مدل پنج عاملی و ادبیات تجربی اختلال شخصیت: بررسی فراتحلیلی

Abstract The current meta-analysis reviews research examining the relationships between each of the five-factor model personality dimensions and each of the 10 Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) personality disorder diagnostic categories. Effect sizes representing the relationships between these two constructs were compiled from 15 independent samples. Results were analyzed both within each individual personality disorder category and across personality disorders, indicating how personality disorders are different and similar, respectively, with regard to underlying personality traits. In terms of how personality disorders differ, the results showed that each disorder displays a five-factor model profile that is meaningful and predictable given its unique diagnostic criteria. With regard to their similarities, the findings revealed that the most prominent and consistent personality dimensions underlying a large number of the personality disorders are positive associations with Neuroticism and negative associations with Agreeableness. Extraversion appears to be a more discriminating dimension, as indicated by prominent but directionally variable associations with the personality disorders. The implications of these meta-analytic findings for clinical application and the advancement of future research are discussed.

Introduction Recently, efforts have been made to integrate the previously independent areas of personality theory and personality disorder research Clark et al., 1994, Dyce & O'Connor, 1998, Livesley, 2001 and Schroeder et al., 1994. The bridging of normal and abnormal personality has largely come from examining how the five-factor model of personality may be used as a method for conceptualizing personality disorders. Derived from numerous factor analyses, the five-factor model arguably represents a general consensus as to the structure of normal personality Costa & McCrae, 1992a, Costa & Widiger, 1994a, Digman, 1990, Digman, 1994, Digman, 1996 and McCrae, 1991. The five-factor model provides a dimensional account of the structure of normal personality traits, dividing personality into the five broad dimensions of Neuroticism, Extraversion, Openness to Experience, Agreeableness, and Conscientiousness (Costa & McCrae, 1992b). It is a hierarchical model, where more specific personality traits or facets are classified under each broad trait dimension. For example, the dimension of Neuroticism is composed of the facets of anxiety, angry hostility, depression, self-consciousness, impulsiveness, and vulnerability (Costa & McCrae, 1992b). Studies examining the relationships between the five personality dimensions and the personality disorder diagnostic categories have mostly been correlational and have used a variety of sample types and dependent measures. The hypothesis underlying this research effort is that personality disorders can be conceptualized as extreme variants of these normal personality dimensions Costa & Widiger, 1994b, Widiger & Costa, 1994, Widiger & Trull, 1992 and Wiggins & Pincus, 1989. Thus, many studies have tried to identify the particular combination of five-factor model personality dimensions associated with each personality disorder. Specific relationships between the five trait dimensions and each personality disorder have been hypothesized based on the diagnostic criteria for each disorder (Widiger, Trull, Clarkin, Sanderson, & Costa, 1994). Furthermore, it has been empirically demonstrated that personality disorders are related to these normal personality trait dimensions in meaningful and predictable ways Ball et al., 1997, Blais, 1997, Brieger et al., 2000, Brooner et al., 1994, Cloninger & Svrakic, 1994, Coolidge et al., 1994, Costa & McCrae, 1990, Duijsens & Diekstra, 1996, Dyce & O'Connor, 1998, Hyer et al., 1994, Lehne, 1994, Morey et al., 2000, Schroeder et al., 1992, Shopshire & Craik, 1994, Soldz et al., 1993, Trull, 1992, Trull et al., 1995, Trull et al., 2001, Wiggins & Pincus, 1989 and Yeung et al., 1993. The volume of published data in this area is case enough for conducting a meta-analytic review to summarize the relationships between the five-factor model and personality disorders. However, a quantitative summary of this literature can serve an even more important function. Ultimately, this body of research brings together two classification and research traditions that exist within the personality disorder literature (i.e., the categorical and the dimensional; Widiger, 1991 and Widiger, 1992). It is the present authors' contention that the five-factor model and personality disorder literature reviewed here is essentially a synthesis of these two traditions, and as such, this meta-analysis represents a hybrid system of personality disorder classification Benjamin, 1993, Millon, 1994b and Morey et al., 2000. Both categorical and dimensional approaches are worthwhile and have their own strengths; therefore, a merging of these two approaches could potentially provide a superior conceptualization of personality disorders, which preserves the strengths of both systems. To appreciate the rivalry between categorical and dimensional approaches to personality disorder classification, one must first consider the intrinsic difficulty in classifying personality disorders. Personality disorders continue to present an organizational and conceptual dilemma for mental health professionals. This is in part due to their highly heterogeneous nature, which presents a problem for any diagnostic or classification approach. While the challenge of classifying the complex domain of personality disorders has a long history (Morey, 1997), in contemporary psychology, it has been most explicitly and extensively attempted by each successive Diagnostic and Statistical Manual of Mental Disorders (DSM) Task Force (Millon & Davis, 1995). A categorical system of classification and diagnosis was adopted for all mental disorders in the original DSM, and personality disorders were no exception. The tradition of categorically classifying personality disorders has continued over the years and is present in the most current diagnostic guidelines (DSM-IV by the American Psychiatric Association, 1994 and International Classification of Diseases (ICD-10) by the World Health Organization, 1992). Under these categorical systems, pathological symptomatology is the main focus, as personality disorders are characterised by the presence of a certain number of behavioural criteria. Furthermore, attempting to elucidate the differences among personality disorder categories is a priority. As such, depending on the content and number of criteria met, a particular class of personality disorder may be deemed present in an individual, provided distress or functional impairment accompanies these criteria. It is presumed that these behavioural criteria reflect manifestations of underlying maladaptive personality traits. In essence, if one has sufficiently maladaptive personality functioning in terms of both number and kind, then one is likely to fit into a specified class of abnormal personality. Strengths that are commonly noted for the categorical classification of personality disorders are (a) conceptual ease, (b) communicability, (c) familiarity, and (d) consistency with clinical decision-making and treatment planning Ball, 2001, Widiger, 1991, Widiger, 1993 and Widiger & Frances, 1994. Over the last decade, the classification of personality disorders has become a more controversial issue Clark & Livesley, 1994, Nelson-Gray & Farmer, 1999 and Widiger, 1992. Extensive debate has developed between supporters of a categorical system Benjamin, 1993, Benjamin, 1996, Frances, 1982, Frances, 1993, Millon, 1996 and Millon & Davis, 1993 and proponents of a dimensional system Costa & McCrae, 1992a, Costa & Widiger, 1994a, Livesley, 2001, McCrae, 1994b, Widiger, 1991, Widiger, 1992, Widiger, 1993 and Wiggins & Pincus, 1989. Those favouring a dimensional conceptualization of personality disorders argue that personality disorders do not reflect dichotomous distinctions as is inherent in a categorical approach, but reflect continua of functioning where distinct boundaries between normal and abnormal personalities are not clear. Within the dimensional approach, the chief focus is on underlying personality dimensions rather than on particular sets of symptoms. In doing so, dimensional accounts can highlight the similarities and differences between the personality disorders. Strengths commonly noted for the dimensional classification of personality disorders are (a) resolution of categorical classification problems of comorbidity and heterogeneity, (b) increased information, and (c) greater diagnostic flexibility Widiger, 1991 and Widiger & Frances, 1994. To read the majority of literature on the issue of personality disorder classification, one might think that categorical and dimensional conceptualizations are mutually exclusive Millon, 1994b, Widiger, 1991, Widiger, 1992 and Widiger & Frances, 1994. That is, it could appear as if personality disorders are either traits that are abnormal and qualitatively distinct from normal personality, which fall into natural classes based on descriptive content, or underlying traits that are both variants of and quantitatively distinct from normal personality (Wiggins & Pincus, 1994). However, it is also possible to view the research examining the relationships between the five-factor model and personality disorders as a synthesis of these categorical and dimensional approaches. The research summarized here uses dimensions to describe categories (Benjamin, 1993), maps the five factors onto each personality disorder (Ball, 2001), and in doing so presents a hybrid system for conceptualizing personality disorders. The blending of these two traditionally opposed classification approaches and the implications that arise from this amalgamation can be better understood by reflecting on the type of data obtained from the literature. Summarizing this research creates a table of effect sizes with 10 rows, each representing the different DSM-IV personality disorder diagnostic categories, and five columns, each representing the five dimensions of the five-factor model of personality. The table can be read in two ways, and the major implications of this research come from these two complementary readings. Firstly, the table can be read within disorder (i.e., from left to right). Analyzing the table horizontally, one can examine the five-factor model personality dimension profiles specific to each personality disorder category. Secondly, the table can be read across disorder (i.e., from top to bottom). Analyzing the table vertically, one can examine the five-factor model personality dimensions consistent and inconsistent across personality disorder categories. Research to date has mostly focused on examining what is different and unique to each personality disorder. Studying how personality disorders are similar has been of secondary importance in the literature. Recently, there have been indications that this trend is being reversed Lynam & Widiger, 2001 and Morey et al., 2000, and this meta-analysis seeks to extend this trend. As such, the importance and implications of both types of understanding of personality disorders will be a main focus of Section 4 of this review. This meta-analytic study has three main aims: (1) to provide a comprehensive, objective, and quantitative summary of five-factor model and personality disorder research, (2) to consider the within-disorder and across-disorder implications of the empirical literature, and (3) to provide a clear picture of the current status of research in this area and use this to identify possible future research directions and clinical implications. Essentially, this meta-analytic review presents a summary of what five-factor personality dimensions are maladaptive both within and across personality disorder categories, showing which dimensions define which disorders and how compelling these definitions are.

Results Table 4 and Table 5 display the main meta-analytic findings for the relationships between personality disorders and five-factor model personality dimensions. Table 4 shows the unweighted mean effect sizes and standard deviations and Table 5 shows the sample size weighted mean effect sizes, for all studies included in the meta-analysis. A comparison of unweighted and weighted means reveal minimal differences (Mdifference=.02, Mindifference=.00, Maxdifference=.15). Therefore, only weighted means are reported in subsequent tables. Significance levels reported in all tables reveal very small mean effect sizes as significant at very small P levels. This is expected when combining correlations over numerous studies with large sample sizes. As concentrating analysis and discussion on significance levels would be neither informative nor valuable, it was decided that effect sizes ≥ .20 would be considered meaningful (see binomial effect size display discussion for rationale); therefore, these effect sizes are highlighted in all tables and are a focus of discussion. Table 4. Unweighted mean effect size estimates and corresponding standard deviations for each DSM-IV personality disorder and five-factor model personality dimension combination DSM-IV personality disorders Five-factor model personality dimensions N E O A C Paranoid .24**** (.18) −.06**** (.18) −.01 (.08) −.31**** (.20) −.03** (.14) Schizoid .10**** (.23) −.38**** (.29) −.15**** (.13) −.17**** (.15) −.02 (.11) Schizotypal .34**** (.11) −.31**** (.19) −.05* (.12) −.20**** (.16) −.11**** (.09) Antisocial .09**** (.14) .06** (.13) .05** (.13) −.33**** (.21) −.24**** (.18) Borderline .47**** (.15) −.07**** (.14) .02 (.10) −.23**** (.15) −.19**** (.18) Histrionic .06** (.21) .41**** (.25) .17**** (.13) −.06*** (.14) −.13**** (.11) Narcissistic .03* (.25) .24**** (.23) .14**** (.14) −.24**** (.16) −.06** (.14) Avoidant .45**** (.15) −.44**** (.18) −.12**** (.12) −.13**** (.16) −.06**** (.11) Dependent .37**** (.21) −.11**** (.10) −.11**** (.14) .07** (.23) −.10**** (.13) Obsessive-Compulsive .08**** (.32) −.13**** (.16) −.06*** (.12) −.06** (.26) .18**** (.24) Mean .22 −.08 −.01 −.17 −.08 Median .17 −.09 −.03 −.18 −.08 N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Effect sizes ≥ .20 are in bold. Standard deviations are reported in parentheses. The number of independent effect sizes on which means are based is 15. **** P<.0001 (one-tailed). ** P<.01 (one-tailed). * P<.05 (one-tailed). *** P<.001 (one-tailed). Table options Table 5. Sample size weighted mean effect size estimates for each DSM-IV personality disorder and five-factor model personality dimension combination DSM-IV personality disorders Five-factor model personality dimensions N E O A C Paranoid .28**** −.12**** −.04** −.34**** −.07**** Schizoid .13**** .23**** −.12**** −.17**** −.03* Schizotypal .36**** .28**** −.01 −.21**** −.13**** Antisocial .09**** .04 .05** −.35**** −.26**** Borderline .49**** −.09**** .02 −.23**** −.23**** Histrionic .02 .42**** .15**** −.06** −.09*** Narcissistic .04 .20**** .11**** −.27**** −.05* Avoidant .48**** −.44**** −.09**** −.11**** −.10**** Dependent .41**** −.13**** −.11**** .05** −.14**** Obsessive-Compulsive .08*** −.12**** −.07**** −.04 .23**** Mean .24 −.07 −.01 −.17 −.09 Median .20 −.12 −.02 −.19 −.09 N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Effect sizes ≥.20 are in bold. The total number of participants on which sample size weighted mean effect sizes are based is 1158. The number of independent effect sizes on which means are based is 15. **** P<.0001 (one-tailed). ** P<.01 (one-tailed). * P<.05 (one-tailed). *** P<.001 (one-tailed). Table options As stated previously, there are two ways of reading Table 4 and Table 5. Examining the tables in a within-disorder manner (i.e., a horizontal reading of data tables) reveals that several personality disorders exhibit predictable profile relationships with the five-factor model. Histrionic and Narcissistic personality disorders display a positive relationship with Extraversion, with Narcissistic personality disorder also showing a negative relationship with Agreeableness. Conversely, Schizoid and Schizotypal personality disorders are characterised by negative relationships with Extraversion, with Schizotypal personality disorder showing further positive associations with Neuroticism and negative associations with Agreeableness. Paranoid and Borderline personality disorders are characterised by both positive correlations with Neuroticism and negative correlations with Agreeableness, with Paranoid being more related to Agreeableness and Borderline being more related to Neuroticism. Antisocial personality disorder is characterised by little association with Neuroticism and negative relationships with Agreeableness and Conscientiousness. Avoidant and Dependent personality disorders are both positively correlated with Neuroticism, with Avoidant personality disorder also displaying a negative relationship with Extraversion. Obsessive-Compulsive personality disorder shows a positive relationship with Conscientiousness; however, this finding is not particularly strong, as it does not generalize between weighted and unweighted mean effect sizes. By examining Table 4 and Table 5 in an across-disorder manner (i.e., a vertical reading of data tables), looking at the effect sizes in each five-factor column and particularly the summary means and medians for the effect sizes in each five-factor column, the following observations can be made. Firstly, no prominent relationships between the Openness to Experience dimension of the five-factor model and any of the personality disorder diagnostic categories are reported. Secondly, while the dimension of Conscientiousness shows some meaningful relationships, the dimensions of Neuroticism, Extraversion, and Agreeableness show the greatest number and magnitude of relationships with the various personality disorder categories. Finally, Neuroticism and Agreeableness are the most consistent dimensions across the personality disorders in terms of the direction (positive or negative) of their relatedness to the various personality disorders. Extraversion is mixed in its directional relationship with the personality disorders (e.g., positive mean effect sizes for Histrionic and Narcissistic personality disorders and negative mean effect sizes for Schizoid, Schizotypal, and Avoidant personality disorders). Neuroticism is consistently positively correlated with the disorders, shown particularly by positive mean effect sizes for Paranoid, Schizotypal, Borderline, Avoidant, and Dependent personality disorders. Agreeableness is consistently negatively correlated with the disorders (except for Dependent personality disorder), shown particularly by negative mean effect sizes for Paranoid, Schizotypal, Antisocial, Borderline, and Narcissistic personality disorders. The consistency of Neuroticism and Agreeableness across the personality disorders compared with the other personality dimensions is particularly reflected in the greater magnitude of five-factor column means and medians for Neuroticism and Agreeableness relative to the other dimensions. This is due to there being a mixture of both positive and negative signs among the effect sizes reported in the Extraversion, Openness to Experience, and Conscientiousness columns, which cancel each other out when the central tendency of the effect sizes in these columns is calculated. The consistency of Neuroticism and Agreeableness is further illustrated by Fig. 1, displaying the sample size weighted mean effect sizes for all personality disorders for Neuroticism (x-axis) and Agreeableness (y-axis) personality dimensions. It is evident from Fig. 1 that all personality disorders (except Dependent) cluster in the positive Neuroticism and negative Agreeableness quadrant of the graph. Sample size weighted mean effect sizes for all DSM-IV personality disorders for ... Fig. 1. Sample size weighted mean effect sizes for all DSM-IV personality disorders for Neuroticism (x-axis) and Agreeableness (y-axis) personality dimensions. Figure options Table 6 shows the file drawer analysis conducted on unweighted means reported in Table 4. According to Rosenthal, 1991 and Rosenthal, 1995, the file drawer analysis indicates the robustness of meta-analytic findings, showing how many studies with null results would have to exist that have not been published before the overall level of significance found in the meta-analysis reaches just significant (P=.05). This is often referred to as how tolerant the meta-analytic results are to this possibility of unpublished null findings. Rosenthal specifies that any tolerance level ≥ 5K+10 (where K is the number of studies on which results are based) indicates an unlikely number of existing studies with null results; hence, the meta-analytic results can withstand this possibility. In accordance with this guideline, tolerance levels in Table 6 must be evaluated against 85 (as K=15). Table 6 shows that a large number of the findings do yield adequate levels of tolerance. It should be noted that tolerance levels corresponding to effect sizes ≥ .20 are very large, confirming the strength of these particular findings. Table 6. Tolerance levels yielded by the file drawer analysis conducted on unweighted effect size estimates for each DSM-IV personality disorder and five-factor model personality dimension combination DSM-IV personality disorders Five-factor model personality dimensions N E O A C Paranoid 910.85* 97.07* 5.13 1483.16* 23.85 Schizoid 157.97* 1328.08* 228.85* 389.22* 5.31 Schizotypal 1698.34* 1191.72* 1.78 531.52* 185.00* Antisocial 99.37* 20.21 24.48 1503.29* 851.00* Borderline 3136.04* 69.64 6.78 693.58* 588.25* Histrionic 17.96 2186.74* 348.70* 47.79 159.57* Narcissistic 1.81 605.08* 195.88* 884.55* 27.94 Avoidant 2985.63* 2589.82* 136.69* 180.15* 75.43 Dependent 2041.90* 186.41* 149.75* 30.83 186.08* Obsessive-Compulsive 96.62* 208.65* 50.44 36.20 436.75* N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Tolerance levels based on effect sizes ≥ .20 are in bold. * Tolerance level ≥ 85. Table options Table 7 shows the binomial effect size display for the weighted means reported in Table 5. The binomial effect size display indicates the practical importance of effect sizes obtained, showing out of 100 people with a particular personality disorder, the number that would score high on a particular personality dimension, and the number that would score low on a particular personality dimension Rosenthal, 1991 and Rosenthal, 1995. Correlations at the .20 level yield a binomial effect size display of 60:40. Values at or above this level are highlighted, as this proportional split was considered by the authors to be practically meaningful. The greatest binomial effect size display reported is for the Borderline–Neuroticism relationship, indicating that 75% of people with Borderline personality disorder would score high on Neuroticism. Table 7. Binomial effect size display for each DSM-IV personality disorder and five-factor model personality dimension combination based on sample size weighted mean effect sizes across all studies DSM-IV personality disorders N E O A C High Low High Low High Low High Low High Low Paranoid 64 36 44 56 48 52 33 67 47 53 Schizoid 57 43 39 61 44 56 42 58 49 51 Schizotypal 68 32 36 64 50 50 40 60 44 56 Antisocial 55 45 52 48 53 47 33 67 37 63 Borderline 75 25 46 54 51 49 39 61 39 61 Histrionic 51 49 71 29 58 42 47 53 46 54 Narcissistic 52 48 60 40 56 44 37 63 48 52 Avoidant 74 26 28 72 46 54 45 55 45 55 Dependent 71 29 44 56 45 55 53 47 43 57 Obsessive-Compulsive 54 46 44 56 47 53 48 52 62 38 N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Binomial effect size displays based on effect sizes ≥.20 are in bold. Shows out of 100 people with a particular personality disorder, the number that would score high on a particular personality dimension and the number that would score low on a particular personality dimension. Table options Table 8, Table 9 and Table 10 show the moderator analyses for sample type, personality disorder measure, and five-factor model measure, respectively. Table 8 shows that for all personality disorder/personality dimension combinations highlighted as meaningful in the overall analysis (i.e., Table 4 and Table 5), nearly all emerged as meaningful (≥.20) when the meta-analysis was divided into clinical and nonclinical subgroups (except clinical sample effect sizes for Narcissistic–Extraversion, Borderline–Conscientiousness, and Obsessive Compulsive–Conscientiousness combinations and the nonclinical sample effect size for Schizoid–Extraversion). This shows that the same pattern of mean effect sizes were found across clinical and nonclinical sample types. However, Table 8 also shows that effect sizes identified as meaningful in the overall meta-analysis are consistently greater in the nonclinical sample subgroup compared with the clinical sample subgroup. Table 8. Sample size weighted mean effect sizes for sample type moderator variable DSM-IV personality disorders Five-factor model personality dimensions N E O A C Paranoid Clinical .23**** −.06** −.01 −.27**** −.07** Nonclinical .32**** −.17**** −.07*** −.39**** −.08**** Schizoid Clinical .13*** −.35**** −.15**** −.17**** −.06* Nonclinical .12**** −.15 −.07**** −.17**** −.02 Schizotypal Clinical .28**** −.20**** −.05 −.20**** −.08** Nonclinical .41**** −.34**** .02 −.22**** −.17**** Antisocial Clinical .09*** .06* −.03 −.28**** −.21**** Nonclinical .10**** .02 .11**** −.40**** −.30**** Borderline Clinical .40**** −.04 .05 −.23**** −.17**** Nonclinical .55**** −.13**** .01 −.23**** −.26**** Histrionic Clinical .12**** .28**** .15**** −.08*** −.10*** Nonclinical −.05**** .51**** .15**** −.05 −.08* Narcissistic Clinical .06** .19**** .15**** −.21**** −.05* Nonclinical .02 .21**** .08*** −.32**** −.05 Avoidant Clinical .41**** −.36**** −.09* −.10** −.11**** Nonclinical .54**** −.48**** −.10**** −.12**** −.09**** Dependent Clinical .28**** −.08** −.06* .00 −.12**** Nonclinical .50**** −.16**** −.15**** .08*** −.15**** Obsessive-Compulsive Clinical .08*** −.09** .02 −.06** .06 Nonclinical .09 −.14**** −.14**** −.03 .33**** N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Effect sizes ≥.20 are in bold. The number of independent effect sizes on which clinical sample means are based is 8. The number of independent effect sizes on which nonclinical sample means are based is 7. Breakdown of clinical samples by five-factor model moderator variable: NEO=6 and Other=2. Breakdown of nonclinical samples by five-factor model moderator variable: NEO=5 and Other=2. Breakdown of clinical samples by personality disorder measure moderator variable: Millon=3, Interview=5, and Other=3. Breakdown of nonclinical samples by personality disorder measure moderator variable: Millon=3, Interview=0, and Other=4. **** P<.0001 (one-tailed). ** P<.01 (one-tailed). *** P<.001 (one-tailed). * P<.05 (one-tailed). Table options Table 9. Sample size weighted mean effect sizes for personality disorder measure moderator variable DSM-IV personality disorders Five-factor model personality dimensions N E O A C Paranoid Millon .23**** −.11**** −.07** −.33**** −.03** Interview .24**** −.09** −.01 −.28**** −.11*** Other .37**** −.13**** −.03 −.40**** −.06* Schizoid Millon .23**** −.04**** −.15**** −.18**** −.08** Interview .08* −.28**** −.11** −.13*** −.03 Other .05* −.45**** −.12** −.20**** .04 Schizotypal Millon .45**** −.35**** −.05 −.18**** −.21**** Interview .23**** −.12** −.01 −.18**** −.05 Other .35**** −.34**** .00 −.27**** −.10** Antisocial Millon .05* .08 .10** −.41**** −.22**** Interview .09** .02 −.04 −.27**** −.21**** Other .16**** .01 .08** −.33**** −.35**** Borderline Millon .58**** −.22**** −.02 −.20**** −.24**** Interview .39**** −.02 .06* −.25**** −.20**** Other .48**** .02 .06* −.27**** −.23**** Histrionic Millon −.19**** .60**** .23**** .02 −.04 Interview .23**** .15**** .08* −.14**** −.12** Other .12** .40**** .14**** −.09** −.13**** Narcissistic Millon −.10* .26**** .21**** −.28**** .02 Interview .11*** .07* .12** −.23**** −.08* Other .14*** .23**** .03 −.28**** −.10** Avoidant Millon .56**** −.48**** −.15**** −.12*** −.15**** Interview .36**** −.33**** −.07 −.08* −.12**** Other .48**** −.48**** −.08** −.15**** −.02 Dependent Millon .41**** −.13**** −.20**** .18**** −.10**** Interview .30**** −.05 −.03 −.09** −.16**** Other .50**** −.19**** −.07** −.01 −.16**** Obsessive-Compulsive Millon −.21**** −.04 −.15**** .20**** .52**** Interview .18**** −.14*** .02 −.18**** −.05 Other .35**** −.22**** −.06** −.23**** .03 N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Effect sizes ≥.20 are in bold. The number of effect sizes on which Millon personality disorder measure means are based is 6. The number of effect sizes on which interview personality disorder measure means are based is 5. The number of effect sizes on which other personality disorder measure means are based is 7. Breakdown of Millon measures by sample type moderator variable: clinical=3 and nonclinical=3. Breakdown of Interview measures by sample type moderator variable: clinical=5 and nonclinical=0. Breakdown of Other personality disorder measures by sample type moderator variable: clinical=3 and nonclinical=4. Breakdown of Millon measures by five-factor model measure moderator variable: NEO=5 and Other=0. Breakdown of Interview measures by five-factor model measure moderator variable: NEO=4 and Other=1. Breakdown of Other personality disorder measures by five-factor model measure moderator variable: NEO=3 and Other=3. **** P<.0001 (one-tailed). ** P<.01 (one-tailed). *** P<.001 (one-tailed). * P<.05 (one-tailed). Table options Table 10. Sample size weighted mean effect sizes for five-factor model measure moderator variable DSM-IV personality disorders Five-factor model personality dimensions N E O A C Paranoid NEO .27**** −.14**** −.06*** −.36**** −.10**** Other .31**** −.06 .03 −.28**** .02 Schizoid NEO .11**** −.20*** −.10**** −.15**** −.05** Other .20**** −.40**** −.19**** −.25**** .01 Schizotypal NEO .38**** −.29**** .00 −.21**** −.14**** Other .27**** −.25**** −.05 −.21**** −.09* Antisocial NEO .09**** .02 .05* −.40**** −.25**** Other .09** .09* .06* −.16**** −.28**** Borderline NEO .51**** −.11**** .03 −.23**** −.23**** Other .40**** −.03 −.01 −.25**** −.20**** Histrionic NEO −.02** .47**** .16**** −.05* −.08** Other .16**** .23**** .11** −.14**** −.12** Narcissistic NEO .00 .22**** .13**** −.27**** −.03 Other .17**** .11* .03 −.28**** −.12** Avoidant NEO .50**** −.45**** −.08*** −.09**** −.13**** Other .42**** −.40**** −.13** −.22**** .03 Dependent NEO .43**** −.13**** −.13**** .09**** −.16**** Other .36**** −.13** −.06 −.12*** −.05 Obsessive-Compulsive NEO .04 −.10**** −.08**** −.01 .25**** Other .23**** −.19**** −.03 −.24**** .10* N=Neuroticism. E=Extraversion. O=Openness to experience. A=Agreeableness. C=Conscientiousness. Effect sizes ≥.20 are in bold. The number of independent effect sizes on which NEO five-factor model measure means are based is 11. The number of independent effect sizes on which Other five-factor model measure means are based is 4. Breakdown of NEO measures by sample type moderator variable: clinical=6 and nonclinical=5. Breakdown of Other measures by sample type moderator variable: clinical=2 and nonclinical=2. Breakdown of NEO measures by personality disorder measure moderator variable: Millon=5, Interview=4, and Other=4. Breakdown of Other measures by personality disorder measure moderator variable: Millon=1, Interview=1, and Other=3. **** P<.0001 (one-tailed). *** P<.001 (one-tailed). ** P<.01 (one-tailed). * P<.05 (one-tailed). Table options The main consistent difference among mean effect sizes displayed in Table 9 is that for nearly all personality disorder/personality dimension combinations highlighted as meaningful in the overall analysis, mean effect sizes were lower in the interview personality disorder measure subgroup relative to the Millon questionnaire personality disorder measure and the Other personality disorder measure subgroups (except for Paranoid–Neuroticism, Schizoid–Extraversion, and Borderline–Agreeableness combinations). It is likely these lower effect sizes are attributable to the sample type involved, as only clinical samples were administered interview personality disorder measures, and as already demonstrated by Table 8, clinical samples yielded generally lower effect sizes than nonclinical samples. Table 9 also shows that there are some differences in the strengths of correlations observed across the different personality disorder measures (e.g., Schizoid and Neuroticism and Extraversion, Schizotypal and Extraversion and Conscientiousness, Borderline and Extraversion, etc.). However, the most concerning differences across the personality disorder measures are in terms of variability in the direction of associations. Specifically, inconsistent with the Interview and Other personality disorder measures (hence the main meta-analytic data), when Millon personality disorder measures are used, negative correlations between Neuroticism and Histrionic, Narcissistic, and Compulsive scales emerge. Furthermore, the Compulsive scale of the Millon is positively associated with Agreeableness and shows a substantial correlation with Conscientiousness, both of which do not emerge when using a different personality disorder measure. These findings can be understood in the context of research examining the diagnostic accuracy of the Millon Clinical Multiaxial Inventories, which has found that the Histrionic, Narcissistic, and Compulsive scales of the measure assess adaptive personality styles rather than types of disordered personality (Craig, 1999). Table 10 shows that the only difference in mean effect sizes between the NEO questionnaire five-factor model measure and the Other five-factor model measure subgroups, is that there was a slight bias for larger means in the NEO subgroup for the personality disorder/personality dimension combinations highlighted as meaningful in the overall analysis (i.e., 12 of the 18 highlighted combinations)