Psychopathy versus psychopathies in classifying criminal offenders (original) (raw)
Psychopathy versus psychopathies in classifying criminal offenders
Jasmin Vassileva 1∗{ }^{1 *}, David S. Kosson 2{ }^{2}, Carolyn Abramowitz 2{ }^{2} and Patricia Conrod 3{ }^{3}
1{ }^{1} Department of Psychiatry, University of Illinois, Chicago, USA
2{ }^{2} Department of Psychology, Rosalind Franklin University of Medicine and Science, USA
3{ }^{3} National Addiction Centre, Institute of Psychiatry, King’s College London, UK
Abstract
Purpose. Psychopathy has been shown to be related to the onset, frequency, and course of antisocial behaviour in criminal offenders. The purpose of the present study was to use cluster analysis to explore the existence of subtypes of criminal offenders in male inmates, based on the two empirically validated dimensions of psychopathy and several other dimensions previously proposed for differentiating offender groups. Methods. Two hundred male inmates participated in the study. Scores on the two dimensions of the Psychopathy Checklist-Revised (Hare, 1991), the Interpersonal Measure of Psychopathy (Kosson, Steuerwald, Forth, & Kirkhart, 1997), DSM-IV diagnoses for alcohol and drug abuse/dependence, and anxiety were standardized and were included in two different types of cluster analyses. Both Ward’s hierarchical method and K-means non-hierarchical method revealed the presence of four subtypes of criminal offenders in the sample. The four-cluster solution was replicated when the sample was split in half and identical cluster analyses were performed on the two subsamples. Results. Two types of cluster analyses identified four subtypes of criminal offenders in two samples of jail inmates. Two of the clusters resembled primary and secondary psychopaths described in the literature, the third group exhibited some antisocial and psychopathic features, and the fourth group was non-psychopathic. Conclusions. Findings suggest that considering the individual contributions of the two dimensions of psychopathy in lieu of the construct as a whole may prove useful in identifying relatively homogeneous groups of criminal offenders.
Criminal offenders are a heterogeneous group. Not only do they differ in the frequency and nature of their antisocial behaviour, but they also vary widely in personality, psychiatric, and demographic characteristics. Identifying groups of offenders based on
- Correspondence should be addressed to Jasmin Vassileva, Department of Psychiatry (MC 912), University of Illinois at Chicago, 1601 West Taylor Street, Chicago, IL 60612, USA (e-mail: jvassileva@psych.uic.edu).
↩︎
- Correspondence should be addressed to Jasmin Vassileva, Department of Psychiatry (MC 912), University of Illinois at Chicago, 1601 West Taylor Street, Chicago, IL 60612, USA (e-mail: jvassileva@psych.uic.edu).
similarities in such characteristics is important for determining the aetiological mechanisms involved in antisocial behaviour, suitability for treatment, and for predicting dangerousness and recidivism. In order to address this variability, many distinct approaches to classifying offenders into more homogeneous groups have been proposed (Andrews, Bonta, & Hoge, 1990; Clements, 1996; Megargee & Bohn, 1979; Nagin, Farrington, & Moffitt, 1995).
Several offender taxonomies highlight the importance of a subset of offenders who commit a disproportionate share of the crimes. However, the nature of this group of offenders has been controversial, as evidenced by disagreements about whether it should be referred to as psychopathy, sociopathy, dyssocial, or antisocial personality disorder (ASPD); (Holtzworth-Munroe & Stuart, 1994; Knight & Prentky, 1990; Lykken, 1995). Although the Diagnostic and Statistical Manual of Mental Disorders (American Psychiatric Association, 1980) referred to these as synonyms, it is now evident that they differ in important ways. Whereas ASPD describes 50−75%50-75 \% of all prison inmates (Hare, Hart, & Harpur, 1991; Harpur & Hare, 1994), only 20−30%20-30 \% of inmates meet the diagnostic criteria for psychopathy embodied in the Hare Psychopathy ChecklistRevised (PCL-R; Hare, 1991), the most widely used instrument for measuring psychopathy. Given the high prevalence of ASPD in prison settings, some researchers have argued that psychopathy is more useful than ASPD in classifying criminal offenders (Stevens, 1993).
There are also controversies about whether psychopathy is a unitary syndrome or whether there are distinct subgroups among individuals with psychopathic features. Several researchers have argued that psychopathy is a unitary syndrome (Cooke & Michie, 2001; Harpur, Hare, & Hakstian, 1989). However, several factor analyses suggest that at least two distinct dimensions underlie PCL-R ratings (Hare, 1991; Harpur et al., 1989). One dimension (Factor 1) is said to reflect affective and interpersonal traits such as callousness, lack of empathy, shallow affect, and manipulativeness that are central elements in clinical descriptions of the psychopath (Cleckley, 1941/1988) and are commonly referred to as core personality features of the psychopath. A second dimension (Factor 2) is said to reflect behaviours indicative of an unstable and antisocial lifestyle, such as poor anger control, parasitic lifestyle, impulsivity, and irresponsible behaviour. Because the two factors correlate only .5-.6, they share only 25−36%25-36 \% of their variance. Importantly, these two dimensions correlate differentially with a wide variety of external criteria (Harpur et al., 1989). Factor 1 correlates more highly with narcissism, with some emotional processing anomalies, and with the distinctive interpersonal style of the psychopath than does Factor 2 (Harpur et al., 1989; Kosson et al., 1997). By contrast, Factor 2 is more strongly related to ASPD and to alcohol and substance use disorders than is Factor 1 (Harpur et al., 1989; Smith & Newman, 1990). The different relationships of the two dimensions of psychopathy with external variables raise the possibility that psychopathy may be better conceptualized as a multifaceted syndrome comprising qualitatively distinct subgroups of psychopaths. For example, Blackburn has noted that ‘among those categorized as psychopaths by the checklist, some will be personally deviant, some will be socially deviant, and some will be both. It therefore seems likely that the checklist tends to identify as psychopathic an antisocial group that is heterogeneous in personality deviation’ (Blackburn, 1988, p. 510).
Several theorists have posited distinct subgroups of individuals with psychopathic features on the basis of differences in anxiety. Specifically, ‘true’ or primary psychopaths are usually characterized by a lack of anxiety, whereas secondary psychopaths may
appear psychopathic, yet experience substantial anxiety and negative affect which may contribute to their impulsive antisocial behaviour (Blackburn, 1975; Karpman, 1941; Lykken, 1995). Yet, the relationship between psychopathy and anxiety remains inconsistent at best. PCL-R total scores are typically independent of but sometimes positively correlated with scores on measures of anxiety or fear (Hale, Goldstein, Abramowitz, Calamari, & Kosson, 2004; Hare, 1991, 2003; Schmitt & Newman, 1999). In addition, whereas some studies have found that the two dimensions underlying psychopathy correlate in opposite directions with measures of anxiety (Hale et al., 2004; Harpur et al., 1989; Patrick & Lang, 1999), others have failed to find consistent differential relationships between measures of anxiety and the dimensions of psychopathy (Schmitt & Newman, 1999). Nonetheless, some cognitive deficits appear specific to psychopathic individuals low in anxiety (Lorenz & Newman, 2002a; Newman, Schmitt, & Voss, 1997), whereas others are specific to psychopathic individuals high in anxiety (Newman, Wallace, Schmitt, & Arnett, 1997).
Similarly, relationships between psychopathy, ASPD, and problems with alcohol and substance abuse are still a subject of debate. Correlations between ASPD and substance use disorders are well documented and sometimes considered important for distinguishing different subtypes of alcoholics (Cloninger, 1987; Lewis, Rice, & Helzer, 1983). Yet, among offenders, substance use disorders are generally unrelated to the core personality traits of the psychopath (Smith & Newman, 1990; Windle, 1999). Thus, there is controversy about whether specific patterns of substance abuse share a genetic diathesis with persistent antisocial behaviour (Reardon, Lang, & Patrick, 2002; Slutske et al., 1998), or whether the pattern of relationships is inconsistent with distinct subtypes of ASPD differing on substance-related problems (Bucholz, Hesselbrock, Heath, Kramer, & Schuckit, 2000).
Based on suggestions that the essence of personality disorders lies in disturbed interactions between an individual and others (e.g. Millon, 1981), Kosson et al. (1997) argued that improved assessments of interpersonal behaviour associated with psychopathy might also be helpful in reducing the heterogeneity of individuals with psychopathic features. To this end, they developed the Interpersonal Measure of Psychopathy (IM-P) in order to code non-verbal behaviours and aspects of interactions suggested to be common in interviews with psychopaths. Initial studies revealed that IM-P scores correlated more highly with the core affective and interpersonal components than with the social deviance components of psychopathy and correlated with non-violent antisocial behaviour in opposite ways from PCL-R Factor 2 (Kosson et al., 1997). Moreover, IM-P scores have been used in conjunction with PCL-based scores to improve prediction of social information processing deficits (Kosson, Suchy, & Cools, 2001). However, IM-P scores have not previously been employed in classification studies.
In summary, there exists ample theoretical and empirical literature both consistent with and inconsistent with the existence of specific subtypes of psychopathic offenders. Although prior studies based on the assumption of the existence of different subgroups of psychopaths have often revealed behavioural differences consistent with this assumption, most such studies have only subdivided participants based on the construct of negative affectivity. It is possible that studies employing other external constructs would lead to a different conclusion. In contrast, studies based on the theoretical assumption that psychopathy is unitary have obtained evidence consistent with this assumption. However, such studies have generally been restricted to internal analyses based on PCL-R items. It is possible that findings would be different if measures of other
constructs had been examined. Consequently, the current study was designed to address the homogeneity of psychopathy from an atheoretical perspective by using cluster analysis and by including a range of relevant variables suggested by theory and research.
Cluster analysis is a multivariate person-oriented approach, which analyses multiple relationships among variables at an individual level by identifying interactions and configurations that go beyond the pairwise correlations of variables (Haapasalo & Pulkkinen, 1992). Whereas most of the existing studies investigating the relationships between psychopathy and other theoretically-relevant variables have utilized a variableoriented approach, relatively few studies of psychopathy have applied the personoriented analysis. In a detailed review of variants of psychopathy, Skeem, Poythress, Edens, Lilienfeld, and Cale (2003) identified only three cluster analytic studies of the PCL/PCL-R. First, a cluster analysis using only the PCL items (Haapasalo & Pulkkinen, 1992) revealed three subtypes of non-violent offenders: a group characterized by the core personality traits of Factor 1, another group distinguished by the antisociality of Factor 2, and a third group whose members scored low on both dimensions. A second cluster analysis of methadone patients (Alterman et al., 1998) used four measures of antisociality for cluster derivation: PCL-R, DSM-IV criteria for conduct disorder; DSM-IV criteria for antisocial personality disorder; and the Socialization (So) Scale from the California Psychological Inventory. The analysis identified six groups, three of which evidenced relatively high scores on the PCL-R in conjunction with high scores on the other measures of antisociality, and only one of which was characterized by high scores on the PCL-R but not on any of the remaining antisociality variables. Third, Herve, Ling, and Hare (2000) conducted a cluster analysis of archival PCL-R data using the three-factor model of psychopathy (Cooke & Michie, 2001). They identified three psychopathic clusters macho, manipulative, and prototypical - and one sociopathic cluster.
Finally, two more cluster analytic studies that did not directly use the PCL-R for cluster derivation are relevant. A cluster analysis of forensic patients based on DSM-III personality disorder criterion sets suggested six clusters, including three with relatively high PCL-R scores, that differed on anxiety, depression, social withdrawal, impulsivity, and aggression (Blackburn & Coid, 1999). Similarly, a cluster analysis of patients based on Cleckley’s psychopathy criteria and the dimensions of aggressiveness and sociabilitywithdrawal yielded two distinct clusters of individuals meeting Cleckley’s criteria: a group characterized by both the affective and behavioural features of psychopathy and a less impulsive and less aggressive group prone to social withdrawal (Blackburn & Maybury, 1985). Thus, several studies suggest that there may be distinct subgroups characterized by different subsets of psychopathic traits. However, to our knowledge, no prior cluster analytic studies have included the two validated dimensions underlying PCL-R scores as separate measures for cluster derivation. Further, none of the previous cluster analytic studies have examined the interactive effects of psychopathy and other theoretically relevant variables in identifying subgroups of criminal offenders.
Therefore, the current study was designed to extend prior cluster analyses of criminal offenders by considering the separate contributions of the two dimensions of psychopathy and several additional measures proposed to distinguish meaningful subgroups of offenders: anxiety, alcohol and drug misuse, and distinct interpersonal behaviours emblematic of psychopathy. Based on evidence that the two validated dimensions underlying the PCL-R correlate differently with a variety of external correlates, we hypothesized that distinct clusters of inmates would emerge based on their association with each of the PCL-R factors. We hypothesized that one cluster would consist of individuals characterized predominantly by the social deviance dimension of
psychopathy (PCL-R Factor 2) in conjunction with pronounced alcohol and/or drug dependence. A second distinct cluster was expected to consist of individuals characterized by higher scores on the personality features dimension of psychopathy (PCL-R Factor 1) as well as by low levels of trait anxiety. This cluster was also expected to exhibit high scores on the IM-P, a relatively new measure used to supplement the PCL-R in assessing the interpersonal affective core of psychopathy (Kosson, Gacono, & Bodholt, 2000; Kosson et al., 1997). We further expected to identify at least one group of offenders lacking psychopathic traits.
Finally, because clustering analyses always yield cluster solutions, we evaluated the validity of the resulting solution following several criteria: (a) we examined the robustness of the solution by splitting the sample into two halves and comparing the solutions, (b) we compared the solutions obtained with two different clustering algorithms (hierarchical and non-hierarchical), and © we examined whether the clusters differed on available external measures of criminal behaviour and demographic variables. We expected that the groups would differ in charges for violent offences; however, due to the inconsistent relationships between specific dimensions of psychopathy and measures of violence (e.g. Harpur & Hare, 1996; Kosson et al., 1997; Salekin, Rogers, & Sewell, 1996; Serin, 1996), no specific hypotheses were made with regard to which cluster(s) would exhibit greater violence.
Method
Participants
Participants were 200 male county jail inmates. All were between 17 and 39 years old, sentenced for felony convictions, incarcerated for 2 weeks or more, and had complete data on the six measures used to derive the clusters. Exclusion criteria included current use of psychotropic medication, psychotic symptoms, and estimated IQ scores below 70. The sample appeared representative of samples in prior psychopathy studies in the USA (mean age =26.10,SD=6.37=26.10, S D=6.37; mean years of education =11.00,SD=2.17=11.00, S D=2.17; mean estimated WAIS-R full scale IQ =91.7=91.7 ). The sample was 54%54 \% African-American, 41%41 \% Caucasian, and 5% Latino. The mean PCL-R score of 24.96(SD=7.06)24.96(S D=7.06) is similar to that reported for inmate samples (Hare, 2003).
Measures used for cluster derivation
The Hare Psychopathy Checklist-Revised (PCL-R; Hare, 1991) is a 20-item checklist designed as an objective measure of psychopathy. PCL-R ratings are typically based on interview information and available file information. Items are scored on a 3-point ordinal scale ( 0=0= definitely does not apply; 1=1= may or may not apply; 2=2= definitely applies). For present purposes, only scores on the eight items comprising Factor 1 and the nine items comprising Factor 2 were used. Ample data attest to the reliability and validity of the two PCL-R factors (Hare, 1991; Harpur et al., 1989). Although a threefactor structure for the PCL-R has been suggested (Cooke & Michie, 2001), external correlates for these three factors are not well established. Moreover, recent analyses suggest four distinct factors that load on two higher-order factors nearly identical to the original two-factor solution (Hare, 2003). For this reason, we selected the two factors that have received substantial empirical validation. In the current sample, inter-rater reliability for PCL-R total scores was adequate, as measured by mean weighted intra-class
correlation (ICC) of .87 across two independent raters, and inter-rater correlations of .69 and .78 for Factor 1 and Factor 2 scores.
The Interpersonal Measure of Psychopathy (IM-P; Kosson et al., 1997) was designed to measure specific interpersonal features of psychopathy based on actual interpersonal interactions between participants and interviewers. It is not a substitute for the PCL-R but an additional measure of the personality core of psychopathy. The interview conducted to complete the PCL-R typically provides the interpersonal data addressed by the IM-P. However, the concrete non-verbal behaviours and interactions on which IM-P scores are based (e.g. interruption of interviewer, incorporation of interviewer into personal stories) are not redundant with scoring criteria for any PCL-R items. For each of the 21 items, raters decide whether the trait or behaviour described the participant not at all (0), somewhat (1), very well (2), or perfectly (3), and scores are summed to yield total scores. The scale has high internal consistency and adequate validity (Kosson et al., 1997, 2000). Importantly, the IM-P correlates more highly with PCL-R Factor 1 than with Factor 2 (Kosson et al., 1997). In the current sample, the IM-P correlated .56 with Factor 1 and .24 with Factor 2, and the ICC for IM-P scores was .85 . The inter-rater reliability for IM-P total scores in the current sample was .76 .
Alcohol and drug abuse and dependence were assessed with the substance use module of the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I; First, Gibbon, Spitzer, & Williams, 1995). Each participant was given two ordinal ratings corresponding to DSM-IV diagnoses related to alcohol and to the other psychoactive substance with the greatest negative impact on his life. Alcohol and substance use symptoms were rated: 0=0= no abuse or dependence; 1=1= abuse; 2=2= mild dependence; 3=3= moderate dependence; 4=4= severe dependence (adapted from the DSM-IV; APA, 1994).
The Trait subscale of the State-Trait Anxiety Inventory scale (STAI-T; Spielberger, 1983) measures stable individual differences in anxiety proneness and is a widely used measure of trait anxiety. Its 20 items are rated on 4-point Likert scales, and total scores range from 20 to 80 . The STAI-T has relatively high reliability, internal consistency, and validity for identifying persons with high levels of neurotic anxiety (Spielberger, 1983).
Measures used for external validation
Demographic characteristics
We examined: age; ethnicity; years of education completed; intelligence, estimated by the Shipley Institute of Living Scale-Revised (SILS-R; Zachary, 1991); and handedness (Chapman & Chapman, 1987).
Criminal behaviour
Institutional file review and interview forms were inspected to obtain the following indices of criminal behaviour: number of non-violent charges; number of violent charges; number of incarcerations; number of symptoms of childhood conduct disorder (CD), using DSM-IV criteria; and criminal versatility as assessed by PCL-R item 20, modified to include both adult offences listed in criminal records and other adult offences acknowledged during the interview. This item is not part of either Factor 1 or Factor 2 in the original two-factor model of psychopathy.
Procedures
Inmates were contacted by telephone and informed of the purpose of the study. Those who agreed to participate and provided informed written consent completed a semistructured interview for use in completing the PCL-R and the IM-P, a semistructured interview assessing alcohol and drug abuse/dependence, a brief test of intelligence, and self-report measures of trait anxiety and handedness. Inmates were paid 55\5 5 for their time. After the interview, institutional file records were reviewed. Both interviewers and observers were graduate students in clinical psychology who had received extensive training in psychopathy, as well as substantial practice in administering and scoring the PCL-R and the IM-P.
Results
All statistical analyses were performed on SPSS 9.0 for Windows.
Data screening
Prior to analysis, PCL-R Factor 1, PCL-R Factor 2, IM-P, alcohol abuse/dependence, drug abuse/dependence, and trait anxiety scores were examined for multicollinearity. No variables exhibited a conditioning index greater than 30 nor were two or more variance proportions greater than .50 per root number, suggesting there were no redundant variables (Tabachnick & Fidell, 1996). There were no outliers, defined as cases with standardized scores in excess of 3.29(p<.001)3.29(p<.001) on any measures included in cluster derivation (Tabachnick & Fidell, 1996).
Cluster derivation
Initial clusters for the full sample were derived with Ward’s agglomerative hierarchical squared Euclidean distance method, using PCL-R Factor 1, PCL-R Factor 2, IM-P, alcohol abuse/dependence, drug abuse/dependence, and STAI-T scores as the variables for cluster derivation. Standardized data were used to minimize bias caused by differences in scale measurement. The number of clusters to be retained was determined using the following criteria: (1) visual examination of the dendrogram, which revealed four clearly identified clusters; (2) the percentage changes in agglomeration coefficient (sum of the squared distances between the clusters) from one step to the next for 2-9 clusters were 23.6, 13.3, 15.2,10.9,10.5,9.6,10.315.2,10.9,10.5,9.6,10.3, and 6.1 , suggesting that a four-cluster solution was optimal, as a rather large change in sum of squared distances occurs at the next step, indicating that no further splitting of clusters should be performed (Hair, Anderson, Tatham, & Black, 1995); (3) plotting the agglomeration coefficients using Morey, Blashfield, and Skinner’s (1983) scree plot procedure, which also suggested that the departure from linearity in the progression of values occurred at the step producing four clusters.
To ascertain whether the cluster solution obtained with the full sample would replicate across samples, the full sample was randomly split in half. Independent samples tt-tests indicated that there were no significant differences between the subsamples on demographic characteristics or on cluster variables (all p s>.05p \mathrm{~s}>.05 ). Ward’s procedures were then performed separately on the two subsamples. Examination of the dendrograms, scree plots, and percentage changes in agglomeration coefficients again suggested a four-cluster solution for both subsamples.
A K-means non-hierarchical cluster analysis was performed next. This two-stage approach results in more reliable solutions and improves interpretability of the clusters
(Borgen & Barnett, 1987; Milligan & Sokal, 1980). Following recommendations of Hair et al. (1995), the cluster centroids derived from Ward’s hierarchical clustering method were used as seed points for the K-means clustering procedure. The K-means analysis on the full sample yielded cluster profiles closely resembling those obtained by Ward’s method, and there was high between-method reliability, measured by classification of cases into clusters with similar profiles (k=.74,p<.05)(k=.74, p<.05). Similar K-means cluster analyses were then performed on the two subsamples. Chi-squared analysis indicated high agreement (87%)(87 \%) and high kappa ( k=.81,p<.05k=.81, p<.05 ) between cluster assignments. Mean scores on the variables within clusters were also compared across subsamples. Of 24 comparisons, only five yielded significant differences in cluster centres between corresponding clusters of the two subsamples. 1{ }^{1}
Description of clusters
As shown in Tables 1 and 2, analyses revealed the presence of two psychopathic groups with relatively distinct profiles, one group of criminals with features of psychopathy, and one non-psychopathic group, which will be described below. Tables 1 and 2 also summarize results of post hoc Newman-Keuls comparisons between the clusters for each of the six variables used to derive the clusters in the full sample and in the two subsamples, following one-way ANOVAs on the variables. In light of the similarity between the Ward’s and the K-means cluster solutions and because K-means analysis is said to provide more interpretable solutions (Hair et al., 1995; Milligan & Sokal, 1980), comparisons based on K-means analyses are reported here. Comparisons based on Ward’s method were similar.
Cluster I: Secondary psychopaths
Members of this group ( 29.5%29.5 \% of the full sample) were characterized by more severe alcohol and drug problems and higher anxiety than all other clusters, as well as the highest PCL-R Factor 2 scores (though only significantly higher than those of Cluster 2). Their Factor 1 and IM-P scores were average relative to other clusters.
Cluster 2: Non-psychopathic criminals with alcohol and drug problems
Men in this group ( 27.5%27.5 \% of the sample) had lower scores on the two PCL-R factors than men in all other clusters and had lower scores on the IM-P than men in Clusters 1 and 3, as well as below average anxiety levels. The presence, on average, of both alcohol and drug abuse and a relative absence of dependence on substances in this group suggest a pattern of bingeing on both alcohol and other drugs.
Cluster 3: Primary psychopaths
This group ( 17.0%17.0 \% of the sample) was comprised of men characterized by higher scores on PCL-R Factor 1 and on the IM-P than men in other clusters. In contrast, their scores on Factor 2 of the PCL-R were average. They were also distinguished from those in the other
- 1{ }^{1} Cluster 1 in Subsample 2 had higher scores on PCL-R Factor 2 than Cluster 1 in Subsample 1(t=12.79,p<.001)1(\mathrm{t}=12.79, \mathrm{p}<.001); Cluster 2 in Subsample 2 had more severe alcohol ( t=4.35,p<.04\mathrm{t}=4.35, \mathrm{p}<.04 ) and drug disorders ( t=14.48,p<.001\mathrm{t}=14.48, \mathrm{p}<.001 ) than Cluster 2 in Subsample 1; Cluster 3 in Subsample 1 had more severe alcohol disorders ( t=7.33,p<.01\mathrm{t}=7.33, \mathrm{p}<.01 ) and anxiety (t=5.63;p<.02)(\mathrm{t}=5.63 ; \mathrm{p}<.02) than the same cluster in Subsample 2. ↩︎
Table I. Cluster characteristics of full sample and two subsamples based on standardized scores obtained by K-means analysis
CLUSTER | |||||||||
---|---|---|---|---|---|---|---|---|---|
Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 | ||||||
M | SD | M | SD | M | SD | M | SD | N−K(p<.05)\mathrm{N}-\mathrm{K}(p<.05) | |
Full sample | |||||||||
N | 59 | 55 | 34 | 52 | |||||
Alcohol | 1.11 | 0.51 | −0.26-0.26 | 0.84 | −0.23-0.23 | 0.80 | −0.84-0.84 | 0.45 | MS=0.43∘1>2,3,4;2,3>4M S=0.43^{\circ} 1>2,3,4 ; 2,3>4 |
Drug | 0.70 | 0.62 | −0.47-0.47 | 0.96 | −06-06 | 1.06 | −0.26-0.26 | 0.94 | MS=0.79∘1>2,3,4M S=0.79^{\circ} 1>2,3,4 |
Factor 1 | 0.18 | 0.86 | −1.11-1.11 | 0.62 | 1.04 | 0.55 | 0.29 | 0.53 | MS=0.45∘3>1,2,4;1,4>2M S=0.45^{\circ} 3>1,2,4 ; 1,4>2 |
Factor 2 | 0.55 | 0.62 | −1.10-1.10 | 0.79 | 0.36 | 0.82 | 0.30 | 0.73 | MS=0.53∘1,3,4>2M S=0.53^{\circ} 1,3,4>2 |
IM-P | −0.11-0.11 | 0.66 | −0.59-0.59 | 0.70 | 1.73 | 0.52 | −0.37-0.37 | 0.43 | MS=0.36∘3>1,2,4;1>2,4M S=0.36^{\circ} 3>1,2,4 ; 1>2,4 |
Anxiety | 0.49 | 0.73 | −0.21-0.21 | 1.01 | −0.17-0.17 | 1.18 | −0.22-0.22 | 0.97 | MS=0.91∘;1>2,3,4M S=0.91^{\circ} ; 1>2,3,4 |
Sample 1 | |||||||||
N | 31 | 25 | 17 | 26 | |||||
Alcohol | 0.83 | 0.55 | −0.43-0.43 | 0.80 | 0.15 | 1.03 | −0.95-0.95 | .28 | MS=0.45∘1>2,3,4;3>2,4;2>4M S=0.45^{\circ} 1>2,3,4 ; 3>2,4 ; 2>4 |
Drug | 0.61 | 0.68 | −1.05-1.05 | 0.51 | −0.93-0.93 | 1.11 | 0.24 | 0.76 | MS=0.57∘1>2,3;3,4>2M S=0.57^{\circ} 1>2,3 ; 3,4>2 |
Factor 1 | −0.32-0.32 | 0.91 | −0.86-0.86 | 0.69 | 1.08 | 0.66 | 0.17 | 0.61 | MS=0.55∘3>1,2,4;4>1,2;1>2M S=0.55^{\circ} 3>1,2,4 ; 4>1,2 ; 1>2 |
Factor 2 | 0.05 | 0.84 | −1.21-1.21 | 0.88 | 0.44 | 0.70 | 0.36 | 0.64 | MS=0.61∘1,3,4>2M S=0.61^{\circ} 1,3,4>2 |
IM-P | −0.19-0.19 | 0.70 | −0.65-0.65 | 0.61 | 1.78 | 0.53 | −0.39-0.39 | 0.46 | MS=0.35∘3>1,2,4;1>2M S=0.35^{\circ} 3>1,2,4 ; 1>2 |
Anxiety | 0.21 | 0.85 | −0.59-0.59 | 0.81 | 0.02 | 0.96 | 0.05 | 0.87 | MS=0.75∘1,3,4>2M S=0.75^{\circ} 1,3,4>2 |
Sample 2 | |||||||||
N | 32 | 23 | 19 | 26 | |||||
Alcohol | 1.19 | 0.49 | −15-15 | 1.00 | −0.39-0.39 | 0.64 | −0.73-0.73 | 0.49 | MS=0.45∘1>2,3,4;2>4M S=0.45^{\circ} 1>2,3,4 ; 2>4 |
Drug | 0.70 | 0.66 | −0.17-0.17 | 1.08 | 0.00 | 1.03 | −0.63-0.63 | 0.88 | MS=0.80∘1>2,3,4M S=0.80^{\circ} 1>2,3,4 |
Factor 1 | 0.41 | 0.70 | −1.35-1.35 | 0.65 | 0.98 | 0.47 | 0.30 | 0.56 | MS=0.38∘3>1,2,4;1,4>2M S=0.38^{\circ} 3>1,2,4 ; 1,4>2 |
Factor 2 | 0.78 | 0.50 | −1.01-1.01 | 0.77 | 0.32 | 0.89 | 0.23 | 0.73 | MS=0.50∘1>2,3,4;3,4>2M S=0.50^{\circ} 1>2,3,4 ; 3,4>2 |
IM-P | −0.13-0.13 | 0.59 | −0.59-0.59 | 0.67 | 1.64 | 0.51 | −0.44-0.44 | 0.48 | MS=0.33∘3>1,2,4;1>2,4M S=0.33^{\circ} 3>1,2,4 ; 1>2,4 |
Anxiety | 0.66 | 0.68 | −0.13-0.13 | 1.08 | −0.28-0.28 | 1.32 | −0.24-0.24 | 1.07 | MS=1.04∘1>2,3,4M S=1.04^{\circ} 1>2,3,4 |
Note. Alcohol and drug ratings refer to severity of abuse or dependence, with higher numbers indicating greater pathology. Factor 1 and Factor 2 refer to dimensions of the Psychopathy Checklist-Revised. IM-P = Interpersonal Measure of Psychopathy. N-K = Newman-Keuls Test for differences between cluster means. >> refers to results from Newman-Keuls analyses, indicating significant differences between clusters. #p<.05\# p<.05.
Table 2. Cluster characteristics of full sample and two subsamples based on raw scores obtained by K-means analysis
CLUSTER | |||||||||
---|---|---|---|---|---|---|---|---|---|
Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 | ||||||
M | SD | M | SD | M | SD | M | SD | N-K ( p<.05p<.05 ) | |
Full sample | |||||||||
N | 59 | 55 | 34 | 52 | |||||
Alcohol | 3.44 | 0.79 | 1.31 | 1.30 | 1.35 | 1.25 | 0.40 | 0.69 | MS=1.04∘1>2,3,4;2,3>4M S=1.04^{\circ} 1>2,3,4 ; 2,3>4 |
Drug | 3.37 | 0.98 | 1.53 | 1.53 | 2.18 | 1.68 | 1.85 | 1.50 | MS=1.99∘1>2,3,4M S=1.99^{\circ} 1>2,3,4 |
Factor 1 | 10.37 | 2.92 | 5.96 | 2.10 | 13.30 | 1.86 | 10.75 | 1.81 | MS=5.18∘3>1,2,4;1,4>2M S=5.18^{\circ} 3>1,2,4 ; 1,4>2 |
Factor 2 | 13.84 | 2.13 | 8.15 | 2.71 | 13.19 | 2.81 | 12.98 | 2.50 | MS=6.32∘1,3,4>2M S=6.32^{\circ} 1,3,4>2 |
IM-P | 29.17 | 4.55 | 25.87 | 4.81 | 41.82 | 3.61 | 27.40 | 2.98 | MS=16.99∘3>1,2,4;1>2,4M S=16.99^{\circ} 3>1,2,4 ; 1>2,4 |
Anxiety | 78.90 | 19.37 | 60.40 | 26.67 | 61.53 | 31.38 | 60.19 | 25.66 | MS=644.37∘1>2,3,4M S=644.37^{\circ} 1>2,3,4 |
Sample I | |||||||||
N | 31 | 25 | 17 | 26 | |||||
Alcohol | 3.00 | 0.86 | 1.00 | 1.23 | 1.94 | 1.60 | 0.23 | 0.43 | MS=1.10∘1>2,3,4;3>2,4;2>4M S=1.10^{\circ} 1>2,3,4 ; 3>2,4 ; 2>4 |
Drug | 3.23 | 1.09 | 0.69 | 0.93 | 2.12 | 1.76 | 2.65 | 1.20 | MS=1.49∘1>2,3;3,4>2M S=1.49^{\circ} 1>2,3 ; 3,4>2 |
Factor 1 | 8.67 | 3.10 | 6.95 | 2.41 | 13.42 | 2.25 | 10.33 | 2.08 | MS=6.48∘3>1,2,4;4>1,2;1>2M S=6.48^{\circ} 3>1,2,4 ; 4>1,2 ; 1>2 |
Factor 2 | 12.50 | 2.87 | 7.69 | 2.30 | 13.46 | 2.40 | 13.77 | 2.22 | MS=7.15∘1,3,4>2M S=7.15^{\circ} 1,3,4>2 |
IM-P | 28.68 | 4.83 | 25.58 | 4.14 | 42.18 | 3.63 | 27.31 | 3.20 | MS=16.60∘3>1,2,4;1>2M S=16.60^{\circ} 3>1,2,4 ; 1>2 |
Anxiety | 71.68 | 22.52 | 50.93 | 21.41 | 66.65 | 25.46 | 67.31 | 23.11 | MS=524.95∘1,3,4>2M S=524.95^{\circ} 1,3,4>2 |
Sample 2 | |||||||||
N | 32 | 23 | 19 | 26 | |||||
Alcohol | 3.56 | 0.78 | 1.48 | 1.56 | 1.11 | 0.99 | 0.58 | 0.76 | MS=1.08∘1>2,3,4;2>4M S=1.08^{\circ} 1>2,3,4 ; 2>4 |
Drug | 3.38 | 1.04 | 2.00 | 1.71 | 2.26 | 1.63 | 1.27 | 1.40 | MS=2.02∘1>2,3,4M S=2.02^{\circ} 1>2,3,4 |
Factor 1 | 11.15 | 2.39 | 5.16 | 2.22 | 13.11 | 1.59 | 10.77 | 1.89 | MS=4.39∘3>1,2,4;1,4>2M S=4.39^{\circ} 3>1,2,4 ; 1,4>2 |
Factor 2 | 14.61 | 1.71 | 8.48 | 2.65 | 13.03 | 3.07 | 12.73 | 2.51 | MS=5.96∘1>2,3,4;3,4>2M S=5.96^{\circ} 1>2,3,4 ; 3,4>2 |
IM-P | 29.06 | 4.06 | 25.91 | 4.61 | 41.26 | 3.54 | 26.92 | 3.30 | MS=15.39∘3>1,2,4;1>2,4M S=15.39^{\circ} 3>1,2,4 ; 1>2,4 |
Anxiety | 83.53 | 18.18 | 62.57 | 28.72 | 58.47 | 35.08 | 59.50 | 28.35 | MS=735.82∘1>2,3,4M S=735.82^{\circ} 1>2,3,4 |
Note. Factor 1 and Factor 2 refer to dimensions of the Psychopathy Checklist-Revised. IM-P = Interpersonal Measure of Psychopathy. N-K = Newman-Keuls Test for differences between cluster means. >> refers to results from Newman-Keuls analyses, indicating significant differences between clusters. ∗p<.05* p<.05.
cluster with high psychopathy factor scores (Cluster 1) by a less severe pattern of alcohol and drug-related problems and by lower anxiety.
Cluster 4: Criminals with features of psychopathy
Members of this cluster ( 26.0%26.0 \% of the sample) were the only participants characterized by neither alcohol abuse nor dependence. They were also characterized by less severe drug problems than men in Cluster 1 and lower IM-P scores than men in Clusters 1 and 3. Their scores on PCL-R Factor 1 were lower than those of men in Cluster 3, but their scores on both factors were higher than those of men in Cluster 2.
They also exhibited the lowest anxiety scores although not significantly lower than those of most other clusters.
External validation of clusters
We compared the clusters on two classes of relevant external variables: demographic characteristics and criminal behaviour. Principal results are summarized below and in Table 3.
Table 3. Differences between clusters on demographic characteristics and criminal behaviour
CLUSTER | |||||||||
---|---|---|---|---|---|---|---|---|---|
Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 | ||||||
M | SD | M | SD | M | SD | M | SD | N−K(p<.05)N-K(p<.05) | |
Age | 28.90 | 6.67 | 25.28 | 5.87 | 26.29 | 5.94 | 23.69 | 5.70 | MS=37.07∘1>2,3,4M S=37.07^{\circ} 1>2,3,4 |
Education | 10.27 | 2.90 | 10.77 | 3.16 | 11.58 | 1.67 | 10.67 | 1.65 | MS=4.67M S=4.67 |
Handedness | 11.08 | 2.60 | 10.94 | 3.08 | 9.97 | 4.02 | 10.98 | 3.06 | MS=9.75M S=9.75 |
IQ | 90.88 | 11.26 | 93.96 | 12.41 | 92.27 | 14.60 | 89.67 | 14.17 | MS=168.17M S=168.17 |
# CD Symptoms | 4.04 | 2.99 | 2.11 | 2.42 | 3.46 | 2.85 | 4.69 | 2.85 | MS=7.79∘4>2M S=7.79^{\circ} 4>2 |
# Non-viol. Ch. | 13.80 | 11.49 | 8.99 | 6.37 | 11.89 | 12.60 | 8.27 | 7.13 | MS=61.07∘1>2,4M S=61.07^{\circ} 1>2,4 |
# Viol. Ch. | 2.42 | 2.57 | 1.84 | 2.52 | 4.32 | 5.61 | 2.16 | 2.33 | MS=10.60∘3>1,2,4M S=10.60^{\circ} 3>1,2,4 |
# Incarcerations | 3.02 | 3.22 | 1.85 | 1.36 | 3.90 | 4.73 | 2.24 | 1.65 | MS=7.90∘3>2,4M S=7.90^{\circ} 3>2,4 |
Crim. Versatility | 1.66 | 0.58 | 1.06 | 0.78 | 1.59 | 0.50 | 1.28 | 0.70 | MS=0.43∘1,3>2,4M S=0.43^{\circ} 1,3>2,4 |
Note. IQ = Wechsler Adult Intelligence Scale IQ estimate based on Shipley Institute of Living ScaleRevised. CD=\mathrm{CD}= Conduct Disorder. # Non-Viol. Ch.=\mathrm{Ch} .= Number of non-violent charges. # Viol. Ch. == Number of violent charges. Crim. Versatility == Criminal versatility. N-K = Newman-Keuls Test for differences between cluster means. >> refers to results from Newman-Keuls analyses, indicating significant differences between clusters (p≤.05).NS=(p \leq .05) . \mathrm{NS}= not statistically significant difference.
Cluster I: Secondary psychopaths
Individuals in this cluster were older (mean age =28.9=28.9 years) than individuals in Clusters 2 and 4. The cluster was composed of 46.6%46.6 \% African-Americans and 51.7%51.7 \% Caucasians. Cluster members were characterized by greater criminal versatility and by more charges for non-violent offences than members of Clusters 2 and 4.
Cluster 2: Non-psychopathic criminals with alcohol and drug problems
Individuals were on average 25.3 years old. The racial composition of this cluster was 43.6%43.6 \% African-American, 47.3%47.3 \% Caucasian, 7.3%7.3 \% Latino, and 1.8%1.8 \% Native American. This cluster was characterized by non-significantly higher intelligence scores
(WAIS-R estimate =94.0=94.0 ) than the other clusters and by significantly fewer charges and less criminal versatility than individuals in the two psychopathic clusters.
Cluster 3: Primary psychopaths
The mean age of individuals in this cluster was 26.3 years. The group was comprised of 60.6%60.6 \% African-Americans and 39.4%39.4 \% Caucasians. Individuals in this cluster were distinguished by the greatest number of charges for violent offences relative to the other three groups and by greater criminal versatility and more incarcerations than individuals in the two non-psychopathic clusters.
Cluster 4: Criminals with features of psychopathy
Individuals in this group were younger than individuals in other clusters (mean age =23.69=23.69 years), but only significantly younger than members of Cluster 1. This cluster included the highest proportion of African-American ( 67.3%67.3 \% ) and Latino individuals ( 9.6%9.6 \% ), with only 23.1%23.1 \% of the members being Caucasian. Although cluster members were characterized by fewer incarcerations and less criminal versatility than individuals in the primary psychopathy cluster, they had, in spite of their young age, been charged with a number of violent offences relatively similar to those in the secondary psychopath cluster. They were also characterized by more childhood CD symptoms than were individuals in the non-psychopathic cluster and non-significantly more CD symptoms than individuals in the primary psychopathic cluster.
Discussion
The current study attempted to address some of the heterogeneity among criminal offenders by means of an empirical/atheoretical classification method. Overall, results revealed the presence of two psychopathic groups with relatively distinct profiles (Clusters 1 and 3), one group with some psychopathic and antisocial features (Cluster 4), and one clearly non-psychopathic group (Cluster 2). Although cluster analytic methods are considered exploratory, the replication of our findings across two relatively large samples and with two different cluster analytic methods provides a basis for increased confidence in the obtained solution. The study extends previous research on subtyping offenders based on psychopathy and other psychological characteristics and lends further evidence to the existence of meaningful subgroups among inmates with elevated scores on the two validated dimensions of psychopathy.
In accord with the first hypothesis, one of the clusters that emerged (Cluster 1: secondary psychopaths) consisted of individuals characterized by severe alcohol and drug dependence, together with elevated scores on the social deviance factor of psychopathy (PCL-R Factor 2). These findings are consistent with prior findings indicating that both alcohol and drug disorders are more highly correlated with the dimension of psychopathy reflecting chronically unstable and antisocial lifestyle (PCL-R Factor 2) than with the dimension reflecting the core personality features of psychopathy (PCL-R Factor 1) (Smith & Newman, 1990). An important additional characteristic of this group was their significantly elevated anxiety. The particular constellation of elevated scores on the social deviance factor of psychopathy (PCL-R Factor 2), high anxiety, and substantial alcohol and drug dependence is reminiscent of the characteristics of the secondary psychopath as conceptualized by several theorists (Blackburn, 1975; Karpman, 1941; Lykken, 1995).
The evidence that high- and low-anxious subgroups of psychopaths also perform differently on a variety of cognitive and emotional measures (Brinkley, Newman, Harpur, & Johnson, 1999; Hiatt, Lorenz, & Newman, 2002; Schmitt, Brinkley, & Newman, 1999) provides additional empirical support for the notion that our secondary psychopaths and primary psychopaths constitute qualitatively different groups of inmates.
Consistent with the second hypothesis, a distinct cluster of criminals (Cluster 3: primary psychopaths) was characterized by elevated scores on the personality features of psychopathy (PCL-R Factor 1), in conjunction with elevated scores on the IM-P, thus revealing interpersonal and affective behaviours considered emblematic of psychopathy. In contrast to Cluster 1, this group of people had average anxiety levels. Together, these characteristics are reminiscent of Cleckley’s (1941/1988) classical description of the psychopath and of what others have referred to as the primary psychopath (Blackburn, 1975; Lykken, 1995). The concurrent elevations on Factor 1 and the IM-P are in line with reports suggesting that the IM-P is more closely associated with the affective and interpersonal aspects of psychopathy than with its antisocial behavioural manifestations (Kosson et al., 1997). Further, the finding that individuals in this cluster were characterized by the greatest number of charges for violent offences is consistent with reports that Factor 1 scores predict violent recidivism better than Factor 2 scores (Serin, 1996) and that primary psychopaths have more convictions for crimes of assault than secondary psychopaths (Blackburn, 1975).
Cluster 4 (criminals with features of psychopathy) comprised a group of criminal offenders characterized by moderate scores both on the dimensions selected for cluster derivation and on the dimensions used for external validation. Current results suggest that individuals in this cluster lack the interpersonal features and criminal versatility of the psychopath but exhibit some of the unstable antisocial lifestyle features and childhood conduct problems associated with ASPD. However, it should be noted that this cluster was characterized by lower rates of criminal behaviour than the psychopathic clusters. Thus, they bear some resemblance to the low-level chronic offenders identified by Moffitt and others (Donnellan, Ge, & Wenk, 2000; Moffitt, Caspi, Harrington, & Milne, 2002).
One additional noteworthy feature of this cluster was its ethnic composition: more than two-thirds of cluster members were African-Americans, whereas Caucasians comprised less than one-quarter of cluster members. The cluster also contained the highest proportion of Latino individuals ( 9%9 \% of all cluster members). Given that few studies have investigated the nature of psychopathy in non-Caucasian samples and that some of these studies suggest important differences in psychopathy and in persistent offenders across ethnicity (Bolt, Lorenz, Vitale, & Newman, 2002; Cooke, Kosson, & Michie, 2001; Donellan et al., 2000; Kosson, Smith, & Newman, 1990; Lorenz & Newman, 2002b), race-related differences may have contributed to current findings.
Unfortunately, because this study was not designed to address ethnicity, our sample sizes were not sufficient to conduct ethnicity-specific analyses. Future studies should include sufficiently large samples to allow separate investigation of subtypes of psychopaths and offenders in different ethnic groups. In fact, one such investigation is already under way in our laboratory (Swogger & Kosson, 2004).
Analyses further identified a group of criminals (Cluster 2: non-psychopathic criminals with alcohol and drug problems) who were clearly non-psychopathic, were not troubled by anxiety, and were prone to alcohol and drug abuse in the absence of physical or psychological dependence on these substances. The latter finding suggests the possibility that acute effects of using illicit substances may have played a primary role in the criminal behaviour of such individuals; a hypothesis that merits further exploration.
The current study also demonstrates the potential value of using an exploratory atheoretical clustering analysis approach for addressing theoretical controversies about the nature of psychopathy. Although it is important to replicate current findings in independent samples, it is interesting that similar findings have recently been reported using a different analysis strategy. Patrick, Hicks, Markon, and Krueger (2002), using a theoretically driven cluster analysis, identified two subgroups of psychopaths, one characterized by low stress reactivity and one characterized by heightened negative emotionality, similar to the primary and secondary psychopaths described in the current study. If the distinction between the primary psychopath and the secondary psychopath is indeed replicable, it may provide a method for addressing the variety of possible aetiological mechanisms underlying the antisocial behaviour of these subgroups.
The current findings also raise some important implications for treatment of subtypes of criminal offenders. It is possible that the psychopaths who have been considered to be notoriously recalcitrant to treatment could be more akin to the primary psychopaths that we identified in our study. In contrast, secondary psychopaths could be more amenable to treatment. For example, as noted by Salekin (2002), the presence of anxiety is considered a positive prognostic sign for treatment efficacy. However, as revealed by Salekin’s meta-analysis of treatment studies of psychopathy, only four studies to date have utilized the PCL-R criteria for operationalizing psychopathy. Therefore, the research on the amenability of PCL-R-classified psychopaths to treatment is still nascent and it would be informative for future studies to explore the response to treatment in primary and secondary psychopaths.
Several limitations of the study should be noted. Although our confidence in the obtained cluster solution was increased by the replication of highly similar groups across two relatively large subsamples, the two samples were selected from the same institution, which may entail a greater similarity between them than if they were selected from different correctional facilities. Similarly, it is possible that the relationships between criminal versatility and membership in the two clusters with high scores on PCL-R Factors 1 and 2 may have been inflated by the use of the same interview and file information to score both the PCL-R factors and the measure of criminal versatility (PCL-R item 20). It is also possible that the ethnic diversity of the sample may have affected the clarity of the obtained solutions; therefore, future studies should perform similar analyses with groups homogeneous in ethnicity. Furthermore, although we tested the two-factor model of psychopathy (Hare, 1991), which has been the most widely used model in research studies to date, recently a three-factor model (Cooke & Michie, 2001) and a four-factor model (Hare, 2003) have been suggested, which may offer a better insight into the multidimensionality of the construct. It is also important to consider different variables when designing replication studies; if similar clusters emerge across different samples and different sets of classification variables, then it will be clear that the clusters identified are robust. Finally, any typology remains of limited clinical utility until its validity is established by repeated replication. Therefore, the current results must be taken as preliminary and should not be liberally generalized until replications with other samples of criminal offenders reproduce the cluster profiles obtained in this study.
References
American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: APA.
Alterman, A. I., McDermott, P. A., Cacciola, J. S., Rutherford, M. J., Boardman, C. R., Mckay, J. R., & Cook., T. G. (1998). A typology of antisociality in methadone patients. Journal of Abnormal Psychology, 107, 412-422.
Andrews, D. A., Bonta, J., & Hoge, R. D. (1990). Classification for effective rehabilitation: Rediscovering psychology. Criminal Justice and Behavior, 17, 19-52.
Blackburn, R. (1975). An empirical classification of psychopathic personality. British Journal of Psychiatry, 127, 456-460.
Blackburn, R. (1988). On moral judgment and personality disorders: The myth of psychopathic personality revisited. British Journal of Psychiatry, 153, 505-512.
Blackburn, R., & Coid, J. W. (1999). Empirical clusters of DSM-III personality disorders in violent offenders. Journal of Personality Disorders, 13, 18-34.
Blackburn, R., & Maybury, C. (1985). Identifying the psychopath: The relation of Cleckley’s criteria to the interpersonal domain. Personality and Individual Differences, 6, 375-386.
Bolt, D. M., Lorenz, A. R., Vitale, J. E., & Newman, J. P (2002). Measurement invariance of the Psychopathy Checklist-Revised (PCI-R) across race and gender in criminal offender populations. Manuscript submitted for publication.
Borgen, F. H., & Barnett, D. C. (1987). Applying cluster analysis in counseling psychology research. Journal of Counseling Psychology, 34, 456-468.
Brinkley, C. A., Newman, J. P., Harpur, T. J., & Johnson, M. M. (1999). Cohesion of texts produced by psychopathic and nonpsychopathic criminal inmates. Personality and Individual Differences, 26, 873-885.
Bucholz, K. K., Hesselbrock, V. M., Heath, A. C., Kramer, J. R., & Schuckit, M. A. (2000). A latent class analysis of antisocial personality disorder symptom data from a multi-centre family study of alcoholism. Addiction, 95, 553-567.
Chapman, L. J., & Chapman, J. P. (1987). The measurement of handedness. Brain and Cognition, 6,175−1836,175-183.
Cleckley, H. (1941/1998). The mask of sanity. St Louis, MO: Mosby.
Clements, C. B. (1996). Offender classification: Two decades of progress. Criminal Justice and Behavior, 23, 121-143.
Cloninger, C. R. (1987). Neurogenetic adaptive mechanisms in alcoholism. Science, 236, 410-416.
Cooke, D. J., Kosson, D. S., & Michie, C. (2001). Psychopathy and ethnicity: Structural, item, and test generalizability of the Psychopathy Checklist-Revised (PCI-R) in Caucasian and African American participants. Psychological Assessment, 13, 531-542.
Cooke, D. J., & Michie, C. (2001). Refining the construct of psychopathy: Towards a hierarchical model. Psychological Assessment, 13, 171-188.
Donnellan, M. B., Ge, X., & Wenk, E. (2000). Cognitive abilities in adolescent-limited and lifecourse persistent criminal offenders. Journal of Abnormal Psychology, 109, 396-402.
First, M. B., Gibbon, M., Spitzer, R. L., & Williams, J. B. W. (1995). User’s guide for the structured clinical interview for DSM-IV axis I disorders. Version 2.0, New York: Biometrics Research.
Haapasalo, J., & Pulkkinen, L. (1992). The psychopathy checklist and non-violent offender groups. Criminal Behaviour and Mental Health, 2, 315-328.
Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. L. (1995). Multivariate data analysis. Inglewood Cliffs, NJ: Prentice Hall.
Hale, L. R., Goldstein, D. S., Abramowitz, C. S., Calamari, J. E., & Kosson, D. S. (2004). Psychopathy is related to negative affectivity but not to anxiety sensitivity. Behaviour Research and Therapy, 42, 697-710.
Hare, R. D. (1991). The Hare Psychopathy Checklist-Revised. Toronto: Multi-Health Systems.
Hare, R. D. (2003). The Hare Psychopathy Checklist-Revised (2nd ed.). Toronto: Multi-Health Systems.
Hare, R. D., Hart, S. D., & Harpur, T. J. (1991). Psychopathy and the DSM-IV criteria for antisocial personality disorder. Journal of Abnormal Psychology, 100, 391-398.
Harpur, T. J., & Hare, R. D. (1994). Assessment of psychopathy as a function of age. Journal of Abnormal Psychology, 103, 604-609.
Harpur, T. J., & Hare, R. D (1996). Psychopathy and violent behaviour: Two factors are better than one. Unpublished manuscript.
Harpur, T. J., Hare, R. D., & Hakstian, R. A. (1989). Two-factor conceptualization of psychopathy: Construct validity and assessment implications. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1, 6-17.
Herve, H., Ling, J., & Hare, R. D (2000, March). Criminal psychopathy and its subtypes. Paper presented at the biennial conference of the American Psychology-Law Society, New Orleans, LA.
Hiatt, K. D., Lorenz, A. R., & Newman, J. P. (2002). Assessment of emotion and language processing in psychopathic offenders: Results from a dichotic listening task. Personality and Individual Differences, 32, 1255-1268.
Holtzworth-Munroe, A., & Stuart, G. L. (1994). Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin, 116, 476-497.
Karpman, B. (1941). On the need for separating psychopathy into two distinct clinical types: Symptomatic and idiopathic. Journal of Criminology and Psychopathology, 3, 112-137.
Knight, R. A., & Prentky, R. A. (1990). Classifying sexual offenders: The development and corroboration of taxonomic models. In W. L. Marshall, D. R. Laws, & H. E. Barbarce (Eds.), (pp. 23-52). Handbook of sexual assault: Issues, theories, and treatment of the offender. New York: Plenum Press.
Kosson, D. S., Gacono, C. B., & Bodholdt, R. H. (2000). Assessing psychopathy: Interpersonal aspects and clinical interviewing. In C. B. Gacono (Ed.), The clinical and forensic assessment of the psychopath: A practitioner’s guide (pp. 203-229). Hillsdale, NJ: Erlbaum.
Kosson, D. S., Smith, S. S., & Newman, J. P. (1990). Evaluating the construct validity of psychopathy in black and white male inmates: Three preliminary studies. Journal of Abnormal Psychology, 99,250−25999,250-259.
Kosson, D. S., Steuerwald, B. L., Forth, A. E., & Kirkhart, K. J. (1997). A new method for assessing the interpersonal behavior of psychopathic individuals: Preliminary validation studies. Psychological Assessment, 9, 1-13.
Kosson, D. S., Suchy, Y., & Cools, J. (2001, November). Social cognitive biases in psychopathic adolescents. Poster presented at the annual meeting of the Society for Research in Psychopathology, Madison, WI.
Lewis, C. E., Rice, J., & Helzer, J. E. (1983). Diagnostic interactions: Alcoholism and antisocial personality. Journal of Nervous and Mental Disease, 171, 105-113.
Lorenz, A. R., & Newman, J. P. (2002a). Deficient response modulation and emotion processing in low-anxious Caucasian psychopathic offenders: Results from a lexical decision task. Emotion, 2, 91-104.
Lorenz, A. R., & Newman, J. P. (2002b). Do emotion and information processing deficiencies found in Caucasian psychopaths generalize to African-American psychopaths? Journal of Abnormal Psychology, 32, 1077-1086.
Lykken, D. T. (1995). The antisocial personalities. Hillsdale, NJ: Lawrence Erlbaum.
Megargee, E. I., & Bohn, M. I., Jr. (with Mayer, J., Jr., & Sink, F.) (1979). Classifying criminal offenders: A new system based on the MMPI. Beverly Hills, CA: Sage.
Milligan, G. W., & Sokal, L. M. (1980). A two-stage clustering algorithm with robust recovery characteristics. Educational and Psychological Measurement, 40, 755-759.
Millon, T. (1981). Disorders of personality. New York: Wiley.
Moffit, T. E., Caspi, A., Harrington, H., & Milne, B. J. (2002). Males on the life-course-persistent and adolescence-limited antisocial pathways: Follow-up at age 26 years. Development and Psychopathology, 14, 179-207.
Morey, L. C., Blashfield, R. K., & Skinner, H. A. (1983). A comparison of cluster analysis techniques within a sequential validation framework. Multivariate Behavioral Research, 18, 309-329.
Nagin, D. S., Farrington, D. P., & Moffitt, T. E. (1995). Life-course trajectories of different types of offenders. Criminology, 33, 111-139.
Newman, J. P., Schmitt, W. A., & Voss, W. D. (1997). The impact of motivationally neutral cues on psychopathic individuals: Assessing the generality of the response modulation hypothesis. Journal of Abnormal Psychology, 106, 563-575.
Newman, J. P., Wallace, J. F., Schmitt, W. A., & Arnett, P. A. (1997). Behavioral inhibition system functioning in anxious, impulsive and psychopathic individuals. Personality and Individual Differences, 23, 583-592.
Patrick, C. J., Hicks, B. M., Markon, K., & Krueger, R. F (2002, September). Profiling the psychopath: Temperament-based subtypes. Paper presented at the annual meeting of the Society for Research in Psychopathology, San Francisco, CA.
Patrick, C. J., & Lang, A. R. (1999). Psychopathic traits and intoxicated states: Affective concomitants and conceptual links. In M. E. Dawson, A. M. Schell, & A. H. Boehmelt (Eds.), Startle modification: Implications for neuroscience, cognitive science, and clinical science (pp. 209-230). New York: Cambridge University Press.
Reardon, M. L., Lang, A. R., & Patrick, C. J. (2002). An evaluation of relations among antisocial behavior, psychopathic traits, and alcohol problems in incarcerated men. Alcoholism: Clinical and Experimental Research, 26, 1188-1197.
Salekin, R. T. (2002). Psychopathy and therapeutic pessimism: Clinical lore or clinical reality? Clinical Psychology Review, 22, 79-112.
Salekin, R. T., Rogers, R., & Sewell, K. W. (1996). A review and meta-analysis of the Psychopathy Checklist and Psychopathy Checklist-Revised: Predictive validity of dangerousness. Clinical Psychology: Science and Practice, 3, 203-215.
Schmitt, W. A., Brinkley, C. A., & Newman, J. P. (1999). Testing Damasio’s somatic marker hypothesis with psychopathic individuals: Risk takers or risk averse? Journal of Abnormal Psychology, 108, 538-543.
Schmitt, W., & Newman, J. P. (1999). Are all psychopathic individuals low-anxious? Journal of Abnormal Psychology, 108, 353-358.
Serin, R. (1996). Violent recidivism in criminal psychopaths. Law and Human Behavior, 20, 207-217.
Skeem, J. L., Poythress, N., Edens, J. F., Lilienfeld, S. O., & Cale, E. M. (2003). Psychopathic personality or personalities? Exploring potential variants of psychopathy and their implications for risk assessment. Aggression and Violent Behavior, 8, 513-546.
Slutske, W. S., Heath, A. C., Dinwiddie, S. H., Madden, P. A., Bucholz, K. K., Dunne, M. P., Statham, D. J., & Martin, N. G. (1998). Common genetic risk factors for conduct disorder and alcohol dependence. Journal of Abnormal Psychology, 107, 363-374.
Smith, S. S., & Newman, J. P. (1990). Alcohol and drug abuse-dependence in psychopathic and nonpsychopathic criminal offenders. Journal of Abnormal Psychology, 99, 430-439.
Spielberger, C. D. (1983). State-Trait Anxiety Inventory (Form Y). Palo Alto, CA: Consulting Psychologists Press.
Stevens, G. F. (1993). Applying the diagnosis of antisocial personality to imprisoned offenders: Looking for hay in a haystack. Journal of Offender Rehabilitation, 19, 1-26.
Swogger, M. T., & Kosson, D. S. (2004). Subtyping criminal offenders: A replication and extension. Manuscript in preparation.
Tabachnick, B. G., & Fidell, L. S. (1996). Using multivariate statistics (3rd ed.). New York: HarperCollins.
Windle, M. (1999). Psychopathy and antisocial personality disorder among alcoholic inpatients. Journal of Studies on Alcohol, 60, 330-336.
Zachary, R. A. (1991). Shipley Institute of Living Scale: Revised manual. Los Angeles: Western Psychological Services.
Received II September 2003; revised version received 16 August 2004