Intro to the
ABEL Assessment of
Sexual Interest
Presented to VSOTA
March 3, 2010
Evan S. Nelson, Ph.D., ABPP, CSOTP
[email protected]
www.psylaw.com
(804) 739-4669
Official Disclaimers
P Dr. Evan Nelson’s test lab has all three:
< AASI-2/ABID (11 years)
< Monarch 21 PPG (2 years)
< Affinity 2.5 (2 years).
P I am not advocating for the purchase or
abandonment of any test.
P I am not being compensated by any of the
tests’ developers.
P The focus of this presentation is on male
adult offenders.
Who Do You Look At?
Pathway to Sexual Behavior
Visual Sexual Sexual
Interest Arousal Behavior
AASI-2 PPGs Legal
Affinity Illegal
Conditions Healthy
Unhealthy
Borrowed with much adaptation from Singer’s theories in the 1980's
Basic Construction of the AASI-2
P 160 images in the body of the test; 15 practice pics
P All of individuals; no nudes; no sexual interactions
P 7 images in each category, for both Caucasians and
African Americans:
< Preschool aged girls, preschool aged boys
< Grade school aged girls, grade school aged boys
< Teenage girls, teenage boys
< Adult women, adult men
P Additional scales: exhibitionism, voyeurism,
fetishism, frotteurism, sadomasochism against
females, sadomasochism against males
Development of the Slides
According to Abel...
P All parties signed informed consent to be
photographed for a sexual test
P Pediatricians and the like were asked to
review images to make sure the body types
were representative of the age
P No data on how many images were in the
initial set, nor item analysis as to why the
specific pictures were selected in the final set
The “Method of Thirds”
As described by Abel . . .
P Find:
< The highest adult/adolescent VRT z-score
< The lowest z-score for all categories
P Z-scores for child scales that are more than
one-third of the way along this line are
considered to be clinically significant
1/3
Adult Men
Adult Women
Preschool Girls
No Test Can Assess Every Form of Deviance
Internal Reliability of the Scales
P Abel claims good internal consistency for the
scales: á = 0.84 to 0.90 (Letourneau found á =
0.60 to 0.90)
P BUT, Abel’s published data appear to be for a
version of the test before it was the AASI, and
collapsed to include both races, and outliers were
excluded
P There seem to be no data on test-retest reliability
over any reasonable interval of time, except
comparing PC to Mac versions; only one internal
study on pre-post treatment VRT changes
Fischer and Smith, 1999; Abel, Warberg, and Holland, 1998; Letourneau, 2002
The Problem with Z-scores Alone
P Z-scores allow ipsative comparisons, but do
not reflect the range of viewing time – and
the AASI-2 does not provide the raw data
P Given this, what constitutes “significantly”
higher scores? If the range is very small
then the method of thirds is not a sensitive
enough way to cut the data
P The method of thirds is not norm based
P What happens with outliers?
Is VRT a Real Measure of Sexual “Something?”
P As erotic content increases, so does viewing
time - but these are not very erotic images
P VRT correlates with image attractiveness,
sexual arousal, and sexual stimulation
P Self-report of the attractiveness of an image
is correlated with viewing time, r=.72
P Homosexuals v. Heterosexuals: each group
preferred to look at pics of their sexual
preference
1. Brown, 1979; Brown, Amoroso, War, Pruesse, Pilkey, 1973; 2. Landolt et al., 1995; 3. Lang et al., 1980,;Quinsey Rice,
Harris, & Reid, 1993; Quinsey et al., 1996; 4. Quinsey et al., 1996; Abel et al., 1990, Wright and Adams, 1994 and 1999;
Wright and Adams, 1999.
VRT and PPG -Harris et al.
Harris, Rice, Quinsey, & Chaplin. (1996). Viewing time as a measure of sexual interest among child
molesters and normal heterosexual men. Behaviour Research and Therapy, 34(4), 389-394.
P 26 child sexual abusers, 25 men no abuse history
P VRT stimuli - nude males and females - but not
AASI-2; and PPG
P Used a deviance index (adult minus child)
P VRT: Child abusers had small index, so looked
longer at kids – but still looked longest at adults
P PPG: Child abusers more arousal to kids than non
abusers AND more arousal to kids than adults on
average, a different result than for VRT
AASI and PPG - Letorneau
Letourneau, 2002. A comparison of objective measures of sexual arousal and interest: Visual
reaction time and penile plethysmograph. Sexual Abuse: A Journal of Research and
Treatment, 14(3), 207-223.
P Used AASI, not AASI-2
P Used PPG with ATSA audiotapes
P Correlated VRT scores (raw, then z-scores)
with maximum change in arousal
P The tests had good agreement for classifying
offenders against boys, but not for abusers of
girls or women; VRT better for teen girl
offenders
AASI-2 v. PPG - Johnson and Listiak
Johnson & Listiak, 1999. The measurement of sexual preference - A preliminary
comparison of phallometry and the Abel Assessment, in Schwartz and Cellini (eds.),
The Sex Offender, Vol. 3, pp 26.21.26.20
P Used rule of thirds for VRT, and erectile
circumference change of 5 mm or greater
P “The authors report that, based on the VRT
portion of the AASI, 94% of the participants
were correctly identified on the basis of the
age and gender of the abuse victim for which
they were convicted.” [retrospective study]
Note: This presenter has not see the original book chapter.
AASI-2 Advantages over PPG
P Subject need not disrobe
P Can be given in many types of settings
P No nudity is less threatening, OK with hyper
religious guys, does not violate probation
prohibition against looking at pornography
P Cheaper to purchase than PPG, and
software much less finicky
P (but costs more in the long run)
P Much faster fast to administer than PPG
AASI-2 S&M Scales are Weak
The AASI-2 Questionnaire
P Background data: demographics, personal
experiences, sexual interests, ideas about
child sexual abuse
P Social skills
P Survey of sexual behaviors, e.g.,
exhibitionism, voyeurism, use of pornography
P Sexual knowledge, attitudes
Validity Scales
P Social Desirability Scale
< Items 29 through 48 of Section IV
< Measures unwillingness to admit to common
violations of social mores
P Cognitive Distortion Score
< Items 93 through 105 of Section I
< Potential justifications for inappropriate sexual
behavior
P Scored 0 to 100%; no research on the utility
of the cut scores; Abel claims Cronbach á >
0.73 for each scale
Understanding AASI-2's
Probability Values
P Developed on a logistical regression: it is the
probability that the subject belongs to the
group of people who have molested children
in the past
P Only useful for criminal defendants who deny
the offense – but really, what can you do with
the result? Testify to the probability that an
offense occurred!? NEVER DO THIS!
Overview of How the
Probability Value was Developed
P Sample of child offenders who admitted, and of
sexual abusers with non-child related offenses
P Used regression to develop model for non-child
abusers, abusers of male children, and abusers of
female children; both intra- and extra familial
P Then, add in group of offenders “thought” to have
offended as evidenced by:
< Had been found guilty of a sexual offense
< Or, accused by at least two different families
Reliability of the Probability of
Past Child Sexual Abuse Score
Cut Score 0.3 Cut Score 0.6
Sensitivity 87% 61%
(actual positives correctly identified)
Specificity 45% 86%
(actual negatives correctly identified)
Prob Past CSA 22.4% 81.3%
(assuming 50% base rate)
N=2,526; ÷2=236.61, p<.0001; ROC 0.81 based on 10-
fold cross validation -- within the development sample.
Incest-only cases were excluded when calculating
validity data.
What Loads on the Probability Scale?
P Ratio of child VRT to frotteurism VRT
(coeff=3.04)
P VRT for frotteurism (coeff=0.86)
P “hobbies and interest subscore” (coeff=0.78)
P Subject was molested as a child (coeff=0.98)
P “I talk to children on their level” (coeff=0.36)
P “I enjoy being around children” (coeff=0.23)
P “I would rather spend my time with children”
(coeff=0.23)
][Note: Abel says nine items – this is only seven; what’s missing?
AASI-2 As a Risk Assessment Tool
P Hanson’s meta analyses say deviant arousal
is correlated with recidivism; Abel says PPG
and VRT have convergent validity, but Abel
has stopped hinting it is predictive
P NO prospective or retrospective studies on
the AASI-2 as a risk assessment tool
P AASI-2 has “Probability Values” - these are
about the risk that a subject molested a child
in the past
What Does VRT Measure?
“...[M]isconception of What VRT Measures:
VRT measures sexual interest. Smith &
Fischer (1999) acted as if there was no
distinction between sexual interest in
children and sexually abusing children.
Since there are other causes of sexual
abuse of children besides having a sexual
interest in children, such as psychopathy
and poor impulse control, it’s wrong to treat
them as being synonymous.”
Take-Home Messages
P Visual Response Time probably is a valid
measure of sexual interest
P It’s not clear that sexual interest is a measure
of sexual behavior
P AASI-2's probability values are intriguing for
hypotheses about past behavior, if that’s a
useful question to ask
P No direct research data on AASI-2's
relationship to future offending
P Clinically, it works! This is an “emerging tool”
Some Pedophiles are Going to
Get Away with It
Affinity 2.5 -Competitor to the AASI-2
P Cheaper: One time $1000 purchase price
P Provides raw viewing time for every picture
P But, limited support
P Virtually no research yet; developed on
adults only but research supports idea of
extending it downwards to adolescents
Another Emerging Competitor?
Banse, Schmidt, and Clarbour, 2010. Indirect measures of sexual interest
in sex offenders. Criminal Justice and Behavior, 37, 319-335.
P “Explicit and Implicit Sexual Interest Profile” (EISIP)
P Combined direct self-report (explicit) with indirect
measures (including viewing time on their own 20
slides) to try to classify child sex offenders
P Unlike Abel, compared child sex offenders to non-
sexual criminals and (non-offenders)
AASI-2 Claims to be Valid with Female Abusers
Are Female Offenders Driven by Visual Cues?
Latinos
Abel Says...
P Recognized that Latino population was
growing. AASI (1) did not have Latino
images.
P Claims research conducted with new images
of Latinos compared to Caucasians.
P Claim: Latino subjects more responsive to
images of Caucasians than of Latinos,
therefore did not change basic set of images.
P No external research.