feiborg-etal-ijmpr-2005

8/7/2019 feiborg-etal-ijmpr-2005

1/15

I n t e r n a t i o n a l J o u r n a l of M e t h o d s in P s y c h i a tr ic R e s e a r c h , 2 0 0 5 , V o l u m e 14 , N u m b e r I, p a g e s 2 9 - 4 2 2 9

Resilience in relation to personality andintelligence

ODDGEIR FRIBORG,' DAG BARLAUG,^ MONICA MARTINUSSEN,' '^ JAN H. ROSENVINGE,' ODINHJEMDAL^

1 University of Troms0, De partm ent of Psychology, Norway2 The Military Institute of Leadership, Oslo, Norway3 Norwegian University of Science and Technology, Department of PsychologyAbstractResilience is a construct of increasing interest, but validated scales measuring resilience factors among adults are scarce. Here, a scalenamed the Resilience Scale for Adults (RSA) was crossvalidated and compared with measures of personality (Big Five/5PFs), cogni-tive abili t ies (Raven's Advanced Matrices, Vocabulary, Number series), and social intell igence (TSIS). All measures were given to482 applicants for the military college.

Confirmatory factor analyses confirmed the fit of the five-factor model, measuring 'personal strength', 'social competence', 'struc-tured style', 'family cohesion' and 'social resources'. Using Big Five to discriminate between well adjusted and more vulnerablepersonality profiles, all resilience factors were positively correlated with the well adjusted personality profile. RSA-personal strengthwas most associated w ith 5PFs-emo tional stabili ty, RSA-social c ompe tence with 5PFs-extroversion and 5PFs-agreeableness, as well asTSlS-social skills, RSA-structured style with 5PFs-conscientiousness. Unexpectedly but interestingly, measures of RSA-family cohe-sion and RSA -social resources were also related to personality. Furthermore, the RSA was unrelated to cognitive abili t ies.

This study supported the convergent and discriminative validity of the scale, and thus the inference that individuals scoring highon this scale are psychologically healthier, better adjusted, and thus more resilient.Key wo rds: resil ience, validation, personality, cognitive abili t ies, social intell igence

Resilience in relation to personality and making it difficult to review and compare results,intelligence Hence, scales that measure the construct more directlyFor centuries, writers have been inspired by the extra- are needed.ordinary capacity some people show in combating Operationalizing and measuring resilience is impor-adversity or misery. In psychology, how ever, the study tan t for two reason s. In clinical psychology, it mayof unexpected positive outcom es, today referred to as provide evidence about which factors are most centralresilience (Garmezy, 1981; Egeland, Sroufe and for regaining and ma intaining me ntal hea lth for differ-Erickson, 1983; Rutter, 1985; Cow en and Work, 1988) ent patien ts. In predicting the ability to tolerate stresshas just started to captu re interes t. In fact, of all the and nega tive affects, it may help in the selection ofpublished articles on resilience, about four-fifths have personnel who will manage tougher job dem ands. Toappeared during the last 10 years. A clear-cut taxon- meet the need for a valid resilience measure for theomy, as well as measurem ent instruments for studying adult popu lation, the authors constructed thethe con struct is, however, still lacking. Research on Resilience Scale for Adu lts (RS A) (Hjemdal, Eriborg,resilience processes has thus relied upon a diversity of Martinussen and Rosenvinge, 2001 ). U sing exp lo-mea surem ent scales only partially related to resilienc e, rative factor analysis, the original 16 resilience factors


2/15

30 Friborgetal.were reduced to five factors: 'personal competence','social competence', 'personal structure', 'family coher-ence', and 'social support'. The RSA has provenreliable and valid (Friborg, Hjemdal, Rosenvinge andMartinussen, 2003). Recently, it was extensivelyrevised by rewriting all items from a Likert-type to asemantic differential-type response format to reduceproblems with acquiescence bias (Friborg, Martinussenand Rosenvinge, 2004)- This was of particular concernas all the items in the original scale were exclusivelypositively worded. However, the resilience scale hasnot yet been submitted to a confirmatory factor analy-sis, which represents a stronger test of the scale's factorstructure. Therefore, the first aim of this study was toconfirm the fit of the measurement model on a sampleexperiencing more stress, followed by investigations ofmodel improvement. The second aim was to investi-gate whether subjects scoring high on the RSA arepsychological healthier and more resourceful than sub-jects scoring low. This was done by studying theconvergent and discriminative validity of the RSAagainst m easures of personality, social intelligence, andcognitive abilities.Resilience and personalityPersonality assessment within the lexical tradition,with t he Big Five model as the most prom inen t, is awidespread and validated method (McCrae and Costa,1997). According to this model, individual differencesin personality may be sufficiently described using fiveoverall factors: neuroticism (also referred to as absenceof'emotional stability'), extroversion, openness, agree-ableness and conscientiousness. Despite criticism(Block, 1995), it has received considerable empiricalsupport across cultures (McCrae and Costa, 1997).Studies relating these personality factors to resilience(faster recovery and less symptomatology after trauma)have been repeatedly associated with a high score onemotional stability, extroversion, openness and consci-entiousness (Riolli, Savicki and Cepani, 2002), as wellas agreeableness (Davey, Eaker and Walters, 2003). Inline with these findings, there is accumulating supportfor applying the Big Five measure to cluster individualsinto well adjusted (resilient) and more vulnerable sub-groups (Asendorpf, Borkenau, Ostendorf and VanAken, 20 01; Ramm stedt, Riem ann, Angleitner andBorkenau, 2004). All studies show evidence for theresilient personality profile being characterized by ahigh score on all the Big Five factors.

Of the five traits, emotional stability (absence ofneuroticism) carries most interest for clinicians. It hasbeen associated with a range of clinically relevantindicators. Individuals low on emotional stability gen-erally report more negative affect, lower self-esteem,and in particular, more symptoms of depression andanxiety (McCrae, 1990). Furnham, Crump andWhelan (1997) validated the NFO inventory usingsubjective ratings that included resilience by trainedassessors. They found a strong negative associationbetween resilience and neuroticism (r = -0. 71 ). As thefactor RSA-personal competence was the most effec-tive in discriminating between psychiatric outpatientsand health controls (Friborg et al., 2003), this factorwas expected to correlate more strongly with emo-tional stability th an the other RSA-factors.

One defining feature of those with high resilience isthe positive social orientation they show towards otherpeople (Werner, 2001). They have good social skills,thrive in social contexts and generally make a positiveimpression of themselves. In Big Five terminology,they are more extroverted. However, not all sub-facetsof extroversion are equally related to resilience.Whereas the extroversion-trait 'sociability' fits wellwith the conception of resilience as a positive socialorientation (Werner, 2001), the extroversion trait'competitiveness' is rather expected to keep others at adistance. Still, 'competitiveness' is not regarded asnegative either, as high levels of drive and energy arefound to increase coping capacity (Cederblad, Dahlin,Hagnell and Hansson, 1995). Furthermore, ifthe socialstyle of the individual is to be experienced as positivelyby others, it should be authentic, empathic and warm(Werner, 2001). A high score on agreeableness, inmeasuring facets like empathy, closeness, trust andcooperation, may indicate such positive qualities. Insummary, RSA-social competence is expected to corre-late strongest with extroversion. Furthermore, inmeasuring positive social traits, RSA-social compe-tence is expected to correlate stronger with 'sociability'than with 'competitiveness', and to correlate equallystrongly with agreeableness and extroversion.

The third and final personality trait that is expectedto covary with resilience is conscientiousness.Individuals high on this trait do not act on impulse,but prefer thorough planning. They generally stick toroutines/habits, prefer order and structure and worksystematically (McCrae and Costa, 1997). Werner andSmith (1992) found that resilient individuals were


3/15

Residence, personality and intelligence 3 1more achievement oriented, pursued more educationand almost all held full-time jobs in their adult life.Their adult career success, despite social adversities,was related to their strong ability to plan and organize,which Clausen (1993) denoted 'planfulness'. In thestudy by Furnham et al. (1997), conscientiousness,especially deliberation, self-discipline and achieve-ment striving, was significantly associated withsubjective ratings of resilience. The Big Five factorconscientiousness was thus expected to correlatestronger with RSA-structured style than any of theother RSA-factors.

A final purpose of validating the RSA against a BigFive inventory was to investigate the degree of vari-ance that is shared with personality (redundancy). Asthe resilience-scores are hypothesized to positivelycovary with measures of personality, some redundancyis certainly expected. However, it should not be toohigh either. According to Co hen (1988), a correlationor factor loading of 0.80 describes as very strong effect,leaving only 36% of the variance unexplained.Consequently, the more the redundancy surpassesthese levels, the less the RSA-factors may contributeuniquely beyond existing measures of personality.Among the five resilience factors, the two externalfactors, 'family cohesion' and 'social resources', were ^expected to share less variance than the other RSA-factors with personality.Residence an d social intelligenceTh e role of interpersonal skills in the treatm ent of psy-chiatric disorders was early recognized by Sullivan(1938). Research on processes leading to resilience,also points to social skills, or the social positive orien-tation, as one of the most protective againstmaladaptation (Murphy and Moriarty, 1976; Werner,2001). Social intelligence may be regarded as an over-all construct for understanding how successfully peoplemanage social relationships. However, the many facetsof social intelligence are not equally important forresilience. In a study by Luthar (1991), 'social expres-siveness' turned out as most protective among sevenother social factors (Riggio, 1986). Recently, Silvera,Martinussen and Dahl (2001) operationalized socialintelligence into a scale (TSIS) containing three dis-tinct components: social information processing,social awareness and social skills. The first two factorsare related to cognitive aspects of understanding andinterpreting ambiguous social information. The last

factor, social skills, is vastly different and relates topositive beliefs about one's social performing abilities.As social skills or expressiveness (Luthar, 1991) havebeen most important for the resilient outcome, thefactor RSA-social competence was expected to corre-late stronger with TSIS-social skills than the otherTSIS-factors.Resilience and cognitive abilitiesHigher intelligence implies better analytical, creative,and practical problem-solving abilities (Sternberg,1998). To some extent, this finding should apply forlife problems too. Intelligent people are more knowl-edgeable, expected to have better self-help skills (Ross,1972) and to cope more actively when faced withstress (Cederblad et al., 1995). Indeed, several longitu-dinal studies have launched evidence for childhoodintelligence being protective against later maladjust-ment (Radke-Yarrow and Sherman, 1990; Egeland,Carlson and Sroufe, 1993; Luthar, D'Avanzo andHites, 2003). For example, Werner (1993) found thatchildren's scholastic competence at age 10 was posi-tively linked with more sources of support as ateena ger a nd a greater sense of self-efficacy at age 18.However, the relation was not strong, as Werner(1993) emphasized by stating that resilient childrenwere not unusually gifted, but rather were adept atusing whatever skills they had available. Contradictiveresults are, however, present. Although Luthar (1991)found that intelligence was the best predictor ofscholastic competence, it turned out as a vulnerabilityfactor for internal affective symptomatology whenlevels of stress increased, thereby seriously challengingthe notion of intelligence as being exclusively protec-tive. Methodological issues are also present, as themany studies linking resilience and intelligence haveused different operationalizations of resilience, both asa process and an outcome variable. Moreover, moststudies have been conducted on individuals withstressful backgrounds, while the present study usedsubjects from the normal population in a stressful test-ing situation. However, taking the majority of resultsinto consideration, the relation between the RSA andintelligence is expected to be positive but no t strong.

MethodSubjectsParticipants were 482 applicants to a military college:


4/15

32 Friborgetal.47 women and 403 men, both M = 24.0 years of age(SD = 2.2 and 3.0, respectively). Gender was notreported for 32 cases. Most of them had finished theofficer training programme and served at least one yearin the military. This group represented a conveniencesample for studying the validity of the scale for subjectsexpected to experience increased work stress and stressrelated to the testing situation.ProcedureThe participants were given the instrument materialsduring the selection programme. They completedthem in a group setting, and returned them to an assis-tant when finished.InstrumentsThe Resilience Scale for Adults (RSA)The current 37-item version of the instrument used afive-point semantic differential scale format in whicheach item had a positive and a negative attribute ateach end of the scale continuum. The positive attrib-utes were keyed to the right for half of the items toreduce acquiescence hiases. Thre e items were added tothe first factor, 'Personal competence', to allow forremodelling of this factor as unpublished data (Friborg,2004) had indicated that a two-factor model mightdescribe the data better.Personality - the Big Five (5PFs)The Norwegian military version of the 5PF (5PF mil2.0) was developed by Engvik (1997) and applies aseven-point Likert scale. It was inspired by the NEO-PI model of Costa and McCrae (1995) and consists of240 items. They are grouped into five second-orderedfactors (the Big Five), each explaining the variance ofsix underlying primary factors. The first second-ordered factor, (I) extroversion, comprises thefollowing primary factors: sociability, social impact,activity, leadership, competitiveness and social bold-ness. Factor II, agreeableness, consists of closeness,empathy, trust, cooperation, helpfulness and positiveemotions. Factor III, conscientiousness, comprisespunctuality, work ethic, order, systematic, ambitionsand determination. Factor IV, emotional stability, isindicated by anxiety, anger, depression, self-conscious-ness, vulnerability and impulsiveness. Factor V,openness, is measured by fantasy, aesthetics, feelings,actions, ideas and values. Cronbach's alpha was high

for all major dimensions, ranging from 0.82 to 0.91(Engvik, 1993).The Troms0 Social Intelligence Scale (TSIS)This instrument purports to measure three facets ofsocial intelligence, one factor con cerning social perfor-mance (social skills) and two factors concerning socialperception (social awareness and social informationprocessing). It contains 21 items, with each factorhaving seven items. Half of the items are reversed.Responses are scored along a seven-point Likert-scalefrom 1 ('describes me extremely poorly') to 7('describes me extremely well'). These semantic labelswere attached to each endpoint. The three-factorstructure, as well as the factors' internal consistency(a = 0.85, 0.72, 0.79, respectively) has proven ade-quate (Silvera et al, 2001)Cognitive intelligence measuresFour measures indicating intelligence were included:(a) Raven Adv anced Matrices (R aven, 1986) as ameasure of non-verbal abstract and analytical intelli-gence; (b) Word Com preh ensio n (from W AIS) as ameasure of verbal intelligence - this being thesubtest with the highest correlation with totalWAIS-score (Engvik, Hjerkinn and Seim, 1978); (c)Number Series is a measure of mathematicalreasoning, in which the task is to fill in the two lastdigits in series of numbers; and (d) M athem atics as aMeasure of Mathematical Knowledge and Skills(Martinussen and Torjussen, 1997). These tests areregularly used for military selection purposes.Data preparationTo improve data quality, participants were removedaccording to three criteria: if more than 10% of thedata were missing (50 cases), a z-score +/- 3 standarddeviations on the acquiescence/nay-saying index (17cases), and cases scoring > 2 SD on the Big Five liescale (three cases). In sum, 60 participants were omit-ted, yielding a final sample of 411 participants.Remaining missing data in the resilience variableswere replaced with the mean composite score for eachindividual to each of the five factors the item belongedto .Skewness and kurtosisThe RSA-items were heavily skewed, showing highkurtosis as well. Negative skew in z-values ranged from


5/15

Resilience, personality and intelligence 33-2.16 to -19.43 (M = -9 .71 , SD = 5.20), w hile for kur-tosis the range was -2.58 to 30.37 (M = 7.35, SD =8.43). Kurtosis was less pronounced, with 12 itemsbeing normally distributed. The covariance matrix wasthus corrected for non-normality by computing theasymptotic covariance matrix in PRELIS 2, which wasincluded as a weight matrix to adjust for non-normal-ity in the poly-choric matrix. It was then possible toestimate all matrices using the Satorra-Bentlerreseated x^ statistics, which correct for biases in stan-dard errors arising from significantly skewed data (Huand Bentler, 1995; DiStefano, 2002).Data analysis and model specificationsSPSS version 11 was used to perform standard descrip-tive, reliability and canonical correlational analyses, aswell as exploratory factor analyses. Structural equationswere solved using LISREL v8.53 (Joreskog and Sorbom,1996). Model fit was evaluated according to Hu andBentler (1995, 1999), with a non-significant Satorra-Bentler x^ RMSEA < 0.06 and CFI/NNFl > 0.95representing a good fit. However, as models are never anexact replica of reality, various degrees of misspecifica-tion are always present. Consequently, if the power of atest is strong enough (large sample and many indica-tors), even minor deviations between model and datawill be significant (Hu and Bentler, 1995), thus reject-ing perfectly acceptable models. The RMSEA-index,which indicates in what degree a model is a reasonableapproximation to the observed data, is therefore usedinstead of absolute fit measures for the evaluation of thecomplete measurement model.Bivariate normalityAs the RSA measure used a five-point scale, the itemindicators were considered as ordinal variables.Bivariate normality was however assumed for the latentvariables. Testing this assumption using PRELIS(Joreskog and Sorbom , 1988), by estimating the fit of allthe poly-choric correlations, proved close fit (RMSEA< 0.05) for 811 of the 946 correlations and fair fit (0.05- 0.081) for the remaining 135 correlations.ReliabilityCronbach's method may underestimate internal con-sistency in case of violations of the tau-equivalenceassumption (Raykov, 2001), or simply as a function offewer items (Cronbach, 1990). As that would inflatethe attenuation-corrected correlations used in this

study, a structural equation modelling approach,according to Raykov (2001), was taken instead to pro-vide more precise coefficient estimations.Resul tsCrossvalidation of the factor modelThe fit of the original 37-item version, specified as fivecorrelated latent factors, was unsatisfactory in absoluteterms (S-B x^ (619) = 1130.74, p < 0.001). Due to thestrong power of this test, the RMSEA-index wasinspected instead, revealing reasonably small degreesof misspecifications (RMSEA = 0.045), thus indicat-ing a promising m odel.Model improvementKnowing that the 37-item version crossvalidated well,the next step implied post hoc modelling within eachresilience factor (now including all the 40 items), toremove poorly fitting items. Post-correctional fitindices for the separate factors, and the complete mea-surement model, are presented in Table 1. The factornames were revised slightly to reflect item contentbetter.

The first factor, 'personal strength', had to beremodelled as a second-ordered factor accounting fortwo underlying primary factors, to improve model fitsignificantly. By removing three additional items dueto correlated errors, the first factor now fitted the datavery well (S-B x^ (34) = 46.92, p = 0.07; RMSEA =0.030). Within the remaining four resilience factors, itwas sufficient to remove one item within each factorto achieve a very strong fit (see Table 1). The degree ofmisspecification within each factor was now very small(RMSEA ranging from 0.013 to 0.042).The complete measurement modelThe correlated resilience model now comprised 33items. Although statistics for absolute fit of this modelwas significant (S-B x^ (516) = 821.99, p < 0.001), theabove modifications reduced the degree of model mis-specification further (RM SEA = 0.037 ), indicatingthat this model will hold reasonably well in the popu-lation, although not exactly.Descriptive statistics and reliabilityThe means, standard deviations and estimates of inter-nal consistency for all the instruments are presented atthe lower part of Table 2. The Cronbach's alphas of the


6/15

3 4 Friborget al.Table 1. Fit indices for each of the RSA-factors and the complete measurement model - estimation method: maximumlikelihood (N = 411)DimensionsNo. itemsS-Bx^d.f.E.-,Ea .90N N F PC F PSRMRR2Note. ** p Od) d>o o o o o o o o o o o

o

4

"nO

6

; ;o

3

roO

3

ONroO

! ^

O N

O

roOO

OOr-1O

.1

InoON

NOo

ro

^ r - " ^ < ^ r - fz: CDvc:!u o ' ^ L n O O r - j . ^ O Oo o o o o o o o o

-^ 'ÔSCOO r ^

o o o o o or o C O O N r O ' O O r OCO O>^ ~ rO Of~'~\ Oc ( i-O"ô o o o o o cSc - ^ O 00r- NOr-Ji n rj- >ro rd > d > d > di d) dii ~ O O N i r - i O O N r o O N ' ^ O N

d di di d)ddi d dd>

(- Tf LnO ^ LOOS ^ Op

o o doI I I Ioo a< o) o

\j 'X os cu fTj- 00ro " Lnd o d d dO> OCOCOCO00 ^ro "^ roddddd

u~i OO ON pd d dro LOrvp p p pd d d d

. - . ^ ^ ^ ^ O N O o o r ^ ' L Op v i - ^ f O u o r v ) p p ' ' Pd d dd > d d d d d

c^ c: > c:> o a oo o oc:> o or-J r~ OroOO"^rO LO-- ro c ciNO ro

O d C

r \OOO pd d > di

ro NOrinNOin ( T1- rod d d d dO OOT- ro ONd d d d do o o o o

O N To pdd

I INO CO Tl- P p pd d d d p p pd ddd

I I I I

00 OT-00 NOroOO ON r.S

p ON p[ r- ro

O N - roin d in oTt^ ON NOin d r r- roin d in

O r~ 00ON NOinro O r

I

,self

c0eee

iue

eee

6.Boae

oCO

iououp

core

oCO

sy-aQJcutu

CO

)o

dco

CO

"so

CO

ne

raa

OCO

o c -^'- - Sro

OJdJOJD

ra g3zr- COON

-5 ain NOr- o6

ONd

OOd

ONd

O N ON[-- NOdd

CO r-ddNO ON1 ^- NOdd

O O O r O T l - T j - Q o c o r ^ fr ^ r ^ O N O r O N O O r ~ - -^N OL n r v] T1-ONi n T t - i n i n O O r o " " ^r o ro r o i n r o r o r x i - H - p . IJ ^" d d d d d d d d d d d d d d d d d d T f - d d ddvoON ^ NO--- o o m inONONtôo mm- o NO^ ^ ' ^ NOCOrororoOOTj-r rvirg pvir r jp p p p p _-.r-jmr^ ^ ' ^d d d d d d d d d d d d d d d d d d " ^'-^ ' d dNO( r^lNOTj-CO Tl- in OO 0 0 O N N O T 1 - N O i n O N r O NO v o _ ' ~ ^ ' ^ " " rvlNOTl-roOOr-jTl-O r v innrroro p ^^ ^ ^ ^ ' ^ ^ ^ ' ~ ^d d d d d d d d d d d d d d d d d d Ti -Trdm d dN O ^ Ô N ' O O r o T 1 - O N O ro ^r -Ji n ^ ^ - ^ O - ^ ^ ^ T 1 - N O C O inNOi n t ^ r ^ r o r O T l - r- r."! r o r o i n r r r ^ p p p p ^ - r - l i n r l - t ^^d d d d d d d d d d d d d d d d d d ^ T t ^ d N o doO r o l O ' i n O O O Nr ^ r O N O r ^J T l - NO i n - r O " ^ N O - C O T l - r o CDOO O T T r o r O T j - r ^ - ^r O T } ^ T ^r O Tj - NO r ^j p p p p - ^r ^l T l^T ly ^ ^d d d d d d d d d d d d d d d d d d ' ^ T i ^ d d d

oZZ CO CO

82dS V

c 6

. 0

i -

SM

Q .raIcu03Q

Cro

c0raQ

C-

Noe

cron

Ate


8/15

36 Friborgetal.strongly supported (r = 0.83) as the other RSA-factorshad significantly lower correlations, ranging from 0.33(family cohesion) to 0.63 (perception of future).

Correlating the RSA-factors with the TSlS-socialintelligence scale indicated that RSA-personalstrength, RSA -social comp etence and RSA-social sup-port were more strongly related to social intelligencethan RSA-family cohesion and RSA-structured style.Furthermore, and as expected, RSA-social compe-tence correlated strongest with TSIS-social skills (r =0.88) and significantly lower with 'social informationprocessing' and 'social awareness', r = 0.29 and r = 0.35respectively.

The correlational patterns hetween the resiliencefactors and intelligence factors (Raven's AdvancedMatrices, mathematics, vocahulary and number series)were non-significant, except for 'social competence',which correlated slightly negatively with 'mathemat-ics'(r =-0.16, p < 0.001).Joint explorative factor analysisCombining the resilience and personality factors in ajoint explorative factor analysis, we investigated whichpersonality factors accounted most for the resiliencefactors. A principal component analysis using varimaxrotation was conducted, prespecifying six factors, fivefor the Big Five and one for the RSA. Despite severalsubstantial side loadings, the Big Five model was rea-sonably reproduced (see Table 3). The modelaccounted for 57% of the variance.

All the resilience factors were accounted for by thethree first principal components. The first componentconsisted mostly of facets related to agreeableness,'sociability' and 'feelings', thus representing a factorindicating a positive social orientation towards otherpeople. Two resilience factors loaded highly on thiscomponent, respectively RSA-social competence(0.69) and RSA-social resources (0.68). Unexpectedly,RSA-social competence had a much smaller side load-ing (0.38) on the fourth component (extroversion).Interestingly, 'social resources' also side loaded (0.30)on the third component (emotional stability).

The second component was most distinct as it con-tained all the conscientiousness factors. It accountedfor two resilience factors, with RSA-structured styleloading most strongly (0.82), and RSA-perception offuture loading moderately (0.47). Of these, 'perceptionof future' side loaded roughly equally on em otional sta-bility (0.43).

RSA-perception of self loaded exclusively andhighly (0.67) upon the third component, which con-tained mainly factors related to emotional stability.This finding was expected as items related to internalstrength was assumed to be best accounted for byabsence of neuroticism. Interestingly, RSA-familycohesion also loaded on this component (0.43), indi-cating that a coherent family life goes along with lessneuroticism. However, this factor side loaded signifi-cantly (0.41) on the first component (agreeableness)as well, thus representing a factor also associated witha positive social orientation.Shared variance between resilience and personalityTo investigate total shared variance between resilienceand personality, a canonical correlation analysis wasrun. The average of the squared crossloadings betweenthe two sets indicated that the five personality canoni-cal variates explained 37.8% of the variance in the sixresilience scores. Conversely, the six resilience canoni-cal variates explained 42.2% of the variance in thepersonality scores. They were thus approximatelyequal in explaining power, though resilience turnedout as slightly stronger.DiscussionThe original correlated five-factor model (37 items)crossvalidated well by indicating a close fit to the pop-ulation parameters. Having confirmed the factorstructure, the model was further improved by post hocmodelling. The first resilience factor, 'personalstrength' had to be remodelled as a second-orderedfactor, containing two primary factors, to achieve agood fit. For the remaining four resilience factors, itwas sufficient to remove the worst fitting item fromeach factor to achieve similar good fit. The finalmodel, now containing 33 items, indicated evensmaller discrepancies between the implied and esti-mated covariance matrix. Due to the smallerdiscrepancies, the tolerance for sample variations isgreater, and it should thus generalize better to newsamples.

The overall factor structure was thus similar to theoriginal one (Hjemdal et al., 2001). However, thefactor names were revised slightly to reflect item con-tent better. The model comprises three resiliencefactors measuring various aspects of 'personal compe-tence' (personal strength, social competence andstructured style), one factor measuring degree of


9/15

Resilience, personality and intelligence 37Table 3.A principal component analysis including the six factors of the RSA and the 30 primary factors of the 5PFs measure,using varimax ro tation (N = 411)Primary factors

11-ClosenessRSA-Soc. competence

11-Positive emotionsRSA -Soc. resources

1-SociabilityIl-Coope rationV-Feelings11-Trust

RSA-Structured styleIll-SystematicIll-OrderIll-Work moralIll-PunctualityIll-DeterminationRSA-Perception, future111-AmhitionIV-Depression

RSA-Perception, selflV-AnxietyIV-SelfconsciousnessIV-ImpulsivenessIV-Vulnerability

RSA-Fam. cohesion1-Social impact1-Leadership1-Social boldness1-Activlty1-CompetitivenessV-ActionsV-AestheticsV-ldcasV-FantasyIl-EmpathyV-Valucsll-HelpfulnesslV-Anger

Variance explained %

10.720.690.680.680.610.580.560.52

0.41

0.37

0.400.48

12

2

0.820.810.740.680.650.490.470.45

0.350.32

-0.32

12

3

0.30

0.380.430.790.670.670.630.620.580.43

0.360.32

0.33

12

4

0.38

0.39

0.42

0.330.310.31

-0.310.780.760.640.470.380.37

10

5

0.48

0.770.670.630.47

6

6

0.41

0.400.610.53

.525

Note. Total variance explained: 57%.Factor loadings below < 0.30 are omitted.

'fatnily cohesion' and a final factor tneasuring how theitidividuals view their own 'social resources'. The lasttwo factors tnay be regarded as individual sources ofsupport that tnay reinforce and strengthen the copingability of the individual. The first factor, 'personal

strength', now contains two prirnary factors tneasuringindividuals' views of their own current strengths andabilities ('perception of self) and their beliefs aboutopportunities for realizing future plans and goals('perception of future').


10/15

38 Friborgetal.Convergent validity of the RSA-scaleFirstly, all the resilience factors were them selves highlycorrelated, implying that if individuals experiencedstrong personal strength or social competence, theymost prohably also experienced a coherent and stablefamily or good social resources. All the expected rela-tions between the Big Five personality factors and theresilience factors were confirmed by showing positivecorrelations of moderate to strong association. Thesame finding was evident for the relation betweenresilience and social intelligence as well, all being posi-tively related. Contrary to expectations, the relationbetween cognitive intelligence and resilience wasslightly negative, but largely insignificant and negligi-ble. As this hypothesis was less strong, as discussed inthe introduction, they were partly regarded as anexploration of this relation. Consequently, theseresults were not exclusively disconfirmative, especiallywith reference to Werner (1993) who stated thatresilient individuals were not necessarily intellectuallystrong but rather adept at effectively using whateverskills they had available to help them cope. This wasalso a finding of Vaillant and Davis (2000) who fol-lowed economically and educationally disadvantagedmen from age 14 to age 65, finding that half of the lowIQ men (115). They had similarincomes, equally educated children, and had devel-oped even more mature defences and warmer objectrelations tha n the high-IQ group. The null correlationbetween the RSA and the intelligence measures thusconfirms the problems with using IQ as direct a indica-tor of resilience. Taken as a whole, the results gavereasonably strong support for the convergent validityof the RSA measurement.

Discriminate validity of the RSA-scaleDiscriminative validity was equally well supported -the expected differences in the magnitude of the corre-lations between resilience and personality, andresilience and social intelligence were confirmed. Firstof all, emotional stability, as the most clinical mean-ingful factor, was strongest related to the resiliencefactor 'personal strength' and its primary factor 'per-ception of self. Individuals scoring low on emotionalstability (high neuroticism) generally report lower self-esteem (Engvik, 1993), and have a ruminative andnegative style of thinking. They often experience peri-

ods of anxious and depressed feelings. The high corre-lation with 'perception of self concurs with previousfindings of this factor as the strongest in discriminatingbetween patients and health controls (Friborg et al.,2003), thus supporting the interpretation of'personalstrength/perception of self as the foremost resourceindicator that may counteract psychological vulnera-bility. However, the other primary factor 'personalstrength/perception of future' correlated equallystrongly with conscientiousness and emotional stabil-ity, thus representing a blend of these factors. Theorthogonally rotated factor solution, maximizing dif-ferences in factor loadings, and thus factor uniqueness,confirmed that 'perception of self loaded uniquelyupon emotional stability, while 'perception of future'loaded equally strongly with 'conscientiousness'. Thisfinding is interesting as Clausen (1993) claimed thatresilient individuals possess a certain kind of planful-ness for their life that helped them realize future plansand goals. Werner (2001) also employed planfulness asan explanation for the adult successful adaptationdespite social hardships (for example, acquiring aneducation and getting oneself a full-time job). A studythat investigated this relation more directly (Nezlek,2001) also supported the notion that psychologicalhealthier individuals were more efficient than vulnera-ble individuals in realizing social and achievementrelated plans on a day-to-day basis as well.RSA-social competence was expected to correlatestrongest with 5P Fs-extroversion andTSIS-social skills,and significantly weaker with the other personality andsocial intelligence factors. The results supported thisdiscriminative interpretation. Furthermore, the inter-pretation of RSA-social competence as a measure ofpositive social qualities was also confirmed, as it corre-lated significantly stronger with the extroversion-trait'sociability' than with 'competitiveness'. Consequently,a high score on RSA-social competence should implysocial skills tha t are exp erienced positively by others aswell. To shed light on that, it is well known that posi-tive social skills (such as knowing how to start aconversation, how to make a good impression) maynot necessarily imply a genuinely warm and empathicpersonality trait. However, taking a resilience perspec-tive on social competence, it is the blend of tbese twoqualities, being socially competent and able to shareexperiences in a trusting, empathic and cooperativeway, that helps establish mutually supportive friendships


11/15

Residence, personality an d inteHigence 39(Werner, 2001). The results supported the interpreta-tion of RSA-social competence as a factor measuringsuch combined positive characteristics, as it correlatedequally strongly with both extroversion and agreeable-ness.

The factor RSA-structured style correlated verystrongly with 5PFs-conscientioiisness, as expected. Inthe factor analysis it also turned out as the most sensi-tive factor (highest loading) for a unit change inconscientiousness. Of the five resilience factors, thisfactor was related to personality most strongly, andamong the six facets of conscientiousness it was signifi-cantly more strongly related to the order andsystematic factors than to ambitions and determina-tion. The high factor loading, as well as the highcorrelations, do not leave much unexplained. Theresilience factor may thus be redundant in explaininganything beyond personality/conscientiousness andmeasurement errors. However, as conscientiousnesshas related positively to recovery after trauma (RioUiet al., 2002), the same would be the case for thisresilience factor.

The two external resilience factors, 'family cohe-sion' and 'social resources', which were expected to beless related to Big Five than the three previous factors,was only partly true as 'social resources' loaded equallystrongly on agreeableness and 'social competence'.Nevertbeless, this result is adequately interpretableand highly interesting, indicating that people whowere trusting, cooperative, emphatic and warm, alsohad a wider social network which could provide themwith support and reinforcement to help overcome psy-chosocial stressors. The interesting question thenbecomes whether it is these personality traits thatenhance the social network of resources, or whether awealth of social support in early life helps developagreeable positive traits. Taken together, the factor'social resources' seems to measure not only degrees ofsocial support and resources, but also the quality of thesupport received. Th is is a promising finding, interest-ing for further studies.Wha t about the mode rate-to-strong redundanc y withpersonality?A moderate-to-strong degree of redundancy betweenresilience and personality was evident from the canon-ical correlation analysis. Moreover, all the resiliencefactors loaded significantly upon the latent personality

factors in the principal component analysis. It is thusincorrect to view the RSA-measure as independent ofpersonality. Rather, the current results speak for theresilience factors as variants of personality factors notaccounted for in the current Big Five model. A posi-tive score on the Big Five factors has been associatedwith a well adjusted personality profile (Asendorpf etal., 2001 ; Ram mstedt et al., 2004) and resilience(Riolli et al, 2002; Davey et al, 2003), so these associ-ations were expected. The redundancy issue was thusnot a qualitative question (either/or), but one of mag-nitude. As more than 60% of the variance wasunshared, further studies are needed to clarify howmuch of this residual variance is unique for resilience,and how well it eventually may explain constructs likestress tolerance, adaptability, adjustment, absence ofpsychiatric symptoms (negative mood), and so forth,beyond personality (incremental validity). Due to therelatively large residual variance, the resilience factorsare certainly expected to explain a significant portionof this residual variance beyond personality. If they addto the incremental predictive validity in prospectivestudies, they will have earned a place in mental healthmeasurement.Reduced reliability in the scores of the current version?One consequence of transforming this scale into asemantic differential response format (Friborg et al,2004) was a slight reduction in internal consistency interms of Cronbach's alpha. However, Cronbach'smethod underestimates true reliability when the tau-equivalence assumption is violated (unequal itemloadings) or when the number of items decreases. Asboth of these factors were present, a structural equa-tion method was used to estimate the true and totalvariance components more directly. Using this method(Raykov, 2001), which is not negatively affected bydifferential item loadings and fewer items, showed thatthe internal consistency of all factors was adequate tohigh.Sample limitations?The applicants for the military academy used in thissample are not entirely representative of theNorwegian population: (a) they were younger than thegeneral adult population; (b) the majority were men;and (c) heterogeneity in subject variance were reduceddue increased homogeneity in personality, educational


12/15

40 Friborgetal.and work preferential background. This may reducethe generalizability for the convergent and discrimina-tive indices but it rather strengthens the validity of thefactor structure. Having an a priori model represents amuch stronger test of the factor structure, especiallywhen the implied model fits well with data from a verydifferent sample. The convergent and discriminatevalidity indices are, however, more sample specific.Still, they are expected to reproduce well on othersamples due to the relatively strong associations foundin this study, and would only represent a problem ifthey showed even stronger associations in other sam-ples. Moreover, the results come from self-reportinventories, with all the constraints that usually imply,such as social desirable responding. That may be ofconcern as the inventories were part of an intake pro-cedure for the academy, thus increasing desirableresponses. However, such problems would representitself as extraneous and systematic variance showingup in the error term. As each factor explained roughlyhalf of the variance in the items, which is normal,extraneous variables were not assumed to affect theresults significantly. Finally, one prime advantage ofutilizing this particular sample was the possibility toinvestigate the construct validity on subjects experi-encing higher degrees of stress. Although it wasreasonable to assume increased activation and stressrelated to the testing situation for acceptance at themilitary college, one obvious drawback was thatexplicit measures of stress were not included. Thisproblem may be met by including measures of stress, orconducting experimental studies inducing stress toinvestigate how well the resilience factors moderatevarious levels of stress.

Utility of the scaleMeasuring protective factors (Friborg et al, 2003) topredict positive adjustment, despite risks and stressors,is an important step in operationalizing scales measur-ing resilience more directly. That is advantageous inthat it provides the possibility of reviewing and com-paring results across studies. Measuring protectivefactors at different levels, both on an intrapersonal(personal/social competence) and an interpersonallevel (family/social resources) is also helpful in reveal-ing new insights in to how different levels of protectiv efactors interact with risk and stress factors. Increasedevidence-based knowledge about wbat kind of

resilience factors increase adjustment or coping capac-ity is relevant for the prediction and selection ofstress-tolerant personnel, as is evidence-based knowl-edge about how they do so, and to what kind ofproblems/life events, as well as for what kinds of indi-viduals. It may also guide practising therapists andresearchers in designing and planning mental healthprevention services.ReferencesAsendorpf JB, Borkenau P, Ostendorf F, van Aken MAG.Carving personality description at its joints; confirma-tion of three replicable personality prototypes for bothchildren and adults. EurJ Pers 2001; 15: 169-98.Block J. A con traria n view of the five-factor approach topersonality description. Psychol Bull 1995; 117:

187-215.Cederblad M, Dahlin L, Hagnelt O, Hansson K.Intelligence and temperament as protective factors formental health: A cross-sectional and prospectiveepidemiological study. Eur Arch Psychiatry ClinNeurosci 1995; 245: 11-19.Clausen JA. American Lives: Looking Back at theChildren of the Great Depression. New York: FreePress, 1993.Cohen J. Statistical Power Analysis for the BehaviouralSciences. New York:Academic Press, 1988.Costa PT, McCrae RR. Domains and facets: hierarchicalpersonality assessment using the Revised NEOPersonality Inventory. J Pers Ass 1995; 64: 21-50.Cowen E, Work W. Resilient children, psychological well-ness, and primary prevention. Am J CommunityPsychol 1988; 16:59 1-60 7.Cronbach LJ. Essentials of Psychological Testing. NewYork: HarperCollins, 1990.Davey M, Eaker DG, Walters LH. Resilience processes inadolescents: personality profiles, self-worth, and coping.J Adolesc Res 2003; 18:347-62.DiStefano C. The impact of categorization with confirma-tory factor analysis. Struct Equation Model 2002; 9:327-346.Egeland BR, Carlson E, Sroufe LA. Resilience as process.Dev Psychopathol 1993; 5: 517-28.Egeland BR, Sroufe LA, Erickson M. The developmentalconsequence of different patterns of maltreatment.Child Abuse Negl 1983; 7: 459-69.Engvik H. 'Big Five' in Norwegian. J Norw Psychol Ass1993; 30: 884-96.Engvik H. 5PFmil 2.0 [Computer software]. Department ofPsychology, University of Oslo, Norway, 1997.Engvik H, Hjerkinn O, Seim S. Handbook of WechslerAdult Intelligence Scale. Jaren: Vigga trykk, 1978.Friborg O. A prospective study of resilience, vulnerability.


13/15

Resilience, personality and intelligence 4 1negative life events and psychiatric symptomatology.Unpublished raw data, 2004.

Friborg O, Hjemdal O, Rosenvinge JH, Martinussen M. Anew rating scale for adult resilience: what are thecentral protective resources behind healthy adjustment?Intj Methods Psychiatr Res 2003; 12: 65-76 .

Friborg O, Martinussen M, Rosenvinge J. Likert-basedversus semantic differential-based scorings of positivepsychological con structs: a psychometric comp arison oftwo versions of a scale measuring resilience. Submittedfor publication, 2004.

Furnham A, Crump J, Whelan J. Validating the NEOpersonality inventory using assessor's ratings. PersIndividual Diff 1997; 22: 669-75.

Garmezy N. Children under stress: perspectives onantecedents and correlates of vulnerability and resis-tance to pathology. In Al Rabin, J Arnoff, AM Barclay,RA Zuckers (eds) Further Explorations in Personality.New York: Wiley Interscience, 1981, pp. 196-26 9.Hjemdal O, Friborg O, Martinussen M, Rosenvinge JH.Preliminary results from the development and valida-tion of a Norwegian scale for measuring adultresilience. J Norw Psychol Ass 2001 ; 38: 31 0-1 7.

Hu L, Bentler PM. Evaluating model fit. In RH Hoyle(ed.) Structural Equation Modeling. Concepts, Issues,and Applications. Thousands Oaks CA: Sage, 1995, pp.76-99 .

Hu L, Bentler PM. Cutoff criteria for fit indexes in covari-ance structure analysis: conventional criteria versusnew alternatives. Struct Equation Model 1999; 6: 1-55.

Joreskog KG, Sorbom D. LISREL 8: Users' ReferenceGuide. MooresviUe IN: Scientific Software Inc, 1996.Luthar SS. Vulnerability and resilience: A study of high-

risk adolescents. Child Dev 19 91; 62: 600-1 6.Luthar SS, D'Avanzo K, Hites S. Maternal drug abuse

versus other psychological disturbances: Risks andresilience among children. In SS Lutbar (ed.)Resil ience and Vulnerabil i ty: Adaptation in theContext of Childhood Adversit ies. New York:Cambridge University Press, 2003, pp. 104-29.

Martinussen M, Torjussen T. Pilot selection in theNorwegian Air Force: a validation and meta-analysis ofthe test battery. Int J Aviat Psychol 1997; 8: 33 -45 .McCrae RR. Controlling neuroticism in the measurementof stress (special issue: II-IV. Advances in measuringlife stress). Stress Med 1990; 6: 237-41 .

McC rae RR, Costa PT. Personality trait structu re as ahuman universal. Am Psychol 1997; 52: 509-16.

Murphy LB, Moriarty AE. Vulnerability, coping andgrowth. New Haven CT: Yale University Press, 1976.

Nezlek JB . Daily psychological adjustment and the planful-ness of day-to-day behavior. J Soc Clin Psychol 2001;20: 452- 75 .

Radke-Yarrow M, Sherman T Hard growing: children whosurvive. In RE Rolf, AS Masten, D Cicch ett i , KHNu echte r lein, S W eintraub (eds) Risk and ProtectiveFactors in the Development of Psychopatbology. NewYork: Cambridge University Press, 1990: 97-119.

Ramm stedt B, Rieman n R, Angleitn er A, Borkenau P.Resil ients, overcontroUers, and undercontrollers: thereplicability of the three personality prototypes acrossinformants. EurJ Pers 2004; 18: 1-14.

Raven JC. Advanced Progressive Matr ices. Oxford:Psychologist Press, 1986.

Raykov T. Estimation of congeneric scale reliability usingcovariance structure analysis with nonlinearconstraints. Br J Math Stat Psychol 2001; 54: 31 5-2 3.

Riggio RE. Assessment of basic social skills. J Pers SocPsychol 1986; 51 : 649 -60.Riolli L, Savicki V, Cep ani A. Re silience in the face ofcatastrophe: optimism, personali ty and coping in theKosovo crisis. J Appl Soc Psychol 2002; 32: 1604-27.

Ross RT. Behavioral correlates of levels of intelligence.Am J Men t Defic 1972; 76: 54 5-9 .

Ru tter M. Resilien ce in the face of adversity. Br JPsychiatry 1985; 147 :598 -611 .

Silvera DH, Martinussen M, Dahl Tl. The Troms0 SocialIntelligence Scale, a self-report measure of social intel-ligence. Scand J Psychol 200 1; 42: 313 -19 .

Stemberg RJ. Applying the tr iarchic theory of humanintell igence in the classroom. In RJ Stemberg, WMWilliams (eds) Intell igence, Instruction, andAssessment: Theory into Practice. Mahwah: LawrenceEribaum Associates Publishers, 1998, pp. 1-15.

Sull ivan HS. Psychiatry: Introduction to the study ofinterpersonal relations. Psychiatry: J Stud InterpersProcess 1938; 1: 121-34.

Vaillant GE, Davis, JT. Social/emotional intelligence andmidlife resilience in schoolboys with low tested intelli-gence. Am J Orthopsychiatry 2000; 70: 215-22.

Werner EE. Risk, resilience, and recovery: Perspectivesfrom the Kauai Longitudinal Study. Dev Psychopathol1993; 5: 503-15.

W erner EE. Journeys from ch ildh ood to midlife: risk,resilience, and recovery. Ithaca NY: Cornell UniversityPress, 2001.

Werner EE, Smith RS. Overcoming the Odds. High RiskChildren from Birth to Adulthood. Ithaca and London:Cornell University Press, 1992.


14/15

42 Friborgetal.

Appendix: the Resilience Scale for Adults, 33 itemsPersonal strengthlPerception ofselfWhen something unforeseen happens I always find a solution n n n D D I often feel bewilderedMy personal problems areunsolvable D n D D D 1 know how to solveMy abilities 1 strongly believe in D D D D D 1 amuncertain aboutMy judgements and decisions I often doubt D D D D D 1 trust completelyIn difficult periods I have a tendency to view everything gloomy D D D D find something good that help me

thriveEvents in my life that I cannot influence 1 manage to come to terms with D D D D D are a constant source of worry/concernPersonal strengt/i/Perception offuntreMy plans for the future are difficult toaccomplish n n D D D possible to accomplishMy future goals I know how to accomplish D D D D D lam imsure how to accomplish1 feel that my future looks very promising D D D D D uncertainMy goals for the future are unclear 0 0 0 0 0 well thought throughSmiciured style1 am at my best when 1 have a clear goal to strive for O O O O D can take one day at a timeWhen I start on new things/projects I rarely plan ahead, just get on with it O O O O O 1 prefer to have a thorough planI am good at organizing my time O O O O O wasting my timeRules and regular routines are abscent in my everyday life O O O O G simplify my everyday lifeSocial competence1 enjoy being together with other people O O O O O by myselfTo be flexible insocial settings is not important to me O O O O O isreally important to meNew friendships are something I make easily O O O O O 1 have difficulty makingMeeting new people is difficult for me O O O O O something I am good atWhen 1 am with others I easily laugh O O O O G 1 seldom laughFor me, thinking of good topics for conversation is difficult O O O O O easyFamily cohesionMy family's understanding of what is important in life is quite different than mine O O O O D very similar to mine' f ^ very happy with my family O O O O O very unhappy with my familyMy family is characterized by disconnection O O O O D healthy coherenceIn difficult periods my family keeps a positive outlook on the future O O O O D Views the future as gloomyFacing other people, our family acts unsupportive of one another O D O O O loyal towards one anotherIn my family we like to do things onour own O O O O O do things togetherSocial resources1 can discuss personal issues with no one O O O O O friends/family-membersThose who are good at encouraging me are some close friends/family members O O O O O nowhereThe bonds among my friends is weak O O O O O strongWhen a family member experiences a crisis/emergency 1 aminformed right away O O O O O it takes quite a while before 1 amtold1 get support from friends/family members O O O D O No oneWhen needed, 1 have noone who can help me O O O O O always someone who can helpmeMy close friends/family members appreciate my qualities O O O O O dislike my qualities

Correspondence: Oddgeir Friborg, University of Troms0,Department ofPsychology, N'9O37 Troms0, Norway.Telephone (+47) 776 45945.Fax (+47) 776 45610.Email ofribor^psyk. uit.no.


15/15

Documents

feiborg-etal-ijmpr-2005