Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Development and validation of a new scale for the assessment of psychopathy Hart, Stephen D. 1992

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1993_spring_phd_hart_stephen.pdf [ 5.38MB ]
Metadata
JSON: 831-1.0086356.json
JSON-LD: 831-1.0086356-ld.json
RDF/XML (Pretty): 831-1.0086356-rdf.xml
RDF/JSON: 831-1.0086356-rdf.json
Turtle: 831-1.0086356-turtle.txt
N-Triples: 831-1.0086356-rdf-ntriples.txt
Original Record: 831-1.0086356-source.json
Full Text
831-1.0086356-fulltext.txt
Citation
831-1.0086356.ris

Full Text

DEVELOPMENT AND VALIDATION OF A NEW SCALE FOR THE ASSESSMENT OF PSYCHOPATHY by STEPHEN DAVID HART B.A., The University of British Columbia, 1984 M.A., The University of British Columbia, 1987 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Psychology)  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA November 1992 © Stephen David Hart, 1992  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission.  (Signature)  Department of ^Psychology The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  ABSTRACT A review of the construct of psychopathy suggested that procedures for assessing the disorder should take into account its two-facet structure, its chronicity, its association with criminality, and its association with deceitfulness. A review of the five most popular assessment procedures currently in use indicated that none of them was completely satisfactory; the Hare Psychopathy Checklist-Revised (PCL-R) appeared to be superior to the other measures in most respects, but it was not well-suited for use outside of forensic settings. It was therefore decided to develop a new scale, based on the PCL-R, that would be suitable for both forensic and nonforensic settings. Pilot testing resulted in a 12-item symptom construct rating scale, named the Psychopathy Checklist: Screening Version (PCL:SV). The PCL:SV was validated in 11 samples (N = 586) from forensic/nonpsychiatric, forensic/psychiatric, civil/psychiatric, and civil/nonpsychiatric settings. Results indicated that the PCL:SV had good internal consistency, interrater reliability, and temporal stability. The scale also appeared to have a two-factor structure, at least in samples with an appreciable base rate of psychopathy. The PCL:SV was highly correlated with other psychopathy-related measures, including the PCL-R, antisocial personality disorder symptom counts, and several self-report scales. It also had a pattern of convergent and discriminant validities that was consistent with both theory and previous research that used the PCL-R. It was concluded that the PCL:SV holds considerable promise as a measure of psychopathy; areas requiring further research were identified.  ii  TABLE OF CONTENTS  ABSTRACT ^  ii  LIST OF TABLES ^  v  ACKNOWLEDGMENTS ^  vii  PREFACE ^  viii  CHAPTER 1: INTRODUCTION ^ A. Assessment Issues: The Nature of the Construct ^  1 2  1. Two-Facet Structure ^  2  2. Chronicity ^  4  3. Association with Criminality ^  5  4. Association with Deceitfulness ^  6  B. Evaluation of Existing Procedures ^  7  1. DSM-III-R Criteria for Antisocial Personality Disorder (APD) ^ 8 2. Hare Psychopathy Checklist-Revised (PCL-R) ^  16  3. Minnesota Multiphasic Personality Inventory (MMPI-2) ^ 22 4. California Psychological Inventory (CPI) ^  27  5. Millon Clinical Multiaxial Inventory (MCMI-II) ^  31  C. Summary of Chapter 1 ^  CHAPTER 2: SCALE DEVELOPMENT AND VALIDATION ^ A. Development of the PCL:SV ^  35  36 36  1. First Draft: The Clinical Version of the PCL-R (CV) ^ 37 2. Final Draft: The Psychopathy Checklist: Screening Version (PCL:SV) ^ 41 iii  B. Validation of the PCL:SV ^  43  1. Overview ^  43  2. Samples ^  44  3. Procedure ^  46  4. Results ^  48  C. Summary of Chapter 2 ^  CHAPTER 3: DISCUSSION ^ A. Implications for Clinical Practice ^  82  83 83  1. The PCL:SV as a Measure of Psychopathic Traits ^  83  2. The PCL:SV as a Screening Test ^  84  3. Psychopathy in Future DSMs ^  85  B. Implications for Research ^  85  1. Norms ^  85  2. Large Samples ^  86  3. Test-retest Reliability ^  86  4. Predictive Validity ^  86  5. Laboratory Studies ^  87  6. Psychopathy and the Big 5 ^  87 89  C. Conclusion ^  REFERENCES ^  90  APPENDIXES ^  101  A. PCL:SV Administration and Scoring Details ^  101  B. Supplementary Tables ^  108  iv  LIST OF TABLES CHAPTER 1 1:  The DSM-III-R Criteria for Antisocial Personality Disorder ^  2:  Items in the PCL-R ^  9 17  CHAPTER 2 3:  Items in the PCL:SV ^  42  4:  Demographic Characteristics of Subjects ^  47  5:  Descriptive Statistics for PCL:SV Total, Part 1, and Part 2 Scores ^ 49  6:  Mean Weighted Interrater Reliabilities and Corrected Item-Total Correlations for the PCL:SV Items in the 11 Samples ^  7:  Internal Consistency (Chronbach's Alpha) of PCL:SV Total, Part 1, and Part 2 Scores ^  8:  52  Item Homogeneity (Mean Inter-Item Correlation) of PCL:SV Total, Part 1, and Part 2 Scores ^  9:  51  53  Interrater Reliability (ICC) of PCL:SV Total, Part 1, and Part 2 Scores ^ 55  10: Correlation Between PCL:SV Part 1 and Part 2 Scores in the 11 Samples ^ 57 11: Coefficients of Congruence Between Factor Loadings for the Oblique, Two-Factor Solution Across the 11 Samples ^  58  12: Coefficients of Congruence Between Factor Loadings for the Oblique, Two-Factor Solution in 7 Samples with Appreciable Base Rates of Psychopathy ^ 60 13: Concurrent Validity: Correlations With PCL-R Total and Factor Scores ^ 62 14: Concurrent Validity: Correlations With APD Adult Symptoms ^ 64 15: Concurrent Validity: Correlations With Self-Report Measures of Psychopathy/APD 65 16: Convergent Validity: Correlations With PDE Ratings in Sample 6 ^ 68 17: Convergent Validity: Correlations With MCMI-II Personality Disorder Scales in Sample 3 ^  69  18: Convergent Validity: Correlations With IAS-R B5 Domain Self-Ratings in Sample 6 ^ 72 19: Convergent Validity: Correlations With IAS-R B5 Octant Self-Ratings in Sample 6 ^ 73  20: Convergent Validity: Correlations With IASR-B5 Domain Observer Ratings in Students and Inmates ^  75  21: Convergent Validity: Correlations With IAS-R B5 Octant Observer Ratings in Students and Inmates ^  76  22: Convergent Validity: Correlations With Measures of Alcohol and Drug Abuse ^ 78 23: Discriminant Validity: Correlations With Self-Reported Depression (Beck Depression Inventory) and Anxiety (State-Trait Anxiety Inventory) at the Time of Assessment ^  80  24: Discriminant Validity: Correlations With Age, Sex, and Race ^ 81 APPENDIXES B-1: Interrater Reliability (r) of PCL:SV Items in 7 Samples ^  108  B-2: Corrected Item-Total Correlations for PCL:SV Total Scores in the 11 Samples ^ 109 B-3: Corrected Item-Total Correlations for PCL:SV Parts 1 and 2 Scores in the 11 Samples ^  110  B-4: Loadings for Oblique, 2-Factor Solution in Samples 1 to 4 ^  111  B-5: Loadings for Oblique, 2-Factor Solution in Samples 5 to 8 ^  112  B-6: Loadings for Oblique, 2-Factor Solution in Samples 9 to 11 ^ 113  vi  ACKNOWLEDGEMENTS I would like to thank David Cox, David Crockett, Don Dutton, and Ron Roesch for their support and encouragement of my graduate training, and to Lynn Alden, David Crockett, Dimitri Papageorgis, and Jerry Wiggins for their help in guiding my Master's and Doctoral research. Special thanks also to Addle Forth, my dear friend and valued colleague over the past ten years. Finally, and most importantly, my deepest thanks go to Robert Hare, whose guidance, loyalty, and companionship I have come to value so highly. Any scholarly virtues evident in this work are due entirely to Bob's influence; any flaws, due to the fact that I still have much to learn from him. I dedicate this thesis to Tammy and Kenzie.  vii  PREFACE In accordance with University guidelines, I would like to clarify the roles played by others in the research described herein. All my research has been conducted under the supervision of Dr. Robert Hare. Many of the ideas were developed in collaboration with Drs. Hare and David Cox, Department of Psychology, Simon Fraser University. However, I must bear sole responsibility for the analyses and comments in this thesis. There is no way that I could have collected the data without financial and logistical support. Financial support came in the form of a grant to Dr. Hare from the John D. and Catherine T. MacArthur Foundation's Research Network on Mental Health and the Law, under the direction of Dr. John Monahan, School of Law, University of Virginia. I wish to express my gratitude to Dr. Monahan and the MacArthur Foundation for this support. Logistical support came primarily from Dr. Adelle Forth, Department of Psychology, Carleton University, who was largely responsible for administering the MacArthur Foundation grant on behalf of Dr. Hare. As required by the granting agency, we submitted a brief report on the development and validation of the PCL:SV to Dr. Monahan upon completion of the research; this report will soon be published by the University of Chicago Press as a chapter in an edited volume (Hart, Hare, & Forth, in press). A number of people administered the PCL:SV as part of independent research projects, and kindly allowed me access to their data. Specifically, I would like to thank Dr. Henry Steadman and Ms. Pamela Clark Robbins of Policy Research Associates, Delmar, NY, for access to the MacArthur Risk Study data; Ms. Catherine Strachan, Department of Psychology, University of British Columbia, for access to her doctoral research data; and Ms. Shelley Brown, Department of Psychology, Carleton University, for access to her undergraduate honor's thesis data. Ms. Brown presented some preliminary analyses of the PCL:SV data for university students at the Annual Meeting of the Canadian Psychological Association (Brown, Forth, Hart, & Hare, 1992). While the data for this project were still being collected, we pilot-tested a procedure for having naive observers make ratings of normal personality. The results of this pilot test, which are described in Chapter 2, were also submitted to a peer-reviewed journal. The paper has since been accepted for publication, and is referred to below as "in press" (Hart & Hare, in press). I apologize in advance for quoting Eliot out of context.  viii  We are the hollow men We are the stuffed men Leaning together Headpiece filled with straw. Alas! Our dried voices, when We whisper together Are quiet and meaningless As wind in dry grass Or rats' feet over broken glass In our dry cellar Shape without form, shade without colour, Paralysed force, gesture without motion; Those who have crossed With direct eyes, to death's other Kingdom Remember us--if at all--not as lost Violent souls, but only As the hollow men The stuffed men...  from T.S. Eliot, The Hollow Men (1925)  ix  CHAPTER 1: INTRODUCTION Observers of human behavior have long argued that people can be classified into "types" on the basis of their personality (Tyrer & Ferguson, 1988). In modern clinical psychology and psychiatry, we refer to abnormal types as personality disorders: characteristic ways of perceiving and relating to the world that result in social dysfunction or disability (e.g., American Psychiatric Association, 1987; Millon, 1981). Psychopathy, or psychopathic personality disorder, can be differentiated from other personality disorders on the basis of its characteristic pattern of interpersonal, affective, and behavioral symptoms (e.g., Cleckley, 1976; Hare, 1991; McCord & McCord, 1964). Interpersonally, psychopaths are grandiose, egocentric, manipulative, dominant, forceful, and cold-hearted. Affectively, they display shallow and labile emotions, are unable to form longlasting bonds to people, principles, or goals, and are lacking in empathy, anxiety, and genuine guilt or remorse. Behaviorally, psychopaths are impulsive and sensation-seeking, and tend to violate social norms; the most obvious expressions of these predispositions involve criminality, substance abuse, and a failure to fulfill social obligations and responsibilities. The assessment of psychopathic personality disorder has been a topic of growing interest over the past decade. There are probably two main reasons for this. The first is the success of diagnostic criteria for psychopathy--specifically, the Psychopathy Checklist (PCL; Hare, 1980) and its recent revision (PCL-R; Hare, 1991). There is now a considerable literature attesting to the reliability and validity of PCL and PCL-R in forensic settings; of particular importance is their predictive validity with respect to criminal behavior (for reviews, see Hare, 1991; Hare, Forth, & Strachan, 1992; Hare & Hart, 1993). The second reason for the growing interest in psychopathy is disappointment with the diagnostic criteria for antisocial personality disorder (APD) contained in the third edition of the American Psychiatric Association's Diagnostic and Statistical Manual of Mental Disorders (DSMIII; American Psychiatric Association, 1980) and its recent revision (DSM-III-R; American Psychiatric Association, 1987). The APD criteria consist largely of a list of overt delinquent and criminal behaviors, and they have been severely criticized for their neglect of interpersonal and 1  affective symptoms historically associated with the construct of psychopathy, such as superficiality, grandiosity, callousness, manipulativeness, lack of remorse, and so forth (e.g., Hare, 1983; Hare, Hart, & Harpur, 1991; Millon, 1981; Rogers & Dion, 1992; Widiger & Corbett, in press). Interestingly, the American Psychiatric Association made a token attempt to include such symptoms in the criteria for APD in the DSM-III-R, and has just completed field trials in anticipation of DSM-IV that may result in a radical swing back towards the traditional construct of psychopathy (Widiger et al., 1992). This dissertation reports the results of an effort to develop and validate a new, PCL-based scale for the assessment of psychopathy. To demonstrate the need for such a scale, in the remainder of this Chapter, I will identify some crucial theoretical issues in the assessment of psychopathy and review existing assessment procedures. A. Assessment Issues: The Nature of the Construct It is quite easy to construct a test or measure and establish that it is more or less reliable (e.g., is internally consistent) and has reasonable criterion-related validity (e.g., is moderately correlated with other measures of the same construct); it is quite another matter to more fully evaluate its construct-related validity. One reason that establishing construct-related validity is so difficult is that it requires a reasonably thorough understanding of the construct being measured (e.g., American Psychological Association, 1985). Thus, any discussion or evaluation of procedures for assessing psychopathy must be guided by theory and research concerning the nature of the disorder. In this section, I will make four fundamental assertions concerning psychopathy and discuss their implications for assessment. 1. Two-Facet Structure My first assertion is that two oblique dimensions are both necessary and reasonably sufficient to provide a comprehensive description of psychopathic symptomatology. The evidence supporting this assertion comes from two sources: first, clinical and empirical studies identifying the key symptoms of psychopathy; and second, research indicating that these key symptoms form two natural clusters. The major clinical description of the psychopath is found in Cleckley's classic text, The Mask of Sanity (1976). He described sixteen characteristics of the disorder: 2  superficial charm and good intelligence; absence of delusions and other signs of irrational thinking; absence of nervousness or psychoneurotic manifestations; unreliability; untruthfulness or insincerity; lack of remorse or shame; inadequately motivated antisocial behavior; poor judgment and failure to learn from experience; pathological egocentricity and incapacity for love; general poverty in major affective relations; specific loss of insight; unresponsiveness in general interpersonal relations; fantastic and uninviting behavior with drink (and sometimes without); suicide rarely carried out; sex life impersonal, trivial, and poorly integrated; and failure to follow any life plan. (Note that this list includes characteristics that in the DSM-III (-R) would be considered symptomatic of antisocial, narcissistic, histrionic, and borderline personality disorder.) Other clinicians (before and after Cleckley) have described longer or shorter lists of characteristics, yet their conceptualization of the disorder is remarkably similar (e.g., Buss, 1961; Craft, 1965; Karpman, 1961; McCord & McCord, 1964; Millon, 1981). Reviews and content analyses of the empirical literature (e.g., Albert, Brigante, & Chase, 1959; Fotheringham, 1957) and surveys of mental health and criminal justice professionals (e.g., Davies & Feldman, 1981; Gray & Hutchinson, 1964; Livesley, 1986; Tennent, Tennent, Prins, & Bedford, 1990) suggest that researchers and practicing clinicians are in close agreement with Cleckley. Several studies indicate that when a reasonably comprehensive set of psychopathic symptoms is factor-analyzed, the resulting structure yields two correlated factors. For example, Harpur, Hakstian, and Hare (1988) factor analyzed the 22 items of the PCL. These items were heavily influenced by Cleckley's list of 16 features (Hare, 1980). Harpur et al. attempted to identify a factor structure underlying the items that was stable across samples, sites, and investigators. They had PCL ratings from six samples, with a total N of 1,119. For each sample, they extracted between 2 and 8 factors, and then subjected the factors to a variety of orthogonal and oblique rotations. The stability of various solutions both within and across samples was determined using split-half cross-validation and congruence. The results strongly supported an oblique, two-factor solution. Factor 1, labeled the "selfish, callous and remorseless use of others," comprised items tapping egocentricity, superficiality, deceitfulness, callousness, and a lack of remorse, empathy, and anxiety--all features that the APD criteria have been criticized for 3  neglecting. On the other hand, Factor 2, labeled a "chronically unstable and antisocial lifestyle," or "social deviance," comprised items tapping impulsivity, sensation-seeking, irresponsibility, aggressiveness, and criminality. The two factors were correlated about r = .50. An identical factor structure has been reported for the 20 items of the PCL-R (Hare et al., 1991). The two factors are differentially correlated with important external variables, such as violence, substance use, and interpersonal style (Harpur, Hare, & Hakstian, 1989; Hare, 1991). In another study, Livesley, Jackson, and Schroeder (1989, 1992) developed self-report scales to measure symptoms of personality disorder (identified via literature review, so as not to limit the domain of traits to those found in DSM-III-R). They conducted factor analyses of the scales in both patient and nonpatient samples. With respect to the prototypical psychopathy/APD symptoms, Livesley et al. found a two-factor structure isomorphic to that reported by Hare and colleagues; they labeled the factors "interpersonal disesteem" and "conduct problems." Finally, Harpur, Hare, Zimmerman, & Coryell (1990) conducted a factor analysis of DSM-III Cluster 2 (Dramatic-Erratic-Emotional) personality disorder symptoms in a large sample of community residents (relatives of psychiatric patients and a control group, consisting of relatives of nonpatients). All subjects were assessed using the Structured Interview for DSM-III Personality (Stangl, Pfohl, Zimmerman, Bowers, & Corenthal, 1985), a reliable and well-validated instrument. Several factors emerged, including two that comprised symptoms of antisocial and narcissistic personality disorder and were isomorphic to the PCL factors. In sum, considerable research suggests that the construct of psychopathy has an underlying structure consisting of two correlated factors. A corollary of this first assertion is that any procedure designed to measure psychopathy should assess both facets of the disorder. 2. Chronicity The second assertion is that psychopathy is a chronic disorder. There is research indicating that the disorder is first evident in early childhood and persists into adulthood (e.g.,  Livesley and Schroeder (1991) have also identified these same two factors in a study of the factorial structure of the existing DSM-III-R APD symptoms. 4  Hare, McPherson, & Forth, 1988; Robins, 1966). Indeed, these characteristics are necessary symptoms in the DSM-III (-R) criteria for APD criteria and contributory symptoms in the PCL criteria for psychopathy; they were also seen as highly prototypical of the disorder in the reviews and surveys described above. Further evidence of chronicity comes from studies indicating that even treatment does little to alter the behavior of criminal psychopaths (Harris, Rice, & Cormier, 1991; Ogloff, Wong, & Greenwood, 1991; Rice, Harris, & Cormier, 1992). 2 There may be a decrease in the frequency of some type of overt antisocial behavior in psychopaths after age 45 or so, particularly property offending (Hare, McPherson, & Forth, 1988); however, there is no evidence that this "burnout" phenomenon (as it is sometimes misleadingly referred to) encompasses a wide range of behavioral symptoms, and no evidence at all that it extends to interpersonal or affective symptoms. One corollary of the assertion that psychopathy is a chronic disorder is that assessment procedures for psychopathy should have high test-retest reliability, at least over relatively brief periods of time. A second is that measurement procedures should be relatively immune to the effects of state variables, such as mood at the time of assessment. 3. Association with Criminality My third assertion is that psychopathy and criminality are distinct but related constructs. (If they are not next-door neighbors, at least they reside on the same conceptual block.) As Hare has remarked, given the characteristics of psychopathy (callousness, remorselessness, impulsivity, and so forth), there is every reason to expect that psychopaths are particularly likely to engage in criminal behavior (Hare & Hart, 1993). This statement should not be interpreted to mean that all psychopaths are criminals (i.e., have official criminal records) or that all criminals are psychopaths; if this was the case, the construct of psychopathy loses its distinctiveness. Rather, we should expect offender populations to have a high base rate of psychopathy relative to other populations, such as community residents or civil psychiatric patients. In addition, within any  In fact, the study by Rice et al. (1992) suggests that treatment may even increase the likelihood of recidivism in psychopaths.  2  5  particular population, psychopaths should be at an increased risk for antisocial behavior (e.g., Hart, Kropp, & Hare, 1988). Two main lines of evidence support this assertion. First, the surveys and reviews cited earlier indicate that repeated antisocial behavior is considered to be a highly prototypical symptom of psychopathy, and indeed is included in the DSM-III (-R) and PCL (-R) criteria. Second, considerable research indicates that psychopathic criminals have a higher density of offending than do nonpsychopaths, even when controlling for previous criminal behavior to avoid circularity in prediction (see reviews cited earlier). To reiterate, psychopathy is related to, but distinct from, criminality. The most important corollary of this assertion is that procedures for the assessment of psychopathy should have significant predictive and/or convergent validity vis-a-vis measures of criminality. A second is that assessment procedures should be suitable for use in both forensic and nonforensic settings. 4. Association with Deceitfulness The fourth and final assertion is that deceitfulness--lying, deception, and manipulation--is closely associated with psychopathy. As was the case with criminality, deceitfulness is considered to be a prototypical symptom of psychopathy and is included in most diagnostic criteria for the disorder. There is also some empirical evidence that psychopaths are more likely than nonpsychopaths to engage in dissimulation, at least in certain contexts (e.g., Kropp, 1992; Hart, Dutton, & Newlove, 1992). The major corollary of this assertion is that assessment procedures for psychopathy should directly assess deceitfulness; a second is that assessment procedures must control for deceitfulness, as this symptom may interfere with the assessment of other features of the disorder.  6  B. Evaluation of Existing Procedures In this section, I will review and critically evaluate the major procedures for assessing psychopathy. In order to identify those procedures, I conducted a computer search of articles indexed in Psychological Abstracts between 1980 and 1992, using "antisocial personality" as the keyword. The result was a list of 442 abstracts. The elimination of dissertations, non-English publications, and non-empirical studies (case histories, reviews, theoretical papers, letters, and so forth) from this list left a total of 183 abstracts. Next, I went through the remaining abstracts and attempted to group them according to the assessment procedure(s) they used. I was able to make concrete determinations concerning 113 (61.7%) of the abstracts; the others did not contain information specific enough to permit their classification.' Among the classifiable abstracts, the most common assessment methods still in use were as follows: the DSM-III or DSM-III-R criteria for antisocial personality disorder, cited in 38 abstracts; the PCL or PCL-R criteria for psychopathy, 30 abstracts; the Psychopathic Deviate (Pd) scale from the Minnesota Multiphasic Personality Inventory (MMPI; Hathaway & McKinley, 1940) and its recent revision (MMPI-2; Butcher, Dahlstrom, Graham, Tellegen, & Kaemmer, 1989), 17 abstracts; and the Socialization  (So) scale from the original and revised California Psychological Inventory (CPI; Gough, 1957, 1987), 5 abstracts. Two procedures used with some degree of frequency were excluded from this review: medico-legal status according to the Mental Health Act of England and Wales, 6 abstracts (excluded as this does not constitute a psychiatric diagnosis in the usual sense of the word); and the Research Diagnostic Criteria (Spitzer, Endicott, & Robins, 1978) and Feighner criteria (Feighner et al., 1972), 12 abstracts (excluded as they have been superseded by the DSM-III and DSM-III-R). Procedures used in fewer than five studies were excluded from this review, with one important exception: the Antisocial scale (6A) of the revised Millon Clinical Multiaxial Inventory,  3 Probably 25% to 50% of these studies used the DSM-III or DSM-III-R criteria for APD, judging from their authorship and/or date of publication.  7  or MCMI-II (Millon, 1987). This inventory was cited in only one abstract that contained the keyword antisocial personality; however, it is currently among the most commonly-used adult assessment devices in clinical practice, and it is also the focus of considerable research in the last few years looking at personality disorders in general, including antisocial personality. In sum, the assessment procedures selected for review included two "clinical-behavioral" measures (the DSM-III-R criteria for APD and the PCL-R criteria for psychopathy) and three self-report measures (the MMPUMMPI-2 Pd scale; the CPI So scale; and the MCMI-II 6A scale). When discussing format, reliability, and norms, I will focus on the most recent version of each procedure. 4 When discussing validity, however, I will also review studies that used earlier versions, as empirical evidence suggests that the most recent and earlier versions of each procedure are highly correlated (Butcher et al., 1989; Gough, 1987; Hare, 1991; Millon, 1987; Morey, 1988b). 1. DSM-III-R Criteria for Antisocial Personality Disorder (APD) Format. The DSM-III-R criteria for APD are fixed and explicit psychiatric diagnostic criteria. There are four criteria, two of which contain multiple subcriteria: (1) current age at least 18; (2) conduct disorder before age 15; (3) antisocial behavior since age 18; and (4) symptoms of the disorder are not limited to periods of active schizophrenia or mania. The criteria are monothetic in nature: each one is necessary, and together they are jointly sufficient, to diagnose APD. The criteria and subcriteria are presented in Table 1. The content of the DSM-III APD criteria was decided by a committee of the American Psychiatric Association's DSM-III Task Force, and was revised slightly by another committee for the DSM-III-R (Widiger et al., 1991). In drafting their criteria, these committees were heavily influenced by the clinical and research traditions at the Washington University in St. Louis, which eschewed the use of inferred personality traits for the diagnosis of APD in favor of the use of specific behavioral indicators of those traits (Robins, 1978).  In each case, the revision represents a minor modification of the original procedure; thus, any criticisms of the revision also apply to the original procedure, mutatis mutandis. 4  8  Table 1 The DSM-III-R Criteria for Antisocial Personality Disorder Criterion/Subcriterion A. Current age at least 18  B. Conduct disorder before age 15, as indicated by at least three of the following: 1.  Truant  7.  Cruel to people  2.  Ran away  8.  Destroyed property  3.  Fought  9.  Set fires  4.  Used weapons  10.  Lied  5.  Forced sex on others  11.  Stole  6.  Cruel to animals  12.  Robbed  C. Antisocial behavior since age 15, as indicated by four or more of the following: 1.  Poor employment record  6.  Lies  2.  Repeated criminal acts  7.  Reckless  3.  Irritable and aggressive  8.  Irresponsible parenting  4.  Poor financial record  9.  No monogamous relationships  5.  Impulsive  10.  Lacks remorse  D. Occurrence of antisocial behavior not exclusively during the course of Schizophrenia or manic episodes  9  The APD criteria do not constitute a scale or test. Their development was not guided by psychometric principles, they do not have a response format per se, and they do not yield a score. Rather, the assessor determines if each (sub-) criterion is present/true or absent/false. The final decision is dichotomous: If the criteria are all present, then a lifetime diagnosis of APD is made; if one or more is absent, no such diagnosis is made. 5 Despite this, many researchers (e.g., Hart, Forth, & Hare, 1991) and some diagnostic interviews (e.g., Loranger, 1988) use the criteria to obtain dimensional APD "scores," such as symptom counts. The DSM-III-R also does not specify a particular method for assessing APD. In the empirical literature, researchers have employed methods ranging from structured interview to semi-structured interview plus a review of case history information to file review alone. Structured interviews should probably be avoided, unless the interviews are supplemented with case history information, as they may be highly susceptible to the effects of deceitfulness. Depending upon the method employed, assessment of APD probably takes 30 to 60 minutes. Reliability. In some respects, it may be unfair to evaluate the APD criteria according to standard psychometric principles, as their construction was not guided by these principles and as diagnostic criteria in general cannot be equated with test items (Blashfield & Livesley, 1991). With this caveat in mind, let us examine the reliability of the APD criteria. Criterion A is typically not analyzed in studies, as it is a simple decision regarding the subject's age. It presumably has near-perfect interrater and short-term test-retest reliability in adults, with the only errors being due to assessor error or deceitfulness on the part of the subject. Criterion B of APD has moderate to high interrater reliability. Kappa coefficients of interrater agreement for the presence versus absence of this criterion, using the interviewerobserver method, have ranged from .34 to .69 (e.g., Hart, Forth, & Hare, 1992; Widiger et al., 1992). Its temporal stability is unknown but must be moderate to high, given the stability of  5 The DSM-III-R does allow the assessor to use certain modifiers to clarify the diagnosis. For example, APD can be diagnosed as present but not currently active (i.e., in full or partial remission), or as probably present (i.e., provisionally diagnosed). Some researchers also structure their assessment methods to yield probabilistic diagnoses (e.g., Loranger, 1988).  10  overall APD diagnoses (see below). Problems have been identified with specific subcriteria, however. Coolidge, Merwin, Wooley, & Hyman (1990) examined APD symptom self-reports in college students and their family members. They found that several subcriteria had extremely low prevalence rates and/or low item-total correlations; overall, the internal consistency of the subcriteria was moderate (alpha = .63). Using the Spearman-Brown prophecy formula, I estimate that item homogeneity was also low to moderate (mean inter-item correlation = .12). It could be argued that the poor performance of the subcriteria directly resulted from the low prevalence of APD diagnoses in the sample (less than 10%); however, similar results were reported by Hart, Forth, & Hare (1992), who used an interview plus file review procedure in samples of incarcerated male offenders, forensic psychiatric patients, and college students. Despite a much higher prevalence of APD, at least in the two forensic samples (64.2% and 15.7%, respectively), a number of symptoms had low prevalence, low item-total correlations, or low interrater reliability. The internal consistency of the subcriteria was moderate (median alpha = .66), as was the item homogeneity (median value of mean inter-item correlation = .12). Symptoms identified as problematic in both studies included Used weapons, Forced Sex, Cruel to animals, Cruel to people, and Robbed. 6 Criterion C has adequate interrater reliability, with researchers reporting kappas of about .50 (Hart, Forth, & Hare, 1992; Stangl et al., 1985). Like criterion B, its temporal stability is unknown but presumably high. In addition to the Coolidge et al. (1990) and Hart, Forth, and Hare (1992) studies, data concerning the C subcriteria comes from the DSM-IV APD field trials (Widiger et al., 1992). All three studies indicate that several subcriteria have low prevalence, poor interrater reliability, or low item-total correlations in both forensic and civil samples. The subcriteria that performed poorly in at least two studies were Irresponsible parenting, No monogamous relationships, and Poor financial record. Internal consistency of the C subcriteria  Morey (1988a,b,c) examined the internal consistency of the 1985 draft DSM-III-R APD subcriteria. Results were not reported separately for B and C subcriteria. The overall alpha was .82 (Morey, 1988a,c); none of the B subcriteria had a corrected item-total correlation of less than .30 (Morey, 1988c).  6  11  was low to moderate in the Hart, Forth, and Hare (1992) study (median alpha = .55); item homogeneity was low (median value of mean inter-item correlation = .10). 7 Criterion D, like Criterion A, has not been analyzed in studies of the APD criteria. This is unfortunate, as its reliability may be substantially lower than that of Criterion A: first, the assessor must diagnose schizophrenia and manic syndrome--diagnoses which themselves are of imperfect reliability--in addition to APD; and second, the assessor must determine whether all the APD symptoms occurred during active periods of schizophrenia or mania. Irrespective of any problems with its constituent criteria, there is general agreement that APD was the most reliable of the DSM-III Axis II disorders; the same appears to be true for DSM-III-R (Widiger, 1992; but cf. Rogers & Dion, 1991). Interrater agreement, using the interviewer-observer method, was moderate in Hart, Forth, and Hare's (1992) inmate sample (kappa = .63) and low to moderate in the DSM-IV field trials (median kappa = .50). Higher interrater reliability has been reported in studies that used structured diagnostic interviews (e.g., Jackson et al., 1991; kappa = 1.00). APD diagnoses also have acceptable test-retest reliability, at least over brief periods of time (e.g., Alterman, Cacciola, & Rutherford, 1992). 8 APD assessments appear to be relatively unaffected by state variables, such as subjects' mood at the time of assessment (Widiger et al., 1992).  Norms. There are no systematic norms concerning the prevalence of DSM-III-R APD symptoms or diagnoses. The recent DSM-IV field trials for APD (Widiger et al., 1992) reported prevalence rates in five different settings, each with approximately 100 subjects. The settings and prevalence rates were as follows: outpatient substance abusers attending a VA clinic, 17%; male prison inmates, 70%; psychiatric and substance abuse inpatients, 36%; adopted-away offspring, 1%; and psychiatric inpatients, 34%. Other research confirms a high prevalence rate (typically  In Morey (1988c), the only C subcriterion that had a corrected item total correlations less than .30 was Impulsive (.27). -  Given the monothetic nature of the criteria, their internal consistency is irrelevant; however, it is of interest to note that the association between the presence versus absence of Criteria B and C was only moderate in the Hart, Forth, & Hare (1992) study, with kappas of .15, .32, and .67 in the inmate, forensic patient, and student samples. 8  12  50% to 75%) in forensic populations using either DSM-III or DSM-III-R criteria (Correctional Service of Canada, 1990; Hare, 1983, 1985; Hart, Forth, & Hare, 1992; Hart & Hare, 1989; Roesch, 1992). We can draw some inferences about the prevalence of APD from the results of the Epidemiologic Catchment Area (ECA) project (Robins & Regier, 1991), which used the DSM-III criteria. In the ECA, a structured interview, the Diagnostic Interview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff, 1981), was administered to a stratified random sample comprising nearly 20,000 adults residing in five large geographic centers in the United States. The respondents included community residents, as well as those institutionalized in psychiatric hospitals, geriatric homes, prisons, and residential substance use programs. According to Robins, Tipp, & Przybeck (1991), the lifetime prevalence of APD was 2.6% (SE = 0.16%). APD prevalence rates were significantly higher in men versus women (by a factor of about 5), in urban versus rural residents (by a factor of about 2), and in those below age 30 versus those above age 64 (by a factor of about 10). There were no significant racial differences in prevalence. Validity. The content-related validity of the APD criteria has been heavily criticized on a number of grounds. First, the content of the criteria and subcriteria is thought by many writers to be too long, overly-specific, and arbitrary--in the words of Millon (1981), "picayunish." Their complexity may give rise to a number of problems, including an over-reliance on retrospective self-reports for assessment, a failure to adhere to the actual criteria in clinical practice, and extreme heterogeneity among those meeting the criteria (Hare, Hart, & Harpur, 1991; Morey & Ochoa, 1989; Rogers & Dion, 1991). Second, the criteria appear to assess primarily the social deviance facet of psychopathy, ignoring many affective and interpersonal symptoms (Hare, 1985; Hare, Hart, & Harpur, 1991; Millon, 1981). Even in their assessment of social deviance, they may focus too much on rare, violent symptoms (Rogers & Dion, 1991). Thus, they may be virtually synonymous with severe and persistent criminality (Hare, 1991). As Widiger (1992) notes, such criticism has led to speculations that the criteria are at once both too broad, overdiagnosing APD in criminal populations, and too narrow, failing to identify true psychopaths  13  in noncriminal populations. 9 Third, some of the criteria can be criticized on logical grounds. For example, Criterion A may be unnecessary; DSM-III-R makes clear in the overview to Axis II that personality disorders persist into adulthood; thus, no APD diagnosis should be made in the case of someone whose antisocial behavior spontaneously remits after adolescence. No other Axis II criteria include an age criterion. Similarly, Criterion B itself may be redundant: DSM-III-R states that symptoms of a personality disorder are usually first evident in childhood, and no other Axis II criteria have specific childhood symptoms that are necessary for the diagnosis. Finally, Criterion D, like other exclusionary criteria in the DSMs, is of unknown validity (Boyd et al., 1984). With respect to concurrent validity, APD diagnoses are correlated about .55 with PCL-R total scores, and have moderate to high levels of agreement with the PCL-R (Hare, 1983, 1985, 1991). Similar levels of agreement were observed between APD and the ICD-10 criteria for dyssocial personality and criteria for psychopathic personality disorder (based on the PCL-R) in the DSM-IV APD field trials (Widiger et al., 1992). Turning to self-report measures, APD diagnoses have low to moderate correlations, typically around .30, with MCMI-II Antisocial/Aggressive (6A) and Sadistic (6B) scales, MMPI Psychopathic Deviate (Pd) and Hypomania (Ma) scales, and the CPI Socialization (So) scale (Hare, 1985; Hart, Forth, & Hare, 1992). Of course, when dimensional measures of APD (e.g., symptom counts) are used, the correlations are somewhat higher, but still only moderate in magnitude (Hart, Forth, & Hare, 1992; Widiger et al., 1992). These relatively low correlations may reflect a problem with the self-report scales rather than with the APD criteria, however. There has been little research looking at the predictive validity of APD. There is limited evidence that APD is associated with poor response to treatment in substance abuse and correctional treatment programs (e.g., Woody & McLellan, 1985; Harris et al., 1991). However,  Interestingly, Morey (1988a) found that symptoms of APD tended to covary with certain symptoms of other personality disorders (e.g., narcissistic, passive-aggressive) to form a cluster that he labelled "psychopathic." This tends to support the view that, although the APD symptoms may be internally consistent, they fail to provide adequate coverage of the domain of the psychopathy construct. 9  14  its predictive efficiency appears to be weak both in absolute terms and relative to that of other measures, such as the PCL-R (e.g., Hare, 1991; Hart et al., 1988; Harris et al., 1991). With respect to construct-related validity, there is a rather large body of literature looking at the association between APD and substance use. In fact, this one topic accounts for over a third of the recent research (14 of 38 abstracts) identified by the computer search. Probably the most common findings are that APD is significantly comorbid with substance use disorders and that substance use patients with APD are more socially deviant or have worse treatment outcomes than other patients (e.g., Liskow & Powell, 1990, 1991; Stabenau, 1990; Woody & McLellan, 1985). Another common finding in the personality disorder literature is that APD is frequently comorbid with other Axis II, Cluster B (Dramatic-Erratic-Emotional) disorders, particularly borderline personality disorder (e.g., Gunderson, Zanarini, & Lisiel, 1991). These findings are not inconsistent with clinical views of psychopathy, and can thus be considered evidence supporting the concurrent validity of APD (although the comorbidity with substance use may be great enough to impede differential diagnosis; Gerstley, Alterman, McLellan, & Woody, 1990). However, there is also evidence of unexpected or theoretically inconsistent comorbidity, such as overlap with schizophrenia and mania when the exclusion criterion (D) is ignored (Boyd et al., 1984; Robins et al., 1991). There is no systematic experimental evidence to support the construct-related validity of the APD criteria. Other reviewers (e.g., Widiger, 1992; Widiger & Corbett, in press) have referred to a body of supportive evidence that includes biochemical, genetic, and adoption studies; but as many (if not most) of the studies cited did not use the actual DSM-III or DSM-III-R criteria, and as the equivalence of DSM-III(-R) and other (e.g., RDC or Feighner) criteria is questionable (Widiger et al., 1992), the relevance of these studies is unclear.  10  Summary. The DSM-III-R criteria for APD have adequate interrater reliability and temporal stability, although some of the subcriteria have extremely low prevalence, poor interrater  It is worth noting that in 5 of the 38 abstracts citing APD, the primary focus of the research was actually psychopathy as assessed by the PCL(-R); in another one, the PCL was used to validate the APD criteria. 10  15  reliability, and/or low item-total correlations. There are no normative data for DSM-III-R APD symptoms or diagnoses. DSM-III (-R) APD diagnoses have adequate concurrent and convergent validity, but the remaining facets of its validity are questionable. Much of the criticism discussed above suggests that the APD criteria focus too much on antisocial behavior, lack a clear two-facet structure, and may be indistinguishable from severe or persistent criminality. These are serious weaknesses that make the APD criteria problematic for research and clinical practice in forensic populations. On the other hand, they do have two strengths. First, with respect to chronicity, they appear to measure a construct that is relatively stable over time. Second, at least when the assessment is based on collateral information in addition to interview data, they may be relatively immune to the effects of deceitfulness. 2. Hare Psychopathy Checklist-Revised (PCL-R)  Format. The PCL-R is a 20-item symptom construct rating scale. To control for the effects of deceitfulness, ratings are made on the basis a semi-structured interview and a review of collateral information (although they can also be based on collateral information alone, if necessary).. Each item consists of a one-page description of a rather complex, high-level trait (e.g., "Shallow Affect" or "Criminal Versatility"); the summary labels of the items are presented in Table 2. The response format is a 3-point scale (0 = item does not apply, 1 = item applies  somewhat, 2 = item definitely applies); items can also be omitted under certain conditions. Individual items are summed (and prorated, if items were omitted) to yield dimensional scores ranging from 0 to 40 that reflect the severity of psychopathic traits; a cutoff score can also be used to yield lifetime diagnoses of psychopathy (< 29 = nonpsychopath; > 30 = psychopath). In addition, the PCL-R yields factor scores reflecting the two facets of psychopathy. Administration and scoring of the PCL-R takes about 2 1/2 to 3 hours. The PCL-R was constructed using a mixture of methods. First, more than a hundred items were generated through a literature review and clinical experience. Second, these items  were piloted; those that were redundant or could not be scored reliably were dropped. Third, the shortened item pool was used on a sample of adult male inmates for whom clinical global ratings of psychopathy were available; items were dropped if they did not discriminate between those 16  Table 2 Items in the PCL-R Item  Description  Factor Loading  1.  Glibness/Superficial Charm  1  2.  Grandiose Sense of Self-Worth  1  3.  Need for Stimulation/Proneness to Boredom  2  4.  Pathological Lying  1  5.  Conning/Manipulative  1  6.  Lack of Remorse or Guilt  1  7.  Shallow Affect  1  8.  Callous/Lack of Empathy  1  9.  Parasitic Lifestyle  2  10.  Poor Behavioral Controls  2  11.  Promiscuous Sexual Behavior  12.  Early Behavioral Problems  2  13.  Lack of Realistic, Long-Term Goals  2  14.  Impulsivity  2  15.  Irresponsibility  2  16.  Failure to Accept Responsibility for Own Actions  1  17.  Many Short-Term Marital Relationships  18.  Juvenile Delinquency  2  19.  Revocation of Conditional Release  2  20.  Criminal Versatility  identified as psychopaths and nonpsychopaths according to the global ratings or if they did not correlate with the other items. The original target population of the PCL-R was incarcerated adult male offenders (Hare, 1991), and most research using the scale has focused on White, North American offenders in federal or state/provincial prisons. However, the PCL-R has also proved useful in research on forensic psychiatric patients (Hart & Hare, 1989; Rice et al., 1992), and has been used with female offenders, young offenders, a variety of ethnic minority groups, offenders in Britain and Europe, and even noncriminals (see Hare, 1991). Reliability. In the PCL-R manual, Hare (1991) presents summary reliability data from 11 forensic samples (N = 1,632). Unlike the APD criteria, construction of the PCL-R was guided by psychometric principles, so it is no surprise that it has superior psychometric properties. The individual PCL-R items have acceptable prevalence, interrater reliability, and itemtotal correlation. The internal consistency reliability is quite high: the median alpha coefficient across the 11 samples was .87, and the median MIC (mean inter-item correlation) was .25. The interrater reliability of total scores is acceptable: the median intraclass correlation coefficient for PCL-R total scores (ICC1; Bartko, 1976) in 6 samples that used multiple raters was .88. For clinical purposes, it is probably best to average two independent ratings; the effective interrater reliability using this procedure (ICC 2 ) was .94. PCL-R diagnoses of psychopathy also have acceptable interrater reliability. Kappa coefficients of agreement between independent raters reported in various studies range between .50 and .80 (e.g., Hart, Forth, & Hare, 1991; Hart & Hare, 1989)." Only one study has looked at the temporal stability of the PCL-R (Alterman et al., 1992). In that study, which looked at 88 adult men attending a methadone maintenance program, the one-month test-retest reliability (r) of total scores was .89. This estimate is similar to that reported for the PCL over a 10-month interval in 42 adult male inmates (Schroeder, Schroeder, &  All the interrater reliability data described here were obtained using the interviewer-observer method. 11  18  Hare, 1983). There is considerable evidence that PCL-R scores are uncorrelated with subjects' emotional states (state anxiety or dysthymia) at the time of assessment (Hare, 1991). The PCL-R factors are less reliable than total scores. This is to be expected, given that the factor scales are shorter in length than the total scale (8 items for Factor 1 and 9 items for Factor 2, versus 20 items for the full scale). Nevertheless, the factor scores are sufficiently reliable for research purposes (Alterman et al., 1992; Hare, 1991). Norms. The PCL-R presents normative data for total and factor scores from 7 samples of adult male prisoners (N = 1192) and 4 samples of adult male forensic psychiatric patients (N = 440). The distribution of scores varies little within the two settings, despite differences between the samples in principle investigator, country of origin, security level of the institution, legal status of subjects, and sampling technique employed. Demographic variables such as age and race appear to have a small but statistically significant association with PCL-R scores; similar results have been reported for the PCL (Hare, 1991). Although the PCL-R has been used with female offenders, young offenders, and noncriminals, no norms are available for these populations at the present time. Validity. The PCL and PCL-R have good content-related validity, as evidenced by their clear two-facet structure. Perhaps the only weakness here is that the items were developed and intended for use in forensic populations. This creates two possible problems for their use with noncriminals. First, the base rate of psychopathy or psychopathic symptoms probably differs greatly from that of the PCL and PCL-R validation samples; consequently, the reliability and validity of the items may be diminished. Second, three items from the PCL and PCL-R are scored on the basis of formal criminal records, making them difficult to score and possibly extremely rare in noncriminals. (Alternatively, these items can be omitted in noncriminals and total scores prorated.) There are preliminary indications that these problems do not render the PCL-R invalid for use with noncriminals, although they may decrease its utility somewhat. The concurrent validity of the PCL and PCL-R is good. They are moderately to highly correlated with clinical global ratings of psychopathy, ratings made using Cleckley's 16 criteria, and APD diagnoses and ratings, typically in the range of .55 to .85 (Hare, 1980, 1985, 1991). 19  Their correlations with the MMPI Pd, MCMI-II 6A, and CPI Pd scales are rather low, averaging about r = .30 in magnitude; however, as noted earlier, this probably reflects problems with the self-report measures (Hare, 1985, 1991; Hart, Forth, & Hare, 1991). The predictive validity of the PCL scales is good, particularly given the rather poor performance of most psychological tests and diagnoses in the prediction of criminal behavior. More than a dozen studies conducted in Canada and the United States indicate that PCL/PCL-R scores are correlated with antisocial and violent behavior both inside and outside of correctional institutions, including recidivism following conditional release from prison, response to correctional treatments, and institutional misconduct (see Hare, 1991; Hare & Hart, 1993). Psychopaths also have criminal careers--patterns of violent and nonviolent offending across the life span--that are quite different from those of nonpsychopaths (Hare, McPherson, & Forth, 1988; Williamson, Hare, & Wong, 1987). In studies that have compared the ability of various measures to predict criminal behavior, the PCL and PCL-R scales perform as well as or better than both other measures of psychopathy (such as APD or the MMPI) and actuarial risk assessment scales (Hare, 1991; Rice et al., 1992; Serin, Barbaree, & Peters, 1990; Simourd, Bonta, Andrews, & Hoge, 1990). The PCL and PCL-R have a clear pattern of convergent and discriminant validities, the interpretation of which is greatly clarified by analysis of the two factors. Like APD, the PCL scales are significantly associated with substance use disorders; however, this association is due entirely to Factor 2 (Hart & Hare, 1989; Hemphill, Hart, & Hare, 1992; Smith & Newman, 1990). Similarly, the PCL and PCL-R correlate positively with DSM-III-R Cluster B personality disorders and negatively with several Cluster C disorders; in this case, however, the association is due primarily to Factor 1 (except for the correlation with APD, which is due primarily to Factor 2; Hare, 1991; Hart & Hare, 1989). The factors also have distinct patterns of correlations with self-report measures of personality: Factor 1 correlates negatively with anxiety and empathy, and positively with narcissism and dominance; Factor 2 correlates positively with sensation-seeking and impulsivity, and negatively with nurturance (Harpur et al., 1989; Hart, Forth, & Hare, 1991; Hare, 1991). Similar results have been found using projective measures (e.g., Gacono, Meloy, & 20  Heaven, 1990). The PCL-R has good clinical specificity with respect to DSM-III-R Axis I and Axis II Cluster C disorders, both in absolute terms (Hart & Hare, 1989; Raine, 1986) and relative to other measures such as the MMPI (Howard, Bailey, & Newman, 1984; but cf. Howard, 1990). Finally, there are more than 20 published experimental investigations supporting the construct validity of the PCL and PCL-R. Although they have no apparent brain damage (at least as measured by standard neuropsychological measures; see Hare, 1984; Hart, Forth, & Hare, 1991), psychopaths have linguistic functions that are abnormal and/or weakly lateralized in the cerebral hemispheres and they give unusual behavioral and physiological responses to affective stimuli (see Hare, Williamson, & Harpur, 1988; Williamson, Harpur, & Hare, 1991). In addition, psychopaths show little physiological arousal in anticipation of noxious stimuli; together with the results of many studies on learning and attentional processes in psychopaths, this has been interpreted as evidence of an adaptive coping response that helps them to selectively ignore cues of impending punishment but that also makes them susceptible to over-focusing on reward cues (for a review, see Harpur & Hare, 1991). It is worth noting here that the construct validity of psychopathy does not seem to be unduly affected by race (Kosson, Nicholls, & Newman, 1990; Wong, 1984).  Summary. The PCL-R has excellent psychometric properties, although there has been relatively little research looking at its temporal stability. It has good normative data for male forensic populations. There is considerable research supporting all facets of the PCL-R's validity. Like APD, the PCL-R appears to measure a chronic disorder. As a measure that is based in large part on file review, it also appears to be relatively immune to deceitfulness. Unlike APD, however, the PCL-R has a clear two-facet structure. In addition, it predicts crime but its content is not too focused on criminality. Perhaps the biggest weakness of the PCL-R is that its target population is adult male  offenders. It has been used with female offenders, young offenders, and noncriminals, but no normative data are available for these groups. Also, some of the PCL-R items may not be relevant for use with noncriminals. Another problem is that completion of the PCL-R is a rather lengthy process that requires access to collateral information (at least in clinical settings). These 21  factors may decrease the PCL-R's attractiveness to clinicians working in civil psychiatric and other noncriminal settings. 3. Minnesota Multiphasic Personality Inventory (MMPI-2)  Format. The MMPI-2 is very familiar to most psychologists as a broad-band, self-report inventory of personality and psychopathology. It contains 567 true-false items, all declarative statements phrased in the first person singular. The items yield scores on 10 basic clinical scales, 4 basic validity (response style) scales, and a number of supplementary scales. In the standard scoring procedure, raw scores are first adjusted for defensiveness (as measured by one of the validity scales, K) and then converted to T-scores; T-scores > 65 are considered to be high. The target population of the MMPI-2 is adults (age 18 or older) who have Grade 8 reading ability or better. The entire MMPI-2 usually takes between 1 and 1 1/2 hours to complete, although patients with serious psychopathology may take 2 hours or longer. Scoring and interpreting the MMPI-2 by hand can be a lengthy process, taking and hour or longer; fortunately, computer scoring and interpretation are available. The MMPI was originally constructed using the criterion groups method, and the items included in the Pd scale were those that reliably discriminated a heterogeneous sample of delinquent adolescents and young adults from normal controls. The MMPI-2 Pd scale contains 50 items that, according to Graham (1990, pp. 61-62), "cover a wide array of topics, including absence of satisfaction in life, family problems, delinquency, sexual problems, and difficulties with authorities. Interestingly, the keyed responses include both admissions of social maladjustment and assertions of social poise and confidence." When the MMPI was revised, 4 Pd items were rephrased to modernize, clarify, or simplify their content. Because of the wide range of items included in the Pd scale, interpretation of high and low scores can be difficult. To aid interpretation, Harris and Lingoes (1955, 1968) constructed five Pd subscales using a rational approach (i.e., based on the apparent content of the items). They labeled the subscales Familial Discord (Pdl), Authority Problems (Pd2), Social Imperturbability (Pd3), Social Alienation (Pd4), and Self Alienation (Pd5).  22  There are actually several ways in which the MMPI-2 can be used to assess psychopathy, including use of the Pd scale alone, Pd in combination with Ma, profile analysis (i.e., two- and three-point code types, such as 4-9 or 4-8-9), or research scales (e.g., Morey, Waugh, & Blashfield, 1985). I will focus on the Pd scale, as it is the most common MMPI-based measure of psychopathy and because its items are a central component of all the other measures. Also, the Pd scale is probably equal or superior to other MMPI-based measures of psychopathy in terms of  its reliability and validity (e.g., Foreman, 1988; Hare, 1985). Reliability. The Pd scale has low internal consistency. In the MMPI-2 normative sample,  alpha was .60 for men and .62 for women. The scale's item homogeneity is extremely poor; using the Spearman-Brown formula, I estimate that it has a MIC of about .03. The Harris-Lingoes subscales have better item homogeneity, but their internal consistency (alpha) is still quite low due to their brevity. The Pd scale has acceptable temporal stability. The MMPI-2 manual reports a one-week test-retest reliability (r) of .81 for 82 male college students and .79 for 111 female college students. The Harris-Lingoes subscales have similar temporal stability (Graham, 1990). These results are consistent with previous research using the 1VIIVIPI, where the typical range of testretest reliability coefficients was .70 to .85 over one day, .60 to .85 over one to two weeks, and .35 to .60 over periods of a year or longer (Graham, 1990); interestingly, temporal stability was lower in criminals than in psychiatric patients or normals. One factor that may limit the temporal stability of the Pd scale is its susceptibility to state variables. Evidence supporting this hypothesis comes from the MMPI-2 manual, which notes that in the normative sample the Pd scale correlated .37 with the Depression (D) scale and .51 with the Psychasthenia (Pt) scale; these scales reflect dysthymic and anxious mood states, respectively. Because the MMPI-2 is a self-report inventory, interrater reliability is not an issue. It has no parallel form. Norms. The MMPI-2 has excellent norms. The normative sample comprised 1,138 adult  men and 1,462 adult women residing in U.S. communities, and was representative of the 1980 national census with respect to sex, age, race, and geographic region of residence. Separate 23  norms are available for men and women. Because the MMPI-2 Pd scale scores are transformed into T-scores, the average T-score for men and women is, by definition, 50. For eight of the MMPI-2 clinical scales, including Pd, a process called "uniform" T-score transformation was employed instead of the usual linear transformation, to insure that the percentile rank for a given T-score was the same on each scale. Because of this innovation, we know with that approximately 8% of both sexes have T-scores > 65 on Pd.  12  Graham (1990) presents norms for  the Pd sub scales based on linear T-scores. No normative data are available for forensic populations at the present time, although they almost certainly will appear in the psychological literature over the next few years. Such data are critical in order to determine if the MMPI-2 is able to make meaningful discriminations among offenders. This may be a problem, as reviews of the use of the MMPI in correctional settings have concluded that "the MMPI profiles of prisoners have seemed remarkably homogeneous. Numerous studies show that scale 4 usually is the most elevated scale in mean profiles of prisoners and identify the 4-2 and 4-9 code types as those most frequently occurring for prisoners" (Graham, 1990, p.196). Validity. The content-related validity of the Pd scale appears to be poor: first, the items focus primarily on the social deviance facet of psychopathy; second, it includes some items whose content is of questionable relevance (e.g., sexual maladjustment); and third, it has poor item homogeneity. A good indication of the heterogeneous content of the Pd scale comes from examination of the correlations among the Harris-Lingoes subscales (Graham, 1990). For men, they range from -.22 to .71, and average about .22; 30% are negative in direction. For women, the range is -.26 to .69, and the average is .21; once again, 30% of the correlations are negative. I should note that these criticisms may be unfair in some respects, as content-related validity was not a concern during the construction and revision of the MMPI.  Note that the test authors do not assume that the actual prevalence of psychopathy in men and women is 8%. The cutoff was selected to identify people with relative, rather than absolute, elevations on the scale; it was not selected because of its efficiency for predicting psychopathy diagnoses. 12  24  The concurrent validity of the Pd scale is low. It correlates about r = .30 with other clinical and self-report measures of psychopathy, including APD, the PCL-R, and the So scale from the CPI (Hare, 1985). Its predictive validity with respect to criminal behavior is also poor, both in absolute terms and in comparison to other psychopathy measures (e.g., Simourd et al., 1990). There is a voluminous research literature on the MMPI and MMPI-2, including a large number of studies that have used the Pd scale to define psychopathy for laboratory research. In general, the evidence supporting the construct-related validity of the Pd scale is not overwhelming. For example, with respect to convergent and discriminant validity, spouse ratings of behavior were collected for 1,644 normals (822 men and 822 women) as part of the MMPI-2 restandardization research; Graham (1988) collected similar ratings, made by attending psychologists and psychiatrists, for 423 patients (232 men and 191 women). The behaviors rated included several dozen symptoms of psychopathology (e.g., "Has trouble sleeping," "Wears unusual clothing"). In normal men, the Pd scale had substantial correlations only with the items "Trouble with the law" (r = .21) and "Uses nonprescription drugs" (r = .16). In normal women, the Pd scale correlated with "Angry, yells" (r = .25), "Cooperative" (r = -.24), and "Irritable, grouchy" (r = .23). In male patients, it correlated with "Guilt feelings" (r = .17) and "Grandiosity" (r = .17); in female patients, it correlated with "Grandiosity" (r = -.21), "Depressive mood" (r = .21), and "Emotional withdrawal" (r = .20). Although the correlations for normals are theoretically consistent, they are different for men and women. The correlations for patients also differ for the sexes; furthermore, most of them are theoretically inconsistent (e.g., negative correlations with grandiosity, positive correlations with depression and guilt). The results of the computer search indicated that recent research on Pd and crime has yielded mixed evidence concerning the construct validity of the scale. On the positive side, Pavelka (1986) and Bayer and Bonta (1985) found that Pd and its subscales were significantly associated with several indexes of criminality (e.g., recidivism) in adult males. Similarly, Walters (1985) found that military offenders diagnosed as APD scored higher on Pd than did non-APD offenders. In contrast, Roger and Gillis (1989) found that Pd scores were not associated with 25  ICD-9 diagnoses of sociopathy in a sample of 470 forensic psychiatric patients. Howard, Bailey, & Newman (1984) found that MMPI scales (including Pd) discriminated better than chance between various groups of mentally disordered offenders, although they performed less well than did the PCL. Two studies looking at learning also yielded somewhat mixed results. Newman and Widom (1985) found that high Pd subjects showed the same deficits on passive avoidance learning deficits tasks that extraverts did, consistent with a disinhibition model of psychopathy. Brown and Gutsch (1985) examined preference for delayed versus immediate reward and the cognitions associated with each preference. They divided high Pd scorers into "primary psychopaths" (high Pd and low self-reported anxiety) and "secondary psychopaths" (high Pd and high anxiety). Primary psychopaths did not differ from nonpsychopaths; however, secondary psychopaths differed from both primary psychopaths and nonpsychopaths. Several studies found results that are inconsistent with theoretical views of psychopathy. A study of community residents found that Pd scores were negatively correlated with measures of socially desirable responding (Ray & Ray, 1982) and positively correlated with anxiety (Ray, 1983). Sutker and Allain (1983) reported that medical students with high Pd scores were more impulsive and reported greater delinquency as juveniles than did matched controls with low scores; however, they were just as empathic on self-report and interview measures. Eisenman (1980) found high school students with high Pd who abused drugs were better liked by teachers than other students (although the author tried to account for this by characterizing the high-Pd drug abusers as better manipulators).  Summary. The MMPI-2 Pd scale has excellent general population norms and acceptable temporal stability. However, possibly as a consequence of the manner in which it was constructed, it has poor internal consistency, item homogeneity, and content-related validity. Both facets of its criterion-related validity are also poor. Its convergent and discriminant validity are questionable, especially in psychiatric populations. It has not proven to be of use in laboratory or experimental research on psychopathy.  26  With respect to the four specific assessment issues discussed earlier, the Pd scale appears to measure a reasonably chronic construct; unfortunately, the identity of that construct is unclear at this time. One thing is clear: The Pd scale is a poor measure of the two-facet model of psychopathy, and is biased towards measurement of social deviance. Although the content of the items is not too closely associated with criminality, the Pd scale may be unable to make meaningful discriminations among offenders. One strength of the MMPI-2 is that it provides a direct appraisal of deceitfulness and testtaking attitude (via the validity scales). One form of deceitfulness--psychologically sophisticated defensiveness--is even corrected for during the scoring process. However, other forms (e.g., malingering, unsophisticated defensiveness) are not controlled; thus, valid completion of the MMPI-2 still requires considerable cooperation from subjects. 4. California Psychological Inventory (CPI) Format. The revised CPI (Gough, 1987) is self-report inventory designed to assess "folk concepts" of personality--"concepts that arise from and are linked to the ineluctable processes of interpersonal life, and that are to be found everywhere that humans congregate into groups and establish social functions" (Gough, 1987, p. 1). The CPI contains 462 true-false items that take the form of declarative statements (not necessarily phrased in the first person). The items form 20 folk concept scales and three "structural" scales (which reflect higher-order factors); a number of special research scales have also been developed. The CPI does not contain scales to assess or correct for response styles, although the manual presents several methods for determining the validity of protocols. Raw item scores on each scale are summed and then transferred into "standard scores" (actually, T-scores); different norms are available for men and women. The target population of the CPI is adolescents and adults (age 13 and older) who have at least Grade 7 reading ability. The test typically takes about 45 to 60 minutes to complete. Computer scoring and interpretation are available. The So scale is intended to measure the internalization of social norms and values, so that those with "very strong rule-breaking and norm-violating proclivities" would receive high scores (Gough, 1987; p. 45). In its original form (Gough & Peterson, 1952), the scale was designed to 27  measure psychopathy as conceptualized in Gough's role-taking theory (Gough, 1948). Later, with the publication of the CPI, the scoring for the scale was reversed to make it consistent with the other scales. Like the MMPI Pd scale, the So scale was constructed using an empirical approach: Items were selected for inclusion in the scale according the magnitude of their correlations with non-test criteria, not on the basis of internal consistency. The current So scale consists of 46 items (some of which were taken from the MIVIPI item pool) that tap such characteristics as delinquency and rejection of traditional values. Some items are intentionally subtle in content. Six items from the original So scale were dropped during the CPI revision process. Given the ratio of items to scales, there must be item overlap on the CPI; however, the manual does not present any information on this subject. Gough has recently developed four subscales for So, based on factor analyses: Optimism, self-confidence, and positive affect (12 items); Self-discipline and cathexis of social norms (15 items); Good memories of home and parents (10 items); and Interpersonal awareness and sensitivity (9 items).' 3 Reliability. The internal consistency reliability of the CPI is moderate. In 400 college students, coefficient alpha was .67 for 200 male college students and .76 for 200 female college students (.71 for the combined sample). Due to the length of the scale, this reflects very low item homogeneity: Using the Spearman-Brown formula, I estimate that MIC in this sample was .05. As was the case with the Pd scale, however, it must be emphasized that Gough does not consider internal consistency to be an important criterion by which to judge the So scale. The So scale has moderate temporal stability. The one-year test-retest reliability was .69 in a sample of 85 high school boys and .74 in 38 high school girls. No information is available in the manual concerning the So scale's susceptibility to state variables. Norms. Gough based the CPI norms on archival data. He selected a total of 1,000 males and 1,000 females from diverse groups--for example, high school students, mathematicians, nurses, prisoners, psychiatric patients, and "San Francisco area residents"--so that the demographic characteristics of the normative sample were similar to those of the general  13  H.G. Gough, personal communication to R.D. Hare, October 19, 1992. 28  population.^Raw scores are apparently converted into standard scores using a linear transformation procedure. The CPI manual does not recommend a specific cutoff score for the So scale. Throughout the manual, though, Gough appears to consider T-scores greater than 65 or 70 to be high, at least in a relative sense. It is likely that somewhere between 5% and 15% of males and females would receive scores at or above these cutoffs. Validity. The content-related validity of the So scale is low to moderate. The manifest content of some So items is of questionable relevance to psychopathy (e.g., "I think Lincoln was greater than Washington" or "My table manners are not quite as good at home as when I am out in company"). Also, the items appear to tap mainly the social deviance facet of psychopathy. Finally, the scale has low to moderate internal consistency and item homogeneity. Recall, however, that these may not be fatal flaws in a scale constructed using the empirical approach. The concurrent validity of the So scale is also low. It is correlated about -.30 with PCL (-R) scores (e.g., Hare, 1985, 1991) and with APD diagnoses (Hare, 1985). Its correlations with other self-report measures, such as the Pd scale, are also typically in the range of -.30 to -.40 (Hare, 1985). One strength of the So scale may be its predictive validity. In a meta-analysis, the So scale was robustly related to antisocial behavior (Simourd et al., 1990). It performed significantly better than did the Pd scale, and at about the same level as the PCL(-R). (However, most of the studies supporting the "predictive" validity of So involved differentiating samples of juvenile delinquents from normal adolescents, whereas those supporting the PCL-R involved the postdiction and prediction of criminal convictions, including violent crime and recidivism after release from prison, in adult male offenders.) With respect to construct-related validity, the CPI manual presents data on the association between So and the results of personality assessments conducted over a period of years at the Institute for Personality Assessment and Research at Berkeley. These personality assessments included three types of ratings (Q-sorts, adjective checklists, and psychosocial history checklists) from three different sources (peers, spouses, and trained staff). Gough presents the items from 29  each assessment method that had the highest positive and negative correlations with each CPI scale. The pattern of results appears to be theoretically consistent, although it is hard to make absolute judgments concerning validity, as Gough does not give the actual magnitude of the correlations. Peers, spouses, and staff were consistent in rating the following adjectives among the 20 strongest correlates of the So scale: reckless (negatively correlated), rebellious (negative), and organized (positive). The results of the computer search found little evidence supporting the construct validity of So in the past 12 years. Barbour-McMullen and Coid (1988) found that So could differentiate APD forensic patients from other patients and from normals. O'Mahony and Murphy (1991) also found that So could differentiate between students and prisoners when the groups were asked to respond honestly; however, the scale was very susceptible to response styles when subjects were asked to fake good. Gorenstein (1982, 1987) found that So-diagnosed psychopaths had performance deficits on neuropsychological tests relative to psychiatric and normal controls; however, this finding contradicts several much larger studies that used other measures of psychopathy (Hare, 1984; Hart, Forth, & Hare, 1990; Sutker, Moan, & Allain, 1983).  Summary. The CPI So scale has many of the same strengths and weaknesses as does the MMPI-2 Pd scale. On the plus side, it has general population norms and moderate temporal stability; on the down side, it has poor internal consistency, item homogeneity, and content-related validity. The criterion-related validity of the So scale may be slightly better than the Pd scale; however, its construct-related validity is just as questionable. With respect to the four specific assessment issues discussed earlier, the So scale appears to measure a reasonably chronic construct; once again, however, the identity of that construct is unclear. The So scale is a poor measure of the two facets of psychopathy, and is biased towards measurement of social deviance. Its content is not unduly associated with criminality. The CPI allows for a limited appraisal of deceitfulness and test-taking attitude, but does not correct for  these; thus, valid completion of the inventory requires the cooperation of subjects.  30  5. Millon Clinical Multiaxial Inventory (MCMI-II) Format. The MCMI-II is a self-report inventory of psychopathology intended for use with clinical populations. It contains 175 true-false items, all declarative statements phrased in the first person singular. It yields scores on 4 validity scales, 13 scales analogous to the DSM-III-R Axis II personality disorders, and 9 clinical syndrome scales analogous to DSM-III-R Axis I disorders. The MCMI scales were constructed using a three-stage procedure. In the first ("theoreticalsubstantive"), Millon's theory of psychopathology was used to identify the relevant constructs to be measured and to generate scale items to measure those constructs. In the second phase ("internal-structural"), the initial item pool was administered to a sample of patients in order to determine the psychometric properties of items and scales; items with poor properties were subsequently dropped. Finally, in the third stage ("external-criterion"), the revised item pool was again administered to a large sample of patients to determine the predictive, convergent, and discriminant validity of scales and items with respect to clinical ratings and diagnoses; these results were used to develop item weights that maximized the MCMI-II's predictive efficiency. Each MCMI-II scale comprises a relatively small number of items (usually between 8 and 10) that are unique to that scale, as well as other items that are shared with other scales. Scale items are given a weight (1, 2, or 3) according to their theoretical importance and empirical validity. Raw scores on some scales are then adjusted according to scores on the validity scales and on two "hidden" (i.e., unreported) correction indexes. Finally, scale scores are transformed into base rate (BR) scores. The BR transformations are complex, and differ according to the sex and race of patients. Their interpretation is also complex: briefly, for the severe personality pathology scales and the clinical syndrome scales, BR scores > 75 indicate that the disorder or syndrome measured by a scale is "present," and scores > 85 that it is "prominent" (Millon, 1987). For the clinical personality pattern scales, Millon recommends a more sophisticated configural analysis that takes into account both absolute and relative scale elevations (although this procedure does not appears to improve significantly over the more simple cutoff method; Rennenberg, Chambless, Dowdall, Fauerbach, & Gracely, 1992). Not surprisingly, computer scoring and interpretation are available (and recommended; Millon, 1987) for the MCMI-II. 31  In the MCMI-I, psychopathy was assessed by the Antisocial/Aggressive (6) scale. The content of that scale was akin to APD in DSM-III-R in many respects, but also included elements of sadism and of the Cleckleyan concept of psychopathy. In the MCMI-II, scale 6 was split into two separate scales: the Antisocial (6A) scale, a more pure measure of APD; and Aggressive/Sadistic (6B) scale, a more focused measure of DSM-III-R sadistic personality disorder traits. Scale 6A contains 45 items: 10 are unique or prototypal (weight factor = 3); 21 are secondary (weight = 2); and 14 are tertiary (weight = 1). Raw scores can range from 0 to 86. 6A has the highest item overlap with scales measuring drug abuse (T; percent overlap = 50), narcissistic personality disorder (5; percent overlap = 38), and sadistic personality disorder (6B; percent overlap = 38). Reliability. According to the MCMI-II manual, the internal consistency of the 6A scale was high (KR20 = .88) in a sample of 825 adult psychiatric patients. As was the case with the MMPI-2, however, this is an overestimate of the scale's item homogeneity, due to the scale length. According to my calculations, item homogeneity is somewhat low (MIC = .14), although Millon does not expect the MCMI-II scales to be unifactorial on theoretical grounds (Millon, 1987). Millon (1987) reports the temporal stability of the MCMI-II scales in six different samples, including a nonclinical sample, two outpatient samples, and three inpatient samples. The stability estimates (r) were moderate, ranging from .64 to .88 (median = .71) over a period of 3 to 5 weeks. Temporal stability may be limited by undue susceptibility to state factors; the MCMI-II manual notes that scale 6A is correlated with scales that reflect mood states, including hypomania (scale N; r = .55) and depression (scale CC; r = .22). Norms. The MCMI-II was normed in two large samples (total N = 1,292) of adult psychiatric patients drawn from a number of clinical settings, including inpatient, outpatient, college guidance, and forensic clinics. (As noted above, it was not intended for use outside such settings, so "normal" norms are not available.) The settings included in the first sample (N = 825) were selected from among those that regularly used the MCMI-I; the degree to which they are representative of patients in general is unknown. Attending clinicians completed a diagnostic 32  checklist, comprising the draft DSM-III-R criteria for the major adult disorders, for each subject. The second sample (N = 467) sampled from settings where clinicians were known to Millon; these clinicians made DSM-III-R Axis I and II diagnoses for each subject. The DSM-III-R diagnostic data were used to construct the MCMI-II BR transformation tables, so that the prevalence of an MCMI-II-diagnosed disorder (according to the decision-making rules outlined in the test manual) was identical to that of the corresponding DSM-III-R disorder. In the combined samples, scale 6A was the most elevated of the 10 clinical personality scales (scales 1 to 8B) for 6% of men and 2% of women; it was the highest or second-highest of the clinical personality scales for 12% of men and 4% of women. At the present time, there are no separate norms for correctional patients. Note that the MCMI-II, unlike the MMPI-2 and CPI, selected cutoff scores so that a scale's selection ratio (the percentage of people that score at or above the cutoff) matches the base rate of the criterion (the prevalence of psychopathy/APD as diagnosed by clinicians) in the normative sample. Put another way, the cutoff identifies people with absolute, rather than relative, elevations. The cutoff for scale 6A had acceptable predictive efficiency characteristics with respect to the criterion diagnoses in the normative sample. Validity. The content-related validity of the 6A scale appears to be acceptable. Although it does not have an explicit two-factor structure, its underlying theory (Millon, 1981) and the manifest content of the items that make up the scale include both social deviance symptoms (e.g., "As a teenager, I got into lots of trouble because of bad school behavior") and affectiveinterpersonal symptoms (e.g., "Pm the kind of person who can walk up to anyone and tell him or her off'). However, recall that the item homogeneity of the scale is somewhat low. The concurrent validity of scale 6A appears to be low to moderate. Hart, Forth, and Hare (1992) looked at the MCMI-II in a sample of 119 adult male offenders who had also been diagnosed using PCL-R and DSM-III-R APD criteria. They reported that scale 6A correlated .45 with PCL-R Total scores, .24 with Factor 1 scores, and .51 with Factor 2 scores; it also  33  correlated .33 with APD diagnoses and .51 with the number of adult APD symptoms.  14  These  correlations, although moderate in magnitude, actually reflected relatively poor diagnostic agreement between the measures. For example, agreement between high scores on scale 6A and PCL-R psychopathy diagnoses was poor, regardless whether a BR cutoff of > 74 or > 84 was used (kappa = .08 and .14, respectively); agreement between 6A and APD diagnoses was slightly higher, but still low (kappa = .25 and .23, respectively). In a sample of 34 court- and self-referred wife assaulters, Hart, Dutton, & Newlove (1992) found a correlation of .53 between 6A and dimensional ratings of APD made using a structured diagnostic interview. In the DSM-IV APD field trials, 6A had generally moderate correlations (around .35 to .40) with symptom counts using the DSM-III-R APD, ICD-10 dyssocial personality disorder, and PCL-R-based psychopathic personality disorder criteria sets. There are apparently no data as yet concerning the predictive validity of 6A with respect to criminal behavior; neither are there data concerning the construct-related validity (e.g., experimental or factorial validity) of the scale.  Summary. Scale 6A of the MCMI-II holds some promise as a measure of psychopathy. It appears to have somewhat better content-related validity than the Pd and So scales; at least, its items have better face validity and are more homogeneous. Its concurrent validity is also as good as or better than other self-reports. Unlike the other self-reports, it has a cutoff that was determined empirically to identify people with absolute, rather than relative, elevations in clinical settings. Finally, the MCMI-II contains scales that attempt to assess and control for response style. Despite its promise, the MCMI-II does have weaknesses. It lacks a clear two-facet structure, and is still biased towards measurement of the social deviance facet of psychopathy. Its concurrent validity is still only moderate in absolute terms; its predictive and construct validity are  This pattern of correlations suggests that the items of scale 6A are biased towards measurement of the social deviance facet of psychopathy, despite appearances to the contrary (Hart et al., 1992). 14  34  unknown. There is no evidence that its controls for response styles are effective. Finally, it has no general population or correctional norms. C. Summary of Chapter 1 Several conclusions can be drawn from the review presented above. First, and most general, it appears that none of the existing assessment procedures for psychopathy is without significant limitations with respect to reliability, validity, or clinical utility. This is important, as it suggests that there is a need for new measures that may complement the existing procedures. Second, we can conclude, as Hare (1985) did, that "clinical-behavioral" procedures (i.e., those that employ expert ratings based on interview and case history data) appear superior to self-report procedures, particularly in terms of validity and ability to control for deceitfulness. Therefore, the development of a new scale may have maximal chance for success if it uses an expert rater format, as opposed to a self-report format. Third, of the procedures reviewed, only the PCL-R has a clear two-facet structure, and this structure has proven extremely useful for clarifying research results. Consequently, new measures should start with an explicit two-facet structure, taking advantage of research on the PCL-R. In Chapter 2, I describe the results of an attempt develop and validate a new scale for the assessment of psychopathy that takes into account the above recommendations.  35  CHAPTER 2: SCALE DEVELOPMENT AND VALIDATION In this Chapter, I will first describe out efforts to develop a new scale for the assessment of psychopathy. Next, I will present the results of efforts to validate the scale in a number of different settings. A. Development of the PCL:SV Several key decisions were made concerning the design of the new scale. First, the scale must be developed according to psychometric theory and evaluated according to standard psychometric criteria. Second, the scale must be conceptually and empirically related to the PCLR. This would help to maximize the scale's chances for success (as discussed in Chapter 1) and would also allow the scale to tap into the extensive empirical literature supporting the validity of the PCL-R. And third, to maximize its utility, the scale must be suitable for use in a wide range of settings (including civil and forensic psychiatric populations) and require relatively little time, effort, and training to administer and score. In order to meet the first two requirements, the new scale retains the format that proved so successful with the PCL-R--namely, an expert-rater, symptom-construct rating scale. Like the PCL-R, the new scale yields both dimensional and categorical indexes of psychopathy. It also has an explicit two-facet structure. Fulfilling the third requirement did not appear to be problematic, as previous research had  indicated that psychopathy, as defined by the PCL-R, could be measured reliably in forensic psychiatric patients (e.g., Harris et al., 1991; Hart & Hare, 1989). One concern was that all the previous research on the PCL (-R) had been conducted in forensic settings; we needed to revise the content of some items to make them appropriate for use with noncriminals. Also, it appeared relatively easy to fulfill at least part of the third requirement--brevity--merely by decreasing the number of items in the new scale, as the high internal consistency of the PCL-R suggested that there was a degree of redundancy among the original items. The main concern here was that decreasing the number of items would decrease interrater reliability. The issue of training was not perceived to be a major problem, as experience suggested that even undergraduates could be taught to make reliable PCL-R assessments of psychopathy. 36  1. First Draft: The Clinical Version of the PCL-R (CV) The first attempt to develop a new, shorter version of the PCL-R resulted in the Clinical Version of the PCL (CV; Cox, Hart, & Hare, 1989). One purpose of the CV was for use in screening jail remands. Because initial assessment interviews in a jail are typically very brief (perhaps 20 to 30 minutes) and because there is often limited access to case history information, the CV consisted of items that were rated primarily on the basis of interview (i.e., stylistic or interpersonal) data. Another purpose was for use in treatment outcome studies, which require measures that may be sensitive to changes in symptom severity over time. It was hoped that the time frame for scoring the CV items could be changed from lifetime to a shorter period (say, the past month). The CV contained only six items: Superficial, Grandiose, Manipulative, Lacks Remorse, Lacks Empathy, and Doesn't Accept Responsibility. The content of these items was derived directly from the PCL-R, but the item descriptions were short, presented in point form, and scored on the basis of a brief interview (10 to 20 minutes). As in the PCL-R, items were scored on a 3-point scale; total scores ranged from 0 to 12. The CV was tested in three different studies. Cox, Hart, and Hare (1989). We tested the CV at the Vancouver Pretrial Services Centre (VPSC), a maximum security jail in Vancouver, Canada. Subjects in this study were 100 males remanded in custody (awaiting trial or a bail hearing) who were referred to staff psychologists for medical, psychological, or security reasons. We did not collect systematic data on the demographic characteristics of these men; however, they ranged in age from 18 to over 60 years, most were White and English-speaking, and most were charged with violent or other serious offenses. Almost all the men had a previous criminal record. Although the sample was not representative of all inmates at VPSC, it was probably very representative of inmates who are monitored or screened by psychologists after admission. One of the researchers was employed as a consulting psychologist at the VPSC. He and  another researcher conducted a series of 100 joint assessments following the usual institutional procedures. Interviews covered the following areas: current charges and past criminal history; educational, occupational, and marital status; and current medical/psychological complaints. Brief 37  counseling followed some interviews. All available case history information was reviewed. Assessments lasted from 5 to 30 minutes in length, and averaged about 20 minutes. After each assessment, both raters independently scored the CV and reviewed any scoring differences. All data were kept confidential and were not released to VPSC medical or security staff. The mean CV score (averaged across the two raters) was 6.24 (SD = 2.45). The intraclass correlation coefficient interrater reliability for the single ratings (ICC j) was .86; the reliability of the averaged ratings (ICC2) was .92. The interrater reliability (r) of the individual items ranged from .52 to .73. Reliability analyses based on the averaged ratings revealed a mean inter-item correlation (MIC) of .51 (range = .40 to .69); internal consistency, as measured by Cronbach's alpha, was .86. The MIC and alpha coefficients observed indicated that the CV could be considered a unidimensional measure. To investigate this issue further, we conducted a principal components analysis (PCA) of the averaged ratings. The PCA yielded one large component (eigenvalue =3.54) accounting for 59% of the common item variance; all other components were much smaller (eigenvalues < .71) and accounted for less than 12% of the remaining variance each. All the CV items had high loadings (> .69) on the first principal component; Item 1 (Superficial) had the highest loading (.84). The cutoff score for a research diagnosis of psychopathy on the PCL-R is 30, a score that is approximately 1 SD above the mean in most samples of male inmates (Hare, 1990). The comparable cutoff score for the CV was 9. We used this cutoff score to divide the two original sets of CV ratings into two groups: those with a score of 9 or above were defined as psychopaths, and those with a score of 8 or less were considered to be nonpsychopaths. Using these categories, the kappa coefficient of diagnostic agreement between the two raters was .93. Applying the above cutoffs to the averaged CV ratings, the base rate of psychopathy in the sample was 15%.  Roy (1988). Roy examined the utility of "present-state" CV ratings, those made in the absence of case history information. He determined the concurrent validity of these ratings in a sample of 60 male federal inmates. All subjects previously had been assessed by independent 38  researchers using PCL(-R) and DSM-III(-R) criteria; Roy reassessed them between 1 and 24 months later, and made CV ratings on the basis of a 30 to 40 minute interview. Subjects ranged in age from 20 to 58 years (mean = 30.5, SD = 8.7); they were serving aggregate sentences of two years or longer, mostly for violent offenses. The mean CV score in the sample was 6.7 (SD = 2.2), the mean PCL score was 26.6 (SD = 7.7), and the mean PCL-R score was 23.6 (SD = 8.2). Using DSM-III criteria, 35% of the subjects met the criteria for APD; for DSM-III-R criteria, the figure was 41.7%. Roy (1988) reported that the interrater reliability of the PCL-R in his sample was r =.74; he did not report reliabilities for the other measures. CV scores were correlated r = .42 with PCL Total scores and r = .38 with PCL-R Total scores. Correlations with Factor 1 scores on the PCL and PCL-R were somewhat higher (r = .54 and .47, respectively), whereas correlations with Factor 2 were lower (r = .23 and .22, respectively). Interestingly, the CV was uncorrelated with DSM-III and DSM-III-R diagnoses of APD (r = -.08 and -.09, respectively). Multiple regression analyses indicated that CV scores in combination with APD diagnoses predicted PCL and PCL-R Total scores significantly better than did either CV scores or APD diagnoses alone (multiple Res ranging between .38 and .51). (It should be noted that none of the above correlations was disattenuated for the unreliability of the various measures across time or raters.) Roesch (1992). Roesch used the CV in another study of pretrial remands at VPSC. Subjects were a random sample of 861 men admitted to the jail over a 12-month period. The subjects ranged from 18 to 71 years in age (M = 30.4, SD = 9.2), and most (82.3%) had a prior criminal record as an adult. Of these subjects, 684 (79.4%) completed an interview-based mental health screening battery that included, in addition to the CV, the following instruments: the Brief Psychiatric Rating Scale (BPRS; Overall & Gorham, 1962), a 19-item symptom construct rating scale of general psychopathology; the Referral Decision Scale (RDS; Teplin & Swartz, 1989), a brief structured interview for major mental disorders; and the Diagnostic Profile (Hart & Hemphill, 1989), a 7-item syndrome construct rating scale. Standard cutoffs were applied to the BPRS and the DP: Subjects with a "hit" on at least one scale were classified as mentally 39  disordered offenders (MDOs); subjects with no hits were designated as non-MDOs. A random subsample of 192 subjects, stratified according to MDO status, were subsequently administered the Diagnostic Interview Schedule (DIS; Robins, Helzer, Croughan, & Ratcliff, 1981). Also, conjoint screening interviews were conducted with 45 subjects to determine the interrater reliability of the rating scales. Complete CV ratings were available for 651 subjects. The mean CV total score was 4.58 (SD = 3.16); using a cutoff of > 9, the base rate of psychopathy was 13.1%. These figures are similar to, although slightly lower than, those reported by Cox et al. (1989); the difference is probably due to the sampling methods employed in the two studies. The internal consistency and item homogeneity of the CV were high (alpha = .88; MIC = .56). In the subsample of 45 subjects, the interrater reliability of the CV was acceptable (ICC1 = .80, ICC2 = .90), as was interrater agreement for psychopathy diagnoses (kappa = .73). A principal components analysis of the CV items once again yielded a single large component (eigenvalue = 3.79) accounting for 63.2% of the common item variance; all other components were much smaller (eigenvalues < .72) and accounted for 12% or less of the remaining variance each. All the CV items had high loadings (> .74) on the first principal component; Item 4 (Lacks Remorse) had the highest loading (.83). With respect to validity, CV scores were negatively correlated with status as a MDO (r = -.11, p = .004). Looking at the DP, the CV was negatively correlated with scores on the Depressed and Organic syndrome scales (r = -.13 and -.12, respectively, both p < .002). On the BPRS, the CV was correlated positively with factors related to grandiosity (r = .43, p < .001) and hostility (r = .30, p < .001), and negatively correlated with factors of dysthymia (r = -.30, p < .001) and psychomotor retardation (r = -.12, p < .002). In the subsample assessed with the DIS, CV scores were positively correlated with APD diagnoses (made ignoring the exclusion criterion) (r = .16, p = .032) and negatively correlated with diagnoses of sexual disorders (r = -.19, p = .008). Summary. The results of these three studies were encouraging, and suggested that it was indeed possible to shorten the PCL assessment procedure. However, two major problems were 40  apparent with the CV. First, factor analyses of the PCL and PCL-R conducted subsequent to the construction of the CV (Harpur et al., 1988; Hare et al., 1990) revealed that all six CV items reflected only Factor 1 of the PCL scales. That is, the CV neglected the social deviance component of psychopathy. The CV's relatively low correlations with PCL(-R) Factor 2 (Roy, 1988) and APD diagnoses (Roesch, 1992; Roy, 1988) support this view. The second problem was that the CV allowed ratings to be made in the absence of case history information. There was some evidence that such a procedure might result in a drop in reliability or validity (Roy, 1988); in addition, it might make the CV unduly susceptible to deceitfulness. Given the arguments I put forward in the Introduction, it is clear that the CV must be considered flawed as a measure of psychopathy. 2. Final Draft: The Screening Version of the Psychopathy Checklist (PCL:SV) Rather than develop a second scale de novo, it was decided to retain the CV's format and expand its content. The six CV items were re-labeled "Part 1," analogous to Factor 1 of the PCL-R. Six new items were then added to tap Factor 2 symptoms: Impulsivity, Poor Behavioral Controls, Lacks Goals, Irresponsibility, Childhood/Adolescent Antisocial Behavior, and Adult Antisocial Behavior. These were labeled "Part 2." In order to make the last two items more suitable for use outside of forensic settings, their content was significantly altered from the original PCL-R descriptions to include actions that did not result in formal contact with the criminal justice system. It was also decided to make the use of case history information a requirement for scoring (at least for clinical purposes). The result of these efforts was a 12-item scale, named the Screening Version of the Psychopathy Checklist (PCL:SV). The scale items are presented in Table 3; details concerning item descriptions, administration, and scoring are presented in Appendix A. The PCL:SV's name is an explicit recognition of its derivation from the PCL-R in content and format. Items are scored using the same 3-point scale as in the PCL-R. Also, raters have the option of omitting as many as three items if they feel there is insufficient information with which to score it; scores are prorated to adjust for the missing items. The PCL:SV yields three dimensional scores. Total scores (the sum of Items 1 through 12) can range from 0 to 24, and 41  Table 3 Items in the PCL:SV Part 1^  Part 2  1. Superficial  7. Impulsive  2. Grandiose  8. Poor Behavior Controls  3. Manipulative  9. Lacks Goals  4. Lacks Remorse  10. Irresponsible  5. Lacks Empathy  11. Adolescent Antisocial Behavior  6. Doesn't Accept Responsibility  12. Adult Antisocial Behavior  42  reflect the degree of overall psychopathic symptomatology exhibited by the individual. Part 1 scores (sum of Items 1 through 6) can range from 0 to 12, and reflect the severity of the interpersonal and affective symptoms of psychopathy (i.e., PCL-R Factor 1). Part 2 scores (sum of Items 7 through 12) can also range from 0 to 12, and reflect the severity of the social deviance symptoms of psychopathy (i.e., PCL-R Factor 2). A Total score of at least 18 (equivalent to a score of at least 30 on the PCL-R) is considered indicative of psychopathy. With respect to ease of administration, scoring, and training, note that the 12-item PCL:SV represents a 40% reduction in length relative to the 20-item PCL-R. In addition, the PCL:SV excludes PCL-R items that are scored on the basis of detailed, highly specific, or difficult-to-confirm information (e.g., marital or sexual history). Pilot testing revealed that the PCL:SV interview could be completed in 30 to 60 minutes, with the case history review and scoring requiring a further 20 to 30 minutes--a 50% reduction in administration time relative to the PCL-R. Also, raters with varied educational and professional backgrounds, from undergraduates to clinical psychologists, were easily trained with a program consisting of a 3-hour lecture and 10 practice ratings; the usual training program for the PCL-R involves 8 to 16 hours of lecture plus the practice ratings. B. Validation of the PCL:SV 1. Overview Research on the validation of the PCL:SV was funded by the MacArthur Risk Study (see Preface). I have collected data from 11 different samples comprising more than 500 subjects. Rather than treat each sample as an independent study, I will present results from all the samples contemporaneously whenever practical. This should facilitate direct comparisons among the samples and reduce redundancy in the descriptions of procedures and results. Certain results are unique to individual samples, however, and will be addressed separately. The validation research focused on several specific issues, not all of which were investigated in each sample. First, I looked at the psychometric properties (distribution and reliability) of the PCL:SV. The facets of reliability examined included internal consistency, item homogeneity, and interrater reliability. Second, I looked at the factor structure of the PCL:SV. 43  Because it was derived from the PCL-R, which has two underlying factors correlated about r = .5, we expected that the PCL:SV would have a parallel structure. Third, I examined the concurrent validity of the PCL:SV with respect to PCL-R scores and adult symptoms of DSM-III-R APD. I expected the correlations among these measures to be high. In order to keep the PCL:SV ratings independent from the PCL-R and APD, different raters assessed them in different sessions, separated by a break of 2 to 7 days. PCL-R data were collected primarily in forensic settings to avoid ambiguity in the interpretation of the results. (Because the PCL-R was not designed for use in civil settings, a low correlation with the PCL:SV might well reflect problems with the PCL-R rather than with the PCL:SV). I also examined concurrent validity with respect to self-report questionnaires related to psychopathy; however, based on past research (e.g., Hare, 1985; Hart, Forth, & Hare, 1991), I expected that self-reports would be only moderately correlated with the PCL:SV. Finally, I examined the convergent and discriminant validity of the PCL:SV. In different samples, the PCL:SV was administered along with interview-based and self-report measures of DSM-III-R Axis II disorders, as well as with measures of normal personality and state measures of psychological distress at the time of assessment. 2. Samples The 10 samples were drawn from four different settings--forensic/nonpsychiatric, forensic/psychiatric, civil/psychiatric, and civil/nonpsychiatric--with a total N of 586. Sample 1. Subjects were 50 adult male inmates at Matsqui Institution, a federal prison in British Columbia. Their average age was 29.0 years (SD = 7.7); 92.0% were White. All were serving aggregate sentences of two years or longer, mostly for violent offenses. Sample 2. Subjects here were 32 adult female inmates at the Burnaby Correctional Centre for Women (BCCW), a provincial prison in British Columbia. This sample included offenders serving provincial sentences (aggregate length of less than two years) and federal sentences (two years or longer). Their average age was 30.0 years (SD = 9.1); 81.3% were White. These data were collected under the supervision of Ms. C. Strachan. Sample 3. Subjects were 67 adult male inmates at various federal prisons in Pacific (British Columbia) Region of the Correctional Service of Canada. Their average age was 33.7 44  years (SD = 7.6); 90.9% were White. All were serving aggregate sentences of two years or longer, mostly for violent offenses. These data were collected under the supervision of Dr. Don Dutton as part of a larger study looking at the correlates of family violence in offenders (Dutton & Hart, 1992a, 1992b). Sample 4. This sample comprised 71 adult offenders (67 men and 4 women) attending the Forensic Psychiatric Outpatient Clinic (FPOC), operated by the British Columbia Forensic Psychiatric Services Commission in Vancouver. Their average age was 37.4 years (SD = 10.0); 90.1% were White. All were on probation or provincial parole and ordered to attend the FPOC for psychiatric treatment (usually medications). Most had a diagnosis of a serious mental disorder (schizophrenia or some other psychotic disorder). Sample 5. Subjects were 49 adult male inpatients at the Kirby Forensic Psychiatric Center in New York, NY. Their average age was 33.9 years (SD = 8.2); 16.3% were White. All were involuntarily hospitalized due to incompetence to stand trial or insanity acquittal. These data were collected under the supervision of Drs. R. Wack and D. Martel. Sample 6. Subjects were 48 adult psychiatric patients (21 men and 27 women) at the University Hospital, UBC Site in Vancouver. Their average age was 32.8 years (SD = 14.7); 93.8% were White. The sample included both inpatients and outpatients who were referred to the Psychology Department for psychodiagnostic assessment. Sample 7. This sample comprised 49 adult psychiatric patients (26 men and 23 women) attending the Western Psychiatric Institute and Clinic at Pittsburgh, PA. Their average age was 34.0 years (SD = 9.9); 63.5% were White. These data were collected under the supervision of Dr. E. Mulvey. Sample 8. Subjects were 80 adult psychiatric patients (54 men and 26 women) at the Greater Kansas City Mental Health Center at Kansas City, MO. Their average age was 31.2 years (SD = 8.8); 66.3% were White. These data were collected under the supervision of Dr. D. Klassen. Sample 9. Subjects were 40 adult psychiatric patients (20 men and 20 women) at the Worcester State Hospital in Worcester, MA. Their average age was 32.6 years (SD = 8.9); 45  91.8% were White. These data were collected under the supervision of Drs. P. Appelbaum and T. Grisso. Sample 10. Subjects were 50 adult undergraduate students (25 men and 25 women) attending various faculties at the University of British Columbia in Vancouver. Their average age was 24.8 years (SD = 6.7); 66.0% were White. Sample 11. Subjects in this sample were 50 adult psychology undergraduates (25 men and 25 women) attending Carleton University in Ottawa, Ontario. Their average age was 20.2 years (SD = 2.2); 86.0% were White. These data were collected by Ms. S. Brown and Dr. A. Forth (Brown, Forth, Hart, & Hare, 1992). The composition of the samples is summarized in Table 4. 3. Procedure Subject recruitment procedures varied from site to site; subjects at some sites were paid, whereas those at others were voluntary. In each case, however, subjects gave informed consent, and all experimental procedures were given ethical approval by the appropriate university and/or institutional committees. The assessment battery completed by subjects also varied from site to site. The single common element in the batteries was the PCL:SV, which was completed on the basis of a semistructured interview and case history information. The interview was similar in structure, but not identical, across the sites. The nature and quantity of case history information varied greatly across settings; indeed, no such information was available in Sample 10, and in Sample 11 it consisted of telephone interviews with a collateral informant (usually a close friend or family member) nominated by the subject. To allow the determination of interrater reliability, PCL:SV interviews were videotaped at several sites, and the case history information was synopsized. An independent rater then completed the PCL:SV on the basis of this information. Where two sets of ratings were available, they were averaged for subsequent analyses. As noted earlier, several sites examined the concurrent validity of the PCL:SV with respect to the PCL-R and DSM-III-R APD. At these sites, the PCL-R and APD ratings were made on the basis of an independent interview, that is, not on the basis of the same interview that 46  Table 4 Demographic Characteristics of Subjects  ^N Setting and Sample  ^  ^Age (yrs.)  Race M^F^All^M (SD)^(% White)  Forensic!Nonpsychiatric 50  29.8 (7.7)  92.0  32  32  30.0 (9.1)  81.3  67  ---  67  33.7 (7.6)  90.9  4. Outpatients  67  4  71  37.4 (10.0)  90.1  5. Inpatients  49  ---  49  33.9 (8.2)  16.3  6. Vancouver, BC  21  27  48  32.8 (14.7)  93.8  7. Pittsburgh, PA  20  20  40  34.0 (9.9)  63.5  8. Kansas City, MO  54  26  80  31.2^(8.8)  66.3  9. Worcester, MA  26  23  49  32.6 (8.9)  91.8  10. UBC undergrads  25  25  50  24.8 (6.7)  66.0  11. Carleton undergrads  25  25  50  20.2 (2.2)  86.0  1. Federal inmates  50  2. Provincial inmates 3. Federal inmates Forensic/Psychiatric  Civil/Psychiatric  Noncriminal/Nonpsychiatric  Note. M = males; F = females. See text for a more detailed description of the samples and settings.  47  was used to score the PCL:SV. To avoid undue contamination of the assessment interviews, they were conducted by different interviewers, blind to each other's ratings, separated by a break of 2 to 7 days. Note that the concurrent validity coefficients are thus attenuated both by temporal unreliability and by unreliability due to raters. From the concurrent validity of the PCL:SV observed in these samples, then, we can estimate the lower bound of its test-retest reliability. All procedures included in the assessment batteries were administered and scored according to the standard instructions, unless otherwise noted. 4. Results Distribution of scores. Table 5 presents the mean and standard deviation of PCL:SV Total, Part 1, and Part 2 scores in the 11 samples. The Table also presents the base rate of psychopathy (i.e., percent of subjects with a PCL:SV Total score > 18) in each sample. It is clear from the Table that PCL:SV scores varied significantly as a function of setting: the two forensic settings had the highest scores, followed by the civil/psychiatric and the civil/nonpsychiatric settings. The base rate of psychopathy in civil settings was very low--10% or less--in five of six samples, and only Sample 8 had a base rate greater than 10%. (Note that these results can be considered preliminary evidence of the concurrent validity of the PCL:SV.) It is interesting that in every sample, the mean Part 2 score was about 2 points higher than the mean Part 1 score. The reason for this is unclear. It may be that Part 1 items are slightly more "difficult" than Part 2 items. That is, Part 1 items may be phrased conservatively, so that even in the case of individuals who exhibit psychopathic symptoms that are all of equal severity, raters are constrained by the framework of the PCL:SV to give them higher scores on Part 1 items than on Part 2 items. Alternatively, it may be that symptoms of psychopathy associated with Part 1 of the PCL:SV are, in general, less severe or less prevalent than symptoms associated with Part 2. There was considerable dispersion of PCL:SV scores within each sample, even in samples where the base rate of psychopathy was very low. This is an important finding, as it suggests that psychopathic traits may prove useful for research or in predicting behavior even in settings where no-one fulfills the diagnostic criteria for psychopathy. 48  Table 5  Descriptive Statistics for PCL:SV Total, Part 1, and Part 2 Scores PCL:SV Score Sample  Total  Part 1  Part 2  Base Rate  Forensic/Nonpsychiatric 1. Federal inmates  15.77 (4.34)  6.95 (2.71)  8.82 (4.99)  34.0  2. Provincial inmates  16.41 (3.49)  7.30 (2.35)  9.11 (2.27)  37.5  3. Federal inmates  12.97 (4.92)  5.16 (2.91)  7.81 (2.94)  17.9  4. Outpatients  13.72 (4.05)  6.01 (2.43)  7.70 (2.19)  19.7  5. Inpatients  16.56 (3.28)  7.55 (2.22)  9.01 (1.73)  34.7  6. Vancouver, BC  5.18 (4.34)  2.01 (2.33)  3.17 (2.64)  2.1  7. Pittsburgh, PA  8.68 (5.92)  3.53 (2.83)  5.15 (3.66)  10.0  13.14 (5.71)  5.33 (3.39)  7.81 (2.99)  23.8  9.63 (4.90)  3.35 (2.72)  6.29 (2.63)  6.1  10. UBC undergrads^3.09 (3.43)  1.41 (1.83)  1.68 (1.85)  0.0  2.88 (2.58)  1.02 (1.33)  1.86 (1.83)  0.0  Forensic/Psychiatric  Civil/Psychiatric  8. Kansas City, MO 9. Worcester, MA  Noncriminal/Nonpsychiatric  11. Carleton undergrads  Note. SD in parentheses. Base rate is percent of subjects with a PCL:SV Total score > 18.  49  PCL:SV items. Appendix B contains the corrected item-total correlations of the PCL:SV items with respect to Total, Part 1, and Part 2 scores in each of the 11 samples; Table 6 presents the mean weighted corrected item-total correlations across the 11 samples. As the Table indicates, the item validities were all acceptable, with none lower than .40. Two independent sets of PCL:SV ratings were collected in seven samples. Appendix B contains the interrater reliability (r) of all 12 items in each of the seven samples; Table 6 also presents the mean weighted interrater reliability of the items across the seven samples. The mean weighted reliabilities were acceptable for all the items, ranging from .50 to .79 and averaging about .60. Although I have not presented the descriptive statistics for the items across the samples, it is clear from the above results that none of the items has an extreme endorsement frequency, at least not one extreme enough to adversely affect its reliability or validity. Internal consistency. Table 7 presents the internal consistency (Chronbach's alpha) of the PCL:SV Total, Part 1, and Part 2 scores across each of the 11 samples. Overall, the internal consistency of PCL:SV Total scores was acceptable for a clinical scale, averaging about .84. This is particularly encouraging given the PCL:SV's relatively short length. Alphas for Part 1 and 2 scores were somewhat lower (mean weighted alphas of .81 and .75, respectively), although this is to be expected as alpha is partially dependent upon scale length. Also, note that alphas for Part 1 were generally higher than those for Part 2; interestingly, the same pattern holds true for PCL-R Factor 1 and 2 scores. Item homogeneity. Table 8 presents the item homogeneity (MIC) of the PCL:SV Total, Part 1, and Part 2 scores across each of the 11 samples. Recall that MIC is not dependent upon scale length; thus, the MICs for the Total and Parts are more directly comparable than are the alphas. The mean weighted MIC for the PCL:SV Total score was .32, indicating a high degree of item homogeneity. Indeed, this value exceeds the cutoff of .20 that is generally interpreted to reflect a unidimensional scale. DeSpite this, the division of the scale into two subscales appears to  50  Table 6 Mean Weighted Interrater Reliabilities and Corrected Item-Total Correlations for the PCL:SV Items in the 11 Samples Item-Total Correlation Item  Interrater Reliability  1.  .60  .50  .57  2.  .65  .50  .63  3.  .62  .48  .47  4.  .59  .65  .64  5.  .57  .63  .63  6.  .57  .55  .58  7.  .59  .54  .58  8.  .69  .43  .41  9.  .60  .47  .44  10.  .53  .56  .58  11.  .79  .45  .54  12.  .50  .52  .53  Median:  .60  .51  .61  .54  Mean:  .61  .53  .59  .52  Total^Part 1^Part 2  51  Table 7 Internal Consistency (Chronbach's Alpha) of PCL:SV Total, Part 1, and Part 2 Scores  PCL:SV Score Total  Part 1  Part 2  1. Federal inmates  .88  .89  .80  2. Provincial inmates  .77  .82  .77  3. Federal inmates  .81  .78  .76  4. Outpatients  .83  .81  .70  5. Inpatients  .72  .73  .52  6. Vancouver, BC  .88  .87  .84  7. Pittsburgh, PA  .87  .78  .80  8. Kansas City, MO  .88  .86  .80  9. Worcester, MA  .83  .77  .67  10. UBC undergrads  .91  .84  .81  11. Carleton undergrads  .69  .58  .66  Median:  .83  .82  .77  Weighted Mean:  .84  .81  .75  Sample Forensic/Nonpsychiatric  Forensic/Psychiatric  Civil/Psychiatric  Noncriminal/Nonpsychiatric  52  Table 8  Item Homogeneity (Mean Inter-Item Correlation) of PCL:SV Total Part 1, and Part 2 Scores PCL:SV Score Sample  Total  Part 1  Part 2  1. Federal inmates  .41  .57  .45  2. Provincial inmates  .24  .44  .40  3. Federal inmates  .27  .37  .35  4. Outpatients  .30  .42  .30  5. Inpatients  .17  .30  .16  6. Vancouver, BC  .40  .54  .46  7. Pittsburgh, PA  .36  .38  .41  8. Kansas City, MO  .39  .51  .40  9. Worcester, MA  .30  .36  .26  10. UBC undergrads  .42  .49  .40  11. Carleton undergrads  .17  .18  .23  Median:  .30  .44  .35  Weighted Mean:  .32  .42  .35  Forensic/Nonpsychiatric  Forensic/Psychiatric  Civil/Psychiatric  Noncriminal/Nonpsychiatric  53  be justifiable, as the mean weighted MICs for Parts 1 and 2 are even higher (.42 and .35, respectively). Interrater reliability. Table 9 presents the interrater reliability (ICC) of Total, Part 1, and Part 2 scores in the seven samples that collected two independent sets of PCL:SV ratings. The Table presents both ICC 1, the reliability of ratings made by one rater, and ICC2, the reliability of ratings averaged across two independent raters. The mean weighted ICC 1 and ICC2 for Total scores were quite high (.84 and .92, respectively) and adequate for research purposes. However, the ICC I may be too low for clinical purposes; those who wish to use the PCL:SV for making important clinical decisions may wish to base such decisions on scores averaged across two independent raters. Note that although Part 1 scores are more internally consistent and have greater item homogeneity than do Part 2 scores, the latter have slightly higher interrater reliability. Once again, the results parallels those found using PCL-R Factor 1 and 2 scores. It may be that psychopathic symptoms related to Part 1 form a cluster that is naturally more cohesive than do Part 2 symptoms, but one that requires greater inference on the part of raters. Alternatively, it may be that ratings of Part 1 symptoms are subject to a "halo effect" that increases their (apparent) internal consistency while decreasing their interrater reliability. The results presented so far in this section focus on the PCL:SV as a dimensional measure; however, the PCL:SV was also intended for use as a diagnostic instrument. I therefore examined the interrater agreement for PCL:SV diagnoses in those samples that had > 5 cases of psychopathy." Kappas were as follows: Sample 1, .70; Sample 2, .36; Sample 4, .36; and Sample 5, .40. The mean weighted kappa across samples was .48, indicating moderate or fair agreement; this is very similar to the figure of .51 that is obtained by calculating kappa on the basis of agreement data that is pooled across samples.  Inclusion of the other samples would have skewed the results, as interrater agreement in them was perfect. However, this agreement was due primarily to the total absence of psychopaths in Samples 10 and 11, and a near-total absence in Sample 6. 15  54  Table 9 Interrater Reliability (ICC) of PCL:SV Total, Part 1, and Part 2 Scores  PCL:SV Score Sample  N  Total  Part 1  Part 2  1. Federal inmates  50  .82 / .91  .84 / .92  .79 / .89  2. Provincial inmates  32  .70 / .85  .66 / .83  .81 / .90  4. Outpatients  59  .81 / .91  .81 / .90  .72 / .86  5. Inpatients  26  .67 / .83  .59 / .80  .76 / .88  28  .86 / .93  .80 / .90  .88 / .94  10. UBC undergrads  50  .92 / .96  .67 / .83  .82 / .91  11. Carleton undergrads  50  .88 / .94  .80 / .90  .90 / .95  Median:  .82 /.91  .80 / .90  .81 / .90  Weighted Mean:  .84 / .92  .77 / .88  .82 / .91  Forensic/Nonpsychiatric  Forensic/Psychiatric  Civil/Psychiatric 6. Vancouver, BC Noncriminal/Nonpsychiatric  Note. ICC = intraclass correlation. Numbers before the oblique stroke are the ICC for single  ratings (ICC1); numbers after, the ICC for averaged ratings (ICC2).  55  Correlation between Part 1 and 2 scores. Table 10 presents the correlation between Part 1 and 2 raw scores in the 11 samples. These correlations varied considerably, ranging from .14 to .73. However, the mean weighted correlation, .53, is very much in line with expectations based on research with the PCL-R (Hare, 1991). Factor structure of the PCL:SV. The PCL:SV is based on an explicit model of psychopathy that posits 2 oblique factors. The results discussed above offer some limited support for this model. A better test would involves the use of factor analytic methods; however, the use of multiple samples, all quite small in factor analytic terms, makes the use of these methods problematic. There are several possible strategies that could be used to investigate the factor structure of the PCL:SV. Given that there is an explicit theoretical structure underlying the items (i.e., two oblique factors correlated about .50), a confirmatory factor analysis makes good sense. However, because this is the first empirical investigation of the PCL:SV, the use of exploratory strategies can also be justified. In light of the above discussion, I decided to use exploratory methods to conduct a preliminary investigation of the factor structure of the PCL:SV. Readers are cautioned not to draw firm conclusions from the following analyses; rather, they should be used to generate hypotheses for future research. I experimented with a variety of methods and found that an oblique two-factor solution, with unweighted least squares extraction and oblimin rotation of the factors, was optimal. Appendix B contains the matrix of rotated factor loadings for each of the 11 samples, as well as the eigenvalues and percentage of variance accounted for by each factor, that were obtained using this procedure. It is difficult to judge the comparability of the factor solutions across samples. To clarify this issue, I calculated Coefficients of Congruence (CCs; Harman, 1976) for corresponding factors in a pairwise manner across the 11 samples. That is, CCs were calculated for each of the two factors, comparing each sample with all of the others. This yielded a total of 110 CCs (55 for each factor), which are presented in Table 11. (The Table does not present CCs between noncorresponding factors.) CCs for the factor that best reflected Part 1 of the PCL:SV are presented 56  Table 10 Correlation Between PCL:SV Part 1 and Part 2 Scores in the 11 Samples Sample Forensic/Nonpsychiatric 1. Federal inmates  .52  2. Provincial inmates  .14  3. Federal inmates  .42  Forensic/Psychiatric 4. Outpatients  .53  5. Inpatients  .37  Civil/Psychiatric 6. Vancouver, BC  .53  7. Pittsburgh, PA  .66  8. Kansas City, MO  .60  9. Worcester, MA  .68  NoncriminaUNonpsychiatric 10. UBC undergrads  .73  11. Carleton undergrads  .33  Median:  .53  Weighted Mean:  .53  Table 11  Coefficients of Congruence Between Factor Loadings for the Oblique, Two-Factor Solution Across the 11 Samples 1  2  3  4  5  6  7  8  9  10  11  1.  (.54)  .98*  .87*  .94*  .91*  .94*  .72  .94*  .74  .64  .73  2.  .95*  (.09)  .86*  .91*  .89*  .91*  .77  .91*  .73  .61  .75  3.  .85*  .79  (.35)  .91*  .82  .82  .54  .83  .62  .48  .37  4.  .89*  .78  .90*  (.50)  .93*  .95*  .70  .94*  .63  .58  .63  5.  .69  .53  .75  .89*  (.16)  .89*  .79  .88*  .78  .72  .72  6.  .95*  .91*  .86*  .93*  .73  (.39)  .65  .97*  .87*  .54  .68  7.  .86*  .87*  .82  .81  .62  .76  (.35)  .62  .69  .62  .83  8.  .88*  .79  .84  .89*  .66  .70  .73  (.53)  .82  .66  .67  9.  .67  .70  .70  .52  .23  .55  .75  .51  (.47)  .47  .61  10.  .68  .76  .65  .67  .50  .64  .90*  .56  .62  (.60)  .64  11.  .82  .90*  .74  .69  .38  .68  .92*  .68  .79  .89*  (.05)  Sample  Note. Congruence coefficients for the factor corresponding to Part 1 are above the diagonal; those for the factor corresponding to Part 2 are below the diagonal; correlations between the two factors are on the diagonal.  58  above the diagonal; those for the factor that best reflected Part 2, below. The correlation between the rotated factors within each sample appear along the diagonal. In general, CCs greater than or equal to .85 are seen to reflect identical (or highly similar) factor structures; in the Table, these are marked with an asterisk. As a gross index of a sample's overall comparability with the other 10 samples, I simply summed the number of CCs > .85 for both factors. The maximum possible value for this index was 20. The value for the samples was as follows: Sample 1 = 12; Sample 2 = 9; Sample 3 = 6; Sample 4 = 11; Sample 5 = 6; Sample 6 = 10; Sample 7 = 4; Sample 8 = 7; Sample 9 = 1; Sample 10 = 2; and Sample 11 = 3. Clearly, a few of the samples had factor solutions that did not correspond well with the others. To further investigate this issue, I correlated the comparability index described above with several characteristics of the samples. For example, the index correlated r = .16 with sample size, -.23 with the percentage of women in the sample, -.23 with the percentage of non-Whites in the sample, and .41 with the mean age of subjects in the sample. All of these correlations were nonsignificant (df = 9, p > .10, two-tailed test), suggesting that the size and demographic characteristics of the sample had little impact on the replicability of its factor structure. On the other hand, the level of psychopathic traits had a significant impact: The index correlated r = .55 with the mean PCL:SV Total score in the sample, .59 with the mean Part 1 score, and .51 with the mean Part 2 score. These correlations approached statistical significance (df = 9, p < .10, two-tailed), and suggested that the two-factor structure of the PCL:SV could not be recovered in samples with a relative absence of psychopathic symptomatology. Consistent with this hypothesis, the index correlated r = .59 with the base rate of psychopathy in the sample and .71 with the presence versus absence of an extreme base rate of psychopathy (i.e., prevalence < 10%). The latter correlation was highly significant (df = 9, p < .01, two-tailed). 16 If the four samples in which the base rate of psychopathy was extreme (Samples 7, 9, 10, and 11) are excluded, the factor solution replicated quite well across samples, as Table 12  In those samples with an extreme base rate, the two factors appeared to collapse into a single factor. 16  59  Table 12  Coefficients of Congruence Between Factor Loadings for the Oblique, Two-Factor Solution in 7 Samples with Appreciable Base Rates of Psychopathy 1  2  3  4  5  6  8  1.  (.54)  .98*  .87*  .94*  .91*  .94*  .94*  2.  .95*  (.09)  .86*  .91*  .89*  .91*  .91*  3.  .85*  .79  (.35)  .91*  .82  .82  .83  4.  .89*  .78  .90*  (.50)  .93*  .95*  .94*  5.  .69  .53  .75  .89*  (.16)  .89*  .88*  6.  .95*  .91*  .86*  .93*  .73  (.39)  .97*  8.  .88*  .79  .84  .89*  .66  .70  (.53)  Sample  Note. Congruence coefficients for the factor corresponding to Part 1 are above the diagonal; those for the factor corresponding to Part 2 are below the diagonal; correlations between the two factors are on the diagonal.  60  indicates. Out of 21 pairwise CCs for the factor corresponding to PCL:SV Part 1, 18 were > .85, and the three others were > .82. For the factor corresponding to Part 2, 11 of 21 CCs were > .85; the others ranged from .53 to .79. The observed factor structures also corresponded quite well to the theoretical structure of the PCL:SV, with most items having high and/or unique loadings on the predicted factor. To summarize this section of the results, factor analysis offers tentative support for the theoretical structure of the PCL:SV, although the oblique two-factor solution is not recovered in samples with an extreme base rate of psychopathy. Concurrent validity: PCL-R. The PCL:SV was correlated with independent PCL-R ratings in five samples; these correlations are presented in Table 13. As discussed earlier, by independent I mean that different individuals made ratings on the basis of different interviews conducted on different occasions, blind with respect to each other's results. Two sets of independent PCL:SV ratings were available for about 75% of the subjects in Table 13; however, only a single PCL-R rating was available for the vast majority of the subjects. The mean weighted correlation between Total scores on the two scales was .80 (range = .55 to .84). PCL:SV Total scores correlated about the same with PCL-R Factor 1 and 2 scores (mean weighted correlations = .67 and .68, respectively), whereas PCL-R Total scores correlated higher with PCL:SV Part 2 scores than with Part 1 scores (mean weighted correlations = .78 and .61, respectively). This latter result is not surprising, given that the content of the PCL-R is slightly biased towards Factor 2 (9 of 20 items, versus 8 of 20 items for Factor 1). The Table also supports the concurrent validity of the PCL:SV subscales. Part 1 scores correlated higher with PCL-R Factor 1 scores than they did with Total or Factor 2 scores (mean weighted correlations of .68 versus .61 and .40, respectively). Similarly, Part 2 scores correlated higher with PCL-R Factor 2 scores than they did with Total or Factor 1 scores (.81 versus .78 and .48, respectively). Diagnostic agreement between the PCL:SV and PCL-R was examined in samples that had > 5 cases of psychopathy according to either scale. Kappas were as follows: Sample 1, .54; Sample 2, .37; Sample 4, .45; and Sample 5, .37. The mean weighted kappa across the four 61  Table 13 Concurrent Validity: Correlations With PCL-R Total and Factor Scores PCL-R Score PCL:SV Score  Sample 1. Federal inmates (N = 50)  Total  Factor 1  Factor 2 .64** .38* .79**  Total Part 1  .78** .62**  .62** .64**  Part 2  .77**  .43**  2. Provincial inmates  Total  .81**  .73**  (N = 32)  Part 1  .45*  .50** .57**  Part 2  .77**  .18  .88**  Total  .84**  .76**  .71**  Part 1  .72**  .79**  .49**  Part 2  .75**  .52**  .76**  Total  .55** .44** .47**  .45** .52**  .39*  Total Part 1 Part 2  .81** .63**  .78** .57**  .88**  .78** .68** .77**  Total  .81  .62  .71  Part 1  .62  .64  Part 2  .77  .43  .38 .79  Total  .80  .67  .68  Part 1  .61  .68  .40  Part 2  .78  .48  .81  4. Outpatients (N = 65) 5. Inpatients (N = 47) 10. UBC undergrads (N = 49) Median:  Weighted Mean:  Part 1 Part 2  .18  .24  .17 .52**  .88**  Note. PCL-R = Revised Psychopathy Checklist (Hare, 1991). * Test-wise p < .05, ** columnwise p < .05 (both one-tailed).  62  samples was .44, indicating moderate or fair agreement; the kappa calculated on the agreement data pooled across the samples was .46. If we consider PCL-R diagnoses to be the criterion, then the pattern of PCL:SV diagnostic errors was asymmetric: false positive decisions outnumbered false negatives by a factor of 5 to 1. (Of course, it could just as easily be argued that the PCL-R underdiagnoses psychopathy.) These concurrent validities are surprisingly high, given that they are attentuated by several sources of unreliability, and suggest that the PCL:SV could be considered a parallel form of the PCL-R. Also, because the two instruments were administered about a week apart, we can conclude that the short-term temporal stability of the PCL:SV must be high (for Total scores, at least r > .81, and probably r > .85). Concurrent validity: APD. The PCL:SV was correlated with independent APD ratings in six samples; these correlations are presented in Table 14. The APD ratings were actually a count of the number of adult (Criterion C) APD symptoms that the subject exhibited. Two sets of independent PCL:SV ratings were available for about 65% of the subjects in Table 14; however, only a single APD rating was available for the vast majority of the subjects. (Note that although independent of PCL:SV ratings, the APD assessments were not necessarily independent of PCL-R ratings.) The mean weighted correlation between PCL:SV Total scores and APD ratings across the six samples was .70 (range = .52 to .85). As expected, Part 1 scores correlated lower with the APD ratings than they did with Part 2 scores (mean weighted correlations of .49 and .72, respectively). Once again, these concurrent validities were surprisingly high in light of their attenuation due to several sources of unreliability, and were very similar to results found using the PCL-R. Concurrent validity: Self-report measures. The association between PCL:SV scores and self-report measures of psychopathy and/or APD was examined in five samples. The self-reports included the CPI So scale, the MCMI-II 6A scale, and Hare's unpublished Self-Report Psychopathy scale (SRP; see Hare, 1985). The concurrent validities are reported in Table 15.  63  Table 14 Concurrent Validity: Correlations With APD Adult Symptoms PCL:SV Score Sample  N  Total  Part 1  Part 2  1. Federal inmates  50  .62**  .46**  .64**  2. Provincial inmates  24  .52**  .06  .68**  4. Outpatients  52  .61**  .51**  .56**  6. Vancouver, BC  27  .85**  .73**  .85**  10. UBC undergrads  49  .67**  .51**  .74**  11. Carleton undergrads  50  .78**  .49**  .75**  Median:  .65  .51  .72  Weighted Mean:  .70  .49  .72  Note. APD = DSM-III-R antisocial personality disorder (American Psychiatric Association, 1987). * Test-wise p < .05, ** column-wise p < .05 (both one-tailed).  64  Table 15 Concurrent Validity: Correlations With Self-Report Measures of Psychopathy/APD PCL-SV Score Sample  N  Total^Part 1  Part 2  CPI Socialization Scale Sample 2  28  -.29^-.14  -.32*  Sample 10  48  -.47**^-.27*  -.59**  Self-Report Psychopathy Scale Sample 2  27  .69**^.41*  .70**  Sample 4  63  .53**^.45**  .48**  Sample 10  49  .63**^.48**  .58**  Sample 11  50  .64**^.48**  .56**  MCMI-II Antisocial Scale Sample 3  40  .68**^.53**  .67**  Note. CPI = California Psychological Inventory (Gough, 1987); MCMI-II = Millon Clinical Multiaxial Inventory-II (Millon, 1987). * Test-wise p < .05, ** test-wise p < .001 (both onetailed).  65  The correlations between the PCL:SV and the three self-reports were moderate to large in magnitude. Indeed, the correlations with 6A and the SRP were larger than those typically reported between clinical-behavioral and self-report measures of psychopathy, as reviewed in the Introduction, a finding that may bode well for the use self-reports in future research. There was a consistent tendency for the self-reports to have higher correlations with Part 2 than with Part 1. This parallels the results of research conducted using the PCL-R factors, and supports the view that the content of self-reports is somewhat biased towards the social deviance facet of psychopathy. Convergent validity: DSM-III-R personality disorders. In past research, the PCL-R has shown an interesting pattern of association with various measures of the DSM-III (-R) personality disorders. In a study that used clinical ratings of the personality disorders in a sample of 80 forensic psychiatric patients (Hart & Hare, 1989), PCL-R Total scores were positively correlated with antisocial, narcissistic, and histrionic personality disorder ratings; and uncorrelated or negatively correlated with ratings of other disorders. Factor 1 scores had a pattern of correlations very similar to that of Total scores, whereas Factor 2 correlated only with APD ratings. In another study that used the MCMI-II in a sample of 119 prison inmates (Hart, Forth, & Hare, 1991), Total scores were positively correlated with scales measuring antisocial, sadistic, paranoid, narcissistic, borderline, passive-aggressive, and schizotypal personality disorder; and uncorrelated or negatively correlated with scales measuring the other disorders. In that study, all the MCMI-II scales correlated higher with Factor 2 than with Factor 1. On the basis of this research, I decided to test the PCL:SV's convergent validity vis-a-vis the DSM-III-R personality disorders in two samples; I predicted that Total scores would be positively correlated with the Cluster B (Dramatic-Erratic-Emotional) antisocial, narcissistic, histrionic, and borderline disorders, as well as with sadistic personality disorder. In Sample 6, we tested the PCL:SV against the Personality Disorder Examination (PDE; Loranger, 1988), a clinical-behavioral measure of the DSM-III-R personality disorders. Evidence is accumulating supporting the PDE's reliability and validity (e.g., Loranger, Susman, Oldham, & Russakoff, 1987; Standage & Ladha, 1988). Although the PDE is not the "gold standard" for 66  personality disorder assessment, it is widely recognized as one of the best available measures, as evidenced by its inclusion in the World Health Organization's International Pilot Study on Personality Disorders (Loranger, Hirschfeld, Sartorius, & Regier, 1991). The PDE is scored on the basis of a semi-structured interview and file review. In the present study, it was administered by graduate students in clinical psychology separately from the PCL:SV assessment procedures; however, these assessments were not blind. That is, in the majority of cases, the researcher who administered the PDE also rated the same subject on the PCL:SV. (The plan was to have a least one set of independent PDE and PCL:SV ratings for each subject; however, a large number of videotaped interviews were accidentally damaged by a research assistant before they could be double-rated.) The PDE interview is used to rate 126 separate items on a 3-point scale (0  —  absent, 1 = subclinical, 2 = present). Item ratings are then transformed to yield a score for each DSM-III-R personality disorder symptom, using the same 3-point scale. The symptoms ratings can then be used to make dimensional ratings (the sum of symptom ratings), symptom counts (the number of symptoms rated present), and diagnoses (according to DSM-III-R diagnostic rules) for each personality disorder. I analyzed dimensional ratings, as the base rate of diagnoses was extremely low (i.e., < 10%) for most disorders. The correlations between the PCL:SV and the PDE dimensional ratings are presented in Table 16. As predicted, Total scores were correlated with the ratings of Cluster B disorders and sadistic personality disorder, as well as with passiveaggressive personality disorder. Part 1 scores were correlated more highly with narcissistic and histrionic ratings than were Part 2 scores, whereas Part 2 scores were more highly correlated with antisocial, borderline, passive-aggressive, and sadistic ratings than were Part 1 scores. The PCL:SV was correlated with the MCMI-II, a self-report measure of psychopathology, in Sample 3. I analyzed base rate (BR) scores on the MCMI-II, rather than raw scores. The pattern of correlations observed, presented in Table 17, was similar in many ways to that obtained using the PDE, although there were some interesting differences. For example, in addition to the predicted correlations, Total scores were also correlated with MCMI-II scales measuring schizotypal, paranoid, and self-defeating personality disorder. The latter results are, in some respects, theoretically incongruous, but consistent with previous research (Hart et al., 1991); they 67  Table 16 Convergent Validity: Correlations With PDE Ratings in Sample 6 PCL:SV Score PDE Rating  Total  Part 1  Part 2  Paranoid  .12  .07  .13  Schizoid  .04  -.04  .11  Schizotypal  -.16  -.23  -.06  Antisocial  .74**  .46**  .83**  Borderline  .48**  .24  .59**  Histrionic  .45**  .45*  .36*  Narcissistic  .58**  .63**  .41*  Avoidant  .06  .03  .07  Dependent  .20  .11  .24  Obsessive-Compulsive  -.03  -.12  .06  Passive-Aggressive  .54* *  .37*  .57**  Sadistic  .47**  .39*  .44*  Self-Defeating  .18  .12  .20  Note. N = 38. PDE = Personality Disorder Examination (Loranger, 1988). * Test-wise p < .05, ** column-wise p < .05 (both one-tailed).  68  Table 17 Convergent Validity: Correlations With MCMI-II Personality Disorder Scales in Sample 3 PCL:SV Score MCMI-II Scale  Total  Part 1  Part 2  Schizoid (1)  .23  .26  .14  Avoidant (2)  .22  .02  .35*  Dependent (3)  -.23  -.26  -.16  Histrionic (4)  .16  .14  .14  Narcissistic (5)  .44**  .48**  .30*  Antisocial (6A)  .68**  .53**  .67**  Aggressive/Sadistic (6B)  .44**  .50**  .29*  Compulsive (7)  .08  .23  Passive-Aggressive (8A)  .48**  .43**  .42**  Self-Defeating (8B)  .28*  .13  .37*  Schizotypal (S)  .50**  .32*  .55*  Borderline (C)  .45**  .34*  .46**  Paranoid (P)  .40*  .39*  .33*  -.09  Note. N = 40. MCMI-II = Millon Clinical Multiaxial Inventory-II (Millon, 1987). * Test-wise p < .05; ** column-wise p < .05 (both one tailed). -  69  may reflect problems with the content-related validity of the MCMI-II, or sophisticated dissimulation or image-management on the part of respondents (Hare, Forth, & Hart, 1989; Hart, Forth, & Hare, 1991). Part 1 scores were correlated more highly than Part 2 scores with the scales measuring sadistic, narcissistic, and paranoid personality disorder; Part 2 scores were correlated more highly than Part 1 scores with scales related to antisocial, schizotypal, borderline, self-defeating, and avoidant personality disorder. None of the PCL:SV scales correlated significantly with the MCMI-II histrionic personality disorder scale. To summarize the results of this section, across two samples and two methods of assessment, PCL:SV Total scores were positively correlated with the Cluster B (DramaticErratic-Emotional) antisocial, narcissistic, and borderline personality disorders; they were also positively correlated with passive-aggressive and sadistic personality disorder. Correlations with other personality disorders were either small, negative, or inconsistent across methods. Convergent validity: Normal personality. There have been several studies looking at the association between psychopathy and various models of normal personality, most notably the five factor (or Big 5) and interpersonal circumplex models (for a review, see Harpur, Hart, & Hare, in press). Most research has used the PCL-R to assess psychopathy and self-reports such as the NEO Personality Inventory (NEO-PI; Costa & McCrae, 1985) or the revised Interpersonal Adjective Scales (IAS-R; Wiggins, Trapnell, & Phillips, 1988) to measure normal personality; however, some research has also used observer ratings of normal personality (e.g., Foreman, 1988). In studies of the five factor model, psychopathy was generally negatively correlated with the dimensions of Agreeableness and Conscientiousness; correlations with Extraversion, Openness, and Neuroticism were smaller in magnitude and inconsistent. In studies of the interpersonal circumplex model, psychopathy was generally positively correlated with Dominance and negatively correlated with Love (or Nurturance). I looked at the association between the PCL:SV and the five-factor model of personality in Sample 6. Patients completed a self-report version of the IAS-R that was expanded to include domain markers for the Big 5 dimensions of Neuroticism, Conscientiousness, and Openness (IASR B5; Trapnell & Wiggins, 1991). The IAS-R B5 is a 124-item adjective rating scale. In the 70  present study, subjects indicated the degree to which each adjective was characteristic of them using an 8-point scale (1 = very uncharacteristic, 8 = very characteristic). Sixty-four items were summed to yield scores for eight octants of the interpersonal circumplex (8 items per scale), and the octant scores were then standardized using norms derived from research with college students. Scores for the interpersonal circumplex dimensions of Dominance and Love were calculated from standardized octant scores using the geometric algorithm recommended by Wiggins et al. (1988). Research suggests that Dominance and Love are equivalent to the Big 5 dimensions of Extraversion and Agreeableness, merely rotated clockwise by about 30° (McCrae & Costa, 1989). Scores for the other three Big 5 dimensions were obtained by summing the remaining items (20 items per scale); these scores were also standardized using college student norms. Based on past research and clinical theory, I predicted that PCL:SV Total scores would be positively correlated with Dominance and negatively correlated with Love, Neuroticism, Conscientiousness, and Openness. The observed correlations are presented in Tables 18 and 19. As predicted, Total scores were negatively correlated with Conscientiousness and Openness. Also, Part 2 scores were more strongly correlated than Part 1 scores with these dimensions. However, Total scores were uncorrelated with the domains of Dominance and Love, and had only a small and marginally significant negative correlation with Unassured-Submissive octant scores. The reasons for the lack of strong results are unclear, but may include the following. First, there was a little variability in the range of self-reported scores on the latter scales, with most subjects rating themselves as very low on Dominance, very high on Neuroticism, and somewhat above average on Love. (This description is not surprising given the composition of the sample, which consisted entirely of civil psychiatric patients.) Second, it appeared that several subjects rated themselves inaccurately on these dimensions. For example, one subject with a high score on the PCL:SV--in fact, the only one in this sample to exceed the cutoff for  psychopathy--rated himself as very low on Dominance and very high on Love. These self-ratings appeared to contradict his behavior prior to hospitalization, as revealed in his PCL:SV and PDE interviews and hospital records. Such (apparently) inaccurate ratings may be the result of conscious attempts to manipulate assessors or of deficits in self-awareness (Hare, Forth, & Hart, 71  Table 18 Convergent Validity: Correlations With IAS-R B5 Domain Self-Ratings in Sample 6 PCL-SV Score IASR-B5 Domain^Total^Part 1^Part 2 Dominance^.12^.07^.13 Nurturance^.04^-.04^.11 Neuroticism^-.16^-.23^-.06 Conscientiousness^-.74**^-.46**^-.83** Openness^-.48**^-.24^-.59**  Note. N = 42. IASR-B5 = Revised Interpersonal Adjective Scales, Big 5 Version (Trapnell & Wiggins, 1991). * Test-wise p < .05; ** column-wise p < .05 (both one-tailed).  72  Table 19 Convergent Validity: Correlations With IAS-R B5 Octant Self-Ratings in Sample 6 PCL:SV Score Total  Part 1  Part 2  PA: Assured-Dominant  .07  .16  -.04  BC: Arrogant-Calculating  .22  .15  .24  DE: Cold-Hearted  .10  -.03  .20  FG: Aloof-Introverted  .09  -.00  .16  HI: Unassured-Submissive  -.30*  -.24  -.30*  JK: Unassuming-Ingenuous  -.15  -.10  -.16  LM: Warm-Agreeable  .01  .06  -.03  NO: Gregarious-Extraverted  .00  .03  -.03  IAS-R Octant Rating  Note. N = 42. IAS-R B5 = Revised Interpersonal Adjective Scales, Big 5 Version (Trapnell & Wiggins, 1991). * Test-wise p < .05, ** column-wise p < .05 (both one-tailed).  73  1989; Hart, Forth, & Hare, 1991). Finally, it may be that the PCL:SV was truly unassociated with Dominance, Love, and Neuroticism. It was my intention to have naive observers make IAS-R B5 ratings of all the subjects in Sample 6. This would have allowed us to determine whether observer ratings of normal personality yielded stronger correlations with the PCL:SV than did self-ratings. Unfortunately, the damage to the videotaped interviews made this impossible. We were able to test this observer rating procedure using 24 videotaped interviews, 12 with male students from Sample 10 and 12 with male inmates from Sample 1 (Hart & Hare, in press). We used a mixed sample to insure that the sample was heterogeneous with respect to personality pathology; however, this also had the undesired effect of making Part 1 and 2 scores highly correlated in the sample, so we analyzed only Total scores. Observers were two female undergraduate students who rated subjects on the IAS-R B5 with only a minor variation in the standard instructions (i.e., "Rate the degree to which the following adjectives are characteristic of the individual..."). The observers were blind to PCL:SV scores, IAS-R B5 scoring, and the experimental hypotheses. Interrater reliability of the IAS-R B5 observer ratings was moderate, but adequate for research purposes. PCL:SV Total scores were then correlated with (unstandardized) scores on the Big 5 domain markers and the interpersonal circumplex octants; these correlations are presented in Tables 20 and 21. As the Table indicates, the results were quite strong and statistically significant, despite the small sample size: Total scores were positively with Dominance and negatively with the remaining domains. The smallest correlation was with Neuroticism (-.44), which may be due to the fact that this is the least publicly observable dimension (Paunonen, 1989). (Alternatively, it may be that psychopathy is negatively associated with most facets of Neuroticism, such as depression, anxiety, and a sense of vulnerability, but positively associated with other facets, such as irritability.) Total scores also correlated most positively with the Arrogant-Calculating octant and most negatively with the Unassuming-Ingenuous octant (.90 and -.92, respectively). Interestingly, much the same pattern of correlations was found within the student and inmate subsamples, suggesting that the findings were not due to stereotyped perceptions of inmates or students. The results of this pilot research lead me to conclude that there may indeed be substantial correspondence between 74  Table 20 Convergent Validity: Correlations With IASR-B5 Domain Observer Ratings in Students and Inmates Sample Students  Inmates  Combined  Dominance  .60*  .52*  .64**  Love  -.07  -.64*  -.83**  Neuroticism  -.57*  -.41  -.44*  Conscientiousness  -.36  -.79**  -.82**  Openness  .14  -.54*  -.77**  N  12  12  24  Domain  Note. IAS-R B5 = Revised Interpersonal Adjective Scales, Big 5 Version (Trapnell & Wiggins, 1991). * Test-wise p < .05, ** column-wise p < .05 (both one-tailed).  75  Table 21 Convergent Validity: Correlations With IAS-R B5 Octant Observer Ratings in Students and Inmates Sample Combined  Students  Inmates  PA: Assured-Dominant  .19  .39  .53**  BC: Arrogant-Calculating  .81**  .76**  .90**  Domain  DE: Cold-hearted  -.04  .73**  .86**  FG: Aloof-Introverted  -.32  .08  .41*  HI: Unassured-Submissive  -.38  -.47  -.66**  JK: Unassuming-Ingenuous  -.66*  -.84**  -.92**  LM: Warm-Agreeable  -.11  -.69*  -.83**  .39  .07  -.18  12  12  24  NO: Gregarious-Extraverted  N  Note. IASR-B5 = Revised Interpersonal Adjective Scales, Big 5 Version (Trapnell & Wiggins, 1991). * Test-wise p < .05, ** column-wise p < .05 (both one-tailed).  76  psychopathy and the Big 5, and that the failure to find such an association in Sample 6 was the result of problems assessing personality pathology via self-report. Convergent validity: Substance use. The PCL-R has been associated with substance use-more specifically, measures of alcohol and drug abuse and dependence--in four different studies (Hart, Forth, & Hare, 1991; Hart & Hare, 1989; Hemphill, Hart, & Hare, 1992; Smith & Newman, 1991). A recent meta-analysis indicated that, across all these studies, Factor 2 had significantly larger correlations with measures of substance use than did Factor 1 (Hemphill et al., 1992). Well-validated self-report measures of substance use were administered to subjects in Samples 3, 10, and 11. Alcohol use was assessed using the Michigan Alcoholism Screening Test (MAST; Selzer, 1971) in all three samples and the Alcohol Dependence (B) scale of the MCMI-II in Sample 3 only. Drug use was assessed using the Drug Abuse Screening Test (DAST; Skinner, 1982) in all three samples and the Drug Dependence (/) scale of the MCMI-II in Sample 3 only. Correlations between these measures and the PCL:SV are presented in Table 22. As expected, correlations with Total and Part 2 scores were moderate to high in magnitude, and all were statistically significant. A few correlation with Part 1 scores were also significant; however, in each instance, the correlations with Part 2 scores were much larger. Discriminant validity: Mood state at assessment. As discussed in the Introduction, psychopathy is a chronic disorder; therefore, psychopathy measures should be stable over time and relatively immune to the effects of state variables, such as mood at the time of assessment. Research indicates that this pattern holds true for the PCL-R (e.g., Harpur et al., 1989). I have already presented some evidence suggesting that the short-term test-retest reliability of the PCL:SV is high; below, I examine the impact of mood states on assessment. The two mood states most often seen as possibly interfering with the assessment of personality disorder are anxiety and depression (e.g., Reich, 1987). Self-report measures of these mood states were administered to subjects in Samples 2, 4, 6, 10, and 11, on the day that they completed the PCL:SV interview. Anxiety was assessed using the State and Trait forms of the State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, & Lushene, 1970); depression was 77  Table 22 Convergent Validity: Correlations With Measures of Alcohol and Drug Abuse PCL-SV Score Sample  Scale  N^Total  Part 1  Part 2  Alcohol Abuse MCMI-II  40^.50**  .23  .65**  MAST  61^.35*  .06  .50**  Sample 10  MAST  48^.39**  .27*  .47**  Sample 11  MAST  50^.51**  .13  .63**  .47**  .70**  .07  .52**  Sample 3  Drug Abuse Sample 3  MCMI-II  40^.64**  DAST  .37*  Sample 10  DAST  48^.62**  .47**  .73**  Sample 11  DAST  50^.59**  .20  .68**  Note. MCMI-II = Millon Clinical Multiaxial Inventory-II (Millon, 1987); MAST = Michigan Alcoholism Screening Test (Selzer, 1971); DAST = Drug Abuse Screening Test (Skinner, 1982). * Test-wise p < .05, ** test-wise p < .001 (both one-tailed).  78  assessed using the Beck Depression Inventory (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). The correlations are presented in Table 23. As expected, PCL:SV Total scores did not correlate significantly with the BDI, STAI-State, or STAI-Trait in any of the samples; the mean weighted correlations across samples were .05, .06, and .06, respectively. Although they are not presented in the Table, correlations with Parts 1 and 2 scores were also small and nonsignificant. I conclude that PCL:SV ratings were not unduly influenced by subjects' mood states at the time of assessment.  Discriminant validity: Age, sex, and race. There are few data available concerning the association between psychopathy and demographic variables such as age, sex, and race. The PCL-R manual (Hare, 1991) summarizes some published and unpublished studies suggesting that race differences within samples appear to be small in terms of effect size, although on occasion they are statistically significant. With respect to age, cross-sectional analyses suggest that PCL-R Total scores have a small, negative correlation with age. There are no systematic data as yet concerning sex differences on the PCL-R; there is no compelling theoretical reason to expect such a difference, although epidemiological research on APD suggests that the prevalence rate in men is higher than that in women (Robins et al., 1991). I examined the association between PCL:SV Total scores and basic demographic characteristics in all 11 samples; the results are presented in Table 24. Consistent with research on the PCL-R, correlations with age were generally small (reaching statistical significance only in Sample 2) and mostly negative in direction (in 8 of 11 samples). The mean weighted correlation with age across samples was -.07. Similarly, the correlations between Total scores and race (dummy coded, 1 = White, 2 = non-White) were mostly small, negative in direction, and nonsignificant; the mean weighted correlation across 8 samples was -.08." However, correlations between Total scores and sex (dummy coded, 1 = male, 2 = female) were significant and negative  17  Subject-by-subject race data were not available for Samples 7, 8, and 9. 79  Table 23 Discriminant Validity: Correlations with Self-Reported Depression (Beck Depression Inventory) and Anxiety (State-Trait Anxiety Inventory) at the Time of Assessment Anxiety Sample  Depression  State  Trait  2. Provincial inmates  -.22  -.12  4. Outpatients  .06  .06  .20  6. Vancouver, BC  .12  .21  .21  10. UBC undergrads  .14  -.02  -.04  11. Carleton undergrads  -.16  Median:  .09  .02  .08  Weighted Mean:  .05  .06  .06  Note. Ns are as follows: Sample 2, 29; Sample 4, 62 for Depression and 64 for Anxiety; Sample 6, 44 for Depression, 41 for State Anxiety, and 43 for Trait Anxiety; Sample 10, 49; and Sample 11, 50. Anxiety scores are percentiles corrected for setting, age, and sex. None of the test-wise correlations are significant at p < .05 (two-tailed).  80  Table 24 Discriminant Validity: Correlations with Age, Sex, and Race Setting and Sample  Age  Sex  Race  Forensic/Nonpsychiatric 1. Federal inmates  -.22  -.26  2. Provincial inmates  -.37*  .00  3. Federal inmates  -.12  -.07  -.10  .05  .17  .01  Forensic/Psychiatric 4. Outpatients 5. Inpatients Civil/Psychiatric 6. Vancouver, BC  .15  -.43**  7. Pittsburgh, PA  -.07  -.35*  8. Kansas City, MO  -.05  -.27**  9. Worcester, MA  -.26  -.23  .17  -.39**  -.14  11. Carleton undergrads  -.09  -.33**  -.06  Median:  -.09  -.34  -.07  Weighted Mean:  -.07  -.33  -.08  -.21  Noncrhninal/Nonpsychiatric 10. UBC undergrads  Note. Coding is as follows: Age, age in years; Sex, 1 = female, 2 = male; and Race, 1 = White, 2 = non-White. * Test-wise p < .05, ** column-wise p < .05 (two-tailed).  81  in 5 of 6 samples's, with a mean weighted correlation of -.33. This suggests that there may indeed be a small but significant sex difference in the prevalence of psychopathic symptoms, with men having more symptoms than women, although possible methodological artifacts such as a selection bias, a sex bias in the content of the PCL:SV items, or a Sex of rater by Sex of subject interaction will need to be ruled out in future research. To summarize the results of this section, PCL:SV Total scores did not appear to be unduly influenced by subjects' age or race; sex appeared to have a modest but robust impact on ratings, with women receiving lower scores than men. Although these findings may have important clinical implications for the use of the PCL:SV, any general conclusions about the association between psychopathy and demographics must await the results of large-scale epidemiological studies. C. Summary of Chapter 2 At the beginning of this Chapter, I set out three requirements for the development and validation of a new scale for the assessment of psychopathy, which in turn were based on the literature review in Chapter 1. If we use these requirements as criteria for judging the results described in the remainder of this Chapter, then it appears that the PCL:SV must be considered a success. First, the PCL:SV has good psychometric properties, and it is clearly reliable enough for use in research (although for clinical purposes, which require a higher standard of reliability, it may be necessary to use PCL:SV scores averaged across two or more independent raters). Second, the scale is conceptually and empirically related to the PCL-R. Like the PCL-R, the PCL:SV appears to have a two-facet structure, and PCL:SV scores correlate highly with their PCL-R counterparts. Indeed, the degree of convergence between the scales is so high that, in many respect, the PCL:SV can be considered a parallel form of the PCL-R. And third, despite its similarity to the PCL-R, the PCL:SV requires less time, effort, and training to administer, and it is suitable for use outside of forensic settings.  No correlation was computed for Sample 4 because so few subjects were female; the remaining samples were either all male or all female. 18  82  CHAPTER 3: DISCUSSION In the final Chapter, I will discuss the implications of this study for research and clinical practice. As I have already summarized and discussed many aspects of the results, I will focus on what is not known about or what remains to be done with the PCL:SV. A. Implications for Clinical Practice Is the PCL:SV ready for clinical use? There is no simple answer to this question. On the one hand, there is a need for more basic information on the PCL:SV's psychometric properties, including its standard errors of measurement, its factor structure, and interrater agreement for psychopathy diagnoses (these will be discussed in detail below). On the other hand, there is considerable evidence that the scale is reliable (internally consistent, stable across time and raters) and valid (appropriate content, associated with other measures of the same construct, theoretically consistent pattern of associations with other constructs). In fact, there is now probably as much or even more evidence supporting the reliability and validity of the PCL:SV than there is supporting that of other psychopathy-related scales, such as 6A of the MCMI-II (which is currently one of the most popular and widely-used psychopathology inventories). Perhaps it would be most appropriate to make a few specific recommendations concerning the clinical use of the PCL: S V. 1. The PCL:SV as a Measure of Psychopathic Traits The available evidence supports the use of PCL:SV Total scores to make statements concerning the relative strength of psychopathic traits in correctional, forensic psychiatric, and civil psychiatric settings. However, there is not enough evidence at present to support the use of PCL:SV diagnoses or subscale scores as the sole criterion by which to make important clinical decisions. This begs the question, what is the utility of PCL:SV scores? High scores should alert clinicians to the possible presence of comorbid disorders, such as substance use or Axis II Cluster  B personality disorders. Clinicians could also use high PCL:SV scores to alert them to further evaluate issues such as potential for violence towards others (i.e., dangerousness), potential for disruptive behavior on hospital wards, poor suitability for interpersonal psychotherapy, potential for abusing prescription medications, and possible malingering of psychiatric or physical 83  symptoms, all of which appear to be associated with psychopathy in forensic settings (see Chapter 1). 2. The PCL:SV as a Screening Test There is evidence that the PCL:SV could be used as a screening test for psychopathy in correctional or forensic psychiatric populations (as its name suggests). That is, the standard PCL:SV cutoff (> 18) can be used to make decisions about further evaluation, rather than actual diagnoses. This conclusion is based on the diagnostic agreement between the PCL:SV and PCLR, and on the pattern of diagnostic errors made by the PCL:SV (designating the PCL-R as a criterion measure). Recall that agreement between the scales was fair, but that the PCL:SV overpredicted psychopathy relative to the PCL-R while making virtually no false negative errors: about half of individuals with high scores on the PCL:SV will also get a high score on the PCL-R, but almost no-one with a low PCL:SV score will get a high score on the PCL-R. Therefore, individuals with a high PCL:SV score could be re-evaluated using the PCL-R, which is a more reliable and better-validated measures of psychopathy, whereas those with low PCL:SV scores could be safely diagnosed as nonpsychopaths. This procedure can results in a significant savings of resources in settings that routinely screen for psychopathy. Let us take the hypothetical example of a forensic psychiatric hospital that receives referrals from local prisons and wants to identify psychopaths among those referrals for the purpose of institutional classification and management decision-making. Let us further assume that 100 prison transfers are evaluated each year, and that the base rate of PCL-R psychopathy in these transfers is 10%. If every transfer was evaluated using the PCL-R, which typically requires about 2'/2 to 3 hours to complete, then the psychopathy assessments would require a total of 275 hours. On the other hand, if the same transfers were evaluated using the PCL: S V, the routine screening would require only 1 to 11/2 hours per transfer, or a total of about 125 hours. About 20% of the transfers would then be given follow-up PCL-R assessments, requiring an additional 11/2 to 2 hours per person, for a total of 35 hours. Thus, using the PCL-R alone would require 275 clinician-hours, whereas using the PCL:SV + PCL-R procedure would require 160 clinician-hours--an annual savings of 115 clinician-hours (about 15 clinicians days) or over 40% in clinical labour costs. 84  3. Psychopathy in Future DSMs The present results have some implications for current and future revisions to the DSM. Perhaps most important implication is that, contrary to previous assumptions (e.g., Robins, 1978), it appears that psychopathic traits can be reliably and validly inferred in both forensic and nonforensic settings. As a result, there is no need to continue using the DSM-III-R's "behavioral checklist" format for the diagnosis of APD; the current criteria can be readily revised to more closely resemble the other Axis II criteria sets without a substantial reduction in reliability. Such a revision would probably result in something that closely resembles Part 2 of the PCL:SV. Another implication concerns the content of the APD criteria, which, as noted in Chapter 1, currently focus on the social deviance facet of psychopathy. The present results suggest that symptoms such as superficiality and grandiosity could easily be included in diagnostic criteria for APD. This would increase the content-related validity of the criteria without significantly reducing their internal consistency. Unfortunately, such expansion also may lead to increased overlap between APD and near-neighbour disorders, such as narcissistic and histrionic personality disorder, unless the criteria for these latter disorders are also revised. B. Implications for Research In validating the PCL:SV, the current study clearly sacrificed depth for breadth. That is, it examined a several facets of validity in a variety of small samples, rather than looking in detail at one or two facets of validity in a large sample. This strategy is a logical one in the early stages of a test's development. In future research, however, greater depth of information will be needed. 1. Norms A priority in future research should be the collection of systematic normative data. This is essential if the PCL:SV is to be maximally useful in research and clinical settings. A major decision that must be made concerns the sampling techniques to be used. For example, one possibility is to collect random-sample norms within specific settings (adult male prisoners, female young offenders, forensic psychiatric outpatients, and so forth). This procedure is relatively simple and inexpensive for any given setting; however, the number of possible settings in which to collect norms is almost limitless. Given that the PCL:SV is intended for use in the general 85  population, not just forensic settings, a stratified random sample of community residents may be more appropriate. Although more expensive and time-consuming, this procedure would yield norms for a reference group that is of general interest. It would also allow the calculation of standard errors of measurement, a determination of the PCL:SV's factor structure (including the use of confirmatory factor analysis), and a determination of the "true" association between psychopathy and demographic characteristics such as age, sex, and race. 2. Large Samples Even if systematic, random-sample norms are not collected, the use of the PCL:SV in a large sample (say, > 100 or 200 subjects) within a given setting would still yield useful data for conducting factor analyses. If two independent raters were used, it would also allow a more robust estimate of interrater agreement for PCL:SV diagnoses. Finally, if other psychopathyrelated measures were administered to the same subjects, diagnostic agreement between the PCL:SV and alternative measures could be determined. 3. Test-retest Reliability The temporal stability of PCL:SV scores was not directly assessed in this study. It will be important to examine this issue further, in both forensic and nonforensic settings, over time periods ranging from short (i.e., one week, one month) to long (one year or longer). An efficient way to examine both interrater and test-retest reliability of PCL:SV dimensional scores and diagnoses would be to conduct a generalizability study based on G- and D-theory (e.g., Schroeder et al., 1983). Unreliability due test items, raters, time, and settings--as well as the interactions of these factors--could be quantified in a single study. 4. Predictive Validity The PCL:SV's predictive validity was not examined in the present study. Of course, development of the PCL:SV was motivated in part by the MacArthur Risk Study, which will examine the scale's ability to predict violence among civil and forensic psychiatric patients released into the community (Hart, Forth, & Hare, in press). It might also be useful to examine the PCL:SV's ability to predict institutional violence in civil or forensic psychiatric hospitals, or in  86  correctional facilities. Of course, predictive validity with respect to variables other than violence, such as general recidivism or property offending, could also be examined. Another type of study that might be of more general interest would look at the predictive validity of PCL:SV scores in job applicants--for example, police, military, or corrections officer recruits. The PCL:SV would be easily administered in such settings, where all applicants undergo personal interviews, criminal record checks, interviews with collateral informants, and even drug screenings. If psychopathy is to be a useful construct outside of forensic settings, then it should predict poor occupational adjustment in employees (negative evaluations by co-workers, low productivity, disciplinary infractions, and so forth). The predictive validity of psychopathic traits should even be observed in settings with a zero or near-zero base rate of psychopathy; that is, employees with significant psychopathic traits should do more poorly than people with few or no psychopathic traits. A study of this sort might receive considerable financial and/or logistical support from large employers. 5. Laboratory Studies As noted in Chapter 1, there is a large and growing literature on the cognitive and psychophysiological correlates of psychopathy. To date, the vast majority of these studies have been conducted with inmates, and the generalizability of the results outside of criminal populations has been questioned. The construction of the PCL:SV may facilitate the replication of key studies (e.g., Williamson et al., 1991) in community residents. If large numbers of "truly psychopathic" community residents cannot be identified (a mixed blessing), it may still be that people with significant psychopathic traits show a pattern of responses similar to those of psychopaths, and that they can therefore serve as analog subjects in laboratory research. Regardless of whether the community subjects are true or analog psychopaths, their use could greatly reduce the costs of future research, as well as increasing confidence in the generalizability of research findings. 6. Psychopathy and the Big 5 The results of the pre-test looking at observer ratings of normal personality were intriguing and deserve further study. Initially, of course, simple replication in a larger sample is 87  required; if the primary findings replicate, then subsequent research should examine the nature of the association between psychopathy and the Big 5. I should emphasize here that it would not be surprising to find a significant association between psychopathy and the Big 5; indeed, several recent studies suggest that omnibus measures of personality pathology have a factor structure highly similar to the Big 5 in both clinical and nonclinical samples (e.g., McCrae & Costa, 1989; Trull, 1992; Wiggins & Pincus, 1989). The real surprise in the present study was magnitude of the association. To be sure, shared method variance may be partly responsible for the magnitude, but the possibility still remains that psychopathy, as measured by the PCL:SV, is simply an additive combination of Big 5 personality dimensions (or, put another way, simply a vector in the Big 5's five-dimensional space). If true, this hypothesis has important implications. For example, it would no longer be valid to treat psychopathy as a categorical construct, to make "diagnoses" of psychopathy or posit etiologic mechanisms underlying the "disorder" (e.g., Livesley et al., 1992). Instead, psychopathy could be decomposed into its constituent Big 5 elements, each of which is dimensional in nature and each of which has its own psychobiological underpinnings. On the other hand, it may be that psychopathy, although composed of Big 5 building blocks, is an emergent construct (one whose whole is greater than the sum of the parts). If this hypothesis is true, then psychopathy can indeed be considered a disorder or taxon. Note that both the summative and multiplicative models are consistent with the correlational results observed here. A comparison of the models could involve multiple regression, using Big 5 ratings to predict PCL:SV scores using either main effects or main effects plus interactions; or taxonometric methods, to investigate whether there is a discontinuity in Big 5 ratings that more-or-less corresponds to PCL:SV psychopathy diagnoses. 19  There is some evidence of a psychopathy taxon underlying the PCL-R (Rice, Harris, & Quinsey, 1992). 19  88  C: Conclusion Although a great deal of basic research remains to be done, the PCL:SV is a promising new measure of psychopathy. There is reason to hope that the scale will help to significantly reduce the costs of psychopathy-related research in forensic settings and facilitate the extension of this research into nonforensic settings. In addition to contributing to the growth of theoretical psychopathology, such research ultimately may help to reduce the tremendous personal, social, and economic hardships that result from psychopathy.  89  REFERENCES Albert, R.S., Brigante, T.R., & Chase, M. (1959). The psychopathic personality: A content analysis of the concept. Journal of General Psychology, 60, 17-28. Alterman, A.I., Cacciola, J.S., & Rutherford, M.J. (1992). The reliability of the Psychopathy Checklist. Manuscript under review. American Psychiatric Association (1980). Diagnostic and Statistical Manual of Mental Disorders, 3rd edition. Washington, DC: Author. American Psychiatric Association (1987). Diagnostic and Statistical Manual of Mental Disorders, 3rd edition, revised. Washington, DC: Author. American Psychological Association (1985). Standards for educational and psychological testing. Washington, DC: Author. Barbour-McMullen, J., & Coid, J. (1988). The psychometric identification of psychopathy in mentally abnormal offenders. Personality and Individual Differences, 9, 817-823. Bartko, J.J. (1976). On various intraclass correlation reliability coefficients. Psychological Bulletin, 83, 762-765. Bayer, B.M., & Bonta, J.L. (1985). The Pd subscales: An empirical evaluation. Journal of Clinical Psychology, 41, 780-788. Beck, A.T., Ward, C.H., Mendelson, M., Mock, J., & Erbaugh, J.K. (1961). An inventory for measuring depression. Archives of General Psychiatry, 4, 561-571. Blashfield, R., & Livesley, W.J. (1991). Metaphorical analysis of psychiatric classification as a psychological test. Journal of Abnormal Psychology, 100, 262-270. Boyd, J.H., Burke, J.D., Gruenberg, E.M., Holzer, C.E. III, Rae, D.S., George, L.K., Karno, M., Stoltzman, R., McEvoy, L., & Nestadt, G. (1984). Exclusion criteria of DSM-III: A study of co-occurrence of hierarchy-free syndromes. Archives of General Psychiatry, 41, 983989. Brown, S.L., Forth, A.E., Hart, S.D., & Hare, R.D. (1992). The assessment of psychopathy in a noncriminal population [Abstract]. Canadian Psychology, 33, 405. Brown, H.J., & Gutsch, K.U. (1985). Cognitions associated with a delay of gratification task: A study with psychopaths and normal prisoners. Criminal Justice and Behavior, 12, 453462. Buss, A.H. (1966). Psychopathology. New York: Wiley. Butcher, J.N, Dahlstrom, W.G., Graham, J.R., Tellegen, A., & Kaemmer, B. (1989). Minnesota Multiphasic Personality Inventory (MMPI-2). Manual for administration and scoring. Minneapolis: University of Minnesota Press. 90  Cleckley, H. (1976). The mask of sanity, 5th edition. St. Louis, MO: Mosby. Coolidge, F.L., Merwin, M.M., Wooley, M.J., & Hyman, J.N. (1990). Some problems with the diagnostic criteria of the antisocial personality disorder in DSM-III-R: A preliminary study. Journal of Personality Disorders, 4, 407-413. Correctional Service of Canada (1990). A mental health profile of federally sentenced offenders. Forum on Corrections Research, 2, 7-8. Costa, P.T., Jr., & McCrae, R.R. (1985). The NEO Personality Inventory manual. Odessa, FL: Psychological Assessment Resources. Cox, D.N., Hart, S.D., & Hare, R.D. (1989). A Clinical Version of the Psychopathy Checklist. Unpublished manuscript. Craft, M.J. (1965). Ten studies into psychopathic personality. Bristol, England: John Wright. Davies, W., & Feldman, P. (1981). The diagnosis of psychopathy by forensic specialists. British Journal of Psychiatry, 138, 329-331. Dutton, D.G., & Hart, S.D. (1992a). Risk markers for family violence in a federally incarcerated population. International Journal of Law and Psychiatry, 15, 101-112. Dutton, D.G., & Hart, S.D. (1992b). Evidence for long-term, specific effects of childhood abuse on criminal behavior in men. International Journal of Offender Therapy and Comparative Criminology, 36, 129-137. Eisenman, R. (1980). Effective manipulation by psychopaths. Corrective and Social Psychiatry and Journal of Behavior Technology, Methods, & Therapy, 26, 116-118. Feighner, J.P., Robins, E., Guze, S.B., Woodruff, R.A., Winokur, G., & Munoz, R. (1972). Diagnostic criteria for use in psychiatric research. Archives of General Psychiatry, 26, 5763. Foreman, M. (1988).^Psychopathy and interpersonal behavior.^Unpublished doctoral dissertation, University of British Columbia, Vancouver, Canada. Forth, A.E., Hart, S.D., & Hare, R.D. (1990). Assessment of psychopathy in male young offenders. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 2, 342-344. Fotheringham, J.B. (1957).^Psychopathic personality: A review. ^Canadian Psychiatric Association Journal, 2, 52-74. Gacono, C., Meloy, J.R., & Heaven, T.R. (1990). A Rorschach investigation of narcissism and hysteria in antisocial personality. Journal of Personality Assessment, 55, 270-279. Gerstley, L.J., Alterman, A.I., McLellan, A.T., & Woody, G.E. (1990). Antisocial personality disorder in substance abusers: A problematic diagnosis? American Journal of Psychiatry, 147, 173-178. 91  Gorenstein, E.E. (1982). Frontal lobe functions in psychopaths. Journal of Abnormal Psychology, 91, 368-379. Gorenstein, E.E. (1987). Cognitive-perceptual deficit in an alcoholism spectrum disorder. Journal of Studies on Alcohol, 48, 310-318. Gough, H. (1948). A sociological theory of psychopathy. American Journal of Sociology, 53, 359-366. Gough, H. (1957). Manual for the California Psychological Inventory. Palo Alto: Consulting Psychologists Press. Gough, H. (1987). California Psychological Inventory: Administrator's guide. Palo Alto: Consulting Psychologists Press. Gough, H., & Peterson, D.R. (1952). The identification and measurement of predispositional factors in crime and delinquency. Journal of Consulting Psychology, 16, 207-212. Graham, J.R. (1988, August). Establishing validity of the revised form of the MMPI. Paper presented at the Annual Meeting of the American Psychological Association, Atlanta, GA. Graham, J.R. (1990). MMPI-2: Assessing personality and psychopathology. New York: Oxford University Press. Gray, K.C., & Hutchinson, H.C. (1964). The psychopathic personality: A survey of Canadian psychiatrists' opinions. Canadian Psychiatric Association Journal, 9, 452-461. Hare, R. D. (1980). A research scale for the assessment of psychopathy in criminal populations. Personality and Individual Differences, 1, 111-119. Hare, R.D. (1983). Diagnosis of antisocial personality disorder in two prison populations. American Journal of Psychiatry, 140, 887-890. Hare, R.D. (1984). Performance of psychopaths on cognitive tasks related to frontal lobe function. Journal of Abnormal Psychology, 93, 133-140. Hare, R.D. (1985). Comparison of procedures for the assessment of psychopathy. Journal of Consulting and Clinical Psychology, 53, 7-16. Hare, R.D. (1991). Manual for the Hare Psychopathy Checklist-Revised. Toronto: Multi-Health Systems. Hare, R.D., Forth, A.F., & Hart, S.D. (1989). The psychopath as prototype for pathological lying and deception. In J.C. Yuille (Ed.), Credibility assessment (pp. 24-49). Dordrecht, The Netherlands: Kluwer. Hare, R.D., Forth, A.F., & Strachan, K. (1992). Psychopathy and crime across the lifespan. In R.DeV. Peters, R.J. McMahon, & V.L. Quinsey (Eds.), Aggression and violence throughout the life span (pp. 285-300). Newbury Park, CA: Sage. 92  Hare, R.D., Harpur, T.J., Hakstian, A.R., Forth, A.E., Hart, S.D., & Newman, J.P. (1990). The Revised Psychopathy Checklist: Descriptive statistics, reliability, and factor structure. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 2, 338-341. Hare, R.D., & Hart, S.D. (1993). Psychopathy, mental disorder, and crime. In S. Hodgins (Ed.), Mental disorder and crime (pp. 104-115). Newbury Park, CA: Sage. Hare, R.D., Hart, S.D., & Harpur, T.J. (1991). Psychopathy and the proposed DSM-IV criteria for antisocial personality disorder. Journal of Abnormal Psychology, 100, 391-398. Hare, R.D., & McPherson, L.M. (1984). Violent and aggressive behavior by criminal psychopaths. International Journal of Law and Psychiatry, 7, 35-50. Hare, R.D., McPherson, L.M., & Forth, A.E. (1988). Male psychopaths and their criminal careers. Journal of Consulting and Clinical Psychology, 56, 710-714. Hare, R.D., Williamson, S.E., & Harpur, T.J. (1991). Psychopathy and language. In T.E. Moffitt and S.A. Mednick (Eds.), Biological contributions to crime causation. Dordrecht, The Netherlands: Martinus Nijhoff. Harman, H.H. (1976). Modern factor analysis (3rd ed.). Chicago: University of Chicago Press. Harpur, T.J., Hakstian, A.R., & Hare, R.D. (1988). Factor structure of the Psychopathy Checklist. Journal of Consulting and Clinical Psychology, 56, 741-747. Harpur, T.J., Hare, R.D., & Hakstian, A.R. (1989). ^Two-factor conceptualization of psychopathy: Construct validity and assessment implications. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1, 6-17. Harpur, T.J., Hare, R.D., Zimmerman, M., & Coryell, W. (1990, August). Dimensions underlying DSM-III personality disorders: Cluster 2. Paper presented at the Annual Meeting of the American Psychological Association, Boston, MA. Harpur, T.J., Hart, S.D., & Hare, R.D. (in press). The personality of the psychopath. In P.T. Costa, Jr., and T.A. Widiger (Eds.), Personality disorders and the five-factor model of personality. Washington, DC: American Psychological Association. Harris, G.T., Rice, M.E., & Cormier, C.A. (1991). Psychopathy and violent recidivism. Law and Human Behavior, 15, 625-637. Harris, R., & Lingoes, J. (1955). Subscales for the Minnesota Multiphasic Personality Inventory. Mimeographed materials, The Langley Porter Clinic. Harris, R., & Lingoes, J. (1968). Subscales for the Minnesota Multiphasic Personality Inventory. Mimeographed materials, The Langley Porter Clinic. Hart, S.D., Dutton, D.G., & Newlove, T.N. (1992). The prevalence of personality disorder among wife assaulters. Manuscript under review.  93  Hart, S.D., Forth, A.E., & Hare, R.D. (1990). Performance of criminal psychopaths on selected neuropsychological tests. Journal of Abnormal Psychology, 99, 374-379. Hart, S.D., Forth, A.E, & Hare, R.D. (1991). The MCMI-II as a measure of psychopathy. Journal of Personality Disorders, 5, 318-327. Hart, S.D., Forth, A.E., & Hare, R.D. (1992, March). Psychometric analysis of the DSM-III-R criteria for antisocial personality disorder. Paper presented at the Mid-Year Meeting of the American Psychology-Law Society (APA Div. 41), San Diego, California. Hart, S.D., & Hare, R.D. (1989). The discriminant validity of the Psychopathy Checklist in a forensic psychiatric population. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1, 211-218. Hart, S.D., & Hare, R.D. (in press). Psychopathy and the Big 5: Correlations between observers' ratings of normal and pathological personality. Journal of Personality Disorders. Hart, S.D., Hare, R.D., & Forth, A.E. (in press). Psychopathy as a risk marker for violence: Development and validation of a Screening Version of the Revised Psychopathy Checklist. In J. Monahan & H.J. Steadman (Eds.), Violence and mental disorder: Advances in risk assessment. Chicago: University of Chicago Press. Hart, S.D., Hare, R.D., & Harpur, T.J. (1992). The Psychopathy Checklist: An overview for researchers and clinicians. In J. Rosen and P. McReynolds (Eds.), Advances in psychological assessment, Vol. 8 (pp. 103-129). New York: Plenum. Hart, S.D., & Hemphill, J.F. (1989). Prevalence of and service utilization by mentally disordered offenders at the Vancouver Pretrial Services Centre: A survey. Report submitted to the British Columbia Corrections Service in fulfillment of contract COR-22153. Hart, S.D., Kropp, P.R., & Hare, R.D. (1988). The performance of male psychopaths following conditional release from prison. Journal of Consulting and Clinical Psychology, 56, 227232. Hathaway, S.R., & McKinley, J.C. (1940). A multiphasic personality schedule (Minnesota): I. Construction of the schedule. Journal of Psychology, 10, 249-254. Hemphill, J.F., Hart, S.D., & Hare, R.D. (1992). Psychopathy and substance use. Manuscript under review. Howard, R.C. (1990). Psychopathy Checklist scores in mentally abnormal offenders: A reexamination. Personality and Individual Differences, 11, 1087-1091. Howard, R.C., Bailey, R., & Newman, F. (1984). A preliminary study of Hare's Research Scale for the Assessment of Psychopathy in mentally abnormal offenders. Personality and Individual Differences, 5, 389-396. Hunt, C., & Andrews, G. (1992). Measuring personality disorder: The use of self-report questionnaires. Journal of Personality Disorders, 6, 125-133. 94  Jackson, H.J., Whiteside, H.L., Bates, G.W., Bell, R., Rudd, R.P., & Edwards, J. (1991). Diagnosing personality disorders in psychiatric inpatients. Acta Psychiatrica Scandanavica, 83, 206-213. Karpman, B. (1961). The structure of neurosis: With special differentials between neurosis, psychosis, homosexuality, alcoholism, psychopathy, and criminality. Archives of Criminal Psychodynamics, 4, 599-646. 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, 250-259. Kropp, P.R. (1992). The relationship between psychopathy and malingering. Unpublished doctoral dissertation, Simon Fraser University, Burnaby, Canada. Liskow, B., & Powell, B.J. (1990). Diagnostic subgroups of alcoholics: Outcome at 1 year. Comprehensive Psychiatry, 31, 549-556. Liskow, B., & Powell, B.J. (1991). Antisocial alcoholics: Are there clinically significant diagnostic subtypes? Journal of Studies on Alcohol, 52, 62-69. Livesley, W.J. (1986). Trait and behavioral prototypes of personality disorder. American Journal of Psychiatry, 143, 728-732. Livesley, W.J., Jackson, D.N., & Schroeder, M. (1989). A study of the factorial structure of personality pathology. Journal of Personality Disorders, 3, 292-306. Livesley, W.J., Jackson, D.N., & Schroeder, M. (1992). Factorial structure of traits delineating personality disorders in clinical and general population samples. Journal of Abnormal Psychology, 101, 432-440. Livesley, W.J., & Schroeder, M. (1991). Dimensions of personality disorder: The DSM-III-R Cluster B diagnoses. Journal of Nervous and Mental Disease, 179, 320-328. Loranger, A.W. (1988). Personality Disorder Examination (PDE) manual. Yonkers, NY: DV Communications. Loranger, A.W., Hirschfeld, R.M., Sartorius, N., & Regier, D.A. (1991). The WHO/ADAMHA international pilot study on personality disorders: Background and purpose. Journal of Personality Disorders, 5, 296-306. Loranger, A.W., Susman, V.L., Oldham, J.M., & Russakoff, L.M. (1987). The Personality Disorder Examination: A preliminary report. Journal of Personality Disorders, 1, 1-13. McCord, W., & McCord, J. (1964). The psychopath: An essay on the criminal mind. Princeton, NJ: Van Nostrand. McCrae, R.R., & Costa, P.T., Jr. (1989). The structure of interpersonal traits: Wiggins's circumplex and the five-factor model. Journal of Personality and Social Psychology, 56, 586-595. 95  Millon, T. (1981). Disorders of personality: DSM-III Axis II. New York: Wiley. Millon, T. (1983). Millon Clinical Multiaxial Inventory manual (3rd ed.). Minneapolis: National Computer Systems. Millon, T. (1987). Manual for the Millon Clinical Multiaxial Inventory-II (2nd ed.). Minneapolis: National Computer Systems. Morey, L.C. (1988a). The categorical representation of personality disorder: A cluster analysis of the DSM-III-R personality features. Journal of Abnormal Psychology, 97, 314-321. Morey, L.C. (1988b). Personality disorders in DSM-III and DSM-III-R: convergence, coverage, and internal consistency. American Journal of Psychiatry, 145, 573-577. Morey, L.C. (1988c). A psychometric analysis of the DSM-III-R personality disorder criteria. Journal of Personality Disorders, 2, 109-124. Morey, L.C., & Ochoa, E.S. (1989). An investigation of adherence to diagnostic criteria: Clinical diagnosis of the DSM-III personality disorders. Journal of Personality Disorders, 3, 180192. Morey, L.C., Waugh, M.H., & Blashfield, R.K. (1985). MMPI scales for DSM-III personality disorders: Their derivation and correlates. Journal of Personality Assessment, 49, 245256. Newman, J.S., Widom, C.S., & Nathan, S. (1985). Passive avoidance in syndromes of disinhibition: Psychopathy and extraversion. Journal of Personality and Social Psychology, 48, 1316-1327. Ogloff, J.R.P., Wong, S., & Greenwood, A. (1990). Treating criminal psychopaths in a therapeutic community program. Behavioral Sciences and the Law, 8, 81-90. O'Mahony, P.D., & Murphy, P.G. (1991). Role-taking ability and Gough's theory of psychopathy. International Journal of Offender Therapy and Comparative Criminology, 35, 107-118. Overall, J.E., & Gorham, D.R. (1962). The Brief Psychiatric Rating Scale. Psychological Reports, 10, 799-812. Paunonen, S.V. (1989). Consensus in personality judgments: Moderating effects of target-rater acquaintanceship and behavior observability. Journal of Personality and Social Psychology, 56, 823-833. Pavelka, F.L. (1986). Psychosocial characteristics of parolees in forensic social work. Journal of Psychiatry and Law, 14, 217-223. Raine, A. (1986). Psychopathy, schizoid personality, and borderline/schizotypal personality disorders. Personality and Individual Differences, 7, 493-501.  96  Ray, J.J. (1983). Psychopathy, anxiety, and malingering. Personality and Individual Differences, 4, 351-353. Ray, J.J., & Ray, J.A.B. (1982). Some apparent advantages of subclinical psychopathy. Journal of Social Psychology, 117, 135-142. Reich, J.H. (1987). Instruments measuring DSM-III and DSM-III-R personality disorders. Journal of Personality Disorders, 1 220-240. Rennenberg, B., Chambless, D.L., Dowdall, D.J., Fauerbach, J.A., & Gracely, E.J. (1992). The Structured Clinical Interview for DSM-III-R, Axis II and the Millon Clinical Multiaxial Inventory: A concurrent validity study of personality disorders among anxious outpatients. Journal of Personality Disorders, 6, 117-124. Rice, M.E., Harris, G.T., & Cormier, C.A. (1992). An evaluation of a maximum security therapeutic community for psychopaths and other mentally disordered offenders. Law and Human Behavior, 16, 399-412. Rice, M.E., Harris, G.T., & Quinsey, V.L. (1990). A follow-up of rapists assessed in a maximum security psychiatric facility. Journal of Interpersonal Violence, 4, 435-448. Rice, M.E., Harris, G.T., & Quinsey, V.L. (1992). Psychopathy as a taxon. Manuscript under review. Robins, L.N. (1966). Deviant children grown up: A sociological and psychiatric study of sociopathic personality. Baltimore, MD: Williams & Wilkins. Robins, L.N. (1978). Aetiological implications in studies of childhood histories relating to antisocial personality. In R.D., Hare & D. Schalling (Eds.), Psychopathic behavior: Approaches to research (pp. 255-271). Chichester, England: Wiley. Robins, L.N., Helzer, J.E., Croughan, J., & Ratcliff, K.S. (1981). National Institute of Mental Health Diagnostic Interview Schedule: Its history, characteristics, and validity. Archives of General Psychiatry, 38, 381-389. Robins, L.N., & Regier, D. (1991). Psychiatric disorders in America: The Epidemiologic Catchment Area study. New York: Free Press. Robins, L., Tipp, J., & Przybeck, T. (1991). Antisocial personality. In L.N. Robins & D. Regier (Eds.), Psychiatric disorders in America: The Epidemiologic Catchment Area study (pp. 258-290). New York: Free Press. Roesch, R. (1992, September). Mental health interventions in jails. Keynote address at the Third European Conference of Law and Psychology, Oxford, England. Rogers, R., & Dion, K. (1991). Rethinking the DSM-III-R diagnosis of antisocial personality disorder. Bulletin of the American Academy of Psychiatry and Law, 19, 21-31. Rogers, R., & Gillis, J.R. (1989). A research note on the MMPI Pd scale and sociopathy. International Journal of Offender Therapy and Comparative Criminology, 33, 21-25. 97  Roy, R. (1988). A Criterion-based Evaluation of an Instrument for the Assessment of Psychopathy. Unpublished Master's thesis, Department of Psychology, Simon Fraser University, Burnaby, British Columbia, Canada. Selzer, M. (1971). The Michigan Alcoholism Screening Test: The quest for a new diagnostic instrument. American Journal of Psychiatry, 127, 1653-1658. Serin, R.C. (1991). Psychopathy and violence in criminals. Journal of Interpersonal Violence, 6, 423-431. Serin, R.C., Peters, R.D., & Barbaree, H.E. (1990). Predictors of psychopathy and release outcome in a criminal population. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 2, 419-422. Schroeder, M., Schroeder, K., & Hare, R.D. (1983). Generalizability of a checklist for the assessment of psychopathy. Journal of Consulting and Clinical Psychology, 51, 511-516. Simourd, D., Bonta, J., Andrews, D., & Hoge, R.D. (1990, May). Psychopathy and criminal behavior: A meta-analysis. Paper presented at the Annual Meeting of the Canadian Psychological Association. Skinner, H.A. (1982). The Drug Abuse Screening Test. Addictive Behavior, 7, 363-371. Smith, S.S., & Newman, J.P. (1990). Alcohol and drug abuse/dependence disorders in psychopathic and nonpsychopathic criminal offenders. Journal of Abnormal Psychology, 99, 430-439. Spielberger, C.D., Gorsuch, R.L., & Lushene, R.E. (1970). Manual for the State-Trait Anxiety Inventory. Palo Alto: Consulting Psychologist Press. Spitzer, R.L., Endicott, J., & Robins, E. (1978). Research Diagnostic Criteria: Rationale and reliability. archives of General Psychiatry, 35, 773-782. Stabenau, J.R. (1990). Additive independent factors that predict risk for alcoholism. Journal of Studies on Alcohol, 51, 164-174. Standage, K., & Ladha, N. (1988). An examination of the reliability of the Personality Disorder Examination and a comparison with other methods of identifying personality disorders in a clinical sample. Journal of Personality Disorders, 2, 267-271. Stang', D., Pfohl, B., Zimmerman, M., Bowers, W., & Corenthal, C. (1985). A structured interview for the DSM-III personality disorders: A preliminary report. Archives of General Psychiatry, 42, 591-596 Sutker, P.A., & Allain, A.N. (1983). Behavior and adaptive behavior in men labeled adaptive sociopaths. Journal of Behavioral Assessment, 5, 65-79. Sutker, P.A., Moan, C.E., & Allain, A.N. (1983). Assessment of cognitive control in psychopathic and normal prisoners. Journal of Behavioral Assessment, 5, 275-287. 98  Tennent, G., Tennent, D., Prins, H., & Bedford, A. (1990). Psychopathic disorder: A useful clinical concept? Medicine, Science, and the Law, 30, 38-44. Teplin, L.A., & Swartz, J. (1989). Screening for severe mental disorder in jails: The development of the Referral Decision Scale. Law and Human Behavior, 13, 1-18. Trapnell, P., & Wiggins, J.S. (1991). Extension of the Interpersonal Adjective Scales to the big five dimensions of personality. Journal of Personality and Social Psychology, 59, 781790. Trull, T.J. (1992). DSM-III-R personality disorders and the five-factor model of personality: An empirical comparison. Journal of Abnormal Psychology, 101, 553-560. Tyrer, P., & Ferguson, B. (1988). Development of the concept of abnormal personality. In P. Tyrer (Ed.), Personality disorders: Diagnosis, management, and course (pp. 1-11). London: Wright. Walters, G.D. (1985). Scale 4 (Pd) of the MMPI and the diagnosis of antisocial personality. Journal of Personality Assessment, 49 474-476. Widiger, T.A., et al. (1992). Results of the DSM-IV Antisocial Personality Disorder Field Trials. Manuscript in preparation. Widiger, T.A., & Corbett, E.M. (in press). Antisocial personality disorder in DSM-IV. In J. Livesley (Ed.), DSM-IV personality disorders. New York: Guilford. Widiger, T.A., Frances, A.J., Pincus, H.A., Davis, W.W., & First, M.B. (1991). Toward an empirical classification for DSM-IV. Journal of Abnormal Psychology, 100, 280-288. Wiggins, J.S., & Pincus, A.L. (1989). Conceptions of personality disorders and dimensions of personality. Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1, 305-316. Wiggins, J.S., Trapnell, P., & Phillips, N. (1988). Psychometric and geometric characteristics of the revised Interpersonal Adjective Scales (IAS-R). Multivariate Behavioral Research, 23, 517-530. Williamson, S.E., Harpur, T.J., & Hare, R.D. (1991). Abnormal processing of affective words by psychopaths. Psychophysiology, 28, 260-273. Williamson, S.E., Hare, R.D., & Wong, S. (1987). Violence: Criminal psychopaths and their victims. Canadian Journal of Behavioral Science, 19, 454-462. Wong, S. (1984). Criminal and institutional behaviours of psychopaths. Ottawa: Ministry of the Solicitor-General of Canada. Woody, G.E., & McLellan, A.T. (1985). Sociopathy and psychotherapy outcome. Archives of General Psychiatry, 42, 1081-1086.  99  APPENDIX A PCL:SV Administration and Scoring Details A. The PCL:SV Assessment Procedure Interviewing. An interview is one of the two key data sources on which the PCL:SV is rated (the other being charts or collateral informants). We use the interview to collect historiodemographic data and to sample the individual's interpersonal style. The former is used primarily to score items in Part 2 of the PCL:SV, whereas the latter is used primarily to score Part 1 items. We recommend the use of semi-structured, rather than a structured, interviews. Some structure aids the interviewer in collecting necessary content-related information, but too much structure can hinder rapport-building and obscure interactional style. For example, with a semistrutured interview, interviewers can use their clinical skills to elicit evidence of emotional bonds, or they can permit an individual to "ramble" and tell stories. In our own research, we use an interview that covers the following areas: presenting problem/current legal status, educational history and goals, vocational history and goals, medical and psychiatric history, family background, marital history, juvenile conduct problems, and adult antisocial behavior (including substance use). Within each area, recommended general questions and follow-up probes are listed; however, interviewers are free to rephrase or even omit questions, or to ask additional questions as they see fit. It is not necessary to complete the interview at one sitting; in fact, multiple interviews may actually be an asset, as they help to insure that the rater obtains a representative sample of the individual's interpersonal style. It should be obvious from the above description that our semi-structured interview is very similar in structure to any reasonably comprehensive clinical interview. Consequently, if the individual has already completed a clinical interview, it may be unnecessary to re-interview that person in order to make PCL:SV ratings. If the clinical interview administered was not sufficiently comprehensive, the rater may wish to ask only selected questions from the PCL:SV interview.  100  In rare instances, it may be impossible to complete an interview due to mental illness, discharge or elopement from the institution, and so forth. In such cases, so long as there is adequate detail in the institutional files, the PCL:SV can be completed on the basis of file information alone. File data can also be supplemented by any limited contacts with or observations of the individual. Charts and collateral informants. As noted above, the second major source of data for scoring the PCL:SV is charts and collateral informants. We do not accept data obtained in the interview at face value. Rather, we attempt to confirm or deny important claims made by the individual. Hospital charts, correctional files, and criminal records may all be used for this purpose, as can interviews with friends or relatives. Of course, interviewers must use their clinical judgement to evaluate the reliability of this collateral information; in general, however, conflicting reports concerning an individual's personality or behavior should alert the interviewer to the possibility that the individual is engaging in impression management. In rare cases, there may be abolsutely no file or collateral information available. The PCL:SV should not be completed on the basis of interview information alone; every attempt should be made to collect at least some file information (e.g., requesting a criminal record for the individual, interviewing a family member, friend, or previous employer). If this is not possible, then PCL:SV ratings should be delayed until collateral information in the form of progress notes, consultant reports, and so forth become available. (In most institutions, such collateral information will be available within two or three weeks.)  B. PCL:SV Scoring Item descriptions. The item descriptions are listed on the PCL:SV scoresheet. These descriptions are brief and in point form. The individual statements under each item are roughly ordered in descending order of importance and frequency (prototypicality). However, raters should not use these item descriptions as a simple checklist. Instead, they should use the entire item description to form an impression (prototype) in their minds and then compare the individual being rated to the prototype. Once raters have assessed someone who matches an item  101  description very well--an exemplar for that item--it may also be helpful to conjure up a mental image of that person while rating the item in question.  Item scores. Each of the PCL:SV items is scored using a 3-point ordinal scale (0, 1, or 2), based on the degree to which the individual matches the item description. Clearly, the scoring is subjective; however, our research indicates that experienced raters can be highly reliable when making judgements of this kind. Scores of 2, 1, and 0 are defined as follows:  2 The item applies to the individual; a reasonably good match in most essential respects; his/her behavior is generally consistent with the flavor and intent of the item.  1 The item applies to a certain extent but not to the degree required for a score of 2; a match in some respects but with too many exceptions or doubts to warrant a score of 2; uncertain about whether or not the item applies; conflicts between interview and file information that cannot be resolved in favor of a score of 2 or 0.  0 The item does not apply to the individual; s/he does not exhibit the trait or behavior in question, or s/he exhibits characteristics that are the opposite of, or inconsistent with, the intent of the item.  In brief: Yes (2); In some respects/maybe (1); No (0). We note again that raters should not use the item description statements as a simple checklist. An individual could receive a score of 2 on an item by displaying one or two of the characteristics to a great degree, or by displaying several of the characteristics to a moderate degree. Take the intensity, frequency, and duration of the individual's symptoms into account when scoring the item. Also, keep in mind that the 102  timeframe for scoring the PCL:SV is the individual's entire life: each item is supposed to reflect a personality trait, rather than a symptom that is present only during brief or rare episodes. Conflicting information regarding the individual can be handled in different ways. If a source of information is deemed to be totally noncredible by the rater, it can be ignored. If a source is seen as less credible than others, information obtained from it can be given less weight. If the sources are equally credible, then the rater can attempt to seek out new information or consider giving the individual a score of 1 on the relevant item(s). Finally, if all information pertaining to an item comes from sources that lack credibility, the item can be omitted. An item may also be omitted if there is insufficient information with which to score it. Up to three items can be omitted for any one subject; assessments requiring more than three item omissions should be considered invalid. Once all the items have been scored, the Part 1, Part 2, Total scores should be calculated and entered on the scoresheet. Where items are omitted, the Part and Total scores should be prorated on the basis of the remaining item scores. If more than two items were omitted on either Part 1 or Part 2, the score for that subscale should be considered invalid. III. PCL:SV Training Program We have found that many raters benefit from formal training program in the use of the PCL:SV, although such training is neither necessary nor sufficient to insure reliable ratings. In our own training, we cover three main topic areas: 1. The nature and assessment of psychopathy. A review of the concept of psychopathy, problems in assessing the disorder, the development of the PCL and PCL:SV, and preliminary research regarding the psychometric properties and validity of the PCL:SV. 2. The PCL:SV assessment procedure. A discussion of interviewing techniques, chart reviews, and collateral informants. 3. PCL:SV scoring. A review of scoring procedures. We also give raters a chance to watch several videotaped interviews with real prisoners to help them establish a set of "internal norms" for scoring individual PCL:SV items.  103  IV. Supplementary Item Scoring Descriptions Below we discuss the individual PCL:SV item descriptions. This section should not be used as an alternative to the descriptions on the PCL:SV; rather, it is intended to clarify the interpretation of these descriptions. Item 1: Superficial. This item describes an individual whose interactional style appears superficial (i.e., glib) to others. Usually, the individual tries to make a favorable impression on others by "shamming" emotions, telling stories that portray him/her in a good light, and making unlikely excuses for undesirable behaviors. S/he may use unnecessary--and, frequently, inappropriate--jargon. Despite its superficiality, the individual's style may be considered engaging. Alternatively, the individual may try to impress others by appearing sullen, hostile, or "macho." Still, the key aspect is that this presentation appears affected and superficial. Both types of individuals are "slippery" in conversation: when challenged with facts that contradict their statements or with inconsistencies in their statements, they simply change their story. Item 2: Grandiose. Individuals who score high on this item are often described as grandiose or as braggarts. They have an inflated view of themselves and their abilities. They appear self-assured and opinionated in the interview situation (a situation where most people are somewhat reticient or deferential). If they are in hospital or prison, they attribute their unfortunate circumstances to external forces (bad luck, the "system") rather than to themselves. Consequently, they are relatively concerned about their present circumstances and worry little about the future. (Note that psychotic delusions are irrelevant to the scoring of this item, unless they are accompanied by the other characteristics listed.) Item 3: Manipulative. People with this characteristic commonly engage in lying, deception, and other manipulations in order to achieve their own personal goals (money, sex, power, etc.). They lie and deceive with self-assurance and no apparent anxiety. They may admit that they enjoy conning and deceiving others; they may even label themselves "fraud artists." Item 4: Lacks Remorse. High scores on this item are given to individuals who appear to lack the capacity for guilt. It is normal to feel justified in having hurt someone on at least a few occasions; however, high scorers on this item appear to have no conscience whatsoever. Some of 104  these latter individuals will verbalize remorse, but in an insincere manner; others will display little emotion about their own actions or the impact they had on others, and will focus instead on their own suffering. (In scoring this item, it is necessary to take the nature of the individual's harmful behaviors into account. Clearly, a lack of remorse concerning relatively trivial acts may not be pathological.) Item 5: Lacks Empathy. This item describes individuals who have little affective bonding with others and are unable to appreciate the emotional consequences (positive or negative) of their actions. As a result, they may appear cold and callous, unable to experience strong emotions and indifferent to the feelings of others. Alternatively, they may express their emotions, but these emotional expressions are shallow and labile. The verbal and nonverbal aspects of their emotion may appear inconsistent. Item 6: Doesn't Accept Responsibility. People who score high on this item avoid taking personal responsibility for their harmful actions by rationalizing their behavior, greatly minimizing the consequences for others, or even denying the actions altogether. Most of their rationalizations involve the projection of blame (or at least partial blame) onto the victim or onto circumstances. Minimizations usually involve denying that the victim suffered any serious or direct physical, emotional, or financial consequences. Denial usually involves claiming innocence, that is, that the victim lied or the individual was framed; alternatively, s/he may claim amnesia due to substance use or to physical or mental illness. Item 7: Impulsive. This item describes people who act without considering the consequences of their ations. They act on the spur of the moment, often as the result of a desire for risk and excitement. They may be easily bored and have a short attention span. Consquently, they lead a lifestyle characterized by instability in school, relationships, employment, and place of residence. Item 8: Poor Behavioral Controls. This item describes people who are easily angered or frustrated; this may be exacerbated by the use of alcohol or drugs. They are frequently verbally abusive (i.e., they swear, insult, or make threats) and physically abusive (i.e., they break or throw  105  things; push, slap, or punch others). The abuse may appear to be sudden and unprovoked. These angry outbursts are often short-lived. Item 9: Lacks Goals. High scores on this item are given to those who do not have realistic long-term plans and commitments. Such people tend to live their lives "day-to-day," not thinking of the future. They may have relied excessively on family, friends, and social assistance for financial support. They often have poor academic and employment records. When asked about their goals for the future, they may describe far-fetched plans or schemes. Item 10: Irresponsible. This item describes people who exhibit behavior that frequently causes hardship to others or puts others risk. They tend to be unreliable as a spouse or parent: they lack commitment to relationships, fail to care adequately for their children, and so forth. Also, their job performance is inadequate: they are frequently late or absent without good reason, etc. Finally, they are untrustworthy with money: they have been in trouble for such things as defaulting on loans, not paying bills, or not paying child support. Item 11: Adolescent Antisocial Behavior. People who score high on this item had serious conduct problems as an adolescent. These problems were not limited to only one setting (i.e., occurred at home, at school, and in the community) and were not simply the result of childhood abuse or neglect (e.g., running away to avoid beatings; stealing food when it wasn't available at home). Such people frequently were in trouble with the law as a youth or minor, and their antisocial activities were varied, frequent, and persistent. Item 12: Adult Antisocial Behavior. This item describes people who frequently violate formal, explicit rules and regulations. They have had legal problems as an adult, including charges with or convictions for criminal offenses. Their antisocial activities are varied, frequent, and persistent.  106  APPENDIX B Supplementary Tables Table B-1 Interrater Reliability (r) of PCL:SV Items in 7 Samples Sample Item  1  2  4  5  6  10  11  1.  .52  .33  .48  .41  .72  .17*  .94  2.  .49  .56  .63  .64  .70  .33  .89  3.  .55  .53  .61  .36  .40  .70  .78  4.  .68  .53  .66  .08*  .56  .55  .67  5.  .43  .56  .66  .51  .40  .41  .77  6.  .67  .57  .21*  .53  .39  .78  .63  7.  .38  .72  .42  .55  .69  .60  .76  8.  .71  .68  .61  .57  .65  .37  .89  9.  .57  -.01*  .72  .19*  .44  .67  .80  10.  .24  .20*  .52  .41  .21*  .74  .80  11.  .83  .76  .73  .87  .60  .62  .83  12.  .23  .46  .27  .59  .65  .42  .80  Note. Ns are as follows: Sample 1, 50; Sample 2, 32; Sample 4, 59; Sample 5, 26; Sample 6, 28; Sample 10, 50; and Sample 11, 50. * p > .05 (one-tailed).  107  Table B-2  Corrected Item-Total Correlations for PCL:SV Total Scores in the 11 Samples Sample Item  1  2  3  4  5  6  7  8  9  10  11  1.  .58  .18*  .59  .51  .58  .52  .55  .53  .51  .55  .10*  2.  .66  .29  .40  .57  .56  .40  .39  .65  .55  .56  .14*  3.  .45  .23*  .35  .46  .50  .55  .53  .59  .50  .58  .29  4.  .79  .73  .51  .61  .38  .69  .67  .63  .70  .82  .43  5.  .77  .74  .62  .50  .13*  .72  .42  .69  .56  .78  .62  6.  .65  .52  .41  .63  .38  .74  .64  .65  .40  .56  .24  7.  .59  .48  .37  .51  .39  .70  .50  .53  .57  .76  .45  8.  .38  .29  .33  .50  .36  .71  .50  .52  .36  .33  .19*  9.  .67  .37  .58  .31  .23*  .29  .48  .53  .49  .63  .34  10.  .58  .36  .43  .53  .30  .77  .60  .65  .54  .72  .39  11.  .52  .42  .59  .35  .13*  .39  .65  .49  .35  .53  .36  12.  .55  .49  .44  .50  .29  .64  .71  .54  .41  .49  .58  Note. Ns are as follows: Sample 1, 50; Sample 2, 32; Sample 3, 67; Sample 4, 71; Sample 5, 49; Sample 6, 48; Sample 7, 40; Sample 8, 80; Sample 9, 49; Sample 10, 50; and Sample 11, 50. * p > .05 (one-tailed).  108  Table B-3  Corrected Item-Total Correlations for PCL:SV Parts 1 and 2 Scores in the 11 Samples Sample Item  1  2  3  4  5  6  7  8  9  10  11  Part 1 1.  .65  .54  .62  .56  .62  .59  .63  .61  .56  .51  .26  2.  .76  .65  .53  .65  .70  .59  .52  .64  .58  .65  .45  3.  .52  .44  .30  .37  .37  .64  .45  .59  .47  .55  .37  4.  .80  .72  .59  .64  .42  .71  .57  .68  .64  .78  .21*  5.  .81  .67  .56  .55  .24  .74  .51  .74  .45  .73  .61  6.  .67  .51  .58  .68  .42  .81  .52  .67  .41  .59  .04*  Part 2 7.  .69  .67  .44  .51  .41  .82  .52  .45  .59  .76  .49  8.  .47  .51  .38  .43  .34  .68  .35  .52  .21*  .32  .17*  9.  .62  .46  .54  .22  .06*  .23*  .57  .48  .47  .65  .39  10.  .56  .42  .57  .57  .49  .77  .67  .61  .51  .67  .30  11.  .59  .54  .64  .49  .19*  .62  .63  .65  .33  .49  .47  12.  .60  .68  .47  .48  .20*  .63  .65  .64  .33  .52  .53  Note. Ns are as follows: Sample 1, 50; Sample 2, 32; Sample 3, 67; Sample 4, 71; Sample 5, 49; Sample 6, 48; Sample 7, 40; Sample 8, 80; Sample 9, 49; Sample 10, 50; and Sample 11, 50. * p > .05 (one-tailed).  109  Table B-4  Loadings for Oblique, 2-Factor Solution in Samples 1 to 4  Sample 1  Sample 2  Sample 3  Sample 4  1^2  1^2  1^2  1^2  1.  .73  -.05  -.27  .70  .32  .42  .55  .07  2.  .90  -.12  -.17  .78  .06  .50  .75  -.03  3.  .55  -.01  -.11  .51  .38  .08  .23  .35  4.  .76  .18  .37  .73  -.01  .84  .72  .04  5.  .83  .08  .44  .71  .41  .43  .69  -.07  6.  .63  .14  .30  .52  -.17  .90  .77  .02  7.  -.03  .79  .71  -.01  .46  .01  .10  .58  8.  -.06  .56  .67  -.14  .56  -.13  .25  .40  9.  .29  .54  .56  .00  .46  .29  .23  .16  10.  .20  .51  .54  .06  .71  -.14  .04  .65  11.  .01  .68  .60  .02  .73  .08  -.21  .75  12.  -.01  .71  .71  -.04  .53  .06  .18  .46  Eigenvalue:  5.17  1.19  3.40  2.31  3.54  1.39  3.84  0.96  % Variance:  43.1  9.9  28.3  19.3  29.5  11.5  32.0  8.0  Factor r:  .54  .09  .35  110  .50  Table B-5  Loadings for Oblique, 2-Factor Solution in Samples 5 to 8 Sample 5  Sample 6  Sample 7  Sample 8  1^2  1^2  1^2  1^2  1.  .77  .18  .60  .04  .21  .73  .67  -.08  2.  .79  .10  .73  -.21  -.01  .86  .67  .08  3.  .30  .51  .73  -.04  .48  .18  .60  .08  4.  .61  -.19  .70  .18  .68  .12  .77  -.05  5.  .36  -.26  .74  .18  .36  .24  .86  -.07  6.  .38  .09  .84  .12  .60  .17  .76  -.03  7.  .14  .57  .12  .80  .57  -.05  .42  .22  8.  .16  .47  .27  .63  .43  .21  .22  .44  9.  .34  .02  .22  .14  .73  -.29  .33  .32  10.  -.02  .56  .32  .69  .74  -.15  .36  .45  11.  -.12  .40  -.30  .89  .69  .06  -.12  .88  12.  .27  .15  .24  .60  .74  .11  -.01  .79  Eigenvalue:  2.61  1.16  5.25  1.55  4.51  1.15  4.84  1.06  % Variance:  21.8  9.7  43.7  12.9  37.6  9.6  40.4  8.8  Factor r:  .16  .39  .35  111  .53  Table B-6  Loadings for Oblique, 2-Factor Solution in Samples 9 to 11 Sample 9  Sample 10  Sample 11  1^2  1^2  1^2  1.  .51  .15  -.16  .93  -.12  .53  2.  .58  .13  .25  .45  -.05  .85  3.  .61  .01  .50  .16  .16  .49  4.  .38  .52  .93  -.00  .64  .00  5.  .12  .61  .45  .48  .67  .47  6.  .65  -.14  .63  .01  .37  -.15  7.  .29  .43  .36  .56  .45  .10  8.  .52  -.07  .28  .11  .17  .06  9.  -.14  .87  .30  .48  .47  -.10  10.  .39  .28  .95  -.13  .51  .09  11.  -.03  .52  .34  .28  .51  -.20  12.  .31  .21  .59  -.05  .81  -.03  Eigenvalue:  3.76  0.87  5.35  0.78  2.73  1.51  % Variance:  31.3  7.3  44.5  6.5  22.8  12.6  Factor r:  .47  .60  .05  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0086356/manifest

Comment

Related Items