wiggins_1979

Upload: ikin-nora

Post on 09-Apr-2018

215 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/7/2019 Wiggins_1979

    1/18

    Journal of Personality and Social Psychology1979, Vol. 37, No. 3, 395-412

    A Psychological Taxonomy of Trait-Descriptive Terms:The Interpersonal DomainJerry S. WigginsUniversity of British Columbia, Vancouver, Canada

    The eventual aim of the research reported is the development of a comprehen-sive taxonomy of trait-descrip tive terms in the English language. Building onearlier work of Allport, Norman, and Goldberg, preliminary a priori distinctionswere made among different domains of trait categories. General procedures fo rdeveloping structured taxonomies within domains were i l lu strated w ith referenceto the interp ers onal domain. Theoretical considerations dic tated th e definitionof th e universe of content, th e choice of measurement model, and the proce-dures for classifying terms within th e domain. Eight adjectival scales were de-veloped as markers of the principal vectors of the interpersonal domain. Thesubstantive, structural, and psychometric characteristics of these scales werefound to be highly satisfactory. Hence, they may prove useful both as assess-ment devices in their ow n r igh t and as reference points for the classification ofvariables in personality and social psychology.

    Personality is that branch of psychologywhich is concerned with providing a syste-matic account of the ways in which individu-al s d i f f e r from one another (Wiggins, Ren-n e r , Clore, & Rose, 1971 ) . Individuals d i f f e rfrom one another in a variety of ways: theiranatomical and physiognomic characteristics;their personal appearance, grooming, andmanner of dress; their social backgrounds,roles, and other demographic characteristics;their effect on others or social stimulus value;and at any given moment in time, theirtemporary states, moods, attitudes, and ac-tivities. But the principal goal of personalitystudy is to provide a systematic account ofindividual differences in human tendencies

    The early stages of this research were supported byGrant MH 12972 from the National Institute ofMental Health, U.S. Public Health Service (LewisR. Goldberg, principal investigator). More recentwork w as suppo r t ed by Canada Council ResearchGrant S75-0109. A n earlier version of this articlewas presented at the meeting of the Society ofMultivariate Experimental Psychology, GlenedenBeach , Oregon, November 1975.Requests fo r reprints should be sent to Jerry S.Wiggins, Department of Psychology, University ofBrit ish Columbia, Vancouver, Brit ish Columbia,Canada V 6T 1WS.

    (proclivities, propensities, dispositions, incli-nations) to act or not to act in certain wayson certain occasions.Although such tendencies are commonlycalled "traits," the use of that term in con-temporary psychological discourse carries withit implications of a pa rticu lar theoretical com-mitment, a preferred method of scientific in -vestigation, and a philosophical preference forcertain kinds of explanation in theory con-struction. Hence, it is necessary to make itclear at the outset that an interest in humantendencies (trai ts) does not imply a theo-retical precommitment to such issues aswhether traits are manifestations of genera-tive or causal mechanisms (Allport, 1 9 3 7 ) ;whether t rai t at tributions reflect specific cog-nitive processes of observers (Heider, 19S8) ;whether trai ts are best construed idiographi-cally or nomothetically (Allport, 1937; Bern& Allen, 1974; Kelly, 1955 ) ; or whether con-sistencies in human tendencies are largely dueto env ironm ental or situational constancies(Mischel, 1968).In my view, consistent patterns of humanconduct consti tute the basic data of person-ality study, which require rather than provideexplanation (Wiggins, Note 1). In approach-

    Copyright 1979 by the American Psychological Association, Inc. 0022-3514/79/3703-0395$00.75395

  • 8/7/2019 Wiggins_1979

    2/18

    396 J E R R Y S . W I G G I N Sing this task of accounting for individualdifferences in human tendencies, I share withothers th e conviction that th e natural lan-guage provides a convenient starting place(Allport, 1937; Cattel l , 1957; Goldberg, Note2 ; Norman, Note 3 ) . The universe of contentof human tendencies is contained within th ecovers of an unabridged dictionary of theEnglish language. I also share with thesewriters the conviction that an adequate tax-onomy of trait-descriptive terms must pre-cede meaningful empirical studies of humantendencies. There have been a num be r ofsystematic efforts to develop a personalitytaxonomy over the past 40 years, and my ownwork has attempted to capitalize on theseearlier efforts.Allport and Odbert (1936) examined theapproximately half million separate entries orderivatives included in Webster's New Inter-national Dictionary ( 1 9 2 4 ) for terms thatappeared "to distinguish the behavior of onehuman being from that of another" (p . 24)and identified a pool of 17,953 terms havingthis characterist ic. Norman (Note 3 ) scannedth e entire contents of Webster's Third N ewInternational Dictionary Unabridged (1961)f o r additional terms that had not been in-cluded in the Allport-Odbert list. The u niverseof content thus denned was approximately27 ,000 terms. Subsequently, Norman andGoldberg were able to reduce this list byeliminating obscure, inappropriate, and ar-chaic terms. Th e focus of all of these investi-gations has been on a subset of approximately3,600 terms that Allport initially called "sta-ble biophysical traits." Allport consideredthese to be "real" traits as opposed to tem-porary states, moods, social roles, physicalcharacterist ics, and so for th.However one conceives of stable biophysicaltraits, I see them as no more nor less realthan any other kinds of human character-istics. I agree with Guilford (1959) tha t "atrait is any distinguishable, relatively en-during way in which one individual differsfrom others" (p. 6). Bu t I se e the majortaxonomic task as that of specifying th e di f -ferent kinds of ways in which individualsdiffer from each other. One kind of way inwhich individuals differ from each other is interms of what they do to each other. Thus, I

    th ink that one of the most important cate-gories of traits is that which may be desig-nated "interpersonal." Within th e realm ofthings that people do to each other, it isdesirable to make a fur ther theoretical dis-tinction between interpersonal exchangesbased on love and status and interpersonalexchanges based on goods, money, and ser-vices (Foa & Foa, 1 9 7 4 ) . The former ar ecalled interpersonal traits and the lat ter ma-terial traits.Another equally "real" way in which indi-viduals are distinguishable from one anotheris in terms of their styles of emotional reac-t ivity, which we refer to as "temperament."It is also possible to distinguish individualson th e basis of the particular roles and statusthey hold within th e framework of our socialinstitutions. Additionally, there are "charac-ter" terms that represent appraisals of anindividual based on a code of proper behav-ior. And finally, there are many words thatrefer to individual differences in qualities ofmind as manifested in thought, perception,and speech.

    Consider the following t rai t-descriptive ad -ject ives: aggressive, miserly, lively, ceremoni-ous, dishonest, and analytical. I would classifythese terms as representative of the six traitcategories just mentioned, namely, interper-sonal traits, material traits, temperamentaltraits, social roles, character, and mentalpredicates. Here I d i f f e r from Allport, Nor-man, and Goldberg, who view all of theseterms as "stable biophysical traits," withoutattempting to differentiate the different kindsof descriptive jobs such terms perform for us.To these earlier authors' important distinc-tion between stable traits and temporarystates, moods, social evaluations, and so forth,I would add finer distinctions within th e cate-gory of stable traits.The present article describes our first ef-forts to develop a taxonomy of personalitytrait-descriptive terms in the English lan-guage. The universe of content is taken to beNorman's (Note 3) total lexicon of 18,125terms. Within this broader framework, I in-tend to focus on the "prime" categories ofNorman, which involve 4,063 relatively fa -miliar and nonobscure terms. Even restrictingou r attention to prime terms, we are still

  • 8/7/2019 Wiggins_1979

    3/18

    TA X O NO MY O F INTE RP E RSO NA L TRAITS 397

    STATUS[ [ LOVE |STATUS] JSTATUS| | LOVE | LOVE [ [STATUS

    Facet IDIRECTION

    Face t IIOBJECT

    Facet IIIRESOURCEINTERPERSONAL VARIABLESNO ( Gregar i ous -Ex t rove r t ed )PA ( A m b i t i o u s - D o m i n a n t )BC ( Arrogant -Ca lcu la t ing )DE ( C o l d - Q u a r r e l s o m e )F G ( A l o o f - I n t r o v e r t e d ) -H I ( L a z y - S u b m i s s i v e ) + - - -JK (Un assumin g - In gen uous ) + + - -LM ( W a r m - A g re e a b l e ) + + +

    Figure 1. Facet composition of interpersonal variables (after Foa & Foa, 1974). (+= acceptance; = rejection.)

    +-t-+

    confronted with a list of staggering size. Ourenterprise required a clear-cut game plan orresearch strategy to guide us through this seaof words.We began by defining a limited taxonomyon an a priori basis. W e attempted to specifydifferent domains of human characteristicsbased on the different descriptive jobs per-formed by the words wi thin th e domains.Next, w e attempted preliminary a priori classi-fications within domains, mainly as a meanso f keeping track of terms and l ightening ou rclerical burdens. The overall strategy fordeveloping a taxonomy within a single domainmay be honorifically described as "iterative,"although "tr ial and error" may be a more aptdescriptor. Our main concern was the avoid-ance of premature "fixing" of a taxonomy.Hence, all preliminary classifications weretentative in nature and subject to continualrevision in the light of other developing cate-gories. This rather complex interactive strat-egy will be illustrated with reference to thedevelopment of a taxonomy within the do-main of interpersonal traits.

    The Domain of Interpersonal TraitsThe line of reasoning to be presented origi-nated in the interpersonal theory of Sullivan( 1953 ) and was later operationalized in a set

    o f personality measurement procedures byLeary ( 1 9 5 7 ) and his colleagues. Foa (1961,1965) integrated this theory with Guttman's(1954) order analysis, and the general orien-tation w as extended to domains other thanth e interpersonal by Rinn (1965). Notableinterpersonal systems based on a circumplexmodel have also been d escribed by Becker andKrug (1964) , Benjamin (1974) , Lor r andM cNair (19 63) , Schaefer (19 59 ), and Stern( 1 9 7 0 ) . This general framework w as inte-grated with interpersonal exchange theory(Homans, 1961) by Carson (1969) and byFoa and Foa ( 1 9 7 4 ) .In their most recent theoretical statement,Foa and Foa (1974 ) describe th e developmentof cognitive categories of social perception asa progressive differentiation of structure in -volving facets of directionality, object, andresource. A somewhat simplified version ofFoa and Foa's representation of this structureis illustrated at the top of Figure 1. The earli-

  • 8/7/2019 Wiggins_1979

    4/18

    398 J E R R Y S. W IGGINSest cognitive schemata are based on the dis-crimination of the directionality of socialevents (give vs. take, accept vs. re ject ) . Withth e acquisition of the concept of social object(self vs. other ) four categories of social mean-ing are discriminated (e.g., give to self, takeaway from othe r ) . Out of an initially undif-ferentiated matrix of resource classes, servicesar e differentiated from love, and the latter isfurther differentiated into love and status.With the distinction between love and status,earlier facets become reorganized, and it ispossible to distinguish eight features of socialmeaning (e.g., granting status to oneself,denying love to another) . These eight featuresmay be thought of as part of a semantic codestrip that provides th e basic discriminationsfor encoding and decoding interpersonal events(Osgood, 1970).Within the above context, interpersonalevents may be denned as dyadic interactionsthat have relatively clear-cut social (status)and emotional (love) consequences fo r bothparticipants ( s e l f and other). This definitionprovides a theoretical basis fo r distinguishinginterpersonal traits from other categories oftrait descriptors, such as temperament, moods,cognitive traits, and physical characteristics.Under th e semantic features in Figure 1are listed eight theoretical interpersonal vari-ables. Th e organization of these variables isthought to be determined by an interrelatedset of societal rules that impart meaning tosocial events (Wiggins, Note 1 ) . Thus, ac -tions that have the same profile of semanticfeatures are categorized as belonging to thesame response class. The semantic fe ature s ofeach of the eight variables appear as rows inFigure 1. Note that the first variable (N O )is coded on all the positive (accept) categoriesfor self and other with respect to both loveand status. The first variable is in markedcontrast to FG for which all values are nega-tive ( re j ec t ) . Variables NO and FG have nofeatures in common, but since acceptance andrejection are logically opposed concepts, itwould be expected that the two variableswould be strongly negatively correlated. Simi-lar relationships can be seen to exist betweenP A and HI, BC and JK , and DE and L M . Notealso that each variable differs from its ad-

    jacent variable by only one element. This isalso true of the first ( N O ) and l as t (LM)variables, and hence the structural relationsamong this set of variables can be representedas a ci rcumplex (G ut tman, 1954) .The labels that have been attached to theinterpersonal variables in Figure 1 (gregari-ous-extraverted, ambitious-dominant, etc.)are meant to capture the flavor of terms thatshare the same profile of semantic featuresand may serve more as tags th an as definitions.Thus, fo r example, th e rather inelegant labelo f "lazy-submissive" is attached to interper-sonal transactions involving incompetence,passive resistance, submission, or obedience.These otherwise diverse attributes share incommon th e semantic features of denyingstatus to self, denying love to both self andother, and granting status to other. A fullerlisting of representative terms in each cate-gory may be found in Table 2 , later in thisarticle.Were we to collect personality measure-ments on the eight variables listed in Figure1, their intercorrelations would, in theory,show the pattern illustrated in Table 1. Thecorrelation of a variable with itself is as-sumed to be unity, so that the principal di-agonal of this matrix contains ones. The cor-relations along this main diagonal are largeand positive, and they decrease across suc-cessive minor diagonals to the w/2nd variable,where they are a minimum. The correlationsthen increase up to a large positive value inth e off-diagonal matrix. Th e circumplexity ofthis or any other set of variables can beevaluated directly from th e intercorrelationmatrix. A rigorous procedu re for assessing th ecircumplexity of a set of variables involvesplotting th e correlations of each variable(ordinate) with the other variables in se-quence (abscissa). This procedure should gen-erate a series of overlapping sine curves thatcan be evaluated for goodness of fit (Stern,1970) .

    An alternative procedure fo r evaluating cir-cumplexity is to extract the first two principalcomponents from th e matrix of intercorrela-tions and to examine th e plot of the variableson the two components. Figure 2 presents th eplot of the eight variables on two principal

  • 8/7/2019 Wiggins_1979

    5/18

    TA X O NO M Y OF INTERPERSONAL TRAITS 399Table 1Hypothetical Correlations Among Eight Interpersonal Variables

    Variable PA B C DE FG H I JK L M NOPAB CDEFGHIJKL MN O

    1.00.50.00-.50-1.00-.50.00.50

    1.00.50.00-.50-1.00-.50.00

    1.00.50.00-.50-1.00-.50

    1.00.50.00-.50-1.001.00.50.00-.50

    1.00.50.00 1.00.50 1.00Note. PA = ambit ious-dominant ; B C = arrogant-calculat ing; D E = cold-quarrelsome; FG = aloof-introverted; H I = lazy-submissive; JK = unassuming-ingenuous; L M = warm-agreeable; N O = gre-garious-extraverted.components extracted from th e intercorrela-tions in Table 1. As can be seen from thisfigure, the intercorrelations among variablesin Table 1 form a perfec t, equally spaced cir-cumplex. This pattern follows, both theoreti-cally and empirically, from th e pattern ofshared semantic features among th e variablesillustrated at the b ottom of Figure 1. Variablesthat share three features in common are ad-jacent to each other on the circle; variablesthat have no features in common are oppositeeach other.

    There are two distinct advantages to therepresentation of interpersonal variables by atwo-dimensional circumplex. The first is thatit provides an explicit conceptual definitionof the universe of content of interpersonal be-havior. Any behavior that meets the defini-tion of a meaningful interpersonal event givenabove m ust be capable of being represented asa vector originating from th e center of thecircle. Thus, the specific system proposedhere is potentially falsifiable. The second ad -vantage of the circumplex model is that it

    AMBITIOUSDOMINANT

    ( P A )oA R R O G A N T -C A L C U L A T I N G 0( B C )

    C O L D -Q U A R R E L S O M E o( D E )

    A L O O F -INTROVERTED o( F G )

    G R E G A R I O U S -o E X T R A V E R T E D( N O )

    W A R M -o AGREEABLE( L M )

    U N A S S U M I N Go INGENUOUS( J K )oL A Z Y -SUBMISSIVE( H I )

    Figure 2. Perfect circumplex of interpersonal variables.

  • 8/7/2019 Wiggins_1979

    6/18

    400 J ERRY S. WIGGINSalerts th e investigator to noticeable "gaps" inthe interpersonal space of a given set of vari-ables. Strictly empirical procedures of vari-able selection are likely to deemphasize th eimportance of certain variables that are im-plied by the logic of the circumplex systembut that are underrepresented in the Englishlanguage. Although th e system of interper-sonal variables under discussion is l imited toeight variables, it could in principle be equallywell represented by 16, 32 , or 64 variables.The thinness with which w e slice th e c i rcum-plex pie is l imited by the reliabili ty withwhich respondents can distinguish betweenclosely synonymous words or phrases.It should be evident that the precedingtheoretical considerations substantially con-strain the final form that a taxonomy of in-terpersonal traits may assume. The definitionof interpersonal events, th e specification ofthe i r underlying facet structure, and the selec-tion of a measurement model to represent re-lationships among variables all express acommitment to a particular, albeit widelyshared, conceptualization of the domain underinvestigation. In this sense, th e eventual tax-onomy of trait-descriptive terms will be a"psychological" taxonomy rather than astrictly "semantic" taxonomy based on dic-tionary definitions. It is assumed that th esemantic structures underlying social percep-tion in this culture cannot be inferred in anyobvious w ay from dictionary definitions. In -vestigators who start with different assump-tions and who ut i l ize different measurementmodels would undoubtedly devise somewhatdifferent taxonomies. Structural relationshipsof the kind at issue here are not "discovered"(Loevinger, 195 7 ) . They are postulated andthen evaluated for goodness of fit.

    Development of Interpersonal ClustersRational Categorization

    In a prel iminary attempt to make th e dis-t inction among kinds of trait terms discussedabove, I classified all of the trait descriptorsin Goldberg 's (Note 2 ) pool of 1,710 adjec-tives into one or another of seven a prioricategories. A t that stage in the development

    of the taxonomy, the principal distinctionswere between interpersonal trait terms, tem-peramental trait terms, and mental predicateterms. Tentative categories of "attitudinalterms" and "constancy" were employed at thispoint, as well as a "miscellaneous" categorythat involved various dimensions that couldnot be clearly classified into other categories.In the initial sorting of the 1,710 adjectives,approximately 800 terms were identified as"interpersonal" under our working definitionof interpersonal traits .For purposes of preliminary categorization,we selected Leary's ( 195 7 ) system of in ter-personal traits, because it appeared to be themost explicit system described in the litera-ture , an entire book b eing devoted to the topic.Tw o colleagues 1 and I thoroughly familiarizedourselves with th e system and its theoreticalframework. Work ing as a team, w e consid-ered each of the approximately 800 termspreviously classified as interpersonal and at-tempted to classify each within one of the 16interpersonal vectors. Adjectives that couldnot be classified within the 16 dimensions andthat did not appear to belong in one of theother five a priori categories were temporarilyset aside for fu r the r analysis. There were re-mark ably few such adjectives. W ith consid-erable effort we were able to distribute 567adjectives across th e 16 categories with unani-mous agreement among three raters.Selection oj Preliminary Markers

    Social desirability scale values for all ad-jectives were available from a previous studyby Norman (Note 2) , and mean self-endorse-ment frequencies were available from a morerecent study by Goldberg (Note 2 ) . On thebasis of these itemmetric data, i tems wereselected to serve as preliminary nuclearclusters within each of the 16 interpersonalcategories. These items were selected in sucha way as to represent th e category unambigu-ously across th e range of endorsement fre-quencies an d desirabili ty values. Approxi-11 am grateful to James M. Kilkowski and Alex-ander Galvin fo r their help an d suppor t in thisenterprise.

  • 8/7/2019 Wiggins_1979

    7/18

    T A X O N O M Y O F INTERPERSONAL TRAITS 401mately 12 items were selected for each cate-gory.In Goldberg's sample of 70 male and 117female University of Oregon undergraduates,self-ratings were obtained on a 9-place scalefor each of the 1,710 trait descriptors. UsingGoldberg's data, ou r next step was to obtainthe intercorrelations among terms selected tobe preliminary markers of each of the cate-gories of interpersonal behavior. By a cluster-ing procedure, a smaller and more homoge-neous subset of i tems was selected within eachcategory, with approximately six i tems percluster.

    The next step was to consider al l i temsthat had been rationally classified as fallingwithin the 16 categories as potential candi-dates for addition to or deletion from eachnuclear interpersonal cluster. For this pur-pose, we scored the p rel im inar y clusters as 16scales and examined the correlations of indi-vidual items with these scales with an eyetoward their circumplex patterning. Thus, w eexamined th e correlation of each of the 567i tems wi th the 16 prelim inary clusters orderedby their hypothetical position in the circum-plex. Items were sought that had positive cor-relations with adjacent clusters, zero correla-tions with orthogonal clusters, an d negativecorrelations with opposite clusters. For mostitems, this circumplex patterning across 16clusters was far from perfect . Item selectionw as fur ther complicated by the fact that someo f the initial nucle ar clusters w ere clearly outo f place on the circumplex, so that we con-stantly had to keep in mind th e inadequaciesof ou r original clusters. By this bootstrapprocedure, we selected a set of eight adjectivesfor each of the 16 vectors that, we hoped,would increase the homogeneity of their vectorand correct the previous shortcomings of thatvector.W e assembled the 128 adjectives chosen tomark the 16 vectors into a test for mat andadministered it to a small group of studentsfrom th e Universi ty of British Columbia. Thestudents were requested to rate the self-ap-plicability of the adjectives on a 9-place Likertscale. Item responses were summed for bothoctant an d sixteenth variables. Intercorrela-tions among the variables were factored by

    th e method of principal components. Figure3 displays th e pattern of octant variablesloading on the first two factors in the Univer -sity of British Columbia sample. Th e patternof loadings for 16 interpersonal variables isconsiderably less orderly than this, due nodoubt to the fact that th e vector scores arebased on fewer items in a relatively smallsample of subjects. But the pattern of octantscores clearly displays some of the difficultieswe encountered in our attempt to map outLeary's ( 1 9 5 7 ) system.The most striking feature of Figure 3 isth e lack of variables in the upper r ight-handquadrant . This gap in the Leary (195 7 ) cir-cumplex has been previously noted by severalauthors including Stern (1970), w ho felt thatthe failure of Octants PA and NO to "close"raises the question of whether the Leary sys-te m is a circum plex. Lo rr and M cNair ( 1965 )noted this gap also and attempted to close itwith additional substantive variables. A n-other notable departure from expectations isth e location of Octant NO, which not onlyfails to appear in the upper quadrant butappears below Octant L M . In a sense, then,we partially succeeded in replicating theLeary system w ith trait-descriptive adjectives,but in so doing we carried over th e faults ofth e system as well.2Development oj Bipolar Clusters

    At this point it occurred to us that a 16-variable circu mple x can be rather easily con-structed from eight genuine bipolar dimen-sions. One of the conceptual difficulties weexperienced in work ing with the Leary sys-tem was a decided lack of bipolarity betweenvectors that appeared opposite each other onthe circle. In particular, the following im-plicitly bipolar contrasts did not make agreat deal of sense: success versus masochism,narcissism ve rsus co nform ity, rebellion versus2

    Juris I . Be rzins (personal communic ation, M arch8, 1977) plot ted th e loadings of the original Leary(1957) Interpersonal Checklist ( ICL ) octants on thefirst two principal components in samples of 685high school students and 1,109 college students. Th eplots fo r both samples ar e indistinguishable fromFigure 3. (See also Rinn, 1965, p. 458, Figure 6.)

  • 8/7/2019 Wiggins_1979

    8/18

    402 JERRY S. WIGGINSs u c cPOV( P

    NARCISSISM -E X P L O I T A T I O N o( B C )

    PUNISHMENT -H O S T I L I T Y( DE )

    R E B E LL I ON -DISTRUST( F G )

    ; E S S -VERA ))

    CONFORMITYo T R U S T

    ( JK)

    COLLABORATION LOVE0(LM)TENDERNESS -GENEROSITY( N O )

    M A S O C H I S M -W E A K N E S S( H I )

    Figure 3. Rotated components of Leary (1957) interpersonal variables.

    tenderness, distrust versus generosity, andpunishment versus collaboration.In Leary's ( 1 9 5 7 ) original system the ten-derness-generosity octant (NO) w as concep-tualized as the bipolar contrast to the rebel-l ious-distrustful octant ( F G ) . This likely ac-counts for the noticeable gap in the upperright quadrant of the system (Figure 3 ).Tenderness and generosity are not bipolarcontrasts to rebellious an d dis t rus t fu l behav-ior. Tenderness and generosity are too weakan d loving to be placed this high on the circle.As Lorr an d McNair ( 1965 ) noted earlier ,the NO octant, which falls between d ominanceand love, reflects a socially exhibitionisticstyle of behaving (gregarious-extraverted).With a little effort, our revised representa-tion of the 16 interpersonal variables can beread from Figure 2 . Each variable representsa bipolar contrast to the variable appearingopposite it on the circle. The contrasts ar edominant-submissive, arrogant-unassuming,calculating-ingenuous, cold-warm, quarrel-some-agreeable, aloof-gregarious, introverted-extraverted, and ambitious-lazy.In developing our bipolar clusters, we at-tempted to select items that were highly nega-tively correlated with their opposite clusteran d that had zero correlations with their

    theoretically orthogonal clusters. Thus, agood dominant item should have a high posi-t ive correlation with a dominant cluster , ahigh negative correlation with a submissivecluster, and essentially zero correlations withquarrelsome and agreeable clusters. We useda tenta t ive set of 16 four-item bipolar clustersas markers to select items having the prop-erties just described. This enabled us to re-vise an d expand our tentative clusters intoeight-item variables. Factor analysis of theseeight-item variables in Goldberg's (Note 2 )sample of 187 subjects revealed the clearestcircumplex structure we had yet encounteredin our own work or in the l i terature. This re-sult was replicated in an additional sampleof 119 subjects from the Universi ty of BritishColumbia.Development of Final Scales

    The items in our eight-item bipolar inter-personal clusters were developed from th epool of 567 adjectives classified as interper-sonal in our initial rational sorting of thetotal list of 1,710 adjectives. The possibilityclearly existed that some adjectives fromamong th e 1,143 no t considered would beappropriately placed within the 16 interper-

  • 8/7/2019 Wiggins_1979

    9/18

    TAXONOMY OF INTERPERSONAL TRAITS 403sonal categories. Consequently, a programwas writ ten to examine th e relationships be-tween all 1,710 adjectives and the 16 bipolarinterpersonal clusters. The program correlateseach adjective w ith the 16 bipolar clusters,selects the cluster with which the adjectiveis most highly correlated, and then prints outan ordered list of variables within each of the16 categories. This listing provided the cor-relation of each item with th e cluster withwhich it was most highly correlated. It alsoprovided the correlation with the bipolar op-posite cluster and with the two orthogonalclusters. This procedure produced 16 liststhat located all 1,710 adjectives w ith respectto the interpersonal cluster with which theywere most highly correlated.Two of us then went through these listsand examined th e correlation of each item withthe cluster from the category to which it hadbeen assigned, as well as the correlation ofth e item with it s opposite and orthogonalclusters.3 Items that clearly belonged in theinterpersonal category to which they had beenassigned were retained. Items that did notappear to belong in the interpersonal categoryto which they were assigned were placed inone of eight other taxonomic categories thatthen existed (e.g., temperament, character,material traits, etc .) .A t this point, an attempt w as made to de-velop more refined subcategories within thebroader taxa. Preliminary groupings of clusterswere formed within th e domains of tempera-ment, character, attitude s, mental pred icates,and social roles. This classification added 38new clus ters to the 1 6 interpersonal clusters,making a total of 54 categories within thepreliminary taxonomy. These 54 categorieswere scored as scales by computing, for eachsubject, the sum of his or her ratings to theitems in that category. In the case of the 16interpe rsonal categories, the 2 5 best items , asdetermined from item-cluster correlations,form th e reference scale for each category.

    Th e next stage of analysis was designed todetermine more exactly th e classification ofdoubtful adjectives in the preceding taxonomicsort. This included adjectives that were in -terpersonal but that seemed to be in thewrong interpersonal category, adjectives that

    were unclassifiable in any category, and ad-jectives that were candidates for deletion onth e grounds of ambiguity, obscurity, or lacko f personological relevance. In addition, alladjectives in the interpersonal category otherthan the 400 ( 1 6 X 2 5 ) marker adjectiveswere classified as doubtfu l for the purpose ofthis analysis. There were 768 such adjectivesdesignated as questionable by this criterion.Each of the 768 adjectives was correlated withall of the 54 taxonomic category scales. Tw oo f us then examined this pattern of correla-tions for all 768 words and assigned them toone of the 54 categories on the basis of bothconceptual and empirical (correlational) con-siderations. Some items were retained in theiroriginal interpersonal category, others movedto other categories, and some deleted fromfur ther consideration. By these procedures, arevised 54-category taxonomy was developed.Within this revised taxonomy, 864 adjectiveswe re classified as interpersonal.The selection of items for inclusion in thefinal interpersonal trait scales involved onemore cycle in our iterative procedures The2 S "best" items previously identified in eachof the 16 interpersonal categories were scoredas reference clusters. These reference clustersprovided a broader representation of the con-tent of the final interpersonal variables thandid our earlier clusters. Within each category,each item was correlated with its own 2 5-itemcluster and with its 25-item opposite andorthogonal clusters. The items were thenordered within each category by their cor-relations with their ow n clusters.Working independently, three of us selectedthe "best" eight items in each of the 16 in-terpersonal categories. One rater made selec-tions strictly on the basis of a summary nu -merical index based on item correlations withsame, opposite, and orthogonal clusters, aprocedure referred to as empirical. Anotherrater attended to the same item correlations,but attempted to choose th e more "meaning-ful" adjectives from pairs that had roughlysimilar empirical characteristics, a procedure

    3 I would like to acknowledge th e substantial helpof Ana Holzmuller in this and many other phases ofth e project.

  • 8/7/2019 Wiggins_1979

    10/18

    404 J E R R Y S. W I G G I N Sdesignated quasi-empirical. The third raterselected sets of eight adjectives that were"psychologically cohesive" in terms of hisconception of the constructs under investiga-tion, th e rational procedure.The empirical an d quasi-empirical i tem-selection procedures yielded sets of 16 scaleswith highly similar psychometric properties.For a specific sixteenth, it was possible toidentify one or the other of the two scalesas being preferable, either on grounds of ahigher alpha coefficient or on the grounds ofa better position in the 16-variable circum-plex plot. Consequently, a combination set of8 variables was formed from th e best of theempirical and quasiempirical sixteenths.An additional procedure for item selectionwas based on a more fine-grained analysis ofthe circumplex properties of individual i tems.Each item was correlated with the 16 25-i temclusters that served as markers of the inter-personal variables. This pattern of correla-tions was then inspected for the number ofdepar tures from perfect circumplex orderingthat occurred. I tems with th e smallest num-ber of departures from this pattern were re-tained in a set of scales labeled "item order-ing."The final selection of a 128-item set of in-terpersonal variables was based on an indexof circumplexity that permits comparisonamong competing scale sets (Wiggins & Mar-ston, Note 4 ) . Very briefly, two factors wereextracted from the intercorrelations amonginterpersonal variables in a given scale set. Ahypothetical factor matrix was constructed torepresent a perfect circumplex solution fo rvariables of this level of reliability (as esti-mated from communalities). The sum of theabsolute differences between elements in thehypothetical factor matrix an d elements inth e obtained factor matr ix w as taken as anindex of circumplexity. Each of the five setsof interpersonal scales was then compared onthis index. Both octants an d sixteenths wereevaluated in Goldberg's (Note 2 ) sample of187 Universi ty of Oregon und ergraduates andin Norman's sample of 123 University ofWestern Australia undergraduates, A n over-all index of circumplexity, based on both 8-and 16-variable solutions in American and

    Australian samples, suggested that th e com-bination set of 128 adjectives had a slightedge over th e item-ordering, empirical, andquasi-empirical procedures and a substantialadvantage over th e rational procedure, whichfared rather badly. This combination set ofadjectives, which has been employed in allsubsequent investigations, is listed in Table 2 .

    Generalizability of Circumplex StructureThe present taxonomy of the interpersonaldomain is based on an explicit structuralmodel ( G u t t m a n , 1954) that follows from afacet analysis of cognitive categories of so-cial percep tion (Foa & Foa, 1974). On thebasis of both theoretical and psychometricconsiderations, a set of eight 16-item scaleswere developed as marker variables of theprincipal vectors of this system. In additionto being potentially useful personality assess-ment measures in their ow n right, these scalesenable one to classify any interpersonal traitdescriptor by establishing it s location withinth e circumplex space. However, th e mean-

    ingfulness of such classification dependsheavily on the generalizability of the presentcircumplex structure to samples other thanthose involved in the derivation of the eightinterpersonal scales. Th e scales were derivedprimarily from Goldberg's (Note 2 ) sampleo f American students and cross-validated atvarious stages in samples of Canadian stu-dents and in Norman's sample of Australianstudents. Although relatively diverse sampleso f students were employed in scale deriva-t ion, evidence for the generalizability of thecircumplex structure must come from othersources. Four samples of subjec ts were testedf o r this purpose.Cross-Validation Samples

    Sample A . Subjects were recruited by ad-vertisements to participate in a "personal-it y study" that involved tw o separate testingsessions of approximately 2 hours each(Marston, Note 5). Subjects were paid $10each for their participation and were providedindividual feedback if requested. The 12 8 in-terpersonal adjectives were embedded in a

  • 8/7/2019 Wiggins_1979

    11/18

    TA X O NO M Y OF INTERPERSONAL TRAITS 405Table 2Final Set o f Interpersonal Adjective Scales

    P. Ambitious(Success)PerseveringPersistentIndust r iousSelf-disciplinedOrganizedDeliberativeStableSteady

    D. Cold(Hate)WarmthlessUnsympathet icIronheartedUnchari tableColdheartedHardheartedCruelRuthless

    H . Lazy(Fai lure)Unproduct iveLazyUnthoroughUnindus t r iousInconsistentDisorganizedUnbusinesslikeImpractical

    L. Warm(Love)TenderheartedGentlehear tedTenderKindEmotionalSympatheticSoftheartedAppreciative

    A. Dominant(Power)DominantAssertiveForcefulDomineeringFi rmSelf-confidentSelf-assuredUn-self-conscious

    E. Quarrelsome(Hostility)Impol i teUncordialDiscourteousUngraciousDisrespectfulUncooperat iveI l l -manneredUncivi l

    I. Submissive(Weakness)Self-doubtingSelf-effacingTimidMeekUnboldUnaggressiveForcelessUnauthor i t a t ive

    M . Agreeable(Collaboration)CourteousCharitableWell-manneredRespectfulCordialCooperativeAccommodat ingForgiving

    B . Arrogant(Narcissism)BigheadedBoisterousConceitedBoastfulOverforwardSwellheadedCockyF l aun t y

    F. Aloof(Disaffiliation)AntisocialUnneighbor lyImpersonalUnsociableDistantDissocialUnsmil ingUncheery

    J. Unassuming(Modesty)NonegotisticalUndemandingUnva inUnwildUnargumenta t iveBoastlessPretenselessConceitless

    N . Gregarious(Affiliation)FriendlyGenialNeighborlyCompanionableApproachableCongenialGood-naturedPleasant

    C. Calculating(Exploitation)SlyTrickyWilyCunningOvercunningCraftyCalculat ingExploitative

    G . Introverted(Withdrawal)SilentShyInt rover t edBashfulInwardUnreveal ingUnspark l ingUndemons t ra t ive

    K. Ingenuous(Trust)U ncunn ingUncalculat ingUncraf tyUnwi lyUnslyGuilelessUndeviousUndecept ive

    O. Extraverted(Outgoingness)OutgoingExtravertedVivaciousJovialEnthusiasticChee r fu lPerkyUnshy

    list of 60S adjectives whose order was ran-domized separately for each subject. Fourpersonality inventories were administered invarying orders along with the total list ofadjectives. The 152 subjects (51 men and101 women) included both undergraduate andgraduate students from a variety of academicmajors at the University of British Columbia.Sample B. Representative samples of vol-unteer and professional workers were recruitedfrom three different social service agencies in

    th e Greater Vancouver area (M erri tt , Note6). The 128 interpersonal adjectives were ad -ministered in a single booklet along with avalue survey and a standardized personalityinventory. Testing w as accomplished indi-vidually, in small groups, or on a take-homebasis. One hundred subjects (29 men and 71women) completed the interpersonal adjec-tive form.Sample C. Students in the second term ofan introductory psychology class at the Uni-

  • 8/7/2019 Wiggins_1979

    12/18

    406 JERRY S. WIGGINSSAMPLE A

    PAo o NO

    SAMPLE BP Ao

    BC o

    DE o

    o LM

    o JKFG o oHI

    P Ao

    BC oDE o

    FG o

    o N O

    o JKoH I

    PAoBC o

    DE oFG o

    o NO

    o LM

    BC o

    DE o

    O JK

    O N O

    o LM

    FG o JKoH I OH IS A M P L E C SAMPLE D

    Figure 4. St ruc tu re of f inal interpersonal variables in f o u r samples. (PA = amb i t iou s -d ominan t ;BC = ar rogan t -c a l cu l a t ing ; DE co ld -quar re l some ; FG =a loof - in t rove r t ed ; HI = lazy-submis-sive ; JK = unassuming- ingenuous ; LM warm-agreeable . )versity of British Columbia were administeredthe 12 8 adjectives, embedded in a larger l istof adject ives, dur ing a regular class period.The students were given a guest lecture onpersonali ty test ing in exchange for their par-t i c ipat ion. Data were collected for 132 sub-jects ( 5 7 men and 75 w o m e n ) .Sample D . Summer students enrolled inan introductory psychology class at the Uni-versi ty of Bri t i sh C olumbia were administeredthe 12 8 adjectives , embed ded in a larger l ist ,d u r in g a regular class period. Th e studentslater received their ow n standardized scoreson the interpersonal variables and a lec tureon interpersonal behavior. Data were collectedfor 139 subjec ts (58 men and 81 w o m e n ) .Circumplex Analyses

    Intercorre la t ions were obtained among theeight interpersonal adjective scales in each ofthe four samples just described. Two prin-

    cipal components were then extracted fromeach of the intercorrelation matrices. In therespective analyses, these tw o components ac-counted for 76.1% (Sample A ), 67.9% (Sam-p le B , 72.0% (Sample C), and 74.8% (SampleD ) of the total variance. F or comparabilityof inspection, th e second component of eachsolut ion w as hand rotated to pass t h roughth e ambitious-dominant vector ( P A ) . Figure4 depic ts th e loadings of the eight interper-sonal scales on two hand -rotate d princip alcomponents in each of four samples.Perfect , evenly spaced circumplexes (Fig-ure 2 ) are not expected in real data, becauseof measurement error, an d hence th e struc-tures in Figure 4 are more properly describedas quasi-circumplexes (G u t tm an , 1954 ) . A l-though the amount of "quasiness" one isprepared to tolerate in empirical data is tosome extent a matter of taste, it seems fairto say that the quasi -c i rcumplex st ructures in

  • 8/7/2019 Wiggins_1979

    13/18

    TAXONOMY OF INTERPERSONAL TRAITS 407Table 3Psychometric Characteristics of Interpersonal Adjective Scales

    Men

    labelPABCDEFGHIJKLMNO

    (n =M

    5.874.302 .954.283.954.646.646.17

    236)SD.951.02.891.19.94.93.80.96

    W o m e n(n =M

    5.723.602.483.874.135.147.086.54

    374 )SD1.011.07.801.171.011.00.76.96

    Total(NM

    5.793.872.664.034.064.956.916.40

    = 610)SD.991.05.841.17.99.97.7 7.96

    Internal consistency(coefficient alpha)Social desirabilityrating(N

    M6.773.662.313.543.375.737.537.52

    = 100)SD1.10.98.81.7 1.61.95.73.66

    Total(N = 610)

    .855.870.877.891.809.801.865.897

    Range ofvaluesfrom thesamples.831-.880.84S-.885.862-.8S9.887-.S94.769-838.743-.840.841-.879.887-.917

    Note. Scale labels stand for ambitious-dominant ( P A ) ; arrogant-calculating ( B C ) ; cold-quarrelsome (DE); aloof-introverted(FG); lazy-submissive ( H I ) ; unassuming-ingenuous (JK); warm-agreeable (LM); and gregarious-extraverted (NO).ft Scale values range f rom 1 (extremely inaccurate description) to 9 (extremely accurate descr ip t ion) .Figure 4 are among the clearest reported inthe personality literature to date. The cir-cumplex ordering obtained in scale derivationsamples is generalizable across relatively di-verse samples of subjects tested in differ ingcontexts. Hence, the eight scales may proveuseful both as measuring instruments and asreference points for the classification of itemsand scales in the interpersonal domain.

    Psychometric CharacteristicsSex Differences

    Table 3 presents normative and psycho-metric data fo r each of the eight interpersonaladject ive scales. The first six columns con-tain means and standard deviations in a com-bined sample of 610 North American univer-sity students. Inspection of the separate meansfor men and women reveals clear-cut, and tosome extent predictable, sex differences inself-report . In comparison with women, menpresent themselves as more ambitious-domi-nant, arrogant-calculating, cold-quarrelsome,and aloof-introverted. Women present them-selves as more lazy-submissive, unassuming-ingenuous , warm-agreeable, and gregarious-extraverted. All of these mean differences ar estatistically reliable. Surprisingly, the smallestdifferences occur on ambitious-dominant (p< .03) and lazy-submissive ( / > < . 0 1 ) ; allother differences are highly reliable (p

  • 8/7/2019 Wiggins_1979

    14/18

    408 J E R R Y S. WIGGINSvariables is open to alternative interpretationso f th e circumplex s tructure in terms of stylisticresponse variables.4A variety of procedures have been sug-gested fo r coping with th e endorsement-de-sirability confound in self-report personalitydata. On the assumption that in principle, re-sponses within al l interpersonal ve ctors shouldbe equip robable, LaForge and Suczek ( 1955 )tried to write desirable phrases for undesir-able dimensions ("Can be strict if necessary")an d undesirable phrases for desirable dimen-sions ("Spoils people with kindness"). Jack-son's (1970) differential reliability index re-moves desirability variance and enhances con-tent saturation at the stage of item selection.Norman (Note 9) has described analysis ofvariance procedures fo r removing desirabilityvariance from th e Subject X Item responsematrix. Finally, several individual differencemeasures of social desirability response stylehave been constructed to permit the assess-ment of desirability responding fo r each sub-ject (e.g., Crowne & Marlowe, 1960; Edwards,1957b; Wiggins, 1959) .

    The extent to which an investigator mayfeel th e need to "correct for" th e ubiquitousendorsement-desirability confound will varywith the purpose of the investigation andwith the investigator's theoretical stance onthe meaning of the evaluative dimension ofaffective meaning. A set of interpersonal vari-ables that did not d i f f e r in desirability (o rdid not reveal sex differences) would be afeeble representation of real-life categorieso f social perception. On the other hand, basicinvestigations of the process of responding topersonality inventory items may require vari-ance decomposition procedures such as thoseof Norman (Note 9). In any event, interpreta-tions of individual or group scores on the in-terpersonal scales should be made with refer-ence to normative data of the kind providedin Table 3 . Whether a person or a groupscores hig h on warm-agreeable or low oncold-quarrelsome can be judged only withreference to the scores of others.Internal Consistency

    In the selection of item sets to mark th evectors of the interpersonal domain, the prin-

    cipal i temmetric criterion was that of struc-tural fidelity (Loevinger , 1957) . B y a varietyo f procedures, we attempted to ensure thateach item selected was properly located withreference to a circumplex that we had adoptedas the most appropriate measurement modelfor representing the relationships among cate-gories of interpersonal perception. I tems werealso selected that had relatively high correla-tions with appropriate preliminary scales.Final scales thus constructed ar e likely tobe internally consistent, although this mustbe demonstrated rather than assumed.The next-to-last column of Table 3 presentscoefficient alpha (Cronbach, 1951) estimateso f internal consistency for each of the inter -personal scales in a combined sample of uni-versity students. The final column of the tableindicates th e range of alpha coefficients ob -tained in the four samples that constituted thecombined student group. Although some ofthe interpersonal scales are more reliablethan others, all of them meet a reasonablystr ingent requirement of acceptable internalconsistency (a > .80). The range of alphacoefficients in the different samples tends tobe small, and this, together with th e repre-sentativeness with which th e universe of con-tent was sampled, makes it reasonable to as-sume that th e mean alpha values reportedare good estimates of the characteristic degreeo f in ter i tem s t ructure for the different inter -personal variables (Loevinger, 1957). In thisrespect, th e extraversion-introversion dimen-sion (NO-FG) is the most internally cohesive,and the variables of unassuming-ingenuous( J K ) and lazy-submissive (HI) are the least.

    DiscussionThe long-range goal of the research re-ported in this article was the development ofa psychological taxonomy that would encom-pass th e approximately 4,000 relatively fa-miliar trait-descriptive terms identified by4 F or example, Douglas N . Jackson (per sonal com-municat ion, November 2 7 , 1975) is convinced thatth e present interpersonal circumplex structure, aswell as others reported in the literature, can beaccounted for in terms of his threshold theory ofdesirability responding (Rogers, 1971; Jackson, Note8 ) .

  • 8/7/2019 Wiggins_1979

    15/18

    TAXONOMY OF INTERPERSONAL TRAITS 409Norman (Note 3). The research strategyadopted for this task relied heavily on cer-tain a priori distinctions between differentdomains of human characteristics. In the pres-ent work, the domain of interpersonal traitswas defined in such a way as to distinguishit from other domains such as temperament,character, moods, and cognitive traits. These,or similar, distinctions have been made bymost personality theorists concerned withtaxonomic issues (e.g., Allport, 1937; Cattell,1950; Murray, 1938). However within th einterpersonal domain, the adoption of atwo-dimensional real-space circumplex model(Benjamin, 1974) to capture the interrela-tionships among trait terms represents a per-sonal methodological preference that is per-haps not as widely shared. On the basis ofth e work reported in this article, it seemsfair to conclude that th e domain of interper-sonal trait descriptors can be classified, quiteprecisely, w ith reference to a circump lex model a conclusion that was not obvious when webegan. Since there are, no doubt, many otherways in which the interpersonal domain couldbe classified, under different definitions anddifferent models, the specific advantages of thepresent framework must be discussed in moredetail.Two-dimensional circular orderings of inter-personal variables have been reported in thel i terature fo r more than 40 years (Schaefer,1961) , and the general conceptual model un-derlying this structure can be traced as farback as Galen (Roback, 1928) . Remarkable"convergences" in conception and structurehave been noted for studies varying widelyin variables, populations, and measurementprocedures (Foa, 1961; Schaefer, 1961; Wig-gins, 1968) . These convergences do not stemfrom th e similarity of generative mechanismspostulated by different theorists (needs, in-terests, dynamisms, etc.) to account fo r sur-face patterns. Instead, the convergences re-flect a set of semantic categories that imparta common meaning to social events observedby a variety of procedures. The categoriesdistinguished by the ordinary language of per-sonality are not sharply demarcated, andthey are perhaps best thought of as "fuzzysets" (Zadeh, Fu, Tanaka, & Shimura, 1975)

    rather than as precise logical distinctions. Thecircumplex model appears to be particularlywell suited fo r representing elements (in thepresent case, adjectives) whose class mem-bership is continuous rather than discrete.Although often working in apparent isola-tion from one another, numerous investigatorshave proposed highly similar models of inter-personal behavior for the study of parents(Chance, 1959; Roe & Siegelman, 1963;Schaefer, 1959), children (Bau mrind & Black,1967; Becker & Krug, 1964; Schaefer & Bay-ley, 1963), parents and children (Benjamin,1974 ; Foa, Triandis, & Katz , 19 66), normaland abnormal adults (Leary, 1957; Lorr &McNair, 1963), psychotics (Lorr, Klett , &McNair , 1963), and college students (Stern,1970) . Within many of these diverse pro-grams of research, the circumplex model hasprovided a nomological network that hasenhanced th e meaning and significance ofth e separate interpersonal variables employed(Schaefer, 1961). U nfortunately, th e potentialof the circ ump lex as an integrativ e conceptualmodel has not been as widely recognized bythose whose research paradigms have focusedon the intensive study of single dimensions inpersonality and social psychology.In a recent book that presents summarieso f research on the major dimensions of per-sonality, the editors state:There obviously has been no overarching plan ortheory, implicit or explicit, guiding th e selection oftopics for trai t researchers. Indeed, the editors wereforced to organize th e book by means of the un-sophisticated tactic of simply placing th e chapters inalphabetical order. (London & Exner, 1978, p. xiv)But surely th e major research topics in con-temporary personality and social psychologysuch as achievement and power (PA), Machi-avellianism (BC), aggression (DE), int ro-version (FG), obedience (HI), interpersonalt rus t ( JK) , a l t ru i sm and helping behavior( L M ) , affiliation and extraversion ( N O ) can be related to each other by schemes moresophisticated than th e alphabet. This is notto claim an established isomorphism betweenth e research topics just enumerated and thepresent interpersonal taxonomy; rather, it isto illustrate th e kinds of questions whoseanswers might bring conceptual clarity to

  • 8/7/2019 Wiggins_1979

    16/18

    410 J E R R Y S. W I G G I N Sboth taxonomic and experimental research.For example, once it is recognized that th edimensions measured by Bern's ( 1 9 7 4 ) mas-culinity and fem inini ty scales are the or-thogonal interpersonal dimensions of ambi-t ious-dominant (PA) and warm-agreeable( L M ) , then the issue of the "independence"of masculinity and feminini ty , so conceived, isseen in a new l ight (Wiggins & Holzmul le r ,1 9 7 8 ) .The alphabetical list of h u m a n "needs"provided by Henry Murray ( 1 9 3 8 ) has provedto be an exceptionally fert i le source of per-sonality dimensions. W ithin th e f ramework ofM urr ay's personology, several of these dimen-sions have been studied in considerable depth(e.g. , Atkinson, 1958; McClel land, 1961;Winte r , 1 9 7 3 ) , although li t t le attention hasbeen given to the interrelationships amongthe needs in the overall system, perhaps be-cause an alphabetical taxonomy provides noguidance in this matter. Needs selected fromMurray 's l is t have formed th e basis for twomultiscale inventories (Edwards , 1959; Jack-son, 1 9 6 7 ) and for scoring keys on an adjec-t ive checklist (Gough & Heilbrun, 1965 ) .There appears to be a consistent factor s t ruc-ture associated with th e dozen needs theseins t ruments happen to share in common, butat tempts to in terpret this s t ructure have notbeen particularly enlightening from a substan-tive point of v iew (Huba & Hamilton, 1976 ) .In contrast, the strangely neglected work ofStern (1958, 1970) presents convincing evi-dence that a circumplex model provides ameaningful representation of the full range ofMurray's need variables.Two-dimensional circumplex models mayhave util i ty outside the realm of ordinarylanguage-trait description associated with per-sonality inventories and adjective checklists.The circular ordering of intercorrelationsamong clinical scales from th e MMPI hasbeen noted (Schaefer, 1961) , and it hasbeen argued that these scales can be inter-preted with in a two-dimensional circularmodel of personality (Kassebaum, Couch,& Slater, 1959). Given the manner in whichth e MMPI clinical scales were constructed,this argument appears rather farfetched onini t ial consideration. However, a recent study

    by Plutchik and Platman ( 1 9 7 7 ) suggests apossible reason why this might be the case.Twelve trait-descriptive adje ctive s w ere se-lected to represent an interpersonal circum-plex (Schaefer , 1961) . Psychiatr is ts werethen asked to rate seven diagnostic categories(e.g., paranoid, schizoid, hysterical) with re-spect to these adjectives. The mean adject iveprofiles of the seven diagnostic categories wereintercorrelated, and two factors were ex-tracted. The circular ordering of the diag-nostic categories (Plutchik & Platman, 1977,p . 4 2 1 ) bears a striking resemblance to thecircular ordering of corresponding MMPIscales reported elsewhere (Schaefer, 1961, p.1 3 5 ) . The constellations of traits impliedby diagnostic labels can be represented bya two-dimensional interpersonal circumplex(Schaefer & Plu tch ik , 1966) . To the extentthat the MMPI clinical scales reflect conven-tional diagnostic labeling, they to o would beexpected to exh ibit a circular ordering .The present taxonomy of interpersonaltraits may prove useful to investigators w hoemploy single adjectives as stimulus materialsin studies of interpersonal perception, impres-sion formation, and trait attribution. Therepresentativeness of the stimulu s m aterials isalways an issue in such studies, and the pres-ent taxonomy provides a systematic basis forsampling th e entire domain of interpersonaltraits as an al ternat ive to, for example, th eless differentiated strategy of sampling "fa-vorable" traits. Elaborate sampling proceduresfor th e s tudy of one or more specific dimen-sions of interpersonal behavior are made pos-sible by the availability of a list of 864 t e rmsthat have been classified within 16 categories.The position of an adjective within a givencategory is indexed by the correlation of thatadjective with it s own, its opposite, and itsorthogonal clusters. In addition, the earlieri temmetric work of Norman (Note 3 ) pro-vides inform ation on these same adjectiveswi th respect to such characteristics as socialdesirabili ty, difficulty level, self-endorsement,and attributions to liked, indifferent, and dis-liked others. Although such itemmetric dataare useful primari ly fo r calibration of s t imu-lus materials, they may themselves be profit-

  • 8/7/2019 Wiggins_1979

    17/18

    T A X O N O M Y OF INTE R PE R S O NA L TRAITS 411ably studied for the light they shed on basicissues of social perception (Goldberg, 1978 ) .

    Reference Notes1. W iggins, J. S. In defense of traits. Invited addressto the Ninth Annual Symposium on Recent D e-velopments in the Use of the MMPI, Los Angeles,Februa ry 2 8, 1974.2 . Goldberg, L. R. Language and personality: De-veloping a taxonomy o f trait-descriptive terms.Invited address to the Division of Evaluation andM easurement at the 86th A nnual Convent ion ofth e American Psychological Association, San Fran-cisco, Augus t 2 7, 1977.3. Norman, W. T. 2,800 personality trait descriptors:Normative operating characteristics for a uni-versity population. Unpubl i shed manuscr ipt , Uni -versi ty of M ichigan, 1967.4. Wiggins, J. S. & Mars ton , B. J. An index fo revaluating circumplexity in two-dimensional real-space models. Unpubl i shed manuscr ipt , Univers i tyof Brit ish Colum bia, 1976.5. Marston, B. J. Unpublished materials, Universityof Brit ish Colu mbia, 1976.6. Merri t t , N. G. Personali ty and value character-istics o f volunteers an d professionals at three dif-

    ferent social service agencies. Unpubl i shed master'sthesis, U niversi ty of Brit ish Columbia, 1977.7. Wiggins, J. S., & Holzm uller, A. Further evidenceo n androgyny and interpersonal flexibility. U n-published manuscript , Universi ty of Bri t ish Co-lumbia, 1978.8 . Jackson, D. N. A threshold model for styl ist ic re-sponding. Paper presented at the 76th AnnualConvention of the A me rican Psychological Associa-tion, San Francisco, September 1968.9. Norman, W. T. On decompo sing personali ty rat-ings. Paper presented at the ann ual meeting of

    th e Society of Mult ivariate Experimental Psy-chology, Londo n, Canada , Novem ber 7 ,19 74.

    ReferencesAllport , G. W. Personali ty: A psychological inter-

    pretation. New York: Holt , 1937.Allport , G. W., & Odbert, H . S . Trait names: Apsycho-lexical study. Psychological Monographs,1936, 47(1 , Whole No. 211) .Atkinson, J. W. (Ed.). Motives in fantasy, actionand society. Prince ton, N.J . : V an Nostrand, 1958.Baumrind , D., & Black, A. E. Socialization practicesassociated with dimensions of competence in pre-school boys and girls. Child Development, 1967,38 , 291-328.Becker, W. C., & Krug , R. S. A circumplex model forsocial behavior in children. Child Development,1964, 35, 371-396.

    Bern, D . J., & Allen, A. On predic t ing some of thepeople some of the t ime: The search fo r cross-situa-tional consistencies in behavior . Psychological Re-view, 1974, 81, 506-520.Bern, S. L. The me asu rem ent of psychological an-drogyny. Journal o f Consulting and Clinical Psy-chology, 1974 , 4 2 , 115-162.Benjamin, L. S. Structural analysis of social be-havior . Psychological Review, 1974, 81, 39 2 -425 .Carson, R. C. Interaction concepts o f personality.Chicago: A ldine, 1969.Cattell, R. B. Personality: A systematic, theoretical,an d factual study. New York: M cGraw-Hi l l, 1950.Cattel l , R. B. Personali ty and motivation structureand measurement. Yonkers-on-Hudson, N. Y.:W orld Book, 1957.Chance, E. Families in treatment; from the view-point of the patient, th e clinician and the re-searcher. New York: Basic Books, 1959.Cronbach , L. J. Coefficient alpha and the internals tructure of tests. Psychometrika, 1951, 16 , 297-334.Crowne , D. P., & Marlowe , D. A new scale of socialdesirability i ndependent of psychopathology. Jour-na l o f Consult ing Psychology, 1960, 24, 349-354.Edwards, A. L. Social desirabil i ty and probabil i tyof endorsement of items in the Interpersonal CheckList. Journal o f Abnormal and Social Psychology,1957, 55 , 394-396. (a)Edwards , A. L. T he social desirability variable inpersonality assessment and research. Ne w Y ork :Hol t, R ineh a r t & W inston, 1957. (b )Edwards , A. L. Edwards Personal Preference Sched-ule. New Y ork: Psychologica l Corp ora t ion, 1959.Foa, U. G. Convergences in the analysis of thes tructure of inte rpersonal behavior . PsychologicalReview, 1961, 68, 341-353.Foa, U . G. New develop ments in facet design andanalysis. Psychological Review, 1965, 72 , 26 2 - 27 4 .Foa, U . G ., & Foa, E. B. Societal structures of themind. Springfield, 111.: Charles C Thomas, 1974.Foa, U. G., Triandis, H. C., & Katz , E. W. Cross-cul tural invariance in the differentiation and or-ganization of family roles. Journal of Personali tyan d Social Psychology, 1966, 4, 316-327.Goldberg, L. R. Differential at t r ibut ion of trai t-descript ive terms to oneself as compared to well-l iked, neutral , and disl iked others: A psychometricanalysis. Journal o f Personali ty and Social Psy-chology, 1978, 36, 1012-1028.Gough, H. G., & He i l b run , A. B. The Adject iveCheck List manual. Palo Alto, Calif.: Consult ingPsychologists Press, 1965.

    Gui l ford , J. P. Personali ty. New York: McGraw-Hi l l ,1959.Gut tman, L. A new approach to factor analysis: Theradex. In P. R. Lazarsfeld (Ed.), Mathematicalthinking in the social sciences. Glencoe, 111.: FreePress, 1954.Heider , F . The psychology o f interpersonal relations.New York: Wiley, 1958.Homans, G. C. Social behavior: It s elementary form.New York: Harcourt , Brace, 1961.

  • 8/7/2019 Wiggins_1979

    18/18

    412 JERRY S. W I G G I N SHu ba, G. J., & Hamilton, D . L. On the g enerality oftrait relationships: Some analyses based on Fiske'spaper. Psychological Bulletin, 1976, 83, 868-876.Jackson, D. N. Personality Research Form manual .Goshen, N.Y.: Research Psychologists Press, 1967.Jackson, D. N. A sequential system for personalityscale development. In C. D. Spie lberger (Ed.) ,Current topics in clinical and comm unity psychol-ogy (Vol. 2). New York: Academic Press, 1970.Kassebaum, G. G., Couch, A. S., & Slater, P. E.The factorial dimensions of the MMPI. Journal

    o f Consulting Psychology, 1959, 23 , 226-236.Kelly, G. A. The psychology of personal constructs(Vols. 1 & 2) . New Y ork : N orton, 19SS.LaForge, R., & Suczek, R. The interpersonal dimen-sion of personali ty: III. A n interpersonal checklist. Journal of Personality, 195S, 24, 94-112.Leary, T. Interpersonal diagnosis o f personality.New York: Ronald Press, 1957.Loevinger, J. Objective tests as instruments of psy-chological theory. Psychological Reports, 1957, 3,635-694 (Monograph No. 9 ) .London, H., & Exner , J . E., Jr . (Eds . ) . Dimensionso f personality. New Y ork : W iley, 1978.Lor r , M ., Klett , C. J., & McNair , D. M . Syndromeso f psychosis. New York: Macmillan, 1963.Lorr , M ., & McNair , D. M. An interpersonal behaviorcircle. Journal of Abnormal an d Social Psychology,1963, 67 , 68-75.Lor r , M ., & McNai r , D. M. Expansion of the inter-personal behavior c i rc le . Journal of Personalityand Social Psychology, 1965, 2 , 823-830.McClel land, D. C. The achieving society. Prince ton,N.J.: V an Nostrand, 1961.Mischel, W . Personality and assessment. New York:Wiley, 1968.M u r r a y , H. A . Explorations in personality. N ewYork: Oxford Universi ty Press, 1938.Osgood, C. E. Speculation on the s t r uc t u r e of inte r-personal intentions. Behavioral Science, 1970, 15,237-254.Plu tch ik , R ., & Pla tman, S. R. Personali ty connota-tions of psychiatric diagnoses. Journal of Nervousand Mental Disease, 1977, 165, 418-422.Rinn , J . L. S t ru c tur e of pheno menal domains. Psy-chological Review, 1965, 72, 445-466.Roback, A. A. The psychology of character. NewY o r k : Harc ourt , B race , 1928.Roe, A., & Siegelman, M . A parent-child quest ion-naire. Child Development, 1963, 34, 355-369.Rogers, T. B. The process of responding to person-ality items: Some issues, a t heory an d some re -

    search. Multivariate Behavioral Research Mono-graphs, 1971, 6, No. 2.Schaefer, E. S. A circumplex model for maternalbehavior . Journal of Abnormal an d Social Psy-chology, 1959, 59 , 226 - 235 .Schaefer, E. S. Converging conceptual models formaternal behavior an d chi ld behavior . In J . C.Glidewell ( Ed . ) , Parental attitudes and child be-havior. Springfield, 111.: Charles C Thomas, 1961.Schaefer, E. S., & Bayley, N. Maternal behavior, childbehavior, an d the i r inte rcorre la t ions from infancythrough adolescence. Monographs o f th e Societyfo r Research in Child Development, 1963, 28(3,Whole N o. 8 7 ) .Schaefer, E. S., & Plutch ik, R. Interrelationships ofemotions, t rai ts, and diagnostic constructs. Psy-chological Reports, 1966, 18 , 399-410.Stern, G. G. Preliminary manual: Activities Index-

    College Characteristics Index. Syracuse, N.Y.:Syracuse Universi ty Psychological Research Cen-ter, 1958.Stern , G. G. People in context: Measuring person-environment congruence in education and indus-try. N ew York: Wiley, 1970.Sullivan, H. S. The interpersonal theory of psychi-atry. Ne w Y ork : Nor t on , 1953.Webster's new international dictionary (2nd ed. ,unabr idged ) . Springfield, Mass. : M err iam, 1924.Webster's third ne w international dictionary of theEnglish language unabridged. Springfield, Mass.:Merriam, 1961.Wiggins, J. S. Interre la t ionships among MMPI mea-sures of dissimulation under standard an d socialdesirability ins t ruc t ions . Journal of ConsultingPsychology, 1959, 23, 419 -427 .Wiggins, J. S. Personal i ty s t ruc ture . Annual Review

    o f Psychology, 1968, 19 , 293-350.Wiggins, J. S., & Holzmuller , A . Psychological an-drogyny and inte rpersonal behavior . Journal o fConsulting and Clinical Psychology, 1978, 46 , 40-52 .Wiggins, J . S., Renne r , K. E., Clore, G. L ., & Rose,R. J. The psychology of personality. Reading,Mass.: Addison-Wesley, 1971.Win te r , D. G. T he power motive. N ew York: FreePress, 1973.Zadeh, L. A., Fu, K.-S., Tanaka, K., & Shimura, M .( Ed s . ) . Fuzzy sets and their applications to cog-nitive and decision processes. New York : Aca -demic Press, 1975.

    Received A pr i l 14, 1978