psychopathy and institutional misbehavior among incarcerated sex offenders: a comparison of the...

10
International Journal of Forensic Mental Health 2002, Vol. 1, No. 1, pages 49-58 ©2002 International Association of Forensic Mental Health Services Psychopathic personality disorder (psychopathy) consists of a distinct pattern of affective, inter- personal, and behavioral symptoms. Emotionally, psychopaths lack empathy for others or guilt for their misdeeds, and they have difficulty forming strong affective bonds. Behaviorally, psychopaths tend to be irresponsible and prone to criminality. Inter- personally, psychopaths are grandiose, callous, and deceitful (Cleckley, 1976; Cooke & Michie, 2001; Hare, 1991). The relevance of the construct of psychopathy to forensic mental health services hardly can be overstated, in that it repeatedly has been shown to identify individuals who are at increased risk for various negative outcomes (e.g., treatment failure, criminal recidivism, violent recidivism) across several forensic and criminal justice populations (Heilbrun et al., 1998; Nolan, Volavka, Mohr, & Czobor, 1999; Rice & Harris, 1995; Seto & Barbaree, 1999; Skeem & Mulvey, 2001; see also Hart & Hare, 1997; Hemphill, Hare, & Wong, 1998). The most widely validated instrument for assessing this disorder is the Psychopathy Checklist- Revised (PCL-R; Hare, 1991), which is scored based on a semi-structured interview and a review of institutional files and other collateral data. Unlike the labor-intensive PCL-R, other commonly used methods for assessing psychopathy (e.g., self-report scales) typically have not met with great success in terms of identifying this personality disorder. For example, several studies using popular psychological tests have found only weak to moderate correlations between the psychopathy-related scales embedded in these inventories and diagnoses of psychopathy based on the PCL-R or diagnoses of Antisocial Personality Disorder based on DSM criteria (e.g., Cooney, Kadden, & Litt, 1990; Hare, 1985; Hart, John F. Edens, Jacqueline K. Buffington-Vollum, Kevin W. Colwell, & Darryl W. Johnson, Department of Psychology and Philosophy, Sam Houston State University; Judy K. Johnson, Texas Department of Criminal Justice. This research was supported by the Texas Department of Criminal Justice under a research agreement with the first author. Points of view are those of the authors and do not necessarily represent the position of the Texas Department of Criminal Justice. We thank Melinda Bozarth at the Texas Department of Criminal Justice and Mary Ross at the Texas Department of Mental Health and Mental Retardation for providing access to the raw data. Correspondence concerning this article should be addressed to John Edens, Department of Psychology and Philosophy, Sam Houston State University, Huntsville, TX 77341-2447 (E-mail: [email protected]). Psychopathy and Institutional Misbehavior Among Incarcerated Sex Offenders: A Comparison of the Psychopathy Checklist-Revised and the Personality Assessment Inventory * John F. Edens, Jacqueline K. Buffington-Vollum, Kevin W. Colwell, Darryl W. Johnson, and Judy K. Johnson The present study compared the utility of two measures of psychopathic traits, the Antisocial Features scale of the Personality Assessment Inventory (Morey, 1991) and the Psychopathy Checklist-Revised (Hare, 1991), to postdict serious institutional misbehavior among incarcerated sex offenders. Archival disciplinary infraction data for 92 offenders were coded as major acts of physical aggression, major acts of verbal aggression/ defiance, or non-aggressive major offenses. Receiver Operating Characteristic (ROC) analyses indicated that ANT performed comparably to the PCL-R in terms of predicting the occurrence of these infractions during the prior two years, although there was a trend for PCL-R scores to be associated somewhat more strongly with physically aggressive offenses. Regression analyses indicated that each measure accounted for unique variance in the criterion variables, suggesting that both may evidence incremental validity in terms of their relation to disciplinary infractions.

Upload: independent

Post on 12-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

International Journal of Forensic Mental Health2002, Vol. 1, No. 1, pages 49-58

©2002 International Association of Forensic Mental Health Services

Psychopathic personality disorder (psychopathy)consists of a distinct pattern of affective, inter-personal, and behavioral symptoms. Emotionally,psychopaths lack empathy for others or guilt for theirmisdeeds, and they have difficulty forming strongaffective bonds. Behaviorally, psychopaths tend tobe irresponsible and prone to criminality. Inter-personally, psychopaths are grandiose, callous, anddeceitful (Cleckley, 1976; Cooke & Michie, 2001;Hare, 1991). The relevance of the construct ofpsychopathy to forensic mental health services hardlycan be overstated, in that it repeatedly has beenshown to identify individuals who are at increasedrisk for various negative outcomes (e.g., treatmentfailure, criminal recidivism, violent recidivism)across several forensic and criminal justicepopulations (Heilbrun et al., 1998; Nolan, Volavka,Mohr, & Czobor, 1999; Rice & Harris, 1995; Seto

& Barbaree, 1999; Skeem & Mulvey, 2001; see alsoHart & Hare, 1997; Hemphill, Hare, & Wong, 1998).

The most widely validated instrument forassessing this disorder is the Psychopathy Checklist-Revised (PCL-R; Hare, 1991), which is scored basedon a semi-structured interview and a review ofinstitutional files and other collateral data. Unlikethe labor-intensive PCL-R, other commonly usedmethods for assessing psychopathy (e.g., self-reportscales) typically have not met with great success interms of identifying this personality disorder. Forexample, several studies using popular psychologicaltests have found only weak to moderate correlationsbetween the psychopathy-related scales embeddedin these inventories and diagnoses of psychopathybased on the PCL-R or diagnoses of AntisocialPersonality Disorder based on DSM criteria (e.g.,Cooney, Kadden, & Litt, 1990; Hare, 1985; Hart,

John F. Edens, Jacqueline K. Buffington-Vollum, Kevin W. Colwell, & Darryl W. Johnson, Department of Psychology andPhilosophy, Sam Houston State University; Judy K. Johnson, Texas Department of Criminal Justice.

This research was supported by the Texas Department of Criminal Justice under a research agreement with the first author.Points of view are those of the authors and do not necessarily represent the position of the Texas Department of Criminal Justice. Wethank Melinda Bozarth at the Texas Department of Criminal Justice and Mary Ross at the Texas Department of Mental Health andMental Retardation for providing access to the raw data.

Correspondence concerning this article should be addressed to John Edens, Department of Psychology and Philosophy, SamHouston State University, Huntsville, TX 77341-2447 (E-mail: [email protected]).

Psychopathy and Institutional Misbehavior AmongIncarcerated Sex Offenders: A Comparison of the

Psychopathy Checklist-Revised and the PersonalityAssessment Inventory*

John F. Edens, Jacqueline K. Buffington-Vollum, Kevin W. Colwell,Darryl W. Johnson, and Judy K. Johnson

The present study compared the utility of two measures of psychopathic traits, the Antisocial Features scale

of the Personality Assessment Inventory (Morey, 1991) and the Psychopathy Checklist-Revised (Hare, 1991),

to postdict serious institutional misbehavior among incarcerated sex offenders. Archival disciplinary infraction

data for 92 offenders were coded as major acts of physical aggression, major acts of verbal aggression/

defiance, or non-aggressive major offenses. Receiver Operating Characteristic (ROC) analyses indicated

that ANT performed comparably to the PCL-R in terms of predicting the occurrence of these infractions

during the prior two years, although there was a trend for PCL-R scores to be associated somewhat more

strongly with physically aggressive offenses. Regression analyses indicated that each measure accounted

for unique variance in the criterion variables, suggesting that both may evidence incremental validity in

terms of their relation to disciplinary infractions.

50 Edens, Buffington-Vollum, Colwell, Johnson, & Johnson

Forth, & Hare, 1991). Moreover, none of the morecommon self-report scales of psychopathy hasdemonstrated criterion-related validity similar to thePCL-R. The failure of self-reports to correlateconsistently with either the PCL-R or importantcriterion measures may reflect a bias in the contentof self-reports, a tendency for psychopaths to be poorobservers or reporters of their interpersonal andemotional styles, a reflection of the “conning andmanipulative” aspects of those with psychopathictendencies, or a combination of these explanations(Edens, Buffington, Tomicic, & Riley, 2001; Edens,Hart, Johnson, Johnson, & Olver, 2000).

Despite the historically poor performance of self-report measures, research attempting to validate suchscales continues. For example, Edens et al. (2000)examined the utility of the Personality AssessmentInventory (PAI; Morey, 1991) to predict psycho-pathy. More specifically, they assessed the concurrentvalidity of the Antisocial Features (ANT) scale ofthe PAI by examining its association with thescreening version of the PCL (PCL:SV; Hart, Cox,& Hare, 1995) in a Canadian sample of forensicpsychiatric patients (Study 1) and its association withthe PCL-R in a U.S. sample of incarcerated sexoffenders (Study 2). In Study 1 (n = 46), thecorrelation between the PCL:SV and ANT was .54,indicating that ANT was a moderately strongdimensional measure of psychopathy. However,ANT was not a particularly good indicator ofdiagnoses of psychopathy (i.e., PCL:SV > 18) in thissample, with Receiver Operating Characteristiccurves (described below) indicating an area underthe curve (AUC) of .85 (SE = .06). In Study 2, theperformance of ANT as a measure of psychopathywas somewhat weaker among sex offenders (n = 55).The PCL-R/ANT correlation was .41, and the AUCfor the ROC curve was .71 (SE = .08).

Although these results suggested that ANT mightnot be a useful dichotomous indicator of psychopathyamong sex offenders, the authors did not comparethe validity of these two measures in terms of theirability to identify various types of “high risk”offenders. As Edens et al. (2000) noted, “despitelimited diagnostic efficiency vis-à-vis…the PCL-R,it is possible that ANT has incremental validity withrespect to clinical outcomes such as institutionalmisbehavior, response to treatment, or violence

following release into the community” (p. 137). Thefew published studies conducted on the criterion-related validity of ANT have suggested that it maybe a useful indicator of institutional misconduct (e.g.,Edens, Poythress, & Watkins, 2001; Salekin, Rogers,& Sewell, 1997; Wang & Diamond, 1999), althoughnone of this research specifically has examined sexoffenders.

In order to address the questions raised by Edenset al. (2000) regarding the potential incrementalvalidity of ANT and the PCL-R, the present researchexamined the association between these twoassessment procedures and one specific type ofcriterion, a history of serious institutional mis-conduct. If such measures are useful in identifyingoffenders who are at increased risk for disciplinaryproblems, then their use in various types of offenderclassification and/or management systems can bejustified. In fact, identification of ‘high risk’offenders has been noted as a significant goal ofinmate management procedures (Clements, 1996;Van Voorhis, 1994).

Several studies have examined the relationshipbetween the PCL-R and disciplinary infractions (forreviews, see Cunningham & Reidy, 1998; Edens,Petrila, & Buffington-Vollum, in press; Edens,Skeem, Cruise, & Cauffman, 2001; Hart & Hare,1997) and have reported a generally positive(although in some studies negligible) association.However, relatively limited research has addressedthis issue specifically in relation to sex offenders,who may present unique challenges in terms ofinstitutional management and treatment (Seto &Barbaree, 1999; Seto & Lalumiere, 2000). Moreover,in terms of the PAI, this is the first study to examineits criterion-related validity with a sex offendersample.

METHOD

Participants

Participants in this study were drawn from twoarchival databases. The first consisted of 55incarcerated male sex offenders who were describedin Study 2 of the Edens et al. (2000) research notedearlier. To augment this sample, data on the measures

Comparison of the PAI and PCL-R 51

described below were obtained for another 371

incarcerated adult males who were being consideredfor possible civil commitment under Texas’s recentlyenacted Sexually Violent Predator Statute (CivilCommitment of Sexually Violent Predators Act,1999) following the completion of their prisonsentences. Initial screening criteria that had to be metprior to being referred for an evaluation includedconviction of at least two prior sex offensescommitted against a non-family member. Becauseexaminees in these two samples were similar in termsof demographic and criminal history variables, weredrawn from the same prison system (Texas Depart-ment of Criminal Justice), and had been administeredthe measures under similar circumstances, their datawere combined into one sample for the analysesdescribed below. The sample was predominantlyCaucasian (58.7%), with relatively smaller per-centages of African American (21.7%) and Hispanic(16.3%) offenders (3.3% “other”). The mean age was37.74 years (SD = 11.09).

Measures

Personality Assessment Inventory. The PAI(Morey, 1991) is a multi-scale self-report inventoryintended to measure “critical clinical variables” (p.1). Examinees rate the 344 items on the extent towhich the statements are true of them on a 4-pointscale (1 = very true, 2 = mainly true, 3 = slightly

true, 4 = false). The items form 22 non-overlappingscales, including: 4 scales for assessing responsebias; 11 scales for assessing clinical syndromes; 5scales for assessing treatment-related characteristics;and 2 scales for assessing interpersonal style.Information published in the test manual (Morey,1991) and in independent reports (e.g., Cashel,Rogers, & Sewell, 1995; Douglas, Hart, & Kropp,2001; Salekin et al., 1997; Salekin, Rogers, Ustad,& Sewell, 1998) supports the validity of several PAIscales for use with forensic and correctionalpopulations (for reviews of this literature, see Edens,Cruise, & Buffington-Vollum, 2001; White, 1996).

Noted earlier, the ANT (Antisocial Features)scale is the primary PAI scale of interest in the current

study. The content of this 24-item scale focuses onpersonality traits and behaviors historicallyassociated with the concept of psychopathy (e.g.,egocentricity, callousness, antisocial acts, sensation-seeking). Although intended to serve as adimensional measure of psychopathic traits, ANTscores at or above 70 T are purported to identifyrespondents who are likely to be impulsive andhostile and who have a history of significantantisocial behavior. ANT comprises three subscalesintended to measure distinct facets of psychopathicsymptomatology. Antisocial Behaviors (ANT-A) tapsthe characteristic history of adolescent conductproblems and adult antisocial behaviors, whereasStimulus-Seeking (ANT-S) reflects a desire to takerisks and need for novelty. Egocentricity (ANT-E)is purported to measure “the pathological ego-centricity and narcissism often thought to lie at thecore of this disorder” (Morey, 1991, p. 72).

Evidence supporting the reliability (test-retestand internal consistency) of ANT has been reportedin the manual (Morey, 1991), as have validationaldata regarding its correlations with other self-reportmeasures of psychopathic traits (see also Edens,Poythress, et al., 2001; Salekin et al., 1997). Notedearlier, partial support for the concurrent validity ofthis scale also has been reported by Edens et al.(2000), in that ANT correlates moderately highlywith PCL-based measures of psychopathy.

Psychopathy Checklist-Revised. The PCL-R(Hare, 1991) is a 20-item rating scale of psychopathicpersonality traits. Items are scored on a 3-point scale,based on the extent to which they are applicable tothe individual (0 = does not apply, 1 = partially

applies, 2 = definitely applies). Items typically arerated on the basis of a clinical interview and a reviewof case history information. A total score, rangingfrom 0 to 40, reflects the extent to which theexaminee matches the description of a prototypicalpsychopath. A cut off of 30 and higher typically isused to delineate “psychopathy.” Items also aresummed to yield two subscale scores. Factor 1represents interpersonal and affective characteristicsof psychopathy such as lack of remorse and empathy,egocentricity, and manipulativeness, whereas Factor2 addresses behavioral and lifestyle factors such asimpulsivity and antisocial behaviors. The psycho-metric properties of the PCL-R are well established(Fulero, 1995; Hare, 1991). Coefficient alphas in the

1 Forty-three inmates had been administered both the PAI andthe PCL-R. However, 6 of the PAI protocols were invalid. Theseoffenders were dropped from further analyses.

52 Edens, Buffington-Vollum, Colwell, Johnson, & Johnson

present sample for the total, Factor 1, and Factor 2scores were .89, .88, and .84, respectively. Meaninter-item correlations were .29, .47, and .39.2

Consistent with research cited in the PCL-R manual,the correlation between Factor 1 and Factor 2 in thissample was .46.

Disciplinary Reports (DRs). The Texas Depart-ment of Criminal Justice (TDCJ) Disciplinary Rules

and Procedures for Offenders manual contains alisting of “Disciplinary Offenses” that define variousforms of (mis)behavior that will result in institutionalsanctions and penalties. These vary in the extent towhich they reflect violence or other behavior that isof major concern to prison staff and officials. Usinga procedure similar to Edens, Poythress, andLilienfeld (1999), we developed with TDCJ staff ana priori classification scheme that categorized eachspecific infraction into one of three categories:Physical Aggression (PA; e.g., “Assaulting anofficer,” “Fighting or assaulting an offender with aweapon”), Verbal Aggression/Acts of Defiance (VA;e.g., “Threatening to inflict harm on an officer,”“Refusing to work,” “Creating a disturbance”), orNon-aggressive infractions (NA; e.g., “Gambling,”“Lying to a staff member,” “Stealing”).3 Moreover,within these categories, offenses are defined byTDCJ as either major or minor, depending onwhether a formal administrative hearing is requiredand the type of penalty that can be imposed for thespecific infraction. For purposes of the presentinvestigation, we chose to focus on major rather thanminor offenses, given that they represent moreserious infractions that result in more severe penalties(e.g., solitary confinement).

PROCEDURE

Noted earlier, this is an archival study in whichtest data were coded from institutional files ofincarcerated sex offenders who had undergone apsychological evaluation in which they wereadministered the PCL-R and the PAI. Test adminis-tration for both samples that were combined for thepresent study involved essentially identical

procedures. PCL-R ratings were made by doctorallevel psychologists under contract with TDCJ whohad extensive training in the assessment ofpsychopathy. Scores were based on a review ofinstitutional files and a detailed clinical interview,in accordance with the recommendations of the testmanual (Hare, 1991). Offenders were administeredthe PAI according to standardized procedures.Protocols were scored using a commercial scoringprogram. A research assistant coded PCL-R, PAI anddemographic data from institutional files and entereddisciplinary reports from a TDCJ computer database.Disciplinary reports were recoded from their originalnumeric values (e.g., 10.1 “Unauthorized contactwith a victim”) into the broader PA, VA, and NAcategories by means of an SPSS conversion program.The time frame of the present study examined thoseoffenses that occurred in the two years prior to thepsychological evaluation.

RESULTS

Preliminary Analyses

The addition of 37 offenders did not substantiallychange the relationship between the PCL-R and ANTthat was reported in the original study (Edens et al.,2000). In the expanded sample of 92 offenders, thePearson r between ANT and the PCL-R was .41 (p <.001), whereas the ROC curve ascertaining thediagnostic accuracy of the ANT scale to predictdiagnoses of psychopathy (i.e., PCL-R scores > 30)also was consistent with earlier findings (AUC = .70,SE = .08, 95% confidence interval = .55 - .86, p =.01).

Regarding the primary criterion measures ofinterest, frequency counts indicated that the totalnumber of major DRs committed during the priortwo years was relatively low (M = 1.34, SD = 3.87).4

2 These analyses are based on 63 examinees, because the filedata did not include scores for individual items in 29 cases. Also,factor scores were not provided in the files of 12 inmates.3 Copies of this coding scheme are available from the first author.

4 Although not the primary focus of the present study, we alsoexamined the mean number of total disciplinary infractions(major and minor) committed during the two years prior to theevaluation, which was 3.69 (SD = 7.13). Both the PCL-R andANT were significantly correlated with this criterion (Pearson r= .40 and .36, respectively, p < .001). Because the distributionof DRs was positively skewed, we computed SpearmanCorrelations as well, which resulted in somewhat differentresults: PCL-R/DR, r = .47, p < .001, ANT/DR, r = .26, p = .01.

Comparison of the PAI and PCL-R 53

Therefore, infraction data were recoded dichoto-mously to indicate whether the offender either hador had not committed at least one offense. Base ratesfor each type of major infraction were: ANY =28.3%; PA = 13.0%; VA = 23.9%; NA = 20.7%.

PRIMARY ANALYSES

In order to assess the relationship between thepsychopathy scales and criterion measures, we firstcomputed point biserial correlations between thePCL-R and its subscales, ANT, and the fourcategories of major disciplinary reports (see Table1).5 Only two correlations were non-significant. ANTfailed to correlate significantly with PA DRs (p =.06, two-tailed), and Factor 1 of the PCL-R did notcorrelate with NA DRs (p = .13, two-tailed). It shouldbe noted that the magnitude of point biserialcorrelations, aside from being influenced by thedistribution of scores on the continuous variable, isalso a function of the base rate of the dichotomousvariable being examined. For example, with a baserate of 10%, the upper limit of this correlation isapproximately .58 (Nunnally & Bernstein, 1994, p.135). As such, other factors held constant, low baserate correlations are lower in magnitude than thosewhere the base rate is closer to .50.

In order to examine whether any of the obtainedcorrelations across the four criterion measures werestronger for ANT or the PCL-R, we conductedCohen’s t tests for dependent correlations (Cohen &Cohen, 1983). Given the small magnitude of thedifferences for the VA, NA, or ANY categories, itwas not surprising that none of these comparisonswas significant (ts between .30 and .86, all p values> .20). Also, despite the difference between the PCL-R/PA and ANT/PA correlations, this comparison onlyapproached statistical significance (t = 1.49, p = .07).Although not the primary focus of this study, similarresults were obtained when comparing Factor 1 and

Factor 2 of the PCL-R, with none of the differencesachieving statistical significance (ts ranging from .81to 1.49, all p values > .05).

Table 1Point Biserial Correlations between the PCL-R, ANT,

and Disciplinary Infractions

ANT PCL-R Factor 1a Factor 2a

Infraction TypeANY .39*** .35*** .25* .35**PA .20 .36*** .25* .34**VA .42*** .39*** .28* .40***NA .39*** .30** .17 .30**

Note. *p < .05. **p < .01. ***p < .001 (all two-tailed).aDue to missing data, n for these analyses was 80.PCL-R = Psychopathy Checklist-Revised. ANT =Antisocial Features. ANY = Any major disciplinaryinfraction. PA = Physically aggressive major disciplinaryinfraction. VA = Verbally aggressive/Act of Defiancemajor disciplinary infraction. NA = Non-aggressive majordisciplinary infraction.

Because of some of the intrinsic limitations ofcorrelations, we also conducted Receiver OperatingCharacteristic (ROC) analyses to examine thepotential utility of both the PCL-R and ANT todifferentiate between those who did versus those whodid not receive major disciplinary infractions. ROCanalyses entail the calculation of the sensitivity andfalse positive (1 – specificity) rates associated withall possible cut off scores on the predictor variable,plotting these values, and connecting them to forma curve (for a more extensive discussion, see Centor,1991). The elevation of the curve, which typically isindexed by the area under the ROC curve (AUC),reflects the diagnostic efficiency of the predictorvariable across its entire range of scores. In brief, itreflects the probability that a randomly selectedoffender who committed an infraction will have ahigher score on the psychopathy measure than arandomly selected offender who did not receive adisciplinary infraction. An AUC of .5 indicates a 50%likelihood that the offender who committed aninfraction will have a higher score than one who didnot commit one. Scores greater than .5 indicateabove-chance accuracy, and an AUC of 1.0represents perfect discrimination between the twogroups.

5 Although separate types of offenses, the VA and NA categorieswere significantly interrelated, in that many offenders who hadcommitted one type of infraction also had committed the other(Kappa = .78, p < .001). As such, correlations with the predictorvariables would be expected to be similar across these twovariables, as well as the ANY DR category, which is an aggregateof all three types of infractions (VA/ANY, Kappa = .89, p <.001; NA/ANY, Kappa = .80, p < .001).

54 Edens, Buffington-Vollum, Colwell, Johnson, & Johnson

Table 2Summary of ROC Analyses for PCL-R and ANT Predictions of Disciplinary Infractions

PCL-R ANT

AUC SE 95% CI AUC SE 95% CI

Infraction TypeANY .72*** .06 .60 - .83 .71** .07 .58 - .84PA .79*** .07 .66 - .91 .66 .09 .48 - .83VA .75*** .06 .63 - .87 .74*** .07 .61 - .87NA .70** .07 .57 - .84 .73** .07 .59 - .87

Note. PCL-R = Psychopathy Checklist-Revised. ANT = Antisocial Features. ANY = Any major disciplinary infraction.PA = Physically aggressive major disciplinary infraction. VA = Verbally aggressive/Act of Defiance major disciplinaryinfraction. NA = Non-aggressive major disciplinary infraction. AUC = Area under the curve. SE = Standard error. CI =Confidence interval. *p < .05. **p < .01. ***p < .001 (all two-tailed).

All ROC analyses were performed using SPSSfor Windows, release 9.0.0. The nonparametricmethod was used to calculate AUCs. Table 2summarizes the results of these analyses for the fourcriterion measures of interest. Except for the ROCcurves for PA DRs, the magnitude of the obtainedAUCs was quite similar for both measures. In orderto examine whether the PCL-R was significantlymore strongly related to PA DRs, the differencebetween the AUC values was compared using aprocedure described by Hanley and McNeil (1983,see formula 3), which corrects for comparisons ofAUCs that are derived from the same sample. Acritical ratio z value of 1.45 was obtained, indicatingthat—consistent with the results of the point biserialcorrelations—the AUC for the PCL-R was notsignificantly higher than for ANT (p = .14). Theseanalyses were repeated for the remaining threecriterion measures, with critical ratio z values rangingfrom .10 to .33 (all p values > .50).

Although these results indicate that both thePCL-R and ANT generally accounted for similaramounts of variance in these criteria, they do notaddress whether either makes a unique contributionin terms of the amount of variance explained. Toaddress the issue of incremental validity (see, e.g.,Sechrest, 1963), we first performed two linearregression analyses in which the PCL-R was used topredict ANT scores and vice versa. The standardizedresiduals from these two analyses represent thevariance in each measure that is independent of theother, that is, variance in PCL-R scores that isunrelated to ANT scores and variance in ANT scores

that is independent of PCL-R scores. Point biserialcorrelations were then computed to determinewhether the residuals for either measure wereassociated with the DR categories. As can be seen inTable 3, once the variance attributable to one measureof psychopathy was partialled out, the correlationsbetween the other measure and DRs were somewhatattenuated, but remained significant in most cases.

Table 3Point Biserial Correlations Between Standardized

Residuals for Each Psychopathy Measure and

Disciplinary Reports

ANT PCL-R

Infraction TypeANY .27** .21*PA .06 .30**VA .29** .23*NA .28** .16

Note. *p < .05. **p < .01. (all two-tailed).Standardized residuals represent the variance in eachpsychopathy measure that is independent of the othermeasure. PCL-R = Psychopathy Checklist-Revised. ANT= Antisocial Features scale. PA = Physical Aggression.VA = Verbal Aggression/Acts of Defiance. NA = Non-aggressive.

Although ROC curves and correlational analysesare informative at a global level, the classificationaccuracy of established cut scores also are relevantin determining the practical utility of a measure.Table 4 summarizes results for commonly used

Comparison of the PAI and PCL-R 55

Table 4Accuracy of PCL-R and ANT Cut Scores in Detecting a History of Disciplinary Infractions

Infraction Sensitivity Specificity PPP NPP Hit Rate KappaType

PCL-R > 30

ANY .35 .89 .56 .78 .74 .27**PA .50 .88 .38 .92 .83 .33**VA .41 .90 .56 .83 .78 .34***NA .37 .88 .44 .84 .77 .26*

ANT T > 70

ANY .67 .83 .54 .89 .79 .46***PA .29 .92 .50 .81 .77 .24*VA .62 .87 .59 .89 .82 .48***NA .48 .87 .53 .85 .78 .36***

Note. *p < .05. **p < .01. ***p < .001 (all two-tailed).PCL-R = Psychopathy Checklist-Revised. ANT = Antisocial Features. ANY = Any major disciplinary infraction. PA =Physically aggressive major disciplinary infraction. VA = Verbally aggressive/Act of Defiance major disciplinaryinfraction. NA = Non-aggressive major disciplinary infraction. PPP = Positive Predictive Power. NPP = NegativePredictive Power.

measures of classification accuracy for both the PCL-R and ANT. As can be seen, Kappa values for all ofthese analyses were significant. Both PCL-R > 30and ANT > 70 generally resulted in relatively fewmisidentifications of offenders as being likely toengage in institutional misconduct, but alsoaccurately detected only 35% to 67% of those whodid in fact commit an offense of some type in thepreceding two years.

DISCUSSION

Consistent with earlier research on other forensicand correctional samples (see, e.g., Edens, Petrila,et al., in press; Hart & Hare, 1997), these resultssupport the validity of psychopathy as a significantcorrelate of serious disciplinary misconductspecifically among sex offenders. Given therelatively limited research that has been conductedon the institutional adjustment of this particularsubgroup of offenders, these findings are importantin that they indicate that psychopathic traits areassociated—not only with recidivism among sexoffenders (e.g., Hemphill et al., 1998; Seto &

Barbaree, 1999)—but also with institutionalmisbehavior while incarcerated. This suggests thatthese measures of psychopathy may provide usefulinformation in terms of better managing ‘high risk’sex offenders while they are institutionalized.Moreover, the fact that these measures wereadministered as part of a forensic evaluation, asopposed to a research study, lends credence to theargument that our findings would generalize to other‘real world’ settings. That is, the instruments andinterviews were not completed as part of aconfidential research project in which the offenders’responses would have no impact on their disposition,which might influence the candor of their responding(see, e.g., Edens, Buffington, et al., 2001).

These results also suggest that, although thePCL-R and ANT generally accounted for similaramounts of variance in the criterion measures, theywere not accounting entirely for the same variance.That is, both evidenced incremental validity over theother measure in terms of their association with threeof the four DR categories. This is perhaps notsurprising, given that the correlation between thePCL-R and ANT was only .41 in the present sample.It also suggests that the combined use of these two

56 Edens, Buffington-Vollum, Colwell, Johnson, & Johnson

measures might lead to greater classificationaccuracy in other samples.

In terms of the efficacy of dichotomousclassifications on the psychopathy measures, thesensitivity of commonly used cut scores (i.e., PCL-R > 30 and ANT > 70) was not exceedingly high foreither scale in this sample, indicating that asignificant number of individuals who obtain scoresabove these cut offs do not have a history of seriousmisbehavior during the preceding two years.However, the high specificity rates for both measuressuggest that they may be particularly useful for rulingout (low scoring) offenders who are not likely toengage in disciplinary infractions. This is generallyconcordant with most research on the PCL-R, in thatlow scores more consistently have been a betterindicator of the absence of various forms ofmisconduct, whereas high scores have been lessconsistently associated with the occurrence ofmisbehavior (Salekin, Rogers, & Sewell, 1996).

If replicated in a prospective study, these resultswould provide strong support for the utility of thePAI as a predictor of important forensic andcorrectional outcome measures, as well as challengethe notion that all self-report measures of psycho-pathy are of limited clinical utility in predictingvarious forms of violence and antisocial conduct, atleast in controlled institutional environments (seealso Edens, Poythress, et al., 2001; Wang &Diamond, 1999). Although the relationship betweenPCL-defined psychopathy and institutional mis-conduct is relatively well established (although,noted earlier, the magnitude of this association insome studies has been quite small), this is one ofonly a few studies to examine the utility of the PAIas an indicator of institutional misbehavior (Salekinet al., 1997; Wang & Diamond, 1999). Collectivelythese studies do suggest, however, that ANT scoresare associated with significant misconduct acrossvarious offender subgroups. For example, Edens,Poythress, et al. (2001) found that the ANT scale isa fairly robust correlate of prison misbehavior amonggeneral population and mentally disordered offenders(total n = 60), with point biserial correlations rangingfrom .26 to .55 for various types of violent and non-violent infractions.

The generally positive conclusions of this studyshould be qualified by several caveats. First, results

are based on a relatively small n, particularly in termsof the total number of inmates who committedphysically violent infractions. Second, this studyexamined a specific subsample of inmates (i.e., malesex offenders), which may limit the generalizabilityof these findings to other forensic and correctionalpopulations. However, as noted earlier, research onother subsamples of offenders (e.g., women, mentallydisordered inmates) suggests that our results are notunique to sex offenders. Third, we used a relativelyshort follow-back period (two years), which mayexplain the relatively low base rate of infractions,particularly physically aggressive infractions.Despite this brief period, however, significant effectswere obtained for most of the correlations betweenthe predictor and criterion measures examined.

A final, and perhaps the most significant,limitation is that this was a retrospective study inwhich the outcome measures occurred prior to theassessment. This leaves open the possibility ofcriterion contamination regarding the PCL-R ratings,in that the examiners’ evaluations of the inmates mayhave been influenced somewhat by their disciplinaryinfraction history. In some ways this places the PAIat a competitive disadvantage because its criterion-related validity is being compared to another measurethat may have been biased somewhat by the criterionof interest (see Cunningham & Reidy, 1998, orEdens, Petrila, et al., in press, for a related discussionof this issue in reference to PCL-R research). Thefact that the PCL-R generally did not performappreciably better than the ANT scale despite thispotential contamination provides rather strongsupport for the utility of ANT, at least in thisparticular study. Nevertheless, prospective researchexamining disciplinary infractions is needed todetermine more definitively the utility of the PAI toidentify sex offenders who are at increased risk forserious institutional misbehavior.

REFERENCES

Cashel, M. L., Rogers, R. R., & Sewell, K. W. (1995). ThePersonality Assessment Inventory (PAI) and the detectionof defensiveness. Assessment, 2, 333-342.

Centor, R. M. (1991). Signal detectability: The use of ROCcurves and their analyses. Medical Decision Making, 11,

102-106.

Comparison of the PAI and PCL-R 57

Civil Commitment of Sexually Violent Predators Act, TexasHealth & Safety Code Ann. §§ 841.001-841.147 (1999).

Cleckley, H. (1976). The mask of sanity (5th ed.). St. Louis, MO:Mosby.

Clements, C.B. (1996). Offender classification: Two decades ofprogress. Criminal Justice and Behavior, 23, 121-143.

Cohen, J., & Cohen, P. (1983). Applied multiple regression/

correlation analysis for the behavioral sciences (2nd ed.).Hillsdale, N.J.: Erlbaum.

Cooke, D. J., & Michie, C. (2001). Refining the construct ofpsychopathy: Towards a hierarchical model. Psychological

Assessment, 13, 171-188.Cooney, N. L., Kadden, R. M., & Litt, M. D. (1990). A

comparison of methods for assessing sociopathy in maleand female alcoholics. Journal of Studies on Alcohol, 51,

42-48.Cunningham, M. D., & Reidy, T. J. (1998). Antisocial personality

disorder and psychopathy: Diagnostic dilemmas inclassifying patterns of antisocial behavior in sentencingevaluations. Behavioral Sciences and the Law, 16, 333-351.

Douglas, K. S., Hart, S. D., & Kropp, P. R. (2001). Validity ofthe Personality Assessment Inventory for forensicassessments. International Journal of Offender Therapy

and Comparative Criminology, 45, 183-197.Edens, J. F., Buffington, J. K., Tomicic, T. L., & Riley, B. D.

(2001). Effects of positive impression management on thePsychopathic Personality Inventory. Law and Human

Behavior, 25, 235-256.Edens, J. F., Cruise, K. R., & Buffington-Vollum, J. K. (2001).

Forensic and correctional applications of the PersonalityAssessment Inventory. Behavioral Sciences and the Law,

19, 519-543.Edens, J. F., Hart, S. D., Johnson, D. W., Johnson, J., & Olver,

M. E. (2000). Use of the Personality Assessment Inventoryto assess psychopathy in offender populations. Psycho-

logical Assessment, 12, 132-139.Edens, J. F., Petrila, J., & Buffington-Vollum, J. K. (in press).

Psychopathy and the death penalty: Can the PsychopathyChecklist-Revised identify offenders who represent “acontinuing threat to society?” Journal of Psychiatry and

Law.

Edens, J. F., Poythress, N. G., & Lilienfeld, S. O. (1999).Identifying inmates at risk for disciplinary infractions: Acomparison of two measures of psychopathy. Behavioral

Sciences and the Law, 17, 435-443.Edens, J. F., Poythress, N. G., & Watkins, M. M. (2001). Further

validation of the Psychopathic Personality Inventory amongoffenders: Personality and behavioral correlates. Journal

of Personality Disorders,15, 403-415.Edens, J. F., Skeem, J. L., Cruise, K. R., & Cauffman, E. (2001).

Assessment of ‘juvenile psychopathy’ and its associationwith violence: A critical review. Behavioral Sciences and

the Law, 19, 53-80.Fulero, S. (1995). Review of the Hare Psychopathy Checklist-

Revised. In J. C. Conoley, J. C., Impara, & L. L. Murphy(Eds.), Twelfth mental measurements yearbook (pp. 453-454). Lincoln, NE: Buros Institute.

Hanley, J. A., & McNeil, B. J. (1983). A method of comparingthe areas under receiver operating characteristic curvesderived from the same cases. Radiology, 148, 839-843.

Hare, R. D. (1985). Comparison of procedures for the assessmentof psychopathy. Journal of Consulting and Clinical

Psychology, 53, 7-16.Hare, R. D. (1991). The Hare Psychopathy Checklist-Revised.

Toronto: Multi-Health Systems.Hart, S. D., Cox, D. N., Hare, R. D. (1995). Manual for the

Screening Version of the Psychopathy Checklist (PCL:SV).Toronto: Multi-Health Systems.

Hart, S. D., Forth, A. E., & Hare, R. D. (1991). The MCMI-IIand psychopathy. Journal of Personality Disorders, 5, 318-327.

Hart, S. D., & Hare, R. D. (1997). Psychopathy: Assessmentand association with criminal conduct. In D. M. Stoff, J.Breiling, & J. D. Maser (Eds.), Handbook of antisocial

behavior (pp. 22-35). New York: Wiley.Heilbrun, K., Hart, S. D., Hare, R. D., Gustafson, D., Nunez,

C., & White, A. J. (1998). Inpatient and postdischargeaggression in mentally disordered offenders: The role ofpsychopathy. Journal of Interpersonal Violence, 13, 514-527.

Hemphill, J. F., Hare, R. D., & Wong, S. (1998). Psychopathyand recidivism: A review. Legal and Criminological

Psychology, 3, 139-170.Morey, L. C. (1991). The Personality Assessment Inventory:

Professional manual. Odessa, FL: Psychological Asses-sment Resources.

Nolan, K. A., Volavka, J., Mohr, P., & Czobor, P. (1999).Psychopathy and violent behavior among patients withschizophrenia or schizoaffective disorder. Psychiatric

Services, 50, 787-792.Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory

(3rd ed.). New York: McGraw-Hill.Rice, M. E., & Harris, G. T. (1995). Psychopathy, schizophrenia,

alcohol abuse, and violent recidivism. International Journal

of Law and Psychiatry, 18, 333-342.Salekin, R. T., Rogers, R., & Sewell, K. W. (1996). A review

and meta-analysis of the Psychopathy Checklist andPsychopathy Checklist-Revised: Predictive validity ofdangerousness. Clinical Psychology: Science and Practice,

3, 203-215.Salekin, R. T., Rogers, R., & Sewell, K. W. (1997). Construct

validity of psychopathy in a female offender sample: Amultitrait-multimethod evaluation. Journal of Abnormal

Psychology, 107, 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, 109-128.Sechrest, L. (1963). Incremental validity: A recommendation.

Educational and Psychological Measurement, 23, 153-158.Seto, M. C., & Barbaree, H. E. (1999). Psychopathy, treatment

behavior, and sex offender recidivism. Journal of

Interpersonal Violence, 14, 1235-1248.Seto, M. C., & Lalumiere, M. L. (2000). Psychopathy and sexual

aggression. In C. Gacono (Ed.), The clinical and forensic

assessment of psychopathy (pp. 333-350). Mahwah, NJ:Erlbaum.

58 Edens, Buffington-Vollum, Colwell, Johnson, & Johnson

Skeem, J. L., & Mulvey, E. P. (2001). Psychopathy and communityviolence among civil psychiatric patients: Results from theMacArthur Violence Risk Assessment Study. Journal of

Consulting and Clinical Psychology, 69, 358-374.Van Voorhis, P. (1994). Psychological classification of the adult

male prison inmate. Albany, NY: State University of NewYork Press.

Wang, E. W., & Diamond, P. M. (1999). Empirically identifyingfactors related to violence risk in corrections. Behavioral

Sciences and the Law, 17, 377-389.White, L. J. (1996). Review of the Personality Assessment

Inventory (PAI): A new psychological test for clinical andforensic assessment. Australian Psychologist, 31, 38-39.