Journal of Personality and Social Psychology1980, Vol. 38, No. 4, 668-678

Influence of Extraversion and Neuroticism on SubjectiveWell-Being: Happy and Unhappy People

Paul T. Costa, Jr. and Robert R. McCraeGerontology Research Center

National Institute on Aging, National Institutes of HealthBaltimore, Maryland

Three studies are reported that examine the relations between personality andhappiness or subjective well-being. It is argued that (a) one set of traits influ-ences positive affect or satisfaction, whereas a different set of traits influencesnegative affect or dissatisfaction; (b) the former set of traits can be viewed ascomponents of extraversion, and the latter as components of neuroticism; and(c) personality differences antedate and predict differences in happiness overa period of 10 years, thus ruling out the rival hypothesis that temporary moodsor states account for the observed relations. A model of individual differencesin happiness is presented, and the separate and complementary roles of traitand adaptation-level theories in explaining happiness are discussed.

Nowhere is the relevance of psychology tohuman concern more evident than in studiesof happiness or subjective well-being. Interestin measuring the quality of life has lead re-searchers (Andrews & Withey, 1976; Brad-burn & Caplovitz, 196S; Campbell, Converse,& Rodgers, 1976; Cantril, 1965) to conductnational surveys of happiness and to examinethe influence of social-structural or demo-graphic variables on perceived well-being.Adaptation-level (AL) theory (Helson, 1964)has been applied to explain individual percep-tions of happiness (Brickman & Campbell,1971; Brickman, Coates, & Janoff-Bulman,1978). As a result of these studies, a numberof issues have been clarified and a few unex-pected findings replicated. The present articleis an attempt to summarize the state of cur-rent knowledge on personality and happinessand to offer a model of happiness that clarifies

This research was supported in part by the Coun-cil for Tobacco Research—U.S.A., Inc., Grant 108SR2and the Medical Research Service of the VeteransAdministration (Normative Aging Study).

Requests for reprints should be sent to Paul T.Costa, Jr., Chief, Section on Stress and Coping,Gerontology Research Center, National Institute onAging, National Institutes of Health, Baltimore CityHospitals, Baltimore, Maryland 21224.

and extends it. Data are provided in supportof several parts of the model.

Conceptualizing and Measuring Happiness

Researchers have used a variety of measuresthat indicate something of the nature anddiversity of conceptions of subjective well-being. Gurin, Veroff, and Feld (1960) adoptedthe most straightforward and intuitive methodof assessing happiness: They asked subjectsif they were "very happy," "pretty happy,"or "not too happy." To obtain a more dif-ferentiated estimate of life satisfaction, Camp-bell, Converse, and Rodgers (1976) requiredsubjects to rate their satisfaction within eachof 10 areas of life—job, marriage, family, andso on. Campbell (1976) regarded this strategyas a "cognitive" appraisal of life satisfaction,since it avoided any direct reference to feel-ings or affects and allowed the subject toassess his or her satisfaction according to hisor her own standards and expectations.

Cantril (1965) used a so-called self-anchor-ing scale, in which individuals defined a"best life" and a "worst life" for themselves,and then rated their present life on this best-to-worst scale. A different approach that hasattained considerable use by researchers wasdeveloped by Bradburn and his colleagues at

In the public domain

668

INFLUENCE OF EXTRAVERSION AND NEUROTICISM ON HAPPINESS 669

the National Opinion Research Center (Brad-burn, 1969; Bradburn & Caplovitz, 196S).Instead of asking for cognitive judgments ofthe adequacy of one's life, he measured feel-ings or affects. Following the old idea thathappiness is the sum of pleasures minuspains, Bradburn developed an Affect BalanceScale by subtracting negative affects (bore-dom, loneliness, depression) experienced dur-ing the past two weeks from positive affects(pride, excitement, pleasure).

Despite the variety of approaches used tothis point, the scope and limits of the domainof measures relevant to subjective well-beingor happiness have not yet been established.In particular, scales originated in clinicalcontexts to measure dysphoric affect, hope-lessness, or insecurity may measure what werecognize within normal limits to be simpleunhappiness. The phrase "psychological well-being" itself carries with it the connotation ofmental health and has been so interpreted bymany researchers (e.g., Moriwaki, 1974; Rob-inson, 1969).

This array of alternative operationalizationsof happiness would be almost unmanageablewere it not for one happy circumstance: Thescales all show reasonably high intercorrela-tions. Bradburn (1969) found that the Nega-tive Affect Scale (NAS), the Positive AffectScale (PAS), and especially the Affect BalanceScale (ABS) scores correlated significantlywith avowals of "very happy," "prettyhappy," or "not too happy." Campbell(1976), in a national probability sample,showed that his index containing judgments oflife satisfaction in several areas of life corre-lated highly (r — .57) with an index of gen-eral affect based on semantic differential rat-ings of life on such scales as interesting-boring and enjoyable-miserable.

Moriwaki (1974) demonstrated the closekinship of morale scales to the subjectivewell-being domain when she reported a corre-lation of .61 between the ABS and the RosowMorale Scale in a small sample of elderlypersons. And in the most comprehensive studyof the subject, Andrews and Withey (1976)correlated 68 measures and indices of happi-ness (including the Gurin, Cantril, and Brad-burn scales) in five national probability sam-ples. They concluded that measures "involv-

ing a general evaluation of the respondents'life-as-a-whole from an absolute perspectivetend to cluster together. . . . Measures thattap life-as-a-whole less generally . . . showpositive relationships of varying strength tothe core cluster" (Andrews & Withey, 1976,p. 76).

In marked contrast to the apparent ease ofmeasurement in the domain of subjectivewell-being are the conceptual enigmas thathave emerged in the course of research. Thefirst of these is the meager relation betweenobjective and subjective indicators of happi-ness or well-being. Common sense suggeststhat wealth, youth, and social privilege shouldcontribute substantially to happiness, andmuch research has been devoted to an investi-gation of this hypothesis. Yet Campbell(1976) reports that only 17% of life satis-faction is predictable from 10 demographicindicators in a national probability sample.Similarly, Andrews and Withey (1976), alsousing national probability samples, accountfor only &% of the variance in life satisfac-tion using age, family cycle stage, family in-come, education, race, and sex as predictors,either singly or in combination. An even moredramatic instance of the apparent irrelevanceof objective circumstances to subjective well-being is provided by Brickman, Coates, andJanoff-Bulman (1978), who report that lot-tery winners were no happier than controls onpresent and estimated future happiness. Para-plegics, although somewhat less happy thancontrols, did not differ from lottery winnersor controls in estimation of future happiness.

The second problem is a paradox that hasnever been fully explained. In 1969, Brad-burn reported that when positive and nega-tive affects are independently measured, theitems form two independent clusters. Al-though positive and negative carry the strongmathematical suggestion of being opposite,Bradburn's PAS and NAS were not opposite(negatively correlated) but independent, vir-tually uncorrelated. Despite this, both positiveand negative affects were found to be associ-ated with overall estimates of happiness.Using a slight modification of the three Brad-burn scales, Lowenthal, Thurner, and Chiri-boga (197S) confirmed that positive and nega-tive affect were independent predictors of

670 PAUL T. COSTA, JR. AND ROBERT R. McCRAE

global happiness. Similarly, Costa and Mc-Crae (Note 1) found a median correlation ofonly —.11 between PAS and NAS across fouradministrations. Andrews and Withey (1976)also replicated the finding, using both thethree Bradburn scales and "cognitive" itemsthat required respondents to assess the"good" and "bad" aspects of their life sepa-rately. It is not surprising that pleasant emo-tions enhance life satisfaction or that un-pleasant emotions diminish it, but the re-peated observation that the pleasantness andunpleasantness of one's life are uncorrelatedis a puzzling phenomenon the explanation forwhich is of considerable theoretical impor-tance.

Personality Correlates oj SubjectiveWell-Being

Personality descriptions of happy personsgenerally resemble descriptions of psychologi-cal and social adjustment. Smith (1961), forexample, lists as correlates of happiness:optimism, warmth, emotional stability, socia-bility, and self-insight. Wessman and Ricks(1966), in their intensive study of a smallsample of Harvard and Radcliffe students,point to large negative correlations with theMinnesota Multiphasic Personality InventoryD scale and the 16 Personality Factor (PF)0 or "guilt-prone" scale in justifying theirconclusion that characteristically happier peo-ple are well-adjusted, high in ego strength,and high in self-esteem, as well as beingsocially involved.

Wilson (1967), in his studies, found socialand family adjustment and self-ideal congru-ence to be important correlates of happinessand concluded his review of the literature byasserting that "happiness is consistently re-lated to successful involvement with people"(p. 304). At the unhappy pole, a number ofinvestigators have found signs of psycho-pathology or neuroticism among unhappypeople. Veroff, Feld, and Gurin (1962) re-ported worry, anxiety, and psychosomaticconcerns among the correlates of unhappiness,as did Bradburn and Caplovitz (1965). Fi-nally, self-rated health has also recurred as animportant predictor of subjective well-being

(Palmore & Kivett, 1977; Wolk & Telleen,1976).

The prominence of the Bradburn scales hasbegun to encourage investigators to look forpersonality correlates of positive and negativeaffect separately, and some revealing trendshave begun to emerge. Moriwaki (1974) re-ported that a nine-item mental health scalewas significantly related to NAS but not toPAS. Beiser (1974) found that reports ofpsychophysiological disorders were associatedwith a negative affect factor but not with apositive affect factor in his instruments. Onthe other hand, role planning and social par-ticipation were associated with positive but notnegative affect factors. Recently, Bradburn(Note 2) has pointed out a similar trend in hisdata. He reports that positive affect exclu-sively is related to social interest, sociability,and activity and that negative affect only isassociated with psychosomatic symptoms,anxiety, poor role adjustment, and worries.

These findings suggest a hypothesis thatmay explain the independence of positive andnegative affect and meaningfully organize thebody of evidence on personality and happi-ness. It is hypothesized that one set of dispo-sitions is responsible for positive affect orsatisfaction, whereas another, independent setof dispositions influences negative affect ordissatisfaction.

In Study 1 the relation between four mea-sures of happiness and seven personality dis-positions hypothesized to be related to posi-tive or negative affect is examined. In Study2 an attempt is made to clarify and organizethe results by testing the original hypothesisusing measures of the broader dimensions ofextraversion (E) and neuroticism (N). Fi-nally, in Study 3 happiness is predicted fromE and N data obtained 10 years previously.

Study 1

From the large number of traits with re-ported associations to subjective well-being,some emerge as more likely to be associateduniquely with one side or the other of theaffect balance formula. Some of the specificfacets of temperament articulated by Bussand Plomin (1975) fall into this category.Buss and Plomin trace negative affects (par-

INFLUENCE OF EXTRAVERSION AND NEUROTICISM ON HAPPINESS 671

ticularly fear and anger) to strength of emo-tional drive, and they speculate that "Ifthere is temperamental input into individualdifferences in positive emotions, it is likely tobe activity (for elation) or sociability (forfriendliness and warmth)" (p. 57). Some evi-dence supporting this idea is offered in termsof daily level of mood ratings, and the divi-sion of traits agrees generally with Bradburn's(Note 2) observations. Study 1 tests thehypothesis that the temperamental traits ofemotionality, tearfulness, hostility, and im-pulsivity will be associated with lower levelsof happiness and especially with high negativeaffect, and that temperamental traits of so-ciability and activity will be associated withhigher levels of happiness and, particularly,with positive affect.

Method

Subjects. The data reported were collected aspart of a project on smoking and personality con-ducted in collaboration with the Normative AgingStudy, an interdisciplinary longitudinal study ofhealth and aging in men (Bell, Rose, & Damon,1972). Participants, volunteers screened for healthand geographical stability, ranged in age from 35 to85 at the time of this research. The sample consistslargely of white veterans, with all but the lowestsocioeconomic groups well represented. A subsampleof 1,100 men was contacted by mail and was askedto complete a series of four questionnaires mailed atintervals of 3 months in 1976. Response rates were79%, 82%, 73%, and 54% for the four mailings.Additional information was obtained on a sub-sample of 172 subjects visiting the study center forregular medical examinations during the data collec-tion period. Results are based on all available cases,with specific ns given in each table.

Measures. Four measures of happiness were col-lected. The principal measure was Bradburn's (1969)scales, which yielded scores for PAS, NAS, and thedifference of these scores, ABS. The Bradburn mea-sures were obtained at each of the four quarterlymailings. An ABS sum score was calculated by sum-ming the ABS scores over the four administrations.

The Hopelessness Scale (Beck, Weissman, Lester,& Trexler, 1974) was given in the third question-naire mailing. The scale was developed to assesshopelessness or pessimism in clinical populations. Ina sample of 294 hospitalized patients, internal con-sistency was found to be .93. Correlations withclinicians' ratings of hopelessness ranged from .62 to.74. Little evidence of the test's applicability tononpsychiatric populations has yet been provided.

The Personal Security Inventory (Knutson, 1952)was designed to measure personal security as thesubjective evaluation of ". . . success, satisfaction,

and surety or confidence" (p. 24) in a number ofareas of life. Knutson reports validation studies inwhich psychiatric patients scored significantly lowerthan normals; personal security was also found to bepositively related to occupational status. A shortened,16-item form of the Personal Security Inventorywas given as part of the fourth mailing.

Those subjects who came to the study for medicalexaminations during the data collection period wereasked to complete the Life Satisfaction Index. Foreach of nine areas (work, health, money, appear-ance, self-respect, getting along with others, love,sex, and religious faith) subjects rated their satis-faction on a 5-point scale. Internal consistency(coefficient alpha) for the summed score was .84 inour sample of 172.

Scales from the EASI-III Temperament Survey(Emotionality/Activity/Sociability/Impulsivity; Buss& Plomin, 1975) were included in the second mail-ing. Each scale consists of five items and was de-veloped through factor analysis and rating valida-tion. Two- to three-month test-retest reliabilitiesaveraged .79 in a sample of 32 women; self-reportscorrelated .51 on the average with spouse ratings ina sample of 137 couples. Scales hypothesized to relateto NAS included general emotionality, fear, anger,and poor inhibition of impulse; scales hypothesizedto relate to PAS included sociability, tempo, andvigor.

Analyses. Pearson correlations among the fourhappiness measures were used to examine evidencefor convergent validity. Correlations were then com-puted between happiness and temperament mea-sures. Finally, correlations between temperamentsand the components of ABS were calculated for eachof the four time points to examine the hypothesizedrelations of the measured traits to PAS and NASover four replications.

Results and Discussion

Correlations between the three Bradburnscales and the alternative happiness measuresat four times follow the pattern reported inthe literature: PAS and NAS scores are sig-nificantly related to happiness measures in 23of 24 cases, but in every case ABS is morehighly correlated with the Beck scale, theKnutson inventory, and the index than eitherof its components is. Of the Bradburn scales,the ABS thus appears to measure happinessbest.

Table 1 presents the intercorrelations of theABS Sum and the three other happiness mea-sures and suggests that different strategies orinstruments for measuring subjective well-being produce similar results. All correlationsare significant and are generally high enoughto suggest convergent validity for the mea-

672 PAUL T. COSTA, JR. AND ROBERT R. McCRAE

Table 1Intercorrelations of Happiness Measures

Measure 1

1.2.3.4.

ABS sumPersonal securityHopelessness"Life Satisfaction

Index

524529

82

.64**

552

93

Note, ns are given below the diagonal.

.61**

.59**

135

ABS =

.40**

.32**

.18*

—

AffectBalance Scale." Scale reflected to "hopefulness."* p < .05. **£<.001.

sures. The Life Satisfaction Index appears tobe a weaker indicator of happiness in thisgroup, but Campbell's (1976) correlation of.57 between a similar index and a happinessmeasure strengthens the argument for treat-ing the index as a measure of happiness.

The temperamental correlates of happiness,shown in Table 2, are also consistent with theliterature. Happiness is positively associatedwith sociability and activity and negativelyassociated with emotionality and impulsivity.

All of the 28 correlations are in the pre-dicted direction, and 27 of them are statisti-cally significant. The median correlation is.24, a value that compares favorably withmultiple correlations of .41 (Campbell, Con-verse, & Rodgers, 1976) or .28 (Andrews &Withey, 1976) reported by survey researcherswhen demographic characteristics are used topredict life satisfaction.

The hypothesis that the scales of generalemotionality, fear, anger, and poor inhibitionof impulse influence primarily negative affect,whereas sociability, tempo, and vigor scales

will influence primarily positive affect, istested in Table 3.

For three of the temperament scales—gen-eral emotionality, anger, and poor inhibitionof impulse—it is clear that only the negativeside of affect balance is substantially related.In each case, correlations with NAS are higherthan with the corresponding ABS score. Thepattern for the fear scale is not so clear: Ateach time, fear is more closely related to NASthan to PAS, but it does show a consistenteffect on lowering PAS as well.

Similarly, two of the positive temperamentscales—tempo and vigor—are associated withPAS but not with NAS, as predicted. Socia-bility is more closely related to PAS than toNAS, as predicted, but it also appears tohave a consistent effect on NAS.

Thus five of the seven scales appear toinfluence happiness by their impact on onlyone side of the affect balance equation. Twoothers show their primary effect on the hy-pothesized affect component but also showsome influence on the other as well.

Study 2

The hypothesis that some traits influencepositive affect and some influence negativeaffect was generally supported by the resultsof Study 1. It is possible to take these resultsone step further by noting that these traitshave an internal organization and coherence.To those familiar with factor models of per-sonality, the list of traits provided by Study1 (and much previous research) begins totake the shape of two established dimensionsof personality: extraversion (E) and neuroti-

Table 2Correlations of Temperament Scales With Happiness Measures

Measure

ABS sum (529)Hopelessness" (757)Personal security (563)Life Satisfaction Index (149)

Generalemotion-

ality

-.33***-.33***-.25***-.21**

Fear

-.40***_-41***- .40***-.32***

Anger

-.21***-.19***-.09*-.12

Poorinhibitionof impulse

-.22***-.23***-.16***-.15*

Socia-bility

.32***

.25***

.32***

.24***

Tempo

.13**

.09**

.12**

.21**

Vigor

.28***

.18***

.28***

.32***

Note, ns are given in parentheses. ABS• Scale reflected to "hopefulness."* p < .05. ** p < .01. *** p < .001.

Affect Balance Scale.

INFLUENCE OF EXTRAVERSION AND NEUROTICISM ON HAPPINESS 673

Table 3Correlations of Temperament Scales With Bradburn Scales at Four Times

Time

1234

1234

1234

n

823903757566

823903757566

823903757566

Generalemotionality

-.07**-.06*-.13***-.05

.30***

.38***

.34***

.32***

-.25***_-29***-.28***-.23***

Fear

-.18***-.20***-.22***-.17***

.26***

.34***

.28***

.33***

-.29***-.36***-.31***-.32***

Anger

Positive Affect

-.00-.02-.02

.00

Negative Affect.24***.26***.21***.20***

Affect Balance

-.17***-.18***_.14***-.12***

Poorinhibitionof impulse

Scale

-.03-.06*-.08*-.04

Scale

.21***

.28***

.22***

.24***

Scale

-.16***-.23***-.18***-.18***

Sociability

.23***

.24***

.24***

.22***

-.13***-.20***-.20***-.13***

.24***

.29***

.28***

.23***

Tempo

.17***

.21***

.14***

.19***

.05

.05

.02

.04

.08*

.10***

.08*

.10**

Vigor

.23***

.24***

.23***

.29***

-.06*-.06*-.06-.08*

.19***

.20***

.18***

.24***

* p < .05. ** p < .01. *** p < .001.

cism (N). And indeed, factor analyses (Costa& McCrae, in press) showed that the EASI-III scales of general emotionality, fear, anger,and poor inhibition of impulse defined an Nfactor, whereas sociability, tempo, and vigorformed part of an E factor. It is now possibleto propose a model of the relations betweenpersonality and happiness. Extraversion, to-gether with its component traits of sociabil-ity, tempo, and vigor, predisposes individualstoward positive affect, whereas neuroticism(and hence general emotionality, impulsivity,fear, and anger) predisposes individuals towardnegative affect. The simplest test of themodel is direct correlation of measures of Eand N with happiness measures.

Method

Subjects. Study 2 employed the same sample andprocedures as Study 1.

Measures. Two measures of N and E were ad-ministered to the subjects by mail, as in Study 1.Cluster analysis of the Cattell Sixteen PersonalityFactor Questionnaire (16 PF) scales (Costa & Mc-Crae, 1976) had shown an anxiety or N cluster andan E cluster that closely resembled the major second-order factors reported for the 16 PF (Cattell, Eber,Si Tatsuoka, 1970). Multiple regression was used toidentify the IS items in Form A that best predicted

full N and E scores in a sample of 969 men. These30 items were used as short-form N and E clusterscales and were included in the third mailing. Addi-tionally, the standard Form A of the Eysenck Per-sonality Inventory (EPI: Eysenck & Eysenck, 1964)was included in the fourth mailing. The theoreticallyindependent dimensions of E and N were empiricallyuncorrelated in the present sample (r = .00, « = 808for short-form 16 PF scales; r = -.02, n = S76 forEPI scales). Evidence of convergent validity is seenin the correlation of .65 (n — 549) between the 16PF and EPI E measures and .68 (n = 553) betweenthe two N measures.

Results and Discussion

Table 4 shows the Pearson correlations ofE and N measures with the Bradburn scalesat four times, 3 months apart.

In all eight (Time X Measures) cases,neuroticism or anxiety is more strongly corre-lated with NAS than with either PAS or ABS.In all eight cases, extraversion is morestrongly correlated with PAS than with NASor, in six of the eight cases, than with ABS.

When E and N measures are correkted withthe three alternative operationalizations ofhappiness—hopelessness, personal security,and the Life Satisfaction Index—11 of the 12correlations are statistically significant, and

674 PAUL T. COSTA, JR. AND ROBERT R. McCRAE

Table 4Correlations of 16 PF and EPI Scales With Bradburn Scales at Four Times

16 PF Short-Form scales

Time N

1 -.11***2 -.063 -.16***4 -.10**

1 .29***2 .41***3 .40***4 .34***

1 -.27***2 -.31***3 -.34***4 -.27***

E n

Positive Affect Scale.16*** 753.22*** 757.19*** 808.25*** 556

Negative Affect Scale

- .03 753-.04 757-.13*** 808-.12** 556

Affect Balance Scale

.12*** 753

.17*** 757

.20*** 808

.25*** 556

EPI scales

N

-.11**-.08**-.17***-.15***

.35***

.38***

.39***

.43***

-.31***-.32***-.34***-.39***

E

.16***

.21***

.17***

.27***

-.01-.01-.05-.07*

.11**

.15***

.15**

.22***

n

554559549575

554559549575

575559549575

Note. PF = Cattell Sixteen Personality Factor Questionnaire. EPI = Eysenck Personality Inventory.N = neuroticism. E = extraversion.* p < .05. ** p < .01. *** p < .001.

all are in the predicted direction. Thus, Eand N not only influence the experience ofpositive or negative affect; they also showconsistent correlations with measures of hap-piness that do not depend on direct reportsof affective experience.

Most factorial-based trait systems recognizeN and E as the broadest and most pervasivedimensions of personality. Eysenck (Eysenck& Eysenck, 1969) has devoted most of hisresearch to an investigation of these two di-mensions directly. Cattell (1973) sees themas second-order traits and has attempted tomeasure the more molecular, first-order as-pects of personality that form them. Guilford(1976) prefers to call the E cluster socialactivity and the N cluster emotional health,but the similarity of these schemes is beyondquestion. The bulk of the literature on thepersonality correlates of happiness can besummarized by saying that more extravertedand more adjusted people are happier. Thecharacteristics listed somewhat indiscrimi-nately under the heading of "psychologicaland social adjustment" can now be brokenapart into two discrete groups. Under the

heading of E come sociability, warmth, in-volvement with people, social participation,and activity. Under N come such character-istics as ego strength, guilt proneness, anxiety,psychosomatic concerns, and worry. Extra-verted traits contribute to one's positive en-joyment or satisfaction in life, although theydo not generally appear to reduce the un-pleasantness of adverse circumstances. Neu-rotic traits predispose one to suffer moreacutely from one's misfortunes, but they donot necessarily diminish one's joy or pleasures.

Study 3

Studies 1 and 2 made the causal assumptionthat personality influenced happiness orsubjective well-being—an interpretation thatsimple correlations cannot themselves sustain.Although it appears unlikely that temporarystates of happiness would substantially alterpersonality, it is plausible to argue that short-term moods or states may affect responses topersonality measures. Perhaps an individual inan upbeat mood will respond like an extravert,whereas the person who is temporarily de-

INFLUENCE OF EXTRAVERSION AND NEUROTICISM ON HAPPINESS 675

pressed will score high on neuroticism. Thelong-term stability of E and N (Costa & Mc-Crae, 1977, in press; Moss & Susman, inpress) argues against this interpretation, but amore direct test is given by an examination ofthe predictive relations between personalitymeasures and levels of subjective well-beingobtained 10 years later. Over this long a timespan, any systematic bias introduced bytemporary moods or states should be elimi-nated. Predictive relations between E, N, andhappiness would thus strengthen the con-temporaneous evidence for the proposed model.

Method

Subjects. Subjects were a subset of those de-scribed in Study 1 who had been given the 16 PFbetween 1965 and 1967. Data were available for 234men.

Measures. An N and an E cluster score wereobtained from analysis of combined A and B formsof the 16 PF (Costa & McCrae, 1976). (These scoresformed the criterion in the selection of items forthe short-form 16 PF scales described in Study 2.)These clusters resemble the second-order factorsreported by Cattell; evidence for their validity andstability is reported elsewhere (Costa & McCrae,1977).

Resets and Discussion

N cluster scores were significantly relatedto NAS (r = .39, p < .001) and to ABS (r= -.30, p < .001) but not to PAS (r =— .08, ns). E cluster scores, by contrast, werenot related to NAS (r = .03, ns) but wererelated to PAS (r = .23, p < .001) and ABS(r = .14, p < .05). Knowing an individual'sstanding on these two personality dimensionsallows a prediction of how happy the personwill be 10 years later. These data effectivelyrule out the alternative explanation thatassociations between happiness and personal-ity result solely from the mediating effect oftemporary moods or states. This finding isalso impressive as indirect evidence of theenduring effects of these dimensions of person-ality.

General Discussion

A Model of Happiness

Figure 1 presents a model of happiness thataccounts for the correlational data reportedhere and in the literature.

EXTRAVERSION:SOCIABILITYTEMPOVIGORSOCIAL

INVOLVEMENT

NEUROTICISM:ANXIETYHOSTILITYIMPULSMTYPSYCHOSOMATIC

COMPLAINTS

*• POSITIVE AFFECT:SATISFACTION

NEGATIVE AFFECT:DISSATISFACTION

Figure 1. A model of personality influences on posi-tive and negative affect on subjective well-being.

The personality traits found to be corre-lates of happiness have been grouped underthe headings of E and N. The direct outcomesof these dispositions, according to the model,are positive affect and negative affect, re-spectively. These two components are sub-jectively "balanced" by the individual to ar-rive at a net sense of subjective well-being,which may be measured as morale, life satis-faction, hopefulness, or simply happiness.

Although it has been known for some timethat positive and negative affect were inde-pendent contributions to global happiness, noone has ever provided a fully satisfactory ex-planation for this phenomenon. It is clearthat there must be two independent sourcesof variation, two sets of causes operating toproduce the two independent effects. In hisinitial attempt at an explanation, Bradburnlooked for objective sources. He suggestedthat the situations that contribute to positiveaffect are separate from those that contributeto negative affect. It is easy to find examplesthat support the plausibility of this position.Thus, poor health makes one unhappy, butgood health is taken for granted, not seen asa source of positive joy. Conversely, a hobbymay bring us considerable pleasure, but fewof us have hobbies that carry much potentialfor pain.

Plausible as the objective sources idea is, itrests more on speculation than on fact. Someof the available facts even contradict it. Forexample, we might expect that when we asksubjects to rate their level of satisfaction withseparate parts of their life—job, marriage,money, religion, and so on—there would belittle correlation between them, even thoughall might contribute to overall happiness.Instead, such items tend to intercorrelate sub-stantially in our own data and that of others

676 PAUL T. COSTA, JR. AND ROBERT R. McCRAE

(e.g., Campbell, Converse, & Rodgers, 1976).Regardless of the area of life, people tend tobe either satisfied or dissatisfied. The twosources of variation must lie within the per-son, and the dimensions of E and N are primecandidates.

Figure 1 calls attention to the separabilityof satisfaction from dissatisfaction, a phe-nomenon somewhat foreign to "common sense"notions of happiness. We tend to assume thatthese two components are opposites and thatmore of one means less of the other. The datashow that reality is more complex. The differ-ence between common sense and the model isseen most clearly when the traits of E and Nare considered in combination. Low N intro-verts and high N extraverts may have similarlevels of life satisfaction or happiness, butthey achieve this result in utterly differentways. The former are seldom depressed butjust as seldom elated. The latter are prone toboth extremes and reach "average" satisfac-tion only because there is as much satisfactionas dissatisfaction in their lives. In some re-spects, the two groups are similar, but futurestudies should also be sensitive to the manydifferences between individuals who may showthe same level of subjective well-being.

Finally, Figure 1 points out a need for con-ceptual clarification of the relation betweenhappiness and mental health. Many research-ers consider the Bradburn scales a measure ofmental health, and our finding that the ABSis strongly related to two clinically validatedscales—the Beck Hopelessness Scale and theKnutson Personal Security Inventory—mightreinforce this notion. If these scales were onlyassociated with N, there would be little reasonto object to considering them as measures ofadjustment, since N is clearly a conceptualcorrelate of mental illness. But the happinessscales also reflect E, whose conceptual relationto mental health is by no means unequivocal.The use of happiness scales as criteria ofadjustment portrays introverts as less men-tally healthy than extraverts; scientific re-searchers should consider whether they wishto so penalize introverts. The independence ofE and N argues that introverts are no moreprone to anxiety, depression, or anger than areextraverts. Whether they should be consideredlower in mental health simply because they

show less zest, vigor, or enthusiasm is athorny question, conceptual rather thanempirical in nature. This caution appliesparticularly to social gerontologists (Lemon,Bengtson, & Peterson, 1972; Neugarten,Havighurst, & Tobin, 1961) who have used ascriteria of adjustment in old-age measuresthat include "zest" as a component of lifesatisfaction.

Happiness, Personality, and Adaptation Level

Few would argue against the position that,for normal people, the major determinant ofmomentary happiness is the specific situationin which the individual finds himself or her-self. Social slights hurt our feelings, tooth-aches make us miserable, compliments raiseour spirits, eating a good meal leaves us satis-fied. The contribution of personality to anyone of these feelings is doubtless small. Yetover time, the small but persistent effects oftraits emerge as a systematic source of varia-tion in happiness, whereas situational deter-minants that vary more or less randomly tendto cancel each other out (cf. Epstein, 1977).

On the other hand, the finding that traitspredict happiness more successfully than suchenduring objective conditions as health,wealth, sex, or race is more problematic. Weall believe that we would be happier if wehad more vigor, money, or power. Surveyresearch data, however, show that these cir-cumstances have very limited impact on sub-jective estimates of well-being. Brickman hasproposed that adaptation-level theory canaccount for these facts. Brickman and Camp-bell (1971) state as the fundamental postu-late of AL theory that "the subjective expe-rience of stimulus input is a function not ofthe absolute level of that input but of thediscrepancy between the input and past levels"(p. 287). As applied to happiness, this meansthat the standards by which people judge thepleasantness or unpleasantness of events orcircumstances are not absolute but relative, setand reset by the positive and negative experi-ences of the individual. According to this view,habituation makes extreme circumstances (likegreat wealth or great poverty) appear morenormal to the individual concerned, who comesto take advantages for granted or learns to

INFLUENCE OF EXTRAVERSION AND NEUROTICISM ON HAPPINESS 677

live with misfortunes. Additionally, AL theorypredicts that in contrast to extreme events,more mundane experiences will be devalued.In this way, highly favorable circumstancesdeprive the individual of many routine plea-sures. Brickman found that lottery winnerstook less pleasure in such small matters aswatching television than did control subjects(although the correlative prediction that para-plegics would take more pleasure in mundaneactivities was not confirmed).

AL theory can thus be used to explain thesmall magnitude of the effects of objectivecircumstances on well-being. As a theory ofhappiness, however, it fails to account for thelarge observable individual differences in hap-piness. Indeed, if happiness were solely theoutcome of processes of adaptation, we wouldexpect that all individuals would answer thatthey were "neutral" on the dimension of hap-piness. Brickman and Campbell speak gloom-ily about the "hedonic treadmill" and give theimpression that no one can remain happy forlong. Yet Gurin reports that 35% of thepopulation considers itself to be "very happy,"and the stability coefficients of well-beingmeasures, which range from .4 to .5 (An-drews & Withey, 1976; Costa & McCrae, Note1; Palmore & Kivett, 1977), indicate thatpeople tend to stay at the same relative levelof happiness over long periods of time. Thisrelative stability in well-being is most easilyinterpreted as an outcome of the stability ofpersonality dimensions that underlie charac-teristic levels of happiness.

AL theory predicts that subjective judg-ments will be a function of the discrepancy ofpresent stimuli from a neutral point deter-mined by past experience. In the case ofhappiness, it appears that this formulation isinsufficient. Constants representing the contri-butions of E and N must be added to thevalue predicted by AL theory. People willadapt to changing circumstances, which in thelong run will neither add to nor detract fromtheir happiness. But throughout these changes,the absolute advantages of being more extra-verted or less neurotic will continue. We mayall be on hedonic treadmills, but the tread-mills of adjusted extraverts are much happierplaces to be.

Reference Notes

1. Costa, P. T., Jr., & McCrae, R. R. The relationsbetween smoking motives, personality, and feel-ings, Progress Report 111. Boston: University ofMassachusetts at Boston, 1977.

2. Bradburn, N. The measurement of psychologicalwell-being. In Jack Ellinson (Chair), Health goalsand health indicators. Symposium presented at themeeting of the American Association for theAdvancement of Science, Denver, 1977.

References

Andrews, F. M., & Withey, S. B. Social indicators ofwell-being: Americans' perceptions of life quality.New York: Plenum, 1976.

Beck, A. T,, Weissman, A., Lester, D., & Trexler, L.The measurement of pessimism: The HopelessnessScale. Journal of Consulting and Clinical Psychol-ogy, 1974, 42, 861-86S.

Beiser, M. Components and correlates of mentalwell-being. Journal of Health and Social Behavior,1974, IS, 320-327.

Bell, B., Rose, C. L., & Damon, A. The normativeaging study: An interdisciplinary and longitudinalstudy of health and aging. Aging and Human De-velopment, 1972, 3, 5-17.

Bradburn, N. M. The structure of psychological well-being. Chicago: Aldine, 1969.

Bradburn, N. M., & Caplovitz, D. Reports on hap-piness: A pilot study of behavior related to mentalhealth. Chicago: Aldine, 196S.

Brickman, P., & Campbell, D. T. Hedonic relativismand planning the good society. In M. H. Appley(Ed.), Adaptation level theory: A symposium.New York: Academic Press, 1971.

Brickman, P., Coates, D., & Janoff-Bulman, R.Lottery winners and accident victims: Is happinessrelative? Journal of Personality and Social Psy-chology, 1978, 36, 917-927.

Buss, A. H., & Plomin, R. A temperament theory ofpersonality development. New York: Wiley, 1975.

Campbell, A. Subjective measures of well-being.American Psychologist, 1976, 31, 117-124.

Campbell, A., Converse, P. E., & Rodgers, W. L.The quality of American life: Perceptions, evalua-tions, and satisfactions. New York: Russell SageFoundation, 1976.

Cantril, H. The pattern of human concerns. NewBrunswick, N.J.: Rutgers University Press, 1965.

Cattell, R. B. Personality and mood by question-naire. San Francisco: Jossey-Bass, 1973.

Cattell, R. B., Eber, H. W., & Tatsuoka, M. M.The handbook for the Sixteen Personality FactorQuestionnaire. Champaign, 111.: Institute for Per-sonality and Ability Testing, 1970.

Costa, P. T., Jr., & McCrae, R. R. Age differencesin personality structure: A cluster analytic ap-proach. Journal of Gerontology, 1976, 31, 564-570.

678 PAUL T. COSTA, JR. AND ROBERT R. McCRAE

Costa, P. T., Jr., & McCrae, R. R. Age differences inpersonality structure revisited: Studies in validity,stability, and change. Aging and Human Develop-ment, 1977, 8, 261-275.

Costa, P. T., Jr., & McCrae, R. R. Still stable afterall these years: Personality as a key to some issuesin aging. In P. B. Baltes & 0. G. Brim, Jr. (Eds.),Life span development and behavior (Vol. 3).New York: Academic Press, in press.

Epstein, S. Traits are alive and well. In P. Magnus-son & N. S. Endler (Eds.), Personality at thecrossroads: Current issues in interactional psy-chology. Hillsdale, N.J.: Erlbaum, 1977.

Eysenck, H. J., & Eysenck, S. B. G. Manual of theEysenck personality inventory. San Diego, Calif.:Educational & Industrial Testing Service, 1964.

Eysenck, H. J., & Eysenck, S. B. G. Personalitystructure and measurement. San Diego, Calif.:Robert R. Knapp, 1969.

Guilford, J. S., Zimmerman, W. S., & Guilford, J.P. The Guilford-Zimmerman temperament surveyhandbook: Twenty-jive years of research andapplication. San Diego, Calif.: Educational & In-dustrial Testing Service, 1976.

Gurin, G., Veroff, J., & Feld, S. Americans view theirmental health. New York: Basic Books, 1960.

Helson, H. Adaptation-level theory. New York:Harper & Row, 1964.

Knutson, A. Personal security as related to station inlife. Psychological Monographs, 1952, (5(5(4, WholeNo. 336) pp. 1-31.

Lemon, B. W., Bengtson, V. L., & Peterson, J. A. Anexploration of the activity theory of aging: Ac-tivity types and life satisfaction among in-moversto a retirement community. Journal of Geron-tology, 1972, 27, 511-523.

Lowenthal, M. F., Thurner, M., & Chiriboga, D.

Four stages of life. San Francisco: Jossey-Bass,1975.

Moriwaki, S. Y. The affect balance scale: A valid-ity study with aged samples. Journal of Geron-tology, 1974, 29, 73-78.

Moss, H. A., & Susman, E. J. Constancy and changein personality development. In 0. G. Brim, Jr., &J. Kagan (Eds.), Constancy and change in humandevelopment. Cambridge, Mass.: Harvard Uni-versity Press, in press.

Neugarten, B. L., Havighurst, R. J., & Tobin, S. S.Measurement of life satisfaction. Journal of Geron-tology, 1961, 16, 134-143.

Palmore, E., & Kivett, V. Change in life satisfaction:A longitudinal study of persons aged 46-70.Journal of Gerontology, 1977, 32, 311-316.

Robinson, J. P. Life satisfaction and happiness. InJ. P. Robinson & P. R. Shaver (Eds.), Measuresof social psychological attitudes. Ann Arbor, Mich.:Institute for Social Research, 1969.

Smith, H. C. Personality adjustment. New York:McGraw-Hill, 1961.

Veroff, J., Feld., & Gurin, G. Dimensions of sub-jective adjustment. Journal of Abnormal and So-cial Psychology, 1962, 64, 192-205.

Wessman, A. E., & Ricks, D. F. Mood and person-ality. New York: Holt, Rinehart & Winston,

1966.Wilson, W. Correlates of avowed happiness. Psycho-

logical Bulletin, 1967, 67, 294-306.Wolk, S., & Telleen, S. Psychological and social

correlates of life satisfaction as a function ofresidential constraint. Journal of Gerontology,1976, 31, 89-98.

Received April 4, 1979