these effects do not appear to be specific to criminal behavior.
Interestingly, psychological abuse, seldom mentioned in theories that emphasize
victimization, was the only type of abuse that played even a minor role in the
prediction of general recidivism.
31
Figure 1
Female Recidivism Survival Curves
Note: The upper curve represents subjects above the median of the Cox coefficient function, and
the lower curve represents subjects below the median.
32
Table 5
Contingency Table for Predictors from Logistic Regression
Predicted
Success
Failure
Total
Success
32
11
43
Failure
29
37
Total
40
40
Actual
33
34
35
(Leblanc, 1992; Marquis, 1992), as well as female juvenile offenders (Dornfeld &
Kruttschnitt, 1992).
Finally, the role of early sexual abuse appears somewhat equivocal. Preadolescent abuse was a good predictor of violent convictions, consistent with
evidence for incarcerated males (Dutton & Hart, 1992), but inconsistent with
Widoms (1989) findings for females in a community sample. At the same time,
it did not appear to be related to the overall number of offences. It would be very
useful to examine this in connection with a more detailed taxonomy of offence
types. For example, it may be that sexual abuse is specifically linked with violent
offences that occur in a domestic context.
The pattern of results we have obtained has considerable implications for
theories of female offending. Contemporary views of female offending continue
to emphasize the experience of abuse, poverty, and substance abuse as being
the most common pathways to crime (Bloom, 2000). The results seen here,
however, are generally inconsistent with class-based theories that rely on socialstructural factors to explain female involvement in serious crime. More generally,
and more important, the current results do not provide empirical support for
theories of female criminal behavior based on separate or unique determinative
paths, such as those emphasizing the direct causative role of victimization for all
female offenders. Although such experiences may play an important role in the
origins of antisocial behavior, they do not appear to have much power in
explaining serious or sustained criminality.
This suggests that a more productive approach to theory development would be
to expand the focus beyond factors felt to be uniquely important for female
offenders to include factors that have been found to be predictive for male
offenders. Such a theory might be best developed within the framework of social
learning theory, because it allows the inclusion of experiential factors that are in
common across gender, as well as some that differ. Research incorporating
measures of both types of factors can provide an increased understanding of
36
both male and female criminal behavior and, ultimately, contribute to the
development of a more powerful comprehensive theory regarding criminal
behavior in general.
Our conclusions must, of course, be tempered by the limitations of the present
study. Although half of the data for all institutional outcomes were gathered
post-test, the study measures cannot be regarded as truly prospective, and one
must be cautious about causal inferences. The data on recidivism help to
address these concerns, but they still do not allow us to address issues of
causation very well.
There are other limitations related to the choice of measures, especially in the
types of information that were omitted. It might have been useful to look at
psychiatric diagnoses, particularly borderline personality disorder, because this
diagnosis has been implicated in violent behavior in psychiatric samples (Raine,
1993), and appears to be a frequent diagnosis in female convicted and remand
populations (Wilkins & Coid, 1991; Coid et al., 1992). Also, the role of proximal
social stability factors, such as employment, marital status, and criminal
associates was largely neglected. These have been found to be important
predictors in male offender populations (Andrews & Bonta, 1994), and have
been described as important in the genesis of female violent offending
(Sommers & Baskin, 1993). While they were omitted in order to make data
acquisition more manageable, they need to be added into the set of predictors
for future work, because more distal measures, such as lack of cohesiveness in
the family of origin, did not demonstrate predictive utility here. In addition,
measures of more recent exposure to physical violence may prove a more
powerful predictor than the degree of exposure for the entire post-adolescent
period.
Although they should be regarded tentatively, the results do have some
implications for assessment and treatment targets in this particular population,
as well as for the mode of treatment delivery. Some of the targets that have
37
been identified as relevant to criminal risk for male offenders, including anger
expression and aspects of psychopathy, are also relevant for this population of
serious female offenders. Because our respondents display many of the
characteristics of male serious offenders, and consistent with the literature
concerning effective treatment modalities for male offenders (Andrews & Bonta,
1994), we would suggest that cognitive-behavioral techniques would be the most
effective. This conclusion is supported by a recent meta-analysis of female
offender treatment studies, which suggests that the most effective interventions
are based upon cognitive-behavioral techniques (Andrews & Dowden, 2000).
These authors caution, however, that the studies analyzed involved primarily
juvenile offenders, and very few incarcerated or adult female offenders.
Thus, interventions for high-risk female offenders should be able to profitably
import many of the components of effective treatments for high-risk male
offenders, when the target is the prevention of recidivism. Of course, this does
not mean that assessment and treatment of all female offenders should be
identical to that for males. Our results point out a number of potentially important
differences even in this population of the most serious female offenders; for
example, the role of abuse variables is still important in this population,
especially if violent behavior is targeted. Moreover, we should emphasize that
the type of offender studied here may not be representative of women offenders
in general, but only those with the most extensive and most serious criminal
histories.
In considering treatment plans, one must also distinguish between treatment
designed to reduce recidivism and treatment designed to meet the mental health
needs of offenders. Some of the factors that appear in this investigation to be
unrelated to criminal behavior, such as the high incidence of depression, are
nevertheless needful of clinical attention. When all of the differences are
considered, it is likely that the most fruitful strategies for the treatment of serious
female offenders will differ somewhat from those most effective with male
offenders, despite the commonalities.
38
REFERENCES
Adler, F. (1975). Sisters in Crime: The Rise of the New Female Criminal.
New York, NY: McGraw-Hill.
Andrews, D.A. & Bonta, J. (1995). Manual for the LSI-R: The Level of
Service Inventory-Revised. Toronto, Ont: Multi-Health Systems, Inc.
Andrews, D.A. & Bonta, J. (1994). The Psychology of Criminal Conduct.
Cincinnati, OH: Anderson Publishing Co.
Andrews, D.A. & Dowden, C. (2000). A meta-analytic investigation into
effective correctional intervention for female offenders. Forum on Corrections
Research, 11(3), 18-21.
Beck, A.T., Rush, A.J., Shaw, B.F. & Emery, G. (1979). Cognitive Therapy
of Depression. New York, NY: Guilford Press.
Beck, A.T., Brown, G., Berchick, R.J., Stewart, B.L. & Steer, R.A. (1990).
Relationship between hopelessness and ultimate suicide: A replication with
psychiatric outpatients. American Journal of Psychiatry, 147(2), 190-195.
Beck, A.T., Weismann, A., Lester, D. & Trexler, L. (1974). The
measurement of pessimism: The Hopelessness Scale. Journal of Consulting and
Clinical Psychology, 42, 861-865.
Belcourt, R., Nouwens, T. & Lefebvre, L. (1993). Examining the
unexamined: Recidivism among female offenders. Forum On Corrections
Research, 5(6), 10-14.
Blanchette, K. & Motiuk, L. (1995). Female offender risk assessment: The
Case Management Strategies approach. Paper presented to the Canadian
Psychological Association Annual Conference, Charlottetown, P.E.I.
Blishen, B.R., Carroll, W.K. & Moore, C. (1981). The 1981 socioeconomic
index for occupations in Canada. Canadian Review of Sociology & Anthropology,
24(4), 465-485.
39
40
41
42
43
44
Robertson, R.G., Bankier, R.G. & Schwartz, M.A. (1987). The female
offender: A Canadian study. Canadian Journal of Psychiatry, 32(9), 749-755.
Robinson, D. & Taylor, J. (1994). The incidence of family violence
perpetrated by federal offenders: A file review study. (Research Report,
Correctional Service of Canada). Ottawa, Ont.
Rosenbaum, J.L. (1989). Family dysfunction and female delinquency.
Crime and Delinquency, 35(1), 31-44.
Rosenblatt, E. & Greenland, C. (1974). Female crimes of violence.
Canadian Journal of Criminology and Corrections, 16(2), 173-180.
Roundtree, G.A., Parker, A.D., Edwards, D.W. & Teddlie, C.B. (1982). A
survey of the types of crimes commited by incarcerated females in two states
who reported being battered. Corrective and Social Psychiatry and Journal of
Behavior Technology, Methods, and Therapy, 28(1), 23-26.
Salekin, R.T., Rogers, R. & Sewell, K.W. (1997). Construct validity of
psychopathy in a female offender sample: A multitrait-mutimethod evaluation.
Journal of Abnormal Psychology, 106(4), 576-585.
Salekin, R.T., Rogers, R., Ustad, K.L. & Sewell, K.W. (1998). Psychopathy
and recidivism among female inmates. Law and Human Behavior, 22(1), 109128.
Saunders, B., Villeponteaux, L., Lipovsky, J., Kilpatrick, D. & Veronen, L.
(1992). Child sexual assault as a risk factor for mental disorders among women:
A community survey. Journal of Interpersonal Violence, 7(2), 189-204.
Serin, R., Peters, R. & Barbaree, H. (1989). Predictors of psychopathy and
release outcome in a criminal population. Psychological Assessment: A Journal
of Consulting & Clinical Psychology, 2, 419-422.
45
46
47