DIFFERENTIAL PREDICTIVE POWER OF THE ...

PsychologicalRepor~,1994, 75, 1507-1515.

O Psychological Reports

1994

DIFFERENTIAL PREDICTIVE POWER O F T H E POSITIVELY VERSUS T H E NEGATIVELY WORDED ITEMS O F T H E LIFE ORIENTATION TEST' JULIAN C. L. LA1

City Polytechnic of Hong Kong Summary.-A self-report measure of dispositional optimism, the Life Orientation Test, was administered to a group of 202 Hong Kong undergraduates. Consistent with prior findings, factor analysis yielded a two-factor solution with all positively worded items loaded on the first factor and all the negatively worded items loaded on the second. Prediction of physical symptom reports from scores on the two subscales was then tested with 85 subjects randomly selected from the original sample. Only the complete test and the subscale defined by the positively phrased items predicted symptom levels concurrently as well as prospectively over 3 wk. The negative subscale suggested by previous research as tapping pessimism rather than dispositional optimism showed no significant correlation with symptom levels. Moreover, when scores of the positive rather than the negative subscale were controlled, the significant correlation between scores on the Life Orientation Test and symptom reports was eliminated. These findings suggested a multidimensional view of the test and that the positive subscale may be sufficient to measure optimism validly. Implications of these for the personality dimensions of positive versus negative affectivity are also discussed.

In the seminal work of Scheier and Carver (1985), dispositional optimism was conceptualized as generalized positive outcome expectancies having important health implications. Operationalized by the Life Orientation Test developed by these two investigators, optimists, in comparison with their pessimistic peers, were less bothered by stress-related physical symptoms (Scheier & Carver, 1987). This has been suggested to be mediated by the tendency of optimists to adopt more effective problem-focused strategies when coping with stress (Scheier, Weintraub, & Carver, 1986). The health benefits associated with dispositional optimism have been further confirmed recently in various stressed populations such as cancer patients (Carver, Pozo, Harris, Noriega, Scheier, Robinson, Ketcham, Moffat, & Clark, 1993; Friedman, Nelson, Baer, & Lane, 1992), coronary patients (Desharnais, Godin, Jobin, Valois, & Ross, 1990; Scheier, Matthews, Owens, Magovern, Lefebvre, Abbot, & Carver, 1989), and gay men at risk for AIDS (Taylor, Kemeny, Aspinwall, Schneider, Rodriguez, & Herbert, 1992). Although the link between optimism and different health measures has been repeatedly reported, controversies concerning the factorial structure of the Life Orientation Test have arisen. In one study using a Dutch-adapted version, Mook, Kleijn, and van der Ploeg (1992) identified two factors with 'Address enquiries to J. C. L. Lai, De artment of Applied Social Studies, City Polytechnic of Hong Kong. Hong Kong, 83 Tat Chee Avenue,

owi ion,

all the positively worded items loaded on the first factor and all the negatively worded items loaded on the first factor and all the negatively worded items on the second. The correlation between these two factors was only .31. The first factor was labelled as "presence of optimism" whereas the second factor was called "absence of pessimism." Mook, et al. (1992) reasoned that separation of the two factors may reflect the independence between the dimensions of positive versus negative affectivity, and "agreeing that one only rarely feels pessimistic is not necessarily equivalent to saying that one feels optimistic nearly all the time" (p. 277). In other words, the positive and negative subscales may measure, respectively, two relatively independent constructs of optimism and pessimism. I n line with these findings, Marshall, Wortman, Kusulas, Hervig, and Vickers (1992) also showed that the items measure two separate dimensions of optimism and pessimism rather than one bipolar optimism-pessimism dimension. Optimism was specifically related to positive affect and extraversion whlle pessimism was primarily associated with negative affect and neuroticism. A two-factor solution had also emerged while the test was being developed by Scheier and Carver (1985); however, contrary to what Mook, et al. (1992) have found, the factor correlation was much higher; r = .64. More importantly, in addition to a two-factor solution, a single-factor solution was shown to fit the data by confirmatory factor analysis, although not as well as the two-factor solution. These findings led Scheier and Carver (1985) to conclude that it is most appropriate to treat the test as unidimensional but the two factors can justifiably be examined separately. In the view of these two investigators, the test is more validly regarded as tapping a single optimismpessimism dimension. Subsequent research on optimism using the test has been dominated by this conception. Despite the prevalence of the unidimensional view, the two-factor alternative is still favored by some researchers who have even operationalized optimism with only the positive half of the test (e.g., Strutton & Lumpkin, 1992). However, one of the most important questions stemming from acceptance of the two-factor structure, which remains relatively unanswered, concerns the relative power of the positive versus the negative subscale or optimism versus pessimism in predicting health outcomes. Different relations between the two dimensions and some health-related criteria would support a multidimensional view. The present study was designed to look at this question in a group of university students in Hong Kong.

METHOD Subjects were 202 undergraduate students taking introductory psychology at the University of Hong Kong. The sample (Sample A ) consisted of 117 women and 85 men, whose mean age was 19.4 and 20.0 yr., respectively. The original English version of the Life Orientation Test (Scheier &

PREDICTION FROM LIFE ORIENTATION TEST

1509

Carver, 1985) was administered. This scale, as listed in Table 1, is composed of four positively phrased items, e.g., "I always look on the bright side of things," four negatively worded statements, e.g., "I hardly ever expect things to go my way," and four filler items, e.g., "It's easy for me to relax." TABLE 1 ITEMSCOMPRISING THE LIFE ORIENTATION TEST* 1. 2. 3. 4. 5. 6. 7.

8. 9. 10. 11. 12.

In uncertain times, I usually expect the best. It's easy for me to relax." If something can go wrong for me, it will. I always look on the bright side of things. I'm always optimistic about my future. I enjoy my friends a lot." It's important for me to keep busy.' I hardly ever expect things to go my way. Things never work out the way I want them to. I don't get upset too easily.' I'm a believer in the idea that "every cloud has a silver lining." I rarely count on good things happening to me.

*From Scheier and Carver (1985) @Health Psychology 1985. T d e r item.

Ratings of the negatively phrased items were reversed prior to scoring of the complete scale. Subjects were asked to indicate on a 5-point scale in Likert format anchored by O (strongly disagree) and 4 (strongly agree) the extent to which they agreed with each of the 12 items. Three separate scores were derived from the test for subsequent analysis. The first was the complete test score computed by adding a l l the eight items, the second derived from the positive half, and the thud from the four negative items (nonreversed item ratings - were used). This was done at the end of a lecture and course credits were given for participation. Data collected during this session were mainly for factor analysis of the test. From Sample A, 100 students were randomly selected to attend another test session three weeks later. However, fifteen did not show up and the final sample completing the second test (Sample B) comprised 52 women and 33 men. I n addition to the Life Orientation Test, a physical symptoms checklist consisting of 35 commonly reported physical symptoms, e.g., sore throat, headaches, constipation, upset stomach, was also administered to Sample B. Subjects were required to indicate how often they had experienced each of the symptoms during the past three weeks on a 4-point scale in Likert format (1 = n o t at all; 4 =very often). Symptom scores were computed by adding ratings of each of the 35 symptoms. The 35 symptoms in the checklist were developed by Lai (1993) from an initial pool of 50 items extracted from the Cohen-Hoberman Inventory of

Physical Symptoms (Cohen & Hoberman, 1983) and the Pennebaker Inventory of Limbic Languidness (Pennebaker, 1982). This pool of items was administered to a group of 60 undergraduates taking final examinations. Thirty-five items were reported as experienced more frequently during examination time than a control nonstressful period. These items had an averaged Cronbach alpha of .78 over two test sessions (M = 43.6, SD = 6.8) and constitute the physical symptoms checklist used in the present study. The checklist's mean, standard deviation, and coefficient alpha for Sample B of the present study were 37.6, 7.7, and .86, respectively. Data collected from this sample were used for evaluating the predictive validity of the Life Orientation Test. Although anonymity could not be attained in the present study as students talung part in both test sessions needed to be identified for allocation of course credits, they were told that the information they provided would be only for research purposes and kept strictly confidential. Previous research indicated that socially desirable responding was minimized by confidentiality of data but not anonymity (Esposito, Agard, & Rosnow, 1984). The test conditions of the present study were therefore not assumed to accentuate biased responding.

RESULTS Descriptive Statistics The test scores of Sample A averaged 17.9, SD = 4.28. Mean score for men was 17.2, SD = 4.5 whereas that for women was slightly higher (M = 18.4, SD = 4.1). However, a t test indicated no significant difference in scores between the two sexes (t,,, = 0.62, p > . l ) . The mean of the positively worded subscale was 9.2 (SD = 3.O), which was higher than that of the negatively worded subscale (M = 8.7, SD = 2.3). The difference between these two means, however, was not statistically significant (t,,, = 1.93, p > .05). The correlations among scores of the total scale, the positive and the negative subscales are all significant, as presented in Table 2. TABLE 2 PEARSON INTERCORRELATIONS OF SCORESOF TOTALLIFE ORIENTATION TEST, POSITNE SUBSCALE,AND NEGATIVE S ~ S C A LFOR E SA~TPLEA Scale

1

2

3 -

1. Total Score 2. Positive Subscale 3 . Negative Subscale

,857 -.73t

---

-

-.25*

* p < .01. t p < ,001, two-tailed. Sample A, N = 202.

Factor Analysis of Life Orientation Test An exploratory factor analysis was performed on test scores (with ratings of negative items reversed) of Sample A using principal axis factoring (SPSS

PREDICTION FROM LIFE ORIENTATION TEST

1511

Version 4.0). Two factors were generated by setting the eigenvalue at 1.0 and using an oblique rotational technique. Moreover, a scree plot of factor variances showed a break after two factors which accounted for 51.5% of the total variance. The first one explained 33% of total common variance whereas the second explained an additional 18.5%. As can be seen in Table 3, the first factor was defined by items worded in a positive direction, and the second defined by items worded negatively. TABLE 3 FACTORLOADINGS FORLIFE ORIENTATIONTESTIN SAMPLEA Factor 1

Factor 2

h2

Item 1 Item 3 Item 4 Item 5 Item 8 Item 9 Item 11 Item 12 Eigenvalues Note.-The above findings are based on the method of principal axis factoring with oblique rotation (N = 202). Reversed ratings of negatively worded items were used.

The Pearson correlation between these two factors was .18. The same two-factor solution also emerged when the data were subsequently analyzed with orthogonal (varimax) rotation. Using the maximum likelihood method and oblique rotation, the same two-factor solution was again an adequate fit to the data [X,,Z(N= 202) = 18.8, p > .05]. Coefficients alpha for the complete Life Orientation Test, the positive subscale, and the negative subscale were .68, .71, and .66, respectively. Although modest, these are a l l within the acceptable range (DeVellis, 1991).

Predictive Validity of the Life Orientation Test With respect to the data of Sample B, scores were relatively stable over three weeks, with test-retest reliability = 0.73. Reliability coefficients for the positive and negative subscales were .71 and .70, respectively. Correlations between physical symptom scores and scores on the Life Orientation Test obtained during the first and second sessions are shown in Table 4. The correlations of symptom reports with full scale scores and positive subscale scores assessed at the same point in time (the second test session) were all significant. This also applied to the relationships between complete test scores obtained at the first session and symptoms reported at the second session. Total scores as well as the positive subscale were significantly correlated with symptom reports; however, scores of the negative subscale ob-

TABLE 4 PEARSONCORRELATIONS OF S ~ O REPORTS M AT SECOND TESTSESSIONWITH LIFE ORIENTATION TEST SCORESOBTAINED A T FIRSTA N D SECOND TEST SESSIONS FOR SAMPLEB Scale

No&.-Levels

Session 1

Session 2

Total Score -.22* -.25* Positive Subscale -.24* -.31t Negative Subscale .15 .12 of physical symptoms were assessed only at the second test session. Sample B,

n = 85. *p< .05. t p < .01, two-tailed.

tained at either the first or second session did not predict symptoms reported at the second test session. Semipartial correlations (Keppel & Zedeck, 1989, p. 430) controlling for scores of either the negative or positive subscale were used to assess which would contribute more to the symptom variance explained by the complete test. Results are presented in Table 5 . TABLE 5 SWV~IPARTIAL CORRELATIONS BETWEENTOTALSCOREON LIFE ORIENTATION TEST AND CONCURRENT AND PROSPECTIVE SYMPTOMSCORESCONTROLLING FOR S~SCALES SCORESOF POSITIVEAND NEGATIVE Measure Positive subscale controlled Negative subscale controlled * p < .05, one-taded.

Concurrent

Prospective

-.I1 -.21*

-.08 -.19*

For the relation between complete Life Orientation Test and symptom reports at the same point in time, significant correlation was eliminated only when scores of the positive subscale were controlled. The same pattern of results was also found for the association between total scores and symptom levels reported three weeks later. Taken together, the above findings pointed to the positive subscale as a more valid predictor of reports of physical symptoms. Moreover, the total test score may owe its predictive power to the positive subscale which has been presumed to measure uniquely the construct of dispositional optimism in prior research (e.g., Marshall, et a]., 1972; Mook, et a]., 1992).

DISCUSSION In the present study, the factorial structure as well as the prehctive validity of the Life Orientation Test was examined. Previous two-factor solution which splits the test into two halves measuring independent constructs of optimism and pessimism was replicated by different methods of factor extraction. Moreover, previous correlation between total scores and symptom reports was also found. When the two factors were examined separately, only optimism which was defined by the four positively worded items predicted

PREDICTION FROM LIFE ORIENTATION TEST

1513

reports of physical symptoms concurrently and over three weeks. Pessimism scores on the four negatively phrased items were not correlated significantly with symptom reports. I n addition, when optimism was controlled, the significant correlation between total scores and symptom levels was eliminated, but this was not observed while pessimism scores were controlled. The predictive power of the complete test is probably conferred by the positive half specifically indexing optimism. O n the other hand, since the correlations between scores on the positive and negative subscales and symptom reports d o not differ impressively, it remains possible that a significant correlation between scores on the negative subscale and symptom reports may be obtained in a larger sample. To assess whether the correlation between positive subscale scores and symptom reports is statistically different from that between scores of negative subscale and symptoms, Hotelling's T tests (Rosenthal & Rosnow, 1991, p. 507) were performed on the prospective and concurrent correlations listed in Table 4. Difference between the two concurrent correlations (second test session) was significant (tH= 2.43, df= 199, p < .01) whereas that between the prospective correlations (first test session) was not (tH = 1.07, df= 199, p > . l ) . Enlarging the sample size may render a significant correlation between scores on the negative subscale and symptom reports more likely to be obtained, but at the same time a larger ~ o t e l l i nT~statistic would also be generated. However, statistical significance should not be pursued without regard to clinical significance. Difference between the correlations .24 and .15 and .32 and .12 is without question of limited clinical implication, but this casts little doubt on the more important finding that the positive subscale contributes more than the negatively worded items to the predictive power of the complete test. This implies that relative to the negative half, the positive subscale may more validly measure what the complete scale was designed to assess, namely, the presence of the construct of optimism. The above findings lend further support to justify separate examination of the two subscales of the Life Orientation Test as optimism versus pessimism. I n the original studies by Scheier and Carver (1985), although a two-factor solution was indicated by nested test to be a better fit to the data than a single-factor one, these investigators still favored the unidimensional view for the high loadings of the items on the first umotated factor extracted and the strong correlation between the two factors generated by oblique rotation. However, they did not go on to show whether the two factors would actually converge on similar health-related criteria and so did not provide the most convincing piece of evidence for the unidimensional conception. Although Marshall, et al. (1992) have reported the validity of the optimism-pessimism distinction by showing that these two constructs related differentially to various mood and personality dimensions, they have not

looked directly at the connections of these two dimensions with health measures. Findings from the present study extend the knowledge regarding the construct validity of the test in that optimism and pessimism correlated differentially with symptom levels. Further, since the positive subscale scores exhibited similar reliability and validity in comparison to the total scores, dispositional optimism may be more validly and parsimoniously measured by the four positive items of the test. A recent attempt to define optimism with only the positive subscale has replicated the previously documented positive association between optimism (indexed by the complete test) and problem-focused coping (Strutton & Lumpkin, 1992). However, the effect of pessimism on coping was not explored in this study so the relative behavioral effects of the two constructs remain unclear. Further investigations are warranted to examine the health or behavioral implications associated separately with optimism and pessimism. Present data, however, should be evaluated in the light of certain limitations. As a health-relevant personality Imension, optimism has been predominantly validated with self-reported measures which may not be related to actual health (Costa & McCrae, 1987). Therefore, f i n l n g a negative correlation between optimism and symptom reports may not have any significant implications for actual health status since dissociation between health complaints and actual physical well-being was observed for other personality dimensions such as neuroticism (Costa & McCrae, 1987) and negative affectivity (Watson & Pennebaker, 1989). Researchers should look at more direct indices of health or physical well-being. The nonsignificant role played by pessimism in predicting somatic complaints also calls into question previous emphasis upon negative affectivity to explain the link between optimism and symptom reports (Hamid, 1990; Smith, Pope, Rhodewalt, & Poulton, 1989). If affectivity has any relation to optimism and health, present findings would suggest a more crucial effect of positive affects which have been shown to be related more to optimism than pessimism (Marshall, et al., 1992). However, the nonsignificant correlation between pessimism and symptom levels should not be taken to dismiss the predictive utility of negative affectivity since these two personality dimensions are only weakly correlated (Marshall, et al., 1992). Whether affects actually mediate the effect of optimism-pessimism on symptom reports has yet to be tested empirically. More fruitful conclusions regarding the health benefits associated with optimism can only be reached when the constructs of optimism and pessimism are studied in a context of positive versus negative affectivity. REFERENCES CARVER, C. S., POZO,C., HARRIS,S. D., NORIEGA, V., SCHEIER, M. F., ROBINSON, D. S., KETCHAM,A. S., MOFFAT,F, L.,

&

CLARK, K. C . (1993) How

coping mediates the ef-

PREDICTION FROM LIFE ORIENTATION TEST

1515

fect of optimism on distress: a study of women with early stage breast cancer. Journal of Personality and Social Psychology, 65, 375-390. COHEN,S , & HOBERMAN, H. (1983) Positive events and social supports as buffer of life changc sr less. Journal of Applied Social Psychology, 13, 99-125. COSTA,P T , & MCCRAE,R. R. (1987) Neuroticism, somatic complaints, and disease: is the bark worse than the bite? JournalofPersonality, 55, 299-316. DESHARNAIS, R., GODIN,G . , JOBIN,J., VALOIS,P., & ROSS,A. (1990) Optimism and health-relevant cognitions after a myocardial infarction. Pr/chological Reports, 67, 1131-1135. DEVELLIS, R. F. (1991) Scale development: theory and applications. Newbury Park, CA: Sage. ESPOSITO,J. L., AGARD,E., & ROSNOW, R. (1984) Can confidentiality of data pay off? Personality and Individual Differences, 5, 477-480. FRIEDMAN, L. C., NELSON,D. V.,BAER,I? E., & LANE,M. (1992) The relationship of dispositional optimism, daily life stress, and domestic environment to coping methods used by cancer patients. Journal of Behavioral Medicine, 15, 127-141. HAMID,I? N. (1990) Optimism and the reporting of flu episodes. Social Behavior and Personality, 18, 225-234. KEPPEL,G., & ZEDECK,S. (1989) Data analysis for research designs. New York: Freeman. LAI, J . C. L. (1993) The effects of examination stress on secretory immunity. Unpublished doctoral dissertation, Univer. of Hong Kong, Hong Kong. MARSHALL, G . N., WORTMAN, C. B., KUSULAS, J . W., HERVIG,L. K., & VICKERS,R. R. (1992) Distinguishing optimism from pessimism: relations to fundamental dimensions of mood and personality. Journal of Personality and Social Psychology, 62, 1067-1074. MOOK,J., K L E ~W. , CHR., & PLOEG,H . M. VAN DER. (1992) Positively and negatively worded items in a self-report measure of dispositional optimism. Psychological Reports, 71, 275-278. PENNEBAKER, J. W. (1982) The psychology of physical symptoms. New York: Springer-Verlag. ROSENTHAL, R., & ROSNOW,R. L. (1991) Essentials of behavioral research: methods and data analysis. New York: McGraw-Hill. SCHELER,M. F., & CARVER, C. S. (1985) Optimism, coping, and health: assessment and implications of generalized outcome expectancies. HeaW Psychology, 4, 219-247. SCHEIER,M. F., & CARVER,C. S. (1987) Dispositional optimism and ph sical well-being: the influence of generalized outcome expectancies on health. Journal ohersonality, 55, 169210. SCHUER,M. F., MAT~HEWS, K. A,, OWENS,J. F., MAGOVERN, G. J., LEFEBVRE, R. C., ABBOT, R. A., & CARVER, C. S. (1989) Dis ositional optimism and recovery from coronary artery bypass surgery: the beneficial efrects on physical and psychological well-being. Journal of Personality and Social Psychology, 57, 1024-1040. SCCIEIER,M. F,, WEINTRAUB, J. K., & CARVER, C. S. (1986) Coping with stress: divergent strategies of optimists and pessimists. Journal of Personality and Social Psychology, 51, 1257-1264. SMITI-I,T. W., POPE, M. K.,RHODEWALT, F.. & POULTON, J. L. (1989) Optimism, neuroticism, coping, and symptom re orts. an alternarlve interpretation of the Life Orientation Test. Journal of Personality a n j ~ o c ; a Psycbolo~y, l 56, 640-648. STRUTTON,D., & LUMPKIN, J. (1992) Relal~onshpbetween optimism and coping strategies in the work environment. Psychological Reports, 71, 1179-1186. TAYLOR,S. E., KEMENY,M. E., ASPINWALL, L. G . , SCHNHDER,S. G., RODRIGUEZ, R., & HERBERT,M. (1992) 0 timism, coping, psychological distress, and high-risk sexual behavior among men at ris! for Acquired Immunodeficiency Syndrome (AIDS). Journal of Personality and Social Psychology, 63, 460-473. WATSON,D., & PENNEBAKER, J. W. (1989) Health complaints, stress, and distress: exploring the central role of negative affectivity. Psychological Review, 96, 234-254.

Accepted November 9, 1934