MILLON CLINICALMULTIAXIAL
INVENTORY–III (MCMI–III)
The Millon Clinical Multiaxial Inventory–III (MCMI–III)
is a 175-item self-report inventory designed to assess
personality characteristic and psychopathology. It has
4 validity scales, 11 clinical personality pattern scales,
and 3 severe clinical syndrome scales. Although not
originally validated in forensic populations, and in
spite of limited research with forensic subjects, it is
increasingly being used in forensic practice. Extensive
changes were made in the development of the
MCMI–III, limiting the applicability of the research
results from prior MCMI versions.
Detection of malingering, denial, and random
responding and diagnostic accuracy are critical issues
that are relevant to the forensic applicability of the
MCMI–III. A number of issues have implications for
use of the MCMI–III in forensic assessment, including
poor detection of malingering and denial, interpretation
of potentially random protocols, and a significant con-
troversy regarding diagnostic accuracy. The existence
of all these issues is likely to result in vigorous chal-
lenges to expert testimony based on the MCMI–III
because the instrument does not meet the criteria estab-
lished in Daubert v. Merrell Dow Pharmaceuticals
(1993), which require an evaluation of the error rate of
assessment methods on which experts rely.Malingering, Denial, Random
Responding, and the MCMI–III
More research is needed before firm conclusions can
be drawn regarding the ability of the MCMI–III to
detect malingering and denial. The extant research
suggests only moderate accuracy, and there are no
studies that use known groups designs with forensic
populations. Mike Schoenberg and colleagues in 2003
compared students simulating psychiatric disorder
with psychiatric patients and found a sensitivity of
58.5% and 51.9% for a Scale Z and Scale X, respec-
tively. Positive predictive power was 55.6% and
66.3% for X and Z, respectively. They concluded that
“the MCMI–III modifier indices were of minimal
clinical utility in distinguishing college student malin-
gerers from bona fide psychiatric patients.” Somewhat
better results, with higher accuracy in detecting
malingering, were reported by Scott Daubert andApril Metzler, who compared two groups of psychi-
atric patients, one instructed to malinger and one
instructed to respond honestly. In a separate study by
Schoenberg and colleagues, an attempt was made to
develop a discriminant equation to detect malingering.
They found some improvement in detecting malinger-
ing. However, research with other instruments by
Richard Rogers’s group and Kucharski and colleagues
has shown that the accuracy of discriminant equations
developed via simulation designs decreases to near-
chance levels when applied to actual forensic popula-
tions. The results of a study by Richard Charter and
Michael Lopez demonstrated that more than 50% of
those responding randomly, using the VI > 1 criterion
recommended in the MCMI–III manual, would be
viewed as interpretable protocols. Failing to exclude
random protocols potentially confounds the research
on malingering and diagnostic accuracy and in clini-
cal practice is likely to inappropriately characterize
random responders as pathological.Diagnostic Accuracy of the MCMI–III
Probably the most difficult issue confronting the
MCMI–III is the current controversy regarding diag-
nostic accuracy. Two validity studies conducted by the
test author in 1994 and 1997 and reanalyses of the
data from these studies make up the findings on diag-
nostic accuracy. A reanalysis of the 1994 database, by
Richard Rogers and colleagues, demonstrated that the
convergent validities of the personality scales was
“disconcertingly low ranging from .07 to .31” and that
the “discriminant correlations were higher than the
convergent validities.” These findings are consistent
with other studies conducted by Paul Retzlaff. Frank
Dyer and Joseph McCann argued that the Rogers and
colleagues study was flawed due to selection of poor
criterion measures and use of data from the 1994
validation study, where there were obvious deficien-
cies in the diagnostic criterion. The 1997 validation
study attempted to address this limitation by including
a Diagnostic and Statistical Manual of Mental
Disorders(fourth edition; DSM-IV) of the American
Psychiatric Association criterion guide for diagnosis.
Louis Hsu reanalyzed both the 1994 and 1997 data
and found marked improvement in the diagnostic
accuracy for the 1997 data. However, serious method-
ological flaws including criterion contamination, con-
firmatory bias, and availability heuristics led him to
conclude that the results potentially overpredict the504 ———Millon Clinical Multiaxial Inventory–III (MCMI–III)M-Cutler (Encyc)-45463.qxd 11/18/2007 12:43 PM Page 504