Journal of Abnormal Psychology2001, Vol. 110, No. 3,401-412

Copyright 2001 by the American Psychological Association, Inc.0021-843X/01/S5.00 DOI: 10.1037//0021-843X. 110.3.401

Using the Five-Factor Model to Represent the DSM-IV PersonalityDisorders: An Expert Consensus Approach

Donald R. Lynam and Thomas A. WidigerUniversity of Kentucky

This study sought to extend previous work on the five-factor dimensional model (FFM) of personalitydisorder (PD) by developing more comprehensive FFM descriptions of prototypic cases. Specifically, theauthors asked experts in each of the 10 DSM-IV PDs to rate the prototypic case by using all 30 facetsof the FFM. Aggregating across raters of the given disorder generated a prototype for each disorder. Ingeneral, there was good agreement among experts and with previous theoretical and empirical FFMtranslations of DSM diagnostic criteria. Furthermore, the ability of the FFM explanation to reproduce thehigh comorbidity rates among PDs was demonstrated. The authors concluded that, with the possibleexception of schizotypal PD, the DSM PDs can be understood from the dimensional perspective of theFFM. Future directions for research, including the use of the present prototypes to "diagnose" personalitydisorder, are discussed.

The American Psychiatric Association's Diagnostic and Statis-tical Manual'of Mental Disorders, fourth edition (DSM-IV; Amer-ican Psychiatric Association, 1994), provides personality disorderdiagnostic categories. There is good evidence that these diagnosesidentify clinically meaningful maladaptive personality traits; sev-eral diagnoses seem especially well-validated, including antisocial(see Stoff, Breiling, & Maser, 1997), borderline (see Clarkin,Marziali, & Munroe-Blum, 1992), schizotypal (Raine, Lencz, &Mednick, 1995), narcissistic (Ronningstam, 1998) and dependent(see Bornstein, 1992) personality disorders. These disorders havelong clinical traditions, are related to cognitive and perceptualabnormalities, influence the life outcomes of individuals, haveimplications for psychological treatment, and show some geneticcomponent (Clarkin & Lenzenweger, 1996; Livesley, 2001).Nonetheless, the DSM-IV personality disorders have been criti-cized for the fundamental assumption that "Personality Disordersrepresent qualitatively distinct clinical syndromes" (AmericanPsychiatric Association, 1994, p. 633). A variety of studies andmethodologies have raised compelling concerns regarding the va-lidity of the categorical model (Clark, Livesley, & Morey, 1997;Livesley, 1998; Widiger, 1993) and have proposed alternativedimensional models for the personality disorders, including (butnot limited to) the 15-factor model of Clark (1993), the interper-sonal circumplex (Benjamin, 1993), the 7-factor model of Clon-inger (Cloninger, Svrakic, Bay on, & Przybeck, 1999), the 18-factor model of Livesley (1998), and the 5-factor model (Widiger& Costa, 1994). From these dimensional perspectives, the DSM

Donald R. Lynam and Thomas A. Widiger, Department of Psychology,University of Kentucky.

This research was supported in part by National Institute of MentalHealth Grant MH60104.

Correspondence concerning this article should be addressed to DonaldR. Lynam, University of Kentucky, Department of Psychology, 115 KastleHall, Lexington, Kentucky 40506-0044. Electronic mail may be sent todlynal@pop.uky.edu.

personality disorder categories can be seen as configurations ofbasic dimensions of personality.

The five-factor model (FFM) of personality was derived origi-nally from studies of the English language to identify the domainsof personality functioning most important in describing the per-sonality traits of oneself and other persons (Digman, 1990; John &Srivastava, 1999; Wiggins & Pincus, 1992). This lexical researchhas emphasized five broad domains of personality, identified asextraversion (surgency or positive affectivity), agreeableness (vs.antagonism), conscientiousness (or constraint), neuroticism (neg-ative affectivity), and openness to experience (intellect or uncon-ventionality; John & Srivastava, 1999). Each of these five domainscan be further differentiated into underlying facets or components.Costa and McCrae (1995) proposed six facets within each domainon the basis of their research with the NEO Personality Inven-tory—Revised (NEO-PI-R; Costa & McCrae, 1992). For example,they suggested that the domain of agreeableness can be usefullydifferentiated into more specific facets of trust (vs. mistrust orskepticism), straightforwardness (vs. deception or manipulation),altruism (vs. egocentrism or exploitation), compliance (vs. oppo-sitionalism or aggression), modesty (vs. arrogance), and tender-mindedness (vs. toughmindedness or callousness).

Empirical support for the construct validity of the FFM isextensive, at both the domain and the facet levels, includingconvergent and discriminant validation across self, peer, andspouse ratings (Costa & McCrae, 1988); temporal stability across7-10 years (Costa & McCrae, 1994); cross-cultural replication (DeRaad, Perugini, Hrebickova, & Szarota, 1998); and heritability(Jang, McCrae, Angleitner, Reimann, & Livesley, 1998; Plomin &Caspi, 1999). The FFM has been used as an integrative model ofpersonality functioning for children (Halverson, Kohnstamm, &Martin, 1994), adults (McCrae & Costa, 1990), the elderly (Costa& McCrae, 1994), and even animal species (Gosling & John,1999). Instructive and informative critiques of the FFM have beenprovided (e.g., Block, 1995; Westen, 1995), but there does appearto be sufficient empirical support for the "basicness" of the FFM

401

402 LYNAM AND WIDIGER

to consider it for use as a possible dimensional model of person-ality disorder symptomatology (Widiger & Costa, 1994).

From the perspective of the FFM, personality disorders repre-sent configurations of basic dimensions of personality. Much re-search is consistent with this possibility. The dimensions of theFFM are related to personality disorder symptomatology. Widigerand Costa (in press) have identified over 50 published studies thathave shown relations between the FFM and personality disordersymptoms. Most of the studies have examined the relations be-tween scores on FFM inventories and scores on measures ofpersonality disorder symptoms. Almost all of the authors haveconcluded that their findings provided support for understandingpersonality disorder symptomatology from the perspective of theFFM.

Many of these FFM studies addressed specific hypotheses putforth by Widiger, Trull, Clarkin, Sanderson, and Costa (1994),who provided five-factor translations of the DSM-III-R personal-ity disorders (American Psychiatric Association, 1987). Besidesserving as useful heuristics for studies concerned explicitly withDSM-III-R personality disorders, these translations suggest thatthe FFM possesses the language necessary for the description ofthe personality disorders and illustrate the beginnings of an im-portant alternative approach to understanding the personality dis-orders in terms of the FFM. To the extent that the FFM can be usedas a starting point in reproducing the PDs and the nomological netsthat surround them, support is generated for the FFM account.

In the present study, we seek to extend this previous work byusing an alternative approach to the development of FFM person-ality disorder profiles, the use of expert consensus. This studyexamines the degree to which experts agree in their FFM descrip-tions of the personality disorders. In addition, we investigate howwell the expert consensus approach agrees with theoretical andempirical approaches to profile development. Finally, this studyexamines how well the overlap among FFM prototypes can repro-duce the high and differential comorbidity rates among the DSMpersonality disorders—an epidemiological fact that a theory ofpersonality disorders should be able to explain but a fact withwhich the categorical model has had much difficulty (Clark et al.,1997; Lilienfeld, Waldman, & Israel, 1994; Livesley, 1998; Wi-diger, 1993).

The expert consensus approach to be used in the present articlewas used previously by Miller, Lynam, Widiger, and Leukefeld(2001) and Lynam (in press) in their attempts to understandpsychopathy from the perspective of the FFM. Miller et al. (2001)developed an FFM profile of psychopathy by compiling the de-scriptions obtained from 15 nationally recognized psychopathyresearchers of a prototypic case of psychopathy with respect toeach of the 30 Costa and McCrae (1995) facets of the FFM. Thisexpert consensus FFM profile was then compared with Widigerand Lynam's (1998) FFM translation of the 20 items within Hare's(1991) Revised Psychology Checklist (PCL-R) in an attempt toprovide convergent validity for the FFM conceptualization ofpsychopathy. The correlation between the expert consensus profileand the PCL-R FFM profile provided by Widiger and Lynam was.64, indicating good convergence but also suggesting some notabledisagreements reflecting a lack of coverage of the PCL-R.

In a comparison between the expert consensus prototype ap-proach and the criteria translation approach, the consensus ap-proach seems to enjoy some advantages. First, in the case of Miller

et al. (2001), the expert consensus FFM profile was constructed bypersons with varying degrees of familiarity with the FFM ratherthan by investigators with initial expectations of the profile thatshould be obtained. Thus, results should be more compelling whenobtained from personality disorder researchers than from the FFMinvestigators themselves. Second, aggregation across multiple ex-perts blunts idiosyncratic interpretations or judgments concerningthe criterion and brings out in stark contrast the points of agree-ment. In these respects, the consensus approach of Miller et al.could be said to provide a more empirical, objective, and indepen-dent FFM profile. Third, the expert consensus methodology allowsfor the possibility of obtaining elevations on facets of the FFM notincluded explicitly within a particular criteria set. In the case ofpsychopathy, although the PCL-R is the most widely used andcompelling measure of psychopathy, concerns have been raisedwith respect to the adequacy of its coverage and conceptualizationof psychopathy (Lilienfeld, 1994; Lykken, 1995). Fourth, theconsensus methodology allows for an evaluation of the internalconsistency and interrater reliability of the FFM descriptions. If theFFM is to serve as a model of personality disorder symptomology,experts should agree in their FFM parsings of the symptoms. Forexample, Miller et al. obtained good agreement among the psy-chopathy researchers with respect to their FFM ratings. More thanhalf of the 30 facets obtained standard deviations of less than .70,and only 17% had a standard deviation greater than 1.0; interraterreliability ranged from .61 to .84, with a mean of .75. Fifth andfinally, the expert consensus approach provides a quantitativeprofile against which individuals' FFM profiles can be matched todetermine the degree of similarity; individuals who are moresimilar to the prototype possess the disorder to a greater degree.Using data from a longitudinal community sample of 481 adults,Miller et al. found that individuals who more closely approximatedthe psychopathic profile had higher scores on a psychopathyinventory; more symptoms of antisocial personality disorder, sub-stance abuse, and substance dependence; more frequent and variedantisocial activities; and fewer internalizing symptoms. All ofthese relations replicate those in previous studies using incarcer-ated, PCL-R defined psychopaths.

Because of the advantages to this consensus approach, a purposeof the current study is to generate expert consensus FFM profilesfor each of the DSM-FV (American Psychiatric Association, 1994)personality disorders. Internal consistency and interrater reliabilityare examined along with a comparison to the profiles generatedtheoretically by Widiger et al. (1994) and empirically by Dyce andO'Connor (1998) to assess convergent validity, points of disagree-ment, and divergent coverage. In addition, the current study furtherexplores the validity of the expert consensus FFM profiles bydetermining the extent to which the overlap among the profilesmirrors the covariation or co-occurrence among the personalitydisorder diagnostic categories.

Comorbidity of personality disorder diagnoses is the normrather than the exception, despite the hope or aspiration ofDSM-IV that only one personality disorder diagnosis would beapplicable for any particular patient (Gunderson, 1992; Gunder-son, Links, & Reich, 1991). Several studies have demonstrated thatmost patients who meet diagnostic criteria for one personalitydisorder will meet the criteria for a number of other personalitydisorder diagnoses (e.g., Skodol, Rosnick, Kellman, Oldham, &Hyler, 1991; Widiger & Trull, 1998). As noted by Mineka,

FFM PD PROTOTYPES 403

Watson, and Clark (1998), "the greatest challenge that the exten-sive comorbidity data pose to the current nosological systemconcerns the validity of the diagnostic categories themselves—dothese disorders constitute distinct clinical entities?" (p. 380).

Although high comorbidity presents a fundamental challenge tothe validity of the categorical approach, it is easily accommodatedwithin a dimensional model that views the categories as configu-rations of basic dimensions of personality (Clark et al., 1997;Livesley, 1998; Widiger, 1993). From this view, the apparentcomorbidity among the personality disorders is understood as theco-occurrence of diagnoses that have overlapping constellations ofmaladaptive personality traits (Lilienfeld et al., 1994). Much of theco-occurrence among the DSM personality disorders might beexplained by the domains and facets of the FFM; from the FFMperspective, disorders are expected to co-occur to the extent thatthey assess common FFM domains and facets. For example, theco-occurrence of the avoidant and schizoid personality disordersmay reflect the involvement of introversion in both disorders(Widiger et al., 1994). Borderline personality disorder is the mostcommonly diagnosed and co-occurring personality disorder, con-sistent with its FFM conceptualization as being the personalitydisorder defined primarily by the facets of neuroticism (Clarkin,Hull, Cantor, & Sanderson, 1993; Widiger et al., 1994).

Only one prior FFM personality disorder study has addressedpersonality disorder covariation (O'Connor & Dyce, 1998). Al-though the results of this study provided support for the FFMconceptualization of personality disorders, it was limited by notaddressing comorbidity hypotheses with respect to individual per-sonality disorders and by confining the analyses to the five broaddomains of the FFM. Some differentiation among the personalitydisorders can occur with respect to the five domains, but moreprecise differentiation might be obtained at the level of the 30facets (Axelrod, Widiger, Trull, & Corbitt, 1997; Dyce &O'Connor, 1998; Trull, Widiger, & Burr, 2001). The current studyextends the findings of O'Connor and Dyce (1998) by determiningthe extent to which the diagnostic co-occurrence or covariationamong personality disorders identified in prior studies can berecreated within the expert consensus FFM profiles. To the extentthat this can happen, support is generated for the FFM account.

Method

Experts were identified through electronic searches of the psychologicaland psychiatric literature by using the personality disorders as search terms.To be included in the present research, an individual had to have publishedat least one article on the specific personality disorder (although mostparticipants had published more than two articles). Between 25 and 30experts were identified for most disorders. Address information was col-lected from the articles or through a search of the World Wide Web. A totalof 197 experts were identified; the number of experts per disorder rangedfrom a low of 23 for schizotypal personality disorder to a high of 29 forhistrionic and dependent personality disorders. Of these 197 experts, 20were asked to rate three different personality disorders, 33 were asked torate two different personality disorders, and 144 were asked to rate onlyone personality disorder.

Experts were asked to rate prototypic cases of the various personalitydisorders with respect to the 30 facets of the FFM as described by theNEO-PI-R (Costa & McCrae, 1992). The identifying label"for each ofthe 30 facets was provided, along with two to four adjectives that describedboth poles of each of the facets. Adjectives were taken from the NEO-PI-Rtest manual and FFM adjective checklists (Costa & McCrae, 1992; Gold-

berg, 1992; Saucier, 1994). For example, anxiousness was assessed withthe following descriptors: fearful, apprehensive versus relaxed, uncon-cerned, cool. Order was assessed by using these descriptors: organized,methodical, ordered versus haphazard, disorganized, sloppy. For each ofthe 30 traits, experts were asked to rate the prototypic case of a specifiedpersonality disorder on a l-to-5 scale. Specifically, he or she was asked thefollowing:

In particular, we would like you to describe the prototypic case for[one, each of the two, or each of the three] personality disorders on a 1to 5 point scale, where 1 indicates that the prototypic person would beextremely low on the trait (i.e., extremely lower than the averageperson), 2 indicates that the prototypic person would be low, 3indicates that the person would be neither low nor high (i.e., does notdiffer from the average individual), 4 indicates that the prototypicperson would be high on that trait, and 5 indicates that the prototypicperson would be extremely high on that trait. For example, for thevulnerability trait dimension, a score of 1 would indicate that theprototypic case for that respective personality disorder would beextremely low in vulnerability (i.e., stalwart, brave, fearless, unflap-pable), whereas a score of 5 would indicate that the prototypic personwith that respective personality disorder would be extremely high invulnerability (i.e., fragile, helpless). Please rate the prototypic case forthe respective personality disorder on each of the 30 trait dimensions.

Experts were provided with one rating form for each disorder, a demo-graphic questionnaire, and a return, postage-paid envelope with no meansfor the researchers to identify the respondent. Experts were told that theratings they provided would be consolidated with ratings obtained fromother selected experts, and they were assured that their individual ratingswould be kept confidential and not be reported.

We received 170 completed ratings from 120 experts. The averagecompletion rate of 63% compares well with similar previous studies; Milleret al. (2001) obtained a completion rate of 69% in their FFM study ofpsychopathy, and Lynam (in press) obtained a completion rate of 57% inhis study of fledgling psychopathy. The per-disorder completion rateranged from a low of 38% for paranoid PD to over 85% for antisocial andborderline PD. A total of 93% of the 120 individuals completed thedemographic questionnaire. Of these, 77% were male, 81% were Cauca-sian, and the median age was 47 years. Fully 96% had doctorate or medicaldegrees, 56% worked in an academic setting, and 22% worked in medicalcenters. Finally, on a scale ranging from 1 (not at all familiar) to 4 (veryfamiliar), 98% of the experts said they were at least moderately familiar(3), and 77% indicated they were very familiar with the DSM-fV person-ality disorders.

Results

Personality Disorder Prototype Descriptions

Table 1 provides the means and standard deviations for each ofthe 30 facets of the FFM for each of the 10 personality disordersincluded within the DSM-IV. As in Miller et al. (2001), taking anyfacet with a mean score of 2 and lower (low) or 4 and higher (high)as characteristic, it is possible to describe each of the personalitydisorders with the language of the FFM. For example, antisocialpersonality disorder was represented by low scores on all 6 facetsof agreeableness, 3 of 6 facets of conscientiousness (dutifulness,self-discipline, and deliberation), and 2 facets of neuroticism (anx-iousness and self-consciousness); it was also represented by highscores on 3 of the 6 facets of extraversion (assertiveness, activity,and excitement seeking) and 2 facets of neuroticism (angry hos-tility and impulsiveness).

404 LYNAM AND WIDIGER

•

j<uQ

"2O

T3'S§

>— cr. CN -H <N -̂ — en o en r-

— - f N - H t N < N ' - - - ^ ' — 1 —

— ' i n c e ^ o v e n o o a N O t ^ - ^ oi^a^

: [ r i rn^^c^r^

•' O" i— <" bo" r-" ̂ •— ' f oo" ^-* csf bo"

o o ' o o o o ' ^ o o o o o o — ; o — o o o o ^ o o ^ o o ' o ' o o o o o o o o

) r- r- c>^^^

' \Q oo oo oo r-- ON o>O O O O O ' O O O O O O O ' O O O O ' O - H O O O O ' O O ' O O O ' O O O O O O O O

-— — r-ooir-- '— < ̂ ̂ ( - ^ I O N — <T t T t r ^

— <" rj — r- in oo I r- oo <n en

, o, o, o, o^ o, o, o, o, o, o o, o, o,col en

, ̂ o, o, o, ̂ o, o o, o, o, ̂ o, o, o, o, o, o, o, o o,m r- •— < |

r^

r-t oC en ^o in t-~ bo es" bo bo o r~- -— cs

O O o o O O O Q O O i n | » n O O O O O Op t ^ ^ ' ^ . ' O 0 ^lenen

P a o >P en r- T2

g £ s s £-" " 8 S I<u s £

(D

•s•gJ3 -a B .2 -3 .a js .a > >^ e «

^ o « i - t « ^ H ' j 3 . t ^ r t >

11s < -blJ!Q^

.1? 1 g1

^oII< <

* -o1 5?s a S! n S .S § » 2! jS cfl 'aj '-C ^ ^2

i|^l^2>

^ill-itilal^

3A -C S SX) w t**3 "g p -g S§,£«<

•- §•»s 1-s

c e" ausJ|1Sa " =3 ̂'5 -S S3 —

"o

so—

§

FFM PD PROTOTYPES 405

Table 2Measures of Agreement Among Experts for the 10 Personality Disorders

Disorder

ParanoidSchizoidSchizotypalAntisocialBorderlineHistrionicNarcissisticAvoidantDependentCompulsive

No. ofraters

10131222241912211917

Averager a

WS

0.650.750.630.660.700.630.650.670.700.71

AverageSDb

0.820.670.830.800.760.850.790.780.750.75

Averageinterrater rc

0.510.530.480.650.530.560.630.630.560.66

Average correcteditem-total t*

0.680.710.660.750.680.740.780.780.730.80

a forcomposite6

0.910.930.910.970.960.960.950.970.960.97

r withWidiger et al.f

0.790.550.790.770.720.580.540.830.780.58

r with Dyce &O'Connor8

0.720.640.740.780.780.540.750.720.610.83

" rwg = the proportional reduction in error variance relative to a discrete uniform distribution. b Average SD = the arithmetic mean of the standarddeviations. c Average interrater r = the average of the correlations between experts' ratings in which experts are treated as variables and facets ascases. d Average corrected item-total r = the average of the correlations between each rater's profile and the composite profile computed without thatrating. e a for composite = coefficient alpha for the composite prototype in which experts are treated as variables and facets as cases; it does depend,in part, on the number of raters. f This is the correlation with the prototypes taken from Widiger et al. (1994). These prototypes were created by assigninga score, ranging from 4 (high) to —4 (low) for each facet according to the following: a 4 or —4 for facets that are high or low according to DSM-HI-Rcriteria; a 3 or - 3 for facets that are high or low on the basis of clinical literature; a 2 or - 2 for facets that are high or low on the basis of associated featuresin the DSM-IH-R; a 1 or -1 for facets that are somewhat high or low on the basis of the clinical literature; and a 0 for facets that are neither high norlow. g This is the correlation of the consensus prototype with the correlations between the five-factor model of personality and the measures of personalitydisorder obtained by Dyce and O'Connor (1998) in a sample of undergraduates.

Agreement Among Experts

Table 2 presents several measures of agreement among ourexperts. The second column provides information on the averagewithin-group agreement, rwg, for the FFM facets for each person-ality disorder. This coefficient has been offered as a means of"assessing agreement among the judgements made by a singlegroup of judges on a single variable in regard to a single target"(James, Demaree, & Wolf, 1984, as cited in James, Demaree, &Wolf, 1993) and represents the proportional reduction in errorvariance relative to a random process. On average, agreement atthe facet level appeared fairly high; in all cases, the expert ratingsachieved at least a 60% reduction in error. The third column ofTable 2 provides information on agreement at the facet levelthrough examination of the standard deviations of expert ratings.In general, the average standard deviations were good: all werebelow 0.90. For most disorders, fewer than 20% of the facetsachieved standard deviations greater than one; this compares fa-vorably to the 17% of facets with standard deviations greater thanone in the Miller et al. (2001) study.1 According to these criteria(i.e., high rwg and low SD), agreement was best for schizoid,dependent, and compulsive PDs and worst for schizotypal andhistrionic PDs.

The next three columns of Table 2 provide information aboutagreement among raters across the entire profile. The fourth col-umn provides the average interrater correlation (i.e., the averagecorrelation of one rater's profile with every other rater's profile)for each personality disorder. These average interrater correlationsrange from a low of .48 for schizotypal personality disorder to ahigh of .66 for obsessive-compulsive personality disorder; all ofthese correlations are somewhat lower than the average of .75reported by Miller et al. (2001). The fifth column provides theaverage correlation between an individual expert's rating and thecomposite made without his or her rating contributing to it; assuch, it offers information on how well individual ratings agree

with the overall composite. These average, corrected item-totalcorrelations are all relatively high, ranging from a low of .66 forschizotypal personality disorder to a high of .80 for obsessive-compulsive personality disorder. The sixth column provides infor-mation on the reliability of the composite through calculation ofcoefficient alpha.2 These reliability estimates are all relativelyhigh, ranging from .91 for paranoid and schizotypal personalitydisorders to .97 for antisocial, avoidant, and obsessive-compulsive personality disorders. All compare favorably to the .98found for the psychopathy ratings. According to these three indi-ces, agreement was best for compulsive, antisocial, and avoidantPDs and worst for schizotypal, schizoid, and paranoid PDs.

Agreement With Hypothesized FFM-PD Relations(Widiger et al., 1994)

The seventh column in Table 2 provides information on theconvergence between the present expert-generated prototypes anda quantification of the hypotheses put forth by Widiger et al.(1994). The FFM descriptions of the personality disorders byWidiger et al. were hierarchical according to whether they con-cerned core features or associated features and whether they werebased on the DSM-HI-R or the clinical literature. For the purposesof this study, hypotheses based on the diagnostic criteria were

1 The Miller et al. (2001) study on psychopathy provides a fairly goodupper-bound comparison for the present results. We believe the Miller etal. study serves as an upper bound because of its concentration on psy-chopathy, which is (a) a disorder described predominantly in terms ofpersonality traits rather than behaviors, which serve as the bases for manyof the DSM criteria; and (b) studied by a relatively small, homogeneousgroup of investigators. Both of these factors can be expected to contributeto high agreement among raters.

2 For this analysis, raters served as variables and facets served as cases.

406 LYNAM AND WIDIGER

Table 3Examples of Hypothetical (Widiger et al, 1994) and Empirical (Dyce & O'Connor, 1998)FFM-PD Profiles

Domain and facet

Widiger et al.a

Paranoid Narcissistic

Dyce & O'Connorb

Histrionic Compulsive

NeuroticismAnxiousnessAngry hostilityDepressivenessSelf-consciousnessImpulsivenessVulnerability

ExtraversionWarmthGregariousnessAssertivenessActivityExcitement seekingPositive emotions

OpennessFantasyAestheticsFeelingsActionsIdeasValues

AgreeablenessTrustStraightforwardnessAltruismComplianceModestyTendermindedness

ConscientiousnessCompetenceOrderDutifulnessAchievement strivingSelf-disciplineDeliberation

240000

-2-2

0002

0••y

-2-3

00

-4-4

0-4-2

^

200000

04

-2403

003000

400000

0-2-4

0-4-4

200100

-.29-.11-.32-.36

.04-.23

.48

.42

.43

.39

.36

.42

.12

.11

.21

.19

.12-.01

.18-.11

.16-.05-.23

.06

.21-.04

.02

.17

.14-.14

.00-.14-.13

.01-.30-.15

.04-.10

.00

.06-.13

.02

-.28.01

-.04-.12-.01-.14

.12

.18

.23

.19

.09

.11

.38

.48

.45

.43

.51

.43

Note. FFM = five-factor model of personality; PD = measures of personality disorder.a Numbers in these columns represent quantification of qualitative hypotheses regarding relations between FFMand PD put forth by Widiger et al. (1994). b The data in columns 3 and 4 are from Table 2 of "PersonalityDisorders and the Five-Factor Model: A Test of Facet-Level Predictions," by J. A. Dyce and B. P. O'Connor,1998, Journal of Personality Disorders, 12, pp. 37-38. Copyright 1998 by Guilford Press. Reprinted withpermission. Numbers in these columns represent empirical correlations between FFM and PD symptomsreported by Dyce and O'Connor (1998).

given a rating of 4 or —4 (a high or low rating on a facet,respectively); hypotheses based on the core features in the clinicalliterature were given a 3 or —3; hypotheses based on associatedfeatures in the DSM-III-R were given a rating of 2 or -2; andhypotheses based on associated features in the clinical literaturewere given scores of 1 or — 1. A score of 0 was assigned if Widigeret al. provided no hypotheses for the facet in question.3 Table 3provides examples of this quantification; the full quantification isavailable on request. To obtain the degree of convergence, eachexpert prototype was correlated with the respective Widiger et al.prototype in an analysis in which facets were treated as cases.

It is evident from Table 2 that there was considerable conver-gence between our expert prototype ratings and the Widiger et al.(1994) hypotheses for 6 of the 10 personality disorders: paranoid,schizotypal, antisocial, borderline, avoidant, and dependent per-

sonality disorders. For these disorders, the convergent validitycorrelations ranged from .77 (antisocial) to .83 (avoidant).4 Theseconvergent validity coefficients for the DSM-TV personality dis-orders compare favorably with the .64 correlation obtained byMiller et al. (2001) for the FFM description of psychopathy.

3 Results change very little under alternative quantification schemes(e.g., including only facets based on DSM criteria or associated features,making no distinction between DSM criteria and clinical literature, orreversing the coding of clinical core features and DSM associated features).

4 Because these correlations are used descriptively rather than inferen-tially to index the convergence between our profiles and the Widiger et al.profiles, no statistical significance tests were performed.

FFM PD PROTOTYPES 407

Lower but still relatively large convergent relations were alsoobtained for the remaining four personality disorders: schizoid,histrionic, narcissistic, and obsessive-compulsive. For thesedisorders, correlations ranged from .54 (narcissistic) to .58(obsessive-compulsive). Comparisons of the expert and hypo-thetical profiles provide information on the divergences thatcontributed to these relatively lower convergent validity coef-ficients. For example, both the schizoid PD researchers andWidiger et al. (1994) characterized the schizoid PD as beingvery low in gregariousness, positive emotionality, excitementseeking, warmth, and openness to feelings. However, Widigeret al. also described the schizoid PD as being low in self-consciousness and high in hostility, whereas the schizoid expertconsensus prototype was neither high nor low in any facets ofneuroticism. In addition, the consensus prototype included lowscores for schizoid PD on the two additional facets of extra-version (i.e., assertiveness and activity) and openness toactions.

The experts and Widiger et al. (1994) characterized the histri-onic personality disorder as involving facets of extraversion, in-cluding gregariousness, excitement seeking, activity, and positiveemotionality; facets of openness to fantasy, feelings, and actions;and low self-discipline, a facet of conscientiousness. The diver-gence appeared to be driven by extreme differences in the ratingsof self-consciousness (rated high by Widiger et al., 1994, but lowby the experts); the characterization by Widiger et al. (1994) ofhistrionic PD as high in hostility; and by the inclusion in the expertconsensus prototype of two additional facets representing impul-sivity that were not included by Widiger et al. (1994): highimpulsiveness and low deliberation.

The greatest disagreement occurred for the narcissistic person-ality disorder. There was agreement with respect to several facetsof antagonism, particularly low altruism, low modesty, and lowtendermindedness, as well as the neuroticism facet of angry hos-tility and the extraversion facet of assertiveness. Similar to the casefor histrionic PD, Widiger et al. (1994) characterized the narcis-sistic personality disorder as being very high in self-consciousnessand vulnerability, whereas the expert raters described the proto-typic narcissist as being very low in self-consciousness and did notbelieve that vulnerability was particularly defining. The expertraters also described the narcissistic individual as being very lowin all facets of agreeableness (including trust, straightforwardness,and compliance), two facets of extraversion (warmth and excite-ment seeking), and openness to feelings, but high in openness toactions.

The expert raters and Widiger et al. (1994) also disagreed insome respects in their descriptions of the obsessive-compulsivePD. Both descriptions included high scores on several facets ofconscientiousness and low scores on warmth, openness to feel-ings, and actions. However, the prototype derived fromobsessive-compulsive PD experts also included additional fac-ets related to the extreme control of impulses (i.e., low scoreson impulsiveness and excitement seeking, and high scores onself-discipline), low scores on openness to ideas and values, andhigh scores on anxiousness and competence. In addition, Wi-diger et al. (1994), but not the experts, indicated that obsessive-compulsive PD was characterized by low scores on two facetsof agreeableness (altruism and compliance) and high scores onassertiveness.

Agreement With Empirical FFM-PD Relations (Dyce &O'Connor, 1998)

The final column of Table 2 provides the convergence betweenthe present expert-generated prototypes and an empirically derivedprofile taken from a study by Dyce and O'Connor (1998) in whichscores on the NEO-PI-R were correlated with PD symptomologyassessed with the Millon Clinical Multiaxial Inventory—III (Mil-Ion, Millon, & Davis, 1994). In this analysis, facets were againtreated as cases and the FFM profiles were correlated with thecorrelations between the NEO-PI-R facets and personality disor-der symptoms taken from Dyce and O'Connor (1998); Table 3provides examples of the empirical FFM-PD profiles taken fromDyce and O'Connor. As can be seen in Table 2, convergence wasagain quite good. Correlations were above .70 for seven of thedisorders: compulsive, antisocial, borderline, narcissistic, schizo-typal, paranoid, and avoidant; agreement was not as strong for theremaining three, especially for histrionic PD. Five of the disordersthat converged well with the Widiger et al. (1994) profiles alsoconverged well with the Dyce and O'Connor profiles (paranoid,schizotypal, antisocial, borderline, and avoidant), whereas twoconverged moderately for both (schizoid and histrionic).

As before, a comparison of the expert and empirical profilesprovided information on the nature of the divergences. For schiz-oid PD, both sources (the experts and the empirical relations)agreed that this PD was characterized by low levels of all facets ofextraversion and two facets of openness (feelings and actions). Thesources diverged in that the empirical profile, but not the experts,indicated that schizoid PD was also characterized as high in twofacets of neuroticism (depression and self-consciousness) and lowin trust.

Expert and empirical descriptions agreed in the characterizationof histrionic PD as high in all facets of extraversion, two facets ofopenness (feelings and actions), and trust and as low in self-consciousness and deliberation. The sources diverged in that ex-perts, but not the empirical profile, characterized histrionic PD ashigh in impulsiveness and low in self-discipline. Conversely, theempirical profile, but not the experts, suggested that histrionic PDis characterized by low scores on additional facets of neuroticism(anxiety, depression, and vulnerability).

Finally, there was agreement across sources in the characteriza-tion of dependent PD as high in most facets of neuroticism (exceptfor angry hostility and impulsiveness) and as low in assertiveness.The divergence appears to be due to dependent PD being charac-terized as high in three facets of agreeableness (trust, compliance,and modesty) by the experts but not by the empirical profile. Inaddition, the empirical profile, but not the expert raters, suggestedthat dependent PD is characterized as low in competence.

Using the FFM Prototypes to Reproduce DSMComorbidities

To examine how well the FFM model could reproduce thecomorbidities among the personality disorders, we first correlatedthe FFM prototypes with one another to generate a matrix repre-senting the conceptual overlap among the prototypes; this wasaccomplished through an analysis that treated facets as cases andtreated different PDs as variables. Results are provided in Table 4.

408 LYNAM AND WIDIGER

Table 4Correlations Among Five-Factor Model Expert Prototypes

10

1. Paranoid2. Schizoid3. Schizotypal4. Antisocial5. Borderline6. Histrionic7. Narcissistic8. Avoidant9. Dependent

10. Compulsive

.51

.48

.08

.27-.36

.29

.41-.23

.53

.64-.49

•JC,L~J

-.69-.32

.72

.33

.50

-.17.34

-.20-.15

.66

.28

.01

—.63.67.80

-.67-.74-.52

—.56.39

-.03-.18-.49

—.51

-.57-.28-.75

—-.56 —-.81 .69 —-.16 .43 .12 —

Note. These correlations were obtained through an analysis that treated facets as cases and personalitydisorders as variables.

Correlations ranged from —.81 between dependent PD and nar-cissistic PD to .80 between antisocial PD and narcissistic PD.5

Next, we assessed the congruence between this matrix and twopreviously generated comorbidity matrices aggregated across mul-tiple studies. This examination was accomplished by correlatingthe comorbidity estimates for a given PD taken from one matrixwith those estimates taken from another matrix. For example, thenumbers in the first column of Table 4 representing the overlapbetween the FFM paranoid profile and the other FFM profiles werecorrelated with comorbidity estimates for paranoid PD taken fromone of the other comorbidity matrices to give an estimate of thecongruence across matrices. To the extent that the .matrix gener-ated by the expert consensus prototypes mirrors the observedcomorbidities (expressed as high correlations between the esti-mates), evidence is generated for the FFM account. Matricesaggregated across multiple studies were used because of the lowagreement between comorbidity estimates taken from any twosingle studies (Oldham et al., 1992).6

The first aggregate matrix of comorbidities was a correlationmatrix reported previously by Widiger et al. (1991), who aggre-gated the covariation among the DSM-II1 personality disordercriterion sets provided in nine previously published studies. Thesecond matrix came out of work by the DSM-IV PersonalityDisorders Work Group, who obtained data on the criterion sets forall of the DSM-IU-R personality disorders from six independentresearch sites. The co-occurrence rates among the DSM-III-Rpersonality disorders were then aggregated across the data sets andreported by Widiger and Trull (1998) in the fourth volume of theDSM-IV Sourcebook.

For each disorder, its association with the other 9 disorders in agiven data set (FFM, aggregation of the 9 previously publishedDSM-HI studies, and the DSM-IV MacArthur comorbidities) wascorrelated with its degree of association with the other 9 disordersin each of the other two data sets. This correlation provides anindex of how well the various comorbidity estimates for a givendisorder compare across the data sets; the higher the correlation,the greater the congruence. Table 5 provides these results. The firsttwo columns provide the correlations between the FFM prototypematrix and the comorbidities from the 9 previously publishedDSM-III studies and the 6 data sets from the DSM-IV MacArthurdata. The third column provides the correlation between the twoaggregated DSM comorbidity matrices. The convergences between

the overlap from the FFM profiles and the DSM-III comorbiditiesranged from a low of .58 for dependent PD to a high of .95 fornarcissistic PD. Convergence was highest for antisocial and nar-cissistic PDs, with correlations greater than .90; five additionalcorrelations were greater than .70 (schizoid, schizotypal, border-line, histrionic, and avoidant). Three disorders showed weakerconvergence (paranoid, dependent, and compulsive), with correla-tions less than .70. Over half of these correlations equal or exceedthose obtained from a comparison of the two aggregated matrices,which provides an important benchmark.7

The congruence between the association of FFM profiles andthe comorbidity estimates from the DSM-IV MacArthur studieswas generally less than what was observed for the DSM-III Co-morbidities. These correlations ranged from a low of .09 forparanoid PD to a high of .89 for schizotypal PD. The three highestconvergences were found for schizotypal, antisocial, and narcis-sistic PDs, with all correlations greater than .75. Somewhat lowerconvergence was found for schizoid, histrionic, avoidant, andcompulsive PDs, with correlations ranging from .60 to .70. Finally,frankly poor convergence was obtained for dependent and para-noid PDs. Almost half of all correlations were as high or higherthan the convergences obtained for the two aggregated matrices.

Discussion

This study sought to extend the five-factor model understandingof personality disorders as configurations of extreme scores oncommon dimensions of personality. First, it extended understand-

5 Note that negative correlations are expected from the FFM account,although they are infrequently observed in empirical comorbidity estimatesand are actually impossible with co-occurrence data. We believe thesedifferences are due to the fact the FFM allows for individuals to score ateither pole of a trait but PD symptoms assess only one pole. For example,individuals can be either suspicious or gullible on the trust facet ofagreeableness but only suspicious or not in the paranoid criteria.

6 For example, using data from a study of Oldham et al. (1992) thatassessed the same patieats with two different structured interviews, wefound that the correlations between the co-occurrences for a given disordercalculated from each interview ranged from .07 to .86 with a mean of .51.

7 Again, because these correlations are used descriptively rather thaninferentially, no statistical significance tests were performed.

FFM PD PROTOTYPES 409

Table 5Similarity Among Comorbidity of FFM Prototype and Two Other DSM Comorbidity Matrices byPersonality Disorder

FFMa and ninePersonality disorder DSM-HI studies6

ParanoidSchizoidSchizotypalAntisocialBorderlineHistrionicNarcissisticAvoidantDependentCompulsive

AH personality disorders

.60

.75

.83

.92

.75

.84

.95

.84

.58

.65

.76

FFM and DSM-IV DSM-IV MacArthur Project andMacArthur Project0 nine DSM-HI studies

.09

.70

.89

.87

.50

.68

.76

.63

.33

.60

.60

.06

.72

.70

.92

.80

.85

.89

.84

.90

.84

.78

Note. FFM = five-factor model; DSM-111 and DSM-IV = Diagnostic and Statistical Manual of MentalDisorders (3rd & 4th eds.; American Psychiatric Association, 1987, 1994). The numbers represent correlationsbetween Comorbidity rates for each disorder taken from different sources.a "Comorbidities" for the FFM represent correlations among expert prototypes. b Comorbidities representcorrelations among personality disorders averaged across nine studies reported in Widiger et al. (1991). c Co-morbidities represent co-occurrence rates among personality disorders from the DSM-IV MacArthur Projectreported in Widiger and Trull (1998).

ing by developing comprehensive FFM descriptions of the person-ality disorders provided by published researchers. Agreementamong the expert raters for almost all of the DSM-IV PDs wasquite good; standard deviations were low, proportional reductionsin error variance were high, and average interrater correlations,average corrected item-total correlations, and composite coeffi-cient alphas were all high. Agreement is especially impressive inlight of the fact that experts were rating prototypic cases from theirown diverse experiences and not the same individual case. Theseresults suggest that most of the DSM-IV PDs can be readilydescribed, and usually with a high level of agreement, in terms ofthe 30 facets of the FFM. The lowest interrater and rater-composite agreement were obtained for the schizotypal PD (.48and .66, respectively), which has also obtained relatively weakersupport as a maladaptive variant of common personality traits instudies correlating measures of the FFM to PD symptomatology(e.g., Trull, 1992). Although these results may reflect shortcom-ings in the FFM assessment of the cognitive-perceptual domain,they may also be due to the fact that schizotypal personalitydisorder might be better understood as a variant of schizophreniarather than as a personality disorder (Siever & Davis, 1991; Wi-diger & Costa, 1994).

The expert-generated prototypes converged very well with theFFM translations of DSM-IV symptoms provided by Widiger et al.(1994) and with empirical relations found between the FFM andpersonality disorders (Dyce & O'Connor, 1998). Although someof this convergence may have been due to previous familiaritywith the Widiger et al. prototypes, we believe this was unlikely.Expert raters were selected on the basis of their familiarity with thepersonality disorders not with the FFM; they were instructed torate prototypic cases, and there were interesting divergences thatwe believe were mostly due to the more comprehensive coverageof FFM facets provided by the expert consensus ratings. Forexample, obsessive-compulsive PD experts suggested that this

personality disorder also involves a substantial anxiousness, acontrolled restraint (low impulsiveness), and an excessive cau-tiousness (low excitement seeking) that are not recognized by theDSM-IV criteria for this disorder (American Psychiatric Associa-tion, 1994). Although it is possible that our experts, rather than theDSM descriptions, were mistaken, our approach to aggregationrequires that these mistakes or misconceptions be widely held byexperts; if these misconceptions were held only by a few raters,they would have been blunted in the aggregated profile.

Some disagreement also reflected points of dispute in the de-velopment of the diagnostic criteria. For example, there was con-siderable disagreement between the narcissistic PD researchersand the ZWM-based descriptions with respect to self-consciousnessand vulnerability. Widiger et al. (1994) described the narcissisticPD as being high in self-consciousness and vulnerability, whereasthe narcissistic PD researchers described the prototypic case of thisdisorder as being low in these traits. This disagreement parallels adifficulty the authors of the narcissistic diagnostic criterion setshave had with the apparently inconsistent response of these per-sons to praise, criticism, or rebuke from others (Gunderson, Ron-ningstam, & Smith, 1991; Ronningstam & Gunderson, 1990; Wi-diger, Frances, Spitzer, & Williams, 1988).

Finally, we examined how well the FFM descriptions of thepersonality disorders could reproduce the very high Comorbidityrates reported in the literature. Although these high Comorbiditiesare nettlesome to the categorical approach (Clark et al., 1997;Lilienfeld et al., 1994; Livesley, 1998; Mineka et al., 1998), theyare expected under a dimensional model that views the personalitydisorders as configurations of a finite number of basic personalitydimensions. Under the FFM account, disorders appear comorbid tothe extent that they are characterized by the same facets. Althoughour results cannot prove that this is the case, we believe that, to theextent the high rates of comorbidity among disorders can bereproduced when one begins with these trait dimensions rather

410 LYNAM AND WIDIGER

than the disorders themselves, support is generated for the dimen-sional understanding.

The conceptual overlap among FFM profiles reproduced wellthe covariation obtained for the schizoid, schizotypal, antisocial,borderline, histrionic, narcissistic, avoidant, and compulsive PDsaggregated across several sets of studies. The FFM predictionscorrelated as highly with the co-occurrence rates reported in ninepreviously published studies, as these studies correlated with theDSM-IV Mac Arthur co-occurrence rates for 8 of the 10 personalitydisorders. Poor results were obtained across both data sets for onlyone personality disorder, dependent PD. This may be due to thefact that the FFM profiles are constructed with bipolar descriptorsin contrast to the DSM criterion sets, which assess only one pole ofa dimension, resulting in much greater differentiation within theFFM profiles than within general clinical settings. This will beespecially true when co-occurrence rates are used. The greatestdifferentiation among the personality disorders was, in fact, pre-dicted for the dependent PD, with correlations of —.81 and — .74with narcissistic and antisocial PDs, respectively, whereas thedependent PD was in fact reported to be somewhat comorbid withthe antisocial and narcissistic PDs in most of the prior comorbiditystudies (Widiger & Trull, 1998).

The distinctiveness between PDs in the present study stands incontrast to results from a recent study by Morey, Gunderson,Quigley, and Lyons (2000), who concluded that the FFM wasunsuccessful in providing much differentiation among the person-ality disorders because the FFM profile they obtained was essen-tially the same for each personality disorder. However, the primaryreasons for the failure of Morey et al. to obtain adequate differ-entiation was the inadequate differentiation among the disordersthemselves and their use of the NEO-FFI (Costa & McCrae, 1992),which only assesses the five domains of the FFM. In their study,each of the 11 DSM-III-R PDs met, on average, the diagnosticcriteria for 3 other personality disorders; persons who met thehistrionic and narcissistic criteria met the criteria for 5 otherpersonality disorders on average. Similar FFM profiles will beobtained for different personality disorders if there is substantialco-occurrence among them because the same individuals willcontribute to the ratings of each disorder. Moreover, more precisedifferentiation between disorders can be obtained at the level ofthe 30 facets (Axelrod et al., 1997).

Several researchers (e.g., Clark et al., 1997; Widiger & Frances,1994) have previously argued for the superiority of the dimen-sional approach; we believe the present results underscore thisconclusion. The dimensional approach avoids dichotomous black-white decisions and the cognitive biases that come with suchdecisions (Cantor & Genero, 1986). Under the FFM account, anindividual is described by using his or her standing on the 30 facetsof the NEO-PI-R (which provides a very detailed and specificdescription using all available information) or by their degree ofapproximation to a prototype. Similarly, personality disorders areunderstood as constellations of basic traits and described by therelative standings of the prototypic case on all 30 facets. Informa-tion on all facets is preserved, and prototypic cases of differentdisorders may diverge in degree for a given facet. In addition, asnoted previously, a dimensional model anticipates high rates ofcomorbidity among disorders; present results suggest that shareddimensions across disorders is sufficient to account for the highcomorbidity. Finally, and often underappreciated, a dimensional

model of personality ties applied research into a corpus of knowl-edge derived from basic research on personality. That is, to theextent that personality disorders are understood from a dimen-sional perspective, basic research in personality can be easilybrought to bear on personality disorders. For example, research onthe structure (John & Srivastava, 1999; Watson, Clark, & Hark-ness, 1994), genetics (Bouchard, Lykken, McGue, Segal, & Tel-legen, 1990; Rutter et al., 1997), neurobiology (Depue, 1996;Zuckerman, 1994), and development (Caspi, 1997; Scarr & Mc-Cartney, 1983) of personality can be applied to research on per-sonality disorders to extend the understanding of mechanisms andtreatment and to generate theory.

In sum, the current results provide support for the specific FFMaccount and the more general dimensional approach to understand-ing the personality disorders. Personality disorders do appear toinvolve constellations of maladaptive variants of the personalitytraits included within the more general FFM of personality func-tioning. We note, however, that simply because the personalitydisorders can be understood from a dimensional perspective doesnot render them useless as descriptions of configurations of traits.Antisocial personality disorder is an important construct becausethe particular combination of high antagonism; low deliberation,dutifulness, and self-discipline; low anxiety and self-conscious-ness; and high impulsiveness, excitement seeking, and angry hos-tility is so virulent (Widiger & Lynam, 1998).

Where does one go from here? Up to this point, we havedemonstrated that (a) the FFM can be used to describe DSMpersonality disorders, (b) experts can agree in their descriptions,(c) various approaches to this description agree with one another,and (d) the FFM can incorporate a finding with which the currentcategorical approach has difficulty—excessive comorbidityamong the personality disorders. Next steps will involve furtherdemonstrating the power of this approach by examining its abilityto reproduce the nomological network surrounding the DSM per-sonality disorders. For example, it is possible to ask whetherpersonality disorders "diagnosed" using the present FFM proto-types behave similarly in terms of interpersonal, social, and occu-pational functioning as do traditional PD diagnoses. The proto-types generated here can be matched against an individual's NEO-PI-R profile to yield a set of similarity scores. The more similar anindividual is to a given prototype, the more he or she could be saidto have that particular personality disorder. For example, in theMiller et al. (2001) study on psychopathy, we used an intraclasscorrelation coefficient to assess the degree of similarity betweenthe FFM profile of the prototypic psychopath and an individual'sNEO-PI-R FFM profile.8 This similarity index was then used as anindex of psychopathy. We found that individuals who more closelyapproximated the prototype had higher scores on a psychopathyinventory; more symptoms of antisocial personality disorder, sub-stance abuse, and substance dependence; more frequent and variedantisocial activities; and fewer internalizing symptoms. All ofthese relations replicated those in previous studies using incarcer-ated, PCL-R (Hare, 1991) defined psychopaths. When the FFM

8 The intraclass correlation was used because it takes into account notonly similarity in shape but also similarity in height of the profiles. It iscalculated as a double-entry correlation in which facets serve as partici-pants and participants serve as variables.

FFM PD PROTOTYPES 411

profiles are shown to reproduce the nomological networks sur-rounding the existing diagnostic categories, then a strong argumentmight be made for replacing these diagnostic categories with theFFM.

References

American Psychiatric Association. (1987). Diagnostic and statistical man-ual of mental disorders (3rd ed., rev.). Washington, DC: Author.

American Psychiatric Association. (1994). Diagnostic and statistical man-ual of mental disorders (4th ed.). Washington, DC: Author.

Axelrod, S. R., Widiger, T. A., Trull, T. J., & Corbitt, E. M. (1997).Relation of five-factor model antagonism facets with personality disor-der symptomatology. Journal of Personality Assessment, 69, 297-313.

Benjamin, L. S. (1993). Interpersonal diagnosis and treatment of person-ality disorders. New York: Guilford Press.

Block, J. (1995). A contrarian view of the five-factor approach to person-ality description. Psychological Bulletin, 117, 187-215.

Bornstein, R. F. (1992). The dependent personality: Developmental, social,and clinical perspectives. Psychological Bulletin, 112, 3-23.

Bouchard, T., Lykken, D., McGue, M., Segal, N., & Tellegen, A. (1990).Sources of human psychological differences: The Minnesota Study ofTwins Reared Apart. Science, 250, 223-228.

Cantor, N., & Genero, N. (1986). Psychiatric diagnosis and natural cate-gorization: A close analogy. In T. Millon & G. Klerman (Eds.), Con-temporary directions in psychopathology (pp. 233-256). New York:Guilford Press.

Caspi, A. (1997). Personality development across the life course. In W.Damon (Series Ed.) & N. Eisenberg (Vol. Ed.), Handbook of childpsychology, Vol. 3. Social, emotional, and personality development (pp.311-388). New York: Wiley.

Clark, L. A. (1993). Manual for the Schedule for Nonadaptive and Adap-tive Personality (SNAP). Minneapolis, MN: University of MinnesotaPress.

Clark, L. A., Livesley, W. J., & Morey, L. (1997). Personality disorderassessment: The challenge of construct validity. Journal of PersonalityDisorders, 11, 205-231.

Clarkin, J. F., Hull, J. W., Cantor, J., & Sanderson, C. (1993). Borderlinepersonality disorder and personality traits: A comparison of SCID-1IBPD and NEO-PI. Psychological Assessment, 5, 472-476.

Clarkin, J. F., & Lenzenweger, M. F. (1996). Major theories of personalitydisorder. New York: Guilford Press.

Clarkin, J. F., Marziali, E., & Munroe-Blum, H. (1992). Borderline per-sonality disorder: Clinical and empirical perspectives. New York: Guil-ford Press.

Cloninger, C. R., Svrakic, D. M., Bayon, C., & Przybeck, T. R. (1999).Measurement of psychopathology as variants of personality. In C. R.Cloninger (Ed.), Personality and psychopathology (pp. 33-65). Wash-ington, DC: American Psychiatric Press.

Costa, P. T., Jr., & McCrae, R. R. (1988). Personality in adulthood: Asix-year longitudinal study of self-reports and spouse ratings on the NEOPersonality Inventory. Journal of Personality and Social Psychol-ogy, 54, 853-863.

Costa, P. T., Jr., & McCrae, R. R. (1992). Revised NEO PersonalityInventory (NEO-P1-K) and NEO Five-Factor Inventory (NEO-FFl) pro-fessional manual. Odessa, FL: Psychological Assessment Resources.

Costa, P. T., Jr., & McCrae, R. R. (1994). Set like plaster? Evidence for thestability of adult personality. In T. Heatherton & J. L. Weinberger (Eds.),Can personality change? (pp. 21-40). Washington, DC: American Psy-chological Association.

Costa, P., & McCrae, R. (1995). Domains and facets: Hierarchical person-ality assessment using the Revised NEO Personality Inventory. Journalof Personality Assessment, 64, 21-50.

Depue, R. A. (1996). A neurobiological framework for the structure of

personality and emotion: Implications for personality disorders. In J. F.Clarkin & M. F. Lenzenweger (Eds.), Major theories of personalitydisorder (pp. 347-390). New York: Guilford Press.

De Raad, B., Perugini, M., Hrebickova, M., & Szarota, P. (1998). Linguafranca of personality: Taxonomies and structures based on the psy-cholexical approach. Journal of Cross-Cultural Psychology, 29, 212—232.

Digman, J. M. (1990). Personality structure: Emergence of the five-factormodel. Annual Review of Psychology, 41, 417-440.

Dyce, J. A., & O'Connor, B. P. (1998). Personality disorders and thefive-factor model: A test of facet-level predictions. Journal of Person-ality Disorders, 12, 31-45.

Goldberg, L. R. (1992). The development of markers of the Big Five factorstructure. Psychological Assessment, 4, 26-42.

Gosling, S. D., & John, O. P. (1999). Personality dimensions in nonhumananimals: A cross-species review. Current Directions in PsychologicalScience, 8, 69-75.

Gunderson, J. G. (1992). Diagnostic controversies. In A. Tasman & M. B.Riba (Eds.), Review of psychiatry (Vol. 11, pp. 9-24). Washington, DC:American Psychiatric Press.

Gunderson, J. G., Links, P. S., & Reich, J. H. (1991). Competing modelsof personality disorders. Journal of Personality Disorders, 5, 60-68.

Gunderson, J. G., Ronningstam, E., & Smith, L. (1991). Narcissisticpersonality disorder: A review of data on DSM—llI-R descriptions.Journal of Personality Disorders, 5, 167—177.

Halverson, C. F., Kohnstamm, G. A., & Martin, R. P. (Eds.). (1994). Thedeveloping structure of temperament and personality from infancy toadulthood. Hillsdale, NJ: Erlbaum.

Hare, R. D. (1991). The Revised Psychopathy Checklist. Toronto, Canada:Multi-Health Systems.

James, L. R., Demaree, R. G., & Wolf, G. (1993). rwg: An assessment ofwithin-group interrater agreement. Journal of Applied Psychology, 78,306-309.

Jang, K. L., McCrae, R. R., Angleitner, A., Riemann, R., & Livesley, W. J.(1998). Heritability of facet-level traits in a cross-cultural twin sample:Support for a hierarchical model. Journal of Personality and SocialPsychology, 74, 1556-1565.

John, O. P., & Srivastava, S. (1999). The Big Five trait taxonomy: History,measurement, and theoretical perspectives. In L. A. Pervin & O. P. John(Eds.), Handbook of personality: Theory and research (2nd ed., pp.102-138). New York: Guilford Press.

Lilienfeld, S. O. (1994). Conceptual problems in the assessment of psy-chopathy. Clinical Psychology Review, 14, 17-38.

Lilienfeld, S. O., Waldman, I. D., & Israel, A. C. (1994). A criticalexamination of the use of the term and concept of comorbidity inpsychopathology research. Clinical Psychology: Science and Prac-tice, 1, 71-83.

Livesley, W. J. (1998). Suggestions for a framework for an empiricallybased classification of personality disorder. Canadian Journal of Psy-chiatry, 43, 137-147.

Livesley, W. J. (Ed.). (2001). Handbook of personality disorders: Theory,research, and treatment. New York: Guilford Press.

Lykken, D. T. (1995). The antisocial personalities. Hillsdale, NJ: Erlbaum.Lynam, D. R. (in press). Psychopathy from the perspective of the five-

factor model. In P. T. Costa Jr. & T. A. Widiger (Eds.), Personalitydisorders and the five-factor model of personality (2nd ed.). Washing-ton, DC: American Psychological Association.

McCrae, R. R., & Costa, P. T., Jr. (1990). Personality in adulthood. NewYork: Guilford Press.

Miller, J. D., Lynam, D. R., Widiger, T. A., & Leukefeld, C. (2001).Personality disorders as extreme variants of common personality dimen-sions: Can the five-factor model adequately represent psychopathy?Journal of Personality, 69, 253-276.

412 LYNAM AND WIDIGER

Millon, T., Millon, C., & Davis, R. (1994). MCM1-1I1 manual. Minneap-olis, MN: National Computer Systems.

Mineka, S., Watson, D., & Clark, L. E. A. (1998). Comorbidity of anxietyand unipolar mood disorders. Annual Review of Psychology, 49, 377-412.

Morey, L. C., Gunderson, J., Quigley, B. D., & Lyons, M. (2000). Dimen-sions and categories: The "Big Five" factors and the DSM personalitydisorders. Assessment, 7, 203-216.

O'Connor, B. P., & Dyce, J. A. (1998). A test of models of personalitydisorder configuration. Journal of Abnormal Psychology, 107, 3-16.

Oldham, J. M., Skodol, A. E., Kellman, H. D., Hyler, S. E., Rosnick, L., &Davies, M. (1992). Diagnosis of DSM-1II—R personality disorders bytwo semistructured interviews: Patterns of comorbidity. American Jour-nal of Psychiatry, 149, 213-220.

Plomin, R., & Caspi, A. (1999). Behavioral genetics and personality. InL. A. Pervin & O. P. John (Eds.), Handbook of personality: Theory andresearch (2nd ed., pp. 251-276). New York: Guilford Press.

Raine, A., Lencz, T., & Mednick, S. (Eds.). (1995). Schizotypalpersonalitydisorder. Cambridge, England: Cambridge University Press.

Ronningstam, E. (Ed.). (1998). Disorders of narcissism: Diagnostic, clin-ical and empirical implications. Washington, DC: American PsychiatricPress.

Ronningstam, E., & Gunderson, J. G. (1990). Identifying criteria fornarcissistic personality disorder. American Journal of Psychiatry, 147,585-601.

Rutter, M., Dunn, J., Plomin, R., Simonoff, E., Pickles, A., Maughan, B.,Ormel, J., Meyer, J., & Eaves, L. (1997). Integrating nature and nurture:Implications of person-environment correlations and interactions fordevelopmental psychopathology. Development and Psychopathology, 9,335-364.

Saucier, G. (1994). Mini-markers: A brief version of Goldberg's unipolarBig Five markers. Journal of Personality Assessment, 63, 506-516.

Scarr, S., & McCartney, K. (1983). How people make their own environ-ments: A theory of genotype —> environment effects. Child Develop-ment, 54, 424-435.

Siever, L. J., & Davis, K. L. (1991). A psychobiological perspective on thepersonality disorders. American Journal of Psychiatry, 148, 1647-1658.

Skodol, A. E., Rosnick, L., Kellman, H. D., Oldham, J. M., & Hyler, S. E.(1991). Development of a procedure for validating structured assess-ments of Axis II. In J. Oldham (Ed.), Personality disorders: Newperspectives on diagnostic validity (pp. 41-70). Washington, DC: Amer-ican Psychiatric Press.

Stoff, D., Breiling, J., & Maser, J. D. (Eds.). (1997). Handbook of antiso-cial behavior. New York: Wiley.

Trull, T. J. (1992). DSM-III-R personality disorders and the five-factormodel of personality: An empirical comparison. Journal of AbnormalPsychology, 101, 553-560.

Trull, T. J., Widiger, T. A., & Burr, R. (2001). A structured interview forthe assessment of the five-factor model of personality: 2. Facet-levelrelations to the Axis II personality disorders. Journal of Personality, 69,175-198.

Watson, D., Clark, L. A., & Harkness, A. R. (1994). Structures of person-ality and their revelance to psychopathology. Journal of AbnormalPsychology, 103, 18-31.

Westen, D. (1995). A clinical-empirical model of personality. Life afterthe Mischelian ice age and the NEO-Iithic era. Journal of Personal-ity, 63, 495-524.

Widiger, T. A. (1993). The DSM-III-R categorical personality disorderdiagnoses: A critique and an alternative. Psychological Inquiry, 4, 75-90.

Widiger, T. A., & Costa, P. T. (1994). Personality and personality disor-ders. Journal of Abnormal Psychology, 103, 78-91.

Widiger, T. A., & Costa, P. T. Jr. (in press). Five-factor model personalitydisorder research. In P. T. Costa Jr. & T. A. Widiger (Eds.), Personalitydisorders and the five-factor model of personality (2nd ed.). Washing-ton, DC: American Psychological Association.

Widiger, T. A., & Frances, A. J. (1994). Toward a dimensional model forthe personality disorders. In P. Costa Jr. & T. Widiger (Eds.), Person-ality disorders and the five-factor model of personality (pp. 19-39).Washington, DC: American Psychological Association.

Widiger, T., Frances, A., Harris, M., Jacobsberg, L., Fyer, M., & Manning,D. (1991). Comorbidity among Axis II disorders. In J. Oldham (Ed.),Axis II: New perspectives on validity (pp. 163-194). Washington, DC:American Psychiatric Press.

Widiger, T. A., Frances, A. J., Spitzer, R. L., & Williams, J. B. W. (1988).The DSM-III-R personality disorders: An overview. American Journalof Psychiatry, 145, 786-795.

Widiger, T. A., & Lynam, D. R. (1998). Psychopathy from the perspectiveof the five-factor model of personality. In T. Millon, E. Simonsen, M.Birket-Smith, & R. D. Davis (Eds.), Psychopathy: Antisocial, criminal,and violent behaviors (pp. 171-187). New York: Guilford Press.

Widiger, T. A., & Trull, T. J. (1998). Performance characteristics of theDSM-IH-R personality disorder criteria sets. In T. A. Widiger, A. J.Frances, H. A. Pincus, R. Ross, M. B. First, W. Davis, & M. Kline(Eds.), DSM-IV sourcebook (Vol. 4, pp. 357-373). Washington, DC:American Psychiatric Association.

Widiger, T. A., Trull, T. J., Clarkin, J. F., Sanderson, C., & Costa, P. T. Jr.(1994). A description of the DSM-III-R and DSM-IV personality dis-orders with the five-factor model of personality. In P. T. Costa Jr. &T. A. Widiger (Eds.), Personality disorders and the five-factor model ofpersonality (pp. 41-56). Washington, DC: American Psychological As-sociation.

Wiggins, J. S., & Pincus, A. L. (1992). Personality: Structure and assess-ment. Annual Review of Psychology, 43, 473—504.

Zuckerman, M. (1994). Behavioral expressions and biosocial bases ofsensation seeking. Cambridge, England: Cambridge University Press.

Received August 24, 2000Revision received January 10, 2001

Accepted February 25, 2001