an australian validation study of the temperament and character inventory

8
An Australian validation study of the temperament and character inventory Introduction The temperament and character inventory (TCI) was designed Ôto assess differences between people in seven basic dimensions of temperament and characterÕ (1). It measures four dimensions of temperament or personality: Ônovelty seekingÕ (or the activation of behaviours in response to novelty and signals of reward or relief of punishment); Ôharm avoidanceÕ (or behavioural inhibition in response to signals of punishment or non-reward); Ôreward dependenceÕ (for behaviours maintained without any reinforcement, and with high scorers rating as sentimental, warm, dedica- ted and attached – as against being withdrawn and cold); and ÔpersistenceÕ. In their overview monograph (1), the authors stated that the four temperaments were Ôgenetically independent dimensions that occur in all factorial combina- tions, rather than mutually exclusive categoriesÕ (p 16). The authors also noted that while the Ôfour temperaments provide an excellent description of traditional sub-types of personality disorderÕ, they also proved that it was Ôunable to distinguish whether someone had any personality disorderÕ (p 16), and thus the three character dimen- sions (self-directedness, cooperativeness and self- transcendence) were added to assess individual differences in self-concepts about goals and values. The authorsÕ monograph reported a principal components analysis undertaken in a non-clinical population, where the TCI items identified seven factors with eigenvalues greater than 1.0. Factor loadings supported the seven putative scales of the TCI, imputing that the TCI temperament and the character dimensions occupy similar domains. Intuitively, however, each might be expected to occupy a differing domain, with temperament relating more to psychophysical qualities or traits, and character to have a more moral conno- tation, a disjunction that might or might not influence factor analytic solutions. Aims of the study: To determine if the putative TCI structure can be confirmed in an Australian sample and to examine the TCI concurrent validity. Parker G, Hadzi-Pavlovic D, Parker K, Malhi G, Mitchell P, Wilhelm K, Austin M-P. An Australian validation study of the temperament and character inventory. Acta Psychiatr Scand 2003: 108: 359–366. Ó Blackwell Munksgaard 2003. Objective: To assess the validity of the temperament and character inventory (TCI) measure in an Australian sample. Method: A sample of depressed subjects completed the TCI and a measure assessing personality disorder constructs (PDCs), while family members and psychiatrists also returned PDC ratings. Results: Factor analyses generally supported the TCI constructs, when the temperament and character scales were analysed separately. Self- reported PDC scores were validated against corroborative witness ratings and used to assess the TCIs concurrent validity. Validation analyses supported all TCI temperament scales, but the Self- transcendence Character scale was unassociated with PDC scores. The remaining two character scales (i.e. cooperativeness and self- directedness) were non-specifically associated with all PDC scores. Conclusion: Study results support the validity of the TCI scales, but question the best model for conceptualizing the TCI. It is suggested that two character scales quantify disordered functioning, while temperament scales quantify style, constructs that may be inter-related or independent in individual subjects. G. Parker, D. Hadzi-Pavlovic, K. Parker, G. Malhi, P. Mitchell, K. Wilhelm, M-P. Austin School of Psychiatry, University of New South Wales, and Mood Disorders Unit, Black Dog Institute, Prince of Wales Hospital, Sydney, Australia Key words: temperament and character inventory; temperament; personality Gordon Parker, School of Psychiatry, University of New South Wales, and Mood Disorders Unit, Black Dog Institute, Prince of Wales Hospital, Sydney, Australia E-mail: [email protected] Accepted for publication March 31, 2003 Acta Psychiatr Scand 2003: 108: 359–366 Printed in UK. All rights reserved Copyright Ó Blackwell Munksgaard 2003 ACTA PSYCHIATRICA SCANDINAVICA 359

Upload: g-parker

Post on 06-Jul-2016

214 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: An Australian validation study of the temperament and character inventory

An Australian validation study of thetemperament and character inventory

Introduction

The temperament and character inventory (TCI)was designed �to assess differences between peoplein seven basic dimensions of temperament andcharacter� (1). It measures four dimensions oftemperament or personality: �novelty seeking� (orthe activation of behaviours in response tonovelty and signals of reward or relief ofpunishment); �harm avoidance� (or behaviouralinhibition in response to signals of punishment ornon-reward); �reward dependence� (for behavioursmaintained without any reinforcement, and withhigh scorers rating as sentimental, warm, dedica-ted and attached – as against being withdrawnand cold); and �persistence�. In their overviewmonograph (1), the authors stated that the fourtemperaments were �genetically independentdimensions that occur in all factorial combina-tions, rather than mutually exclusive categories�(p 16). The authors also noted that while the �fourtemperaments provide an excellent description oftraditional sub-types of personality disorder�, they

also proved that it was �unable to distinguishwhether someone had any personality disorder�(p 16), and thus the three character dimen-sions (self-directedness, cooperativeness and self-transcendence) were added to assess individualdifferences in self-concepts about goals andvalues.The authors� monograph reported a principal

components analysis undertaken in a non-clinicalpopulation, where the TCI items identified sevenfactors with eigenvalues greater than 1.0. Factorloadings supported the seven putative scales of theTCI, imputing that the TCI temperament and thecharacter dimensions occupy similar domains.Intuitively, however, each might be expected tooccupy a differing domain, with temperamentrelating more to psychophysical qualities ortraits, and character to have a more moral conno-tation, a disjunction that might or might notinfluence factor analytic solutions.Aims of the study: To determine if the putative

TCI structure can be confirmed in an Australiansample and to examine the TCI concurrent validity.

Parker G, Hadzi-Pavlovic D, Parker K, Malhi G, Mitchell P, WilhelmK, Austin M-P. An Australian validation study of the temperament andcharacter inventory.Acta Psychiatr Scand 2003: 108: 359–366.�BlackwellMunksgaard 2003.

Objective: To assess the validity of the temperament and characterinventory (TCI) measure in an Australian sample.Method: A sample of depressed subjects completed the TCI and ameasure assessing personality disorder constructs (PDCs), while familymembers and psychiatrists also returned PDC ratings.Results: Factor analyses generally supported the TCI constructs, whenthe temperament and character scales were analysed separately. Self-reported PDC scores were validated against corroborative witnessratings and used to assess the TCIs concurrent validity. Validationanalyses supported all TCI temperament scales, but the Self-transcendence Character scale was unassociated with PDC scores. Theremaining two character scales (i.e. cooperativeness and self-directedness) were non-specifically associated with all PDC scores.Conclusion: Study results support the validity of the TCI scales, butquestion the best model for conceptualizing the TCI. It is suggestedthat two character scales quantify disordered functioning, whiletemperament scales quantify style, constructs that may be inter-relatedor independent in individual subjects.

G. Parker, D. Hadzi-Pavlovic,K. Parker, G. Malhi, P. Mitchell,K. Wilhelm, M-P. AustinSchool of Psychiatry, University of New South Wales,and Mood Disorders Unit, Black Dog Institute, Prince ofWales Hospital, Sydney, Australia

Key words: temperament and character inventory;temperament; personality

Gordon Parker, School of Psychiatry, University of NewSouth Wales, and Mood Disorders Unit, Black DogInstitute, Prince of Wales Hospital, Sydney, AustraliaE-mail: [email protected]

Accepted for publication March 31, 2003

Acta Psychiatr Scand 2003: 108: 359–366Printed in UK. All rights reserved

Copyright � Blackwell Munksgaard 2003

ACTA PSYCHIATRICASCANDINAVICA

359

Page 2: An Australian validation study of the temperament and character inventory

Material and methods

Instruments

We assessed the 144-item TCI measure – which has20 items for each scale and four validity checkitems. The study measures included our PERson-ality Measure (PERM) (2) self-report question-naire incorporating descriptors of personality style,underpinning 15 formalized personality disorders,and thus weighted to psychopathological ratherthan normal personality functioning. For each ofthose 15 formal personality disorder categories,development analyses of the PERM had identifiedone or more component personality disorder con-struct (or PDC). For some personality disorders(e.g. avoidant) only one integral PDC scale wasderived. For others (e.g. borderline), there weretwo constructs derived (i.e. volatile attitudes andmood; empty and self-destructive), while for theantisocial style, three constructs (deceitful; irres-ponsible; and frustration-aggression) were derived.The 15 personality disorder categories generated atotal of 30 PDCs, and with developmental studiesestablishing modest to moderate test–retest andinter-rater reliability. Principal components analy-ses identified that the 30 PDCs were underpinnedby four factors – labeled emotional dysregulation,dissocial, inhibition and compulsivity.Analyses in the current study allowed us to

determine the extent to which TCI scales correlatedwith psychopathological personality style compo-nents as measured by the PERM (concurrentvalidation) as well as assess the current TCIstructural model.

Sample and procedure

The study was undertaken on a consecutive seriesof depressed patients referred to our tertiaryreferral Mood Disorders Unit. Prior to baselineassessment, patients received a mailed setof questionnaires (including current studycomponents) and were asked to complete each

questionnaire as they had perceived themselvesgenerally over the years and independent of anymood state effect.As the PERM measure was the principal valid-

ation scale, patients were requested to nominate afamily member who could complete an observer-rated PERM scale (as they had observed thepatient over the years) and thus provide corrobor-ative witness data. The interviewing psychiatristwas asked to rate the patient on all 30 PDCs – onthe basis of historical information obtained atinterview and from their observations. For eachPDC, six-point rating options of �not at all� to �anextreme degree� were allowed for psychiatristratings. The three approaches to rating PDCsallowed us to select the most appropriate raterstrategy for TCI validation.

Results

Of the provisional sample of 212 patients (meanage 39.5, SD 12.5, 59.4% female), personalityquestionnaire data for 156 subjects (currentsample) and corroborative witness reports for 102subjects were available. The mean age of currentsample members was 38.9 years (SD 12.0), with57.7% being female, akin to data for those 102 forwhom corroborative witness data (mean age 38.4,59.8% female) was available. Comparison of thosewho did or did not provide corroborative witnessdata did not identify any difference on self-reportTCI or PERM scores. At baseline assessment, thecurrent sample’s mean Hamilton depression scorewas 16.0 (SD 6.2), suggesting a moderate level ofdepression severity.Table 1 reports relevant TCI data. The Cron-

bach alpha coefficients indicate high internal con-sistency, while the correlation matrix quantifies theextent to which the differing scales were inter-related or independent.Table 2 examines agreement (quantified by

Pearson correlation, so as to make comparisonagainst data from our PERM developmental

Table 1. Means scores, internal consistency and intercorrelation of TCI scales

ScoresInternal consistency

Correlation matrix

TCI scale Mean SD Cronbach alpha NS HA RD Pers SD Coop

Novelty seeking 57.4 9.9 0.73 –Harm avoidance 72.9 13.9 0.90 )0.15Reward dependence 64.9 10.9 0.81 )0.20 )0.22Persistence 61.0 13.9 0.91 )0.02 )0.38 0.08Self-directedness 57.5 13.9 0.89 )0.22 )0.53 0.19 0.32Cooperativeness 76.5 9.9 0.82 )0.09 )0.19 0.57 0.16 0.41Self-transcendence 47.2 16.3 0.91 0.25 )0.24 0.15 0.18 )0.05 0.10

NS, novelty seeking; HA, harm avoidance; RD, reward dependence; Pers, persistence; SD, self-directedness; Coop, cooperativeness; ST, self-transcendence.

Parker et al.

360

Page 3: An Australian validation study of the temperament and character inventory

study) between the three differing raters on PDCscores. As all inter-rater coefficients involving thepsychiatrist ratings were low, it is unlikely that thepsychiatrists were validly assessing such personal-ity constructs. The self-report vs. corroborativewitness inter-rater coefficients were generally mod-erate, as established in the PERM developmentstudy (1) – and recorded again in Table 2 –supporting the validity of self-reported PDCscores. Therefore these are used in subsequentanalyses – rather than the psychiatrist-rated scores.Next, the structure of the TCI was examined via

a series of factor analyses. An unrestricted factoranalysis identified 25 factors with eigenvaluesexceeding 1.0 (and accounting for 69% of thevariance). The first factor accounted for 12% ofthe variance, the next six for 9%, 6%, 4%, 4%, 3%and 2%, respectively, suggesting very few substan-tive factors. When the TCI framework of a seven-factor solution was imposed (with an obliquerotation), relatively clear-cut confirmation wasfound for the �persistence�, �self-transcendence�,�novelty seeking� and �reward dependence�

constructs (in the sense that putative TCI itemsdominated the loadings on identified factors), butthe first factor was a bipolar one (with �self-directedness� items loading positively and �harmavoidance� negatively), while �cooperativeness�items were intermingled with �self-directedness�and (low) �novelty seeking� items on the fourthfactor.Analyses of the TCI temperament and character

scales were undertaken separately. For the tem-perament items, 20 factors with an eigenvalue ofmore than 1.0 were identified, and with the firstfour factors accounting for 36% of the variance.When a four-factor solution was imposed on theitem set, the 16 highest loading items on the firstfactor were all TCI �harm avoidance� ones (withfactor loadings of 0.76 to 0.34), the 17 highestloading items on the second factor were TCI�persistence� items (loadings of 0.77 to 0.30), the 11highest loading on the third factor were TCI�reward dependence� items (loadings of 0.71 to0.31), and seven or the 10 highest loading items onthe fourth factor were TCI �novelty seeking� items

Table 2. Correlations for PERM personality disorderconstructs (PDC) measuring agreement betweenpatient self-report, corroborative witness, andpsychiatrist judgments in current study andpreviously assessed (PERM development study)self-report and corroborative witness agreement

Correlations

Current studyPERM development study (1)

PDC SR with Psych CW with Psych SR with CW SR with CW

Avoidant 0.29* 0.20 0.51* 0.45Easily flustered 0.27* 0.19 0.60* 0.52Nervous/highly strung 0.40* 0.31* 0.47* 0.52Defers to others 0.28* 0.01 0.44* 0.29Avoidance of being alone 0.13 0.02 0.53* 0.58Self-critical 0.46* 0.32* 0.45* 0.45Negative in outlook 0.37* 0.24 0.51* 0.53Labile emotions 0.24* 0.22 0.28 0.45Vain 0.31* 0.35* 0.51* 0.45Exploits others 0.07 0.07 0.47* 0.41Self-important 0.24* 0.29 0.26 0.28Inflexible 0.19 0.15 0.42* 0.30Overly conscientious 0.31* 0.19 0.46* 0.37Dominant over others 0.26* 0.19 0.39* 0.48Sadistic 0.05 0.01 0.39* 0.46Emotionally detached 0.26* 0.08 0.41* 0.37Solitary 0.42* 0.38* 0.51* 0.52Isolative 0.33* 0.29 0.50* 0.52Eccentric 0.21 0.07 0.19 0.26Self-sabotaging 0.21 0.17 0.55* 0.44Sense of inferiority 0.34* 0.22 0.50* 0.43Suspicious 0.28* 0.23 0.43* 0.42Hostile to others 0.21 0.26 0.58* 0.50Negativistic 0.06 0.04 0.11 0.41Passive-aggressive 0.16 0.17 0.37* 0.42Volatile attitudes/moods 0.36* 0.38* 0.37* 0.53Empty and self-destructive 0.33* 0.27 0.49* 0.48Deceitful 0.17 0.01 0.24 0.46Irresponsible 0.27* 0.18 0.61* 0.62Frustration-aggressive 0.30* 0.21 0.63* 0.55

SR, self-report; CW, corroborative witness; Psych, psychiatrist rated. *P < 0.001 (for current study correlation coef-ficients only).

Australian validation study of the TCI

361

Page 4: An Australian validation study of the temperament and character inventory

(loadings 0.59 to 0.36). Thus, the four TCItemperament constructs were quite distinctly con-firmed as dominating the solution.The separate factor analysis of the character

items identified a first factor accounting for 13.7%of the variance, the second for 12.5% and the thirdfor 9.6%, but with a total of 27 factors generatingeigenvalues of more than 1.0. A forced three-factorsolution identified that the 17 highest loading itemson the first factor were from the TCI �Self-transcendence� scale (loadings of 0.80 to 0.32),that the 16 highest loading items on the secondfactor (0.77 to 0.42) were TCI �self-directedness�items and that all 14 items in the third factor(loading 0.80 to 0.34) were from the TCI �cooper-ativeness� scale. Again, the TCI character con-structs were confirmed.The correlation matrix (of all seven TCI scales)

reported in Table 10.5 of the TCI Guide was thencompared (1) with a correlation matrix of our TCIscale scores, using a multiple-group analysis inLISREL. The two matrices were not significantlydifferent [v2 ¼ 37.2, P ¼ 0.11; root mean squareerror of approximation (RMSEA) ¼ 0.026, 95%CI (0.0–0.044)]. This analysis suggests that theassociations between scale scores were comparableacross the independent studies.Published studies of the TCI factorial structure

almost invariably factor analyse TCI scales or sub-scales rather than items. The factor analysis inTable 10.6 of the TCI Guide (1) reports an obliquesolution. There, TCI sub-scales for a particularscale generally loaded highly on a single factor, withthe exception of reward dependence, where two ofits sub-scales (i.e. RD1 and RD4) loaded on thecooperativeness factor. A comparable exploratoryanalysis (oblique rotation of seven factors) of ourdata showed similar findings, but with even lesssupport for the RD factor (see Table 3). Thus, againwe have comparability of findings across studies.A series of confirmatory factor analyses (CFAs)

was undertaken to test how well the factorialstructure of the sub-scales corresponded to themodel postulated by its developers. As for anyCFA, the number of factors was fixed to thenumber in the model and sub-scales were onlyallowed to load on their putative factors. A good fitis indicated by an RMSEA statistic of around 0.05and below, or a CFI statistic of 0.95 or above. Thefirst analysis used all sub-scales and fitted sevenfactors (four temperaments and three characters).The fit was only modest [RMSEA ¼ 0.10, 95% CI(0.09–0.11); CFI ¼ 0.84], with the character sub-scales fitting better than the temperament sub-scales. In the next analysis, we attempted to fit thefour-factor model to the temperament sub-scales

alone, but the model could not be fitted (with anumerical problem associated with the rewarddependence sub-scales). The three-factor modelfor the character sub-scales also did not show goodfit [RMSEA ¼ 0.11, 95% CI (0.09–0.13);CFI ¼ 0.89]. Examination of the loadings indica-ted that, in general, the sub-scales loaded on theirputative scales. The fit statistics, however, sugges-ted that non-trivial associations between the sub-scales remained which could not be explained bythe hypothesised factors – either additional factorsor some re-arrangement of the sub-scales beingneeded.Table 4 reports intercorrelations of TCI scale

and individual PDC scores. Coefficients offerconcurrent validity support for the four TCItemperament scales. Thus, higher �novelty seeking�scale scores were associated with PDC descriptorsof (say) avoidance of being alone, labile emotions,volatile attitudes and moods, and irresponsibility –but unassociated with solitary, isolative andemotionally detached PDC scores. Higher �harmavoidance� scale scores were associated with PDCsindicating being easily flustered, nervous/highlystrung and avoidant, as well as having a negativeoutlook and being self-critical. Higher �rewarddependence� scale scores were negatively associatedwith isolative, solitary, emotionally detacheddescriptors as well as several defining difficulties

Table 3. Factor loadings for exploratory factor analysis of TCI subscales

NS1 0.24 0.29 0.07 0.35 )0.06 )0.01 0.10NS2 0.47 )0.25 0.24 0.04 0.10 0.07 0.10NS3 0.49 0.10 0.08 )0.03 )0.30 )0.08 )0.05NS4 0.62 0.09 0.01 )0.19 )0.03 0.06 )0.07HA1 )0.09 )0.51 0.06 0.04 )0.47 )0.10 )0.06HA2 )0.09 )0.93 0.12 0.10 )0.16 0.09 0.04HA3 0.13 )0.80 )0.40 0.10 0.06 )0.06 0.03HA4 )0.03 )0.24 )0.18 )0.20 )0.25 )0.13 )0.09RD1 0.02 )0.02 0.09 0.01 )0.21 )0.32 0.27RD3 0.22 0.14 0.75 0.07 0.07 )0.16 )0.14RD4 0.09 )0.04 0.14 )0.10 )0.12 )0.76 )0.20P1 )0.14 )0.08 0.11 0.79 0.10 )0.07 )0.01P2 )0.16 )0.14 0.06 0.91 )0.18 0.04 )0.03P3 0.14 )0.01 )0.07 0.83 0.04 0.09 )0.06P4 )0.12 0.08 )0.12 0.45 0.44 )0.03 )0.03SD1 )0.03 )0.04 )0.07 0.03 0.75 )0.10 )0.09SD2 0.09 0.06 0.04 0.06 0.69 )0.07 0.06SD3 )0.08 0.20 )0.01 0.20 0.59 0.01 )0.09SD4 )0.02 )0.13 0.09 )0.16 0.84 0.04 0.06SD5 )0.28 0.11 0.11 )0.14 0.67 0.07 0.01C1 0.17 )0.12 0.01 0.01 0.34 )0.49 0.00C2 0.02 0.23 0.15 0.08 0.08 )0.34 0.21C3 )0.11 0.10 )0.21 0.06 )0.03 )0.86 )0.01C4 )0.07 )0.12 0.13 )0.09 0.14 )0.54 0.03C5 )0.19 0.08 0.03 0.00 )0.01 )0.51 0.08ST1 0.14 0.03 )0.02 0.11 )0.11 0.09 0.69ST2 )0.12 )0.03 )0.12 )0.08 )0.01 0.12 0.99ST3 0.04 )0.04 )0.03 )0.07 0.07 )0.05 0.87

RD2 sub-scale does not exist in the 144-item version of the TCI.

Parker et al.

362

Page 5: An Australian validation study of the temperament and character inventory

in interacting with others. Higher �persistence� scalescores were positively associated with over-consci-entiousness, and negatively associated with passive-aggressive and irresponsibility descriptors. Turningto the TCI character dimensions, there were few(and weak) links between TCI �self-transcendence�scale scores and individual PDC scores. The tworemaining TCI character scales showed a patternquite different from all the other inter-correlationalanalyses (i.e. a non-specificity pattern). Virtually allPDCs were negatively associated with TCI �self-directedness� scale scores, and with many of theassociations quite striking. Similarly, but withassociations not being quite so high or consistent,TCI �cooperativeness� scale scores were negativelycorrelated with all PDC scores.These associations were clarified by generating

four higher-order PERM scales (as detailed in ourearlier paper (1) by summing relevant PDC scoresto create �emotional dysregulation�, �dissocial�,�inhibition� and �compulsivity� scores). Table 5reports correlations between those four higher-order PERM scores and TCI scales. Once again,we can observe concurrent support for three of thefour TCI temperament scales. The TCI novelty

seeking scale is most clearly associated with �dis-social� scores, TCI harm avoidance with �emotionaldysregulation�, TCI reward dependence with �inhi-bition� scores, while TCI persistence scores werenot clearly associated with any higher-orderPERM scales. In terms of TCI character scales,self-transcendence appeared similarly unassociatedwith PERM scales, while TCI self-directedness andcooperativeness scores were quite diffusely associ-ated with higher-order PERM scores.Regression analyses was undertaken to deter-

mine the percentage of the four higher-orderPERM scores accounted for by the four TCItemperament scale scores as a set, and the threeTCI character scale scores as a set, before andafter entering Hamilton depression scores in orderto determine whether there was any impact ofmood state severity. Hamilton scores made asignificant contribution to only one of the sevenmultivariate analyses – and less than 1% in thatanalysis – arguing against a mood state influence.Nevertheless, analyses are reported after includingHamilton scores as a predictor, and we note onlythe variables identified as significant. ExaminingTCI temperament scales first, higher PERM

Table 4. Correlations between self-report PERMpersonality disorder constructs (PDC) and TCI scales TCI scale

PDC scale NS HA RD Per SD Coop ST

Avoidant 0.13 0.61 )0.25 )0.30 )0.67 )0.30 0.02Easily flustered 0.11 0.66 )0.02 )0.33 )0.64 )0.22 0.05Nervous/highly strung )0.02 0.67 )0.12 )0.23 )0.57 )0.15 0.03Defers to others 0.09 0.48 0.04 )0.42 )0.56 )0.08 0.10Avoidance of being alone 0.27 0.42 0.02 )0.25 )0.61 )0.20 0.15Self-critical 0.18 0.59 )0.02 )0.26 )0.71 )0.11 0.01Negative outlook )0.01 0.57 )0.28 )0.36 )0.61 )0.15 )0.06Labile emotions 0.31 0.20 )0.09 0.02 )0.51 )0.41 0.18Vain 0.33 0.02 0.02 0.11 )0.43 )0.22 0.15Exploits others 0.31 0.05 )0.32 )0.08 )0.47 )0.54 0.16Self-important 0.38 )0.01 )0.12 0.15 )0.47 )0.34 0.26Inflexible 0.01 0.52 )0.23 )0.18 )0.59 )0.36 0.04Overly conscientious )0.06 0.15 )0.12 0.35 )0.23 )0.15 0.06Dominant over others 0.32 0.14 )0.26 0.03 )0.50 )0.58 0.21Sadistic 0.33 )0.23 )0.24 0.30 )0.09 )0.26 0.28Emotionally detached )0.01 0.23 )0.51 )0.19 )0.48 )0.35 )0.09Solitary 0.03 0.41 )0.56 )0.31 )0.49 )0.34 )0.01Isolative )0.07 0.53 )0.62 )0.22 )0.49 )0.36 )0.07Eccentric 0.36 0.22 )0.18 )0.06 )0.56 )0.30 0.28Self-sabotaging 0.28 0.38 )0.20 )0.23 )0.71 )0.36 0.15Inferiority 0.13 0.53 )0.24 )0.28 )0.71 )0.22 )0.02Suspicious 0.12 0.45 )0.29 )0.08 )0.64 )0.36 0.11Hostile to others 0.19 0.25 )0.23 0.04 )0.48 )0.54 0.10Negativistic 0.31 0.44 )0.13 )0.11 )0.72 )0.36 0.17Passive-aggressive 0.26 0.25 )0.18 )0.45 )0.60 )0.34 0.18Volatile attitudes/moods 0.35 0.31 )0.06 )0.02 )0.56 )0.34 0.28Empty and self-destructive 0.27 0.47 )0.14 )0.10 )0.70 )0.25 0.17Deceitful 0.34 0.04 )0.32 )0.01 )0.45 )0.54 0.18Irresponsible 0.43 0.22 )0.23 )0.31 )0.61 )0.34 0.12Frustration-aggressive 0.22 0.33 )0.17 )0.01 )0.44 )0.31 0.19

NS, novelty seeking; HA, harm avoidance; RD, reward dependence; Per, persistence; SD, self-directedness; Coop,cooperativeness; ST, self-transcendence. For correlations in bold, P < 0.001.

Australian validation study of the TCI

363

Page 6: An Australian validation study of the temperament and character inventory

�emotional dysregulation� scores were predicted (inorder and with individual variance quantified) byhigh TCI harm avoidance (37.2% variance), highnovelty seeking (8.7%) and low reward depend-ence (0.9%) scores. Higher PERM �dissocial�scores were predicted by high novelty seeking(21.2%), low reward dependence (8.9%) and highharm avoidance (3.8%). Higher PERM �inhibi-tion� scores were predicted by low TCI rewarddependence (41.2%), high harm avoidance(10.3%) and high novelty seeking (1.8%).Higher PERM �compulsivity� scores were predic-ted by high TCI harm avoidance (16.0%), highpersistence (3.5%), high novelty seeking (3.2%),and low reward dependence (1.6%). In terms ofTCI character scales, low self-directness scoresdominated the prediction of all four higher-orderPERM scores – 65.6% of �emotional dysregula-tion� (alone); 41.0% of �dissocial� [joining with lowcooperativeness (7.8%) and high self-transcend-ence (7.4%)]; 31.1% of �inhibition� [joining withlow cooperativeness� (5.1%); and 31.4% of �com-pulsivity� (joining with low co-operativeness(4.7%)].

Discussion

Assessment of personality in depressed samplesmay be problematic because of mood state influ-ences. While we studied depressed subjects, it wasestablished however that the sample’s depressionlevel was moderate (as quantified by the meanHamilton score), and our multivariate analysesexcluded any distinct effect of depressed mood onpersonality measure scores. Our use of corrobor-ative witness data for our PERM measure estab-lished support for using the patient self-report dataset and rejected clinical ratings made by ourpsychiatrists. Another advantage to our use ofself-report data is that patients attending ourtertiary referral service are extremely motivatedto provide an accurate assessment of themselves as

part of their clinical assessment and because self-report questionnaires are pre-posted, are not undertime pressure to complete such measures.The study advances our knowledge about the

TCI. First, it confirms many of its underlyingassumptions and constructs but, secondly, resultsargue for a somewhat reformulated model. Whenboth temperament and character items were ana-lysed together, no clear-cut support for the TCIseven-factor model was found. This may wellreflect a poor subject to item ratio, but we suspectother determinants. A major difficulty in develop-ing measures of personality (not unique to the TCI)lies in determining – from factor analyses – howmany intrinsic factors exist. Orthodox rules (suchas examining for those factors generating eigen-values more than 1.0 or examining a Scree plot)tend to be unhelpful, particularly when largenumbers of factors and are then suggested, withmany presumably reflecting the interdependence ofmany personality descriptors at both lower-orderand higher-order facet and trait levels. Any impo-sition of a set of finite factors will be difficult toreplicate in independent samples. However, in thisstudy, when analyses were undertaken of compo-nent temperament and character TCI scales sepa-rately, support for formalized TCI scale constructswas found, in that there was minimal �spillage� ofitems from their a priori scale status to otherfactors. Such support was less evident when ouranalyses considered both temperament and char-acter items together, which invites consideration asto how the TCI is best modelled or conceptualized.By examining correlations between TCI scale

scores and PDC scores measuring psychopatho-logical personality traits, we were able to examinethe concurrent validity of the TCI, and observeseveral contrasting patterns. In essence, support forthe validity of the TCI temperament scales wasfound by observing a specificity pattern, whereassociations linked lower-order and higher-orderPDC scores with certain TCI scales. Temperament

Table 5. Correlations between TCI scale scores and the four higher-order dimensions underlying PDCs

Higher-order PERM scale

Emotional dysregulation Dissocial Inhibition Compulsivity

TCI scale r p r p r p r p

Novelty seeking 0.22 0.005 0.46 0.001 )0.02 0.86 0.15 0.06Harm avoidance 0.64 0.001 0.18 0.02 0.43 0.001 0.45 0.001Reward dependence )0.18 0.027 )0.23 0.004 )0.64 0.001 )0.17 0.03Persistence )0.27 0.001 )0.07 0.37 )0.27 0.001 0.01 0.97Self-directedness )0.83 0.001 )0.68 0.001 )0.55 0.001 )0.57 0.001Cooperativeness )0.32 0.001 )0.51 0.001 )0.39 0.001 )0.40 0.001Self-transcendence 0.12 0.15 0.28 0.001 )0.06 0.47 0.13 0.11

Parker et al.

364

Page 7: An Australian validation study of the temperament and character inventory

Character Inventory �persistence� scores were,however, the least likely of the TCI temperamentscales to show any patterning with PDC scores.This finding is consistent with the study by Svrakicet al. (3) where, �persistence� (as well as �self-transcendence�) scores appeared largely inde-pendent of the number of personality disordersymptoms. As well, Gutierrez and colleagues (4)found weak internal consistency to the persistencescale in testing a Spanish version of the TCI. In theearlier TPQ measure, �persistence� was a sub-scaleof the �reward dependence� scale but, on the basisof several factor analyses (see 1), it was reconfig-ured as a fourth temperament dimension. In arecent study by the TCI authors (5), TCI �persist-ence� scores were most strongly associated with lowhistrionic personality disorder scores, and weaklyassociated with obsessive and paranoid scores –but again were less likely to be linked withpersonality scores than other TCI temperamentscales. Thus, �persistence�may or may not representa separate and distinct temperament dimension.A specificity pattern was missing, however, when

PDC scores were examined against TCI characterscales. First, PDC scores were essentially unassoci-ated with TCI �self-transcendence� scores. Thisfinding is consistent with an earlier TCI study (3)where �self-transcendence� scores were independentof personality disorder symptoms (whereas �coop-erativeness� and �self-directedness� scores wereclearly linked with personality disorder symptoms).Our results are also strikingly similar to thosereported in a recent paper by TCI authors (5), anddiffer only in these, that authors did establish linksbetween TCI �self-transcendence� scores and several(but not all) personality disorder styles. We suggest,however, that the �self-transcendence� TCI scale isunlikely to be a �character� scale in the sense ofreferring to any moral value (such as definingintrinsic �goodness� or �badness�), or in having anysocially utilitarian connotation. It therefore standsconceptually in contrast with the two other TCIcharacter scales, where �cooperativeness� and �self-directness� traits would appear to have strongpositive attributions – both for the individual andto society. Self-transcendence’s utility as a measureof psychopathology appears limited in that itappears neither to be positively or negativelyassociated with personality psychopathology.The two other TCI character scales (i.e. �Self-

directedness� and �cooperativeness�) showed non-specific links with virtually all PDC scale scores, afinding interesting in and of itself but also invitingthe TCI model to be reviewed.How can we best conceptualize the TCI if some

of its scales link specifically with certain personality

constructs while others are linked non-specificallywith all expressions of psychopathology? The mostparsimonious model allows that the temperamentscales more capture intra-psychic personality com-ponents, while the character scales of self-directed-ness and cooperativeness describe mostly theparameters of disordered personality functioningrather than character per se – unless the definitionof character is widened beyond its usual moralconnotation emphasis. This interpretation is nottoo dissimilar from the view put by the TCIauthors (1) – that those two character scales (butnot the temperament scales) �predict whether ornot someone has a personality disorder� (p 127).Thus, despite our study involving a sample of

depressed patients (as against a non-clinicalsample), this Australian study provides – whenTCI temperament and character scales were ana-lysed separately – validation of most of the TCIconstructs. In essence, analyses suggest only twocaveats, with the status of �persistence� as atemperament construct remaining problematicand the utility of self-transcendence characterscale remaining unclear. Perhaps more import-antly, we concur that, as two of the TCI characterscales have been shown to be efficient discrimina-tors of �personality disorder�, these scales have thepotential to inform us about a first-level issue (i.e.functioning and disordered personality function-ing), while the temperament scales provide second-level information (i.e. personality or temperamentstyle). This view corresponds to a recent report byTCI researchers (5), where the authors concludedthat TCI character scales should be used todiagnose the presence and severity of personalitydisorder, and temperament traits be used to effectdifferential diagnoses. Our independent validationtherefore supports use of the TCI to determineboth the likelihood of disordered personalityfunctioning and central temperament dimensions.

Acknowledgements

We acknowledge the assistance of Kerrie Eyers, ChristineBoyd, Heather Brotchie and Yvonne Foy, as well as anNHMRC Programme Grant (2223708) and an InstituteInfrastructure Grant from the NSW Department of Health.The assistance of Professor Bob Cloninger in providing TCIinformation is gratefully acknowledged.

References:

1. Cloninger CR, Przybeck TR, Svrakic DM, Wetzel RD. TheTemperament and Character Inventory (TCI): A Guide toits Development and Use. St Louis, Missouri: Center forPsychobiology of Personality, 1994.

Australian validation study of the TCI

365

Page 8: An Australian validation study of the temperament and character inventory

2. Parker G, Hadzi-Pavlovic D, Wilhelm K. Modelling andmeasuring the personality disorders. J Pers Disord 2000;14:189–198.

3. Svrakic DM, Whitehead C, Przybeck TR, Cloninger CR.

Differential diagnosis of personality disorders by the sevenfactor model of temperament and character. Arch GenPsychiat 1993;50:991–999.

4. Gutierrez F, Torrens M, Boget R, Martin-Santos R, San-

gorrin J, Perez G, Salamero M. Psychometric properties of

the Temperament and Character Inventory (TCI) ques-tionnaire in a Spanish psychiatric population. Acta Psy-chiatr Scand 2001;103:143–147.

5. Svrakic DM, Draganic S, Hil K, Bayon C, Prybeck TR,Cloninger CR. Temperament, character, and personalitydisorders: etiologic, diagnostic, treatment issues. ActaPsychiatrica Scand 2002;106:189–195.

Parker et al.

366