updated HE PANASX Manual for the Positive and Negative Affect Schedule Expanded Form - PDF document

C. Construct Validity Table 3 Means and Standard Deviations for the General Positive Affect (PA) and Negative Affect (NA) Scales as aFunction of Rated Time Frame and Subject Population ...........................................................................3Table 4 Internal Consistency Reliabilities (Coefficient a) and Intercorrelations of the General PA and NA Scales .........4Table 5 Descriptive Statistics for the General PA and NA Scales Derived from Mean Daily Mood Scores ...................5Table 6 Correlations between the General PA and NA Scales and Regression-Based Scores on the First TwoVarimax Factors in Six Samples Assessed with Tellegen's Set of 60 Mood Descriptors .................................6Table 7 Correlations between the General PA and NA Scales and Regression-Based Scores on the First TwoVarimax Factors in Ten Samples Assessed with the 60 PANAS-X pastfew weeks . Use the following scale to record your answers: ______ excited______ determined______ strong______ timid______ hostile______ frightened______ scornful______ alone______ proud______ astonished______ relaxed______ alert______ jittery______ interested______ irritable______ upset______ lively______ loathing______ delighted______ angry______ ashamed______ confident______ inspired______ bold______ at ease______ energetic______ fearless______ blue______ scared______ concentrating______ disgusted Item Composition of strongBasic Negative Emotion ScalesFear (6)afraid, scared, frightened, nervous, jittery, shakyHostility (6)angry, hostile, irritable, scornful, disgusted, loathingGuilt (6)guilty, ashamed, blameworthy, angry at self, disgusted with self, dissatisfied with selfSadness (5)sad, blue, downhearted, alone, lonelyBasic Positive Emotion ScalesJoviality (8)happy, joyful, delighted, cheerful, excited, enthusiastic, lively, energeticSelf-Assurance (6)proud, strong, confident, bold, daring, fearless ScalesThe goal in developing these scales was to create reliable andvalid measures that were also brief and simple to administer.The primary concern was to select descriptors that wererelatively pure markers of either Negative Affect or PositiveAffect; that is, terms that had a substantial loading on onefactor but a near-zero loading on the other. As a startingpoint, we used the 60 terms included in the factor analysesreported by Zevon and Tellegen (1982). Tellegen concerned that the terms be relatively pure markers of afactor. We therefore specified that the terms not have asecondary loading of I.25I or greater in either at Southern Methodist University(SMU), a private southwestern university. However, we(and others) have also collected data on various student, adultand psychiatric patient samples, and we present these resultsas well.We have obtained PANAS-X ratings using eight differenttemporal instructions. Subjects have rated how they felt:(a) "right now (that is, at the present moment)" (Momentinstructions); (b) "today" (Today); (c) "during the past fewdays" (Past Few Days); (d) "during the past week" (PastWeek); (e) "during the past few weeks (Past Few Weeks);(f) 33.16.817.96.4Past MonthSMU undergraduates1,00634.57.220.27.3Past YearSMU undergraduates 35.17.418.96.4Detroit-area adultsb various samples. We have not foundany large or consistent gender differences on these scales (theissue of gender differences will be discussed in more detailsubsequently), so our own data are collapsed across gender.Nevertheless, we recommend testing for gender differencesin any new (especially nonstudent) sample.Inspecting Table 3, one sees that subjects report time period increases, the probability that a subject willhave experienced a significant amount of a given affect alsoincreases. Finally, it is interesting to compare the meanscores of the college student, adult, and psychiatric patientsamples. Table 3 indicates that the mean Positive Affectscores of college students and adults are generally quitesimilar. The findings for Negative Affect are more variable,but the bulk of the data suggest that adults report slightlylower levels of Negative Affect than college students.Overall, these data suggest that normative data collected oncollege student samples can be applied to community-dwelling adult samples with some confidence, particularlywith regard to Positive Affect.______________________________________________________________________________Table 4 Internal Consistency Reliabilities (Coefficient Alpha) and Intercorrelations of the GeneralPositive Affect and Negative Affect Scales______________________________________________________________________________ Coefficient a______________________Positive .87-.31Past WeekSMU undergraduates1,521.88.85-.14Dallas-area adults 328.90.90-.38Australian adultsa 229.86.79-.07Past Few WeeksSMU undergraduates2,076.87.87-.13SMU employees 164.86.87-.09Past MonthSMU undergraduates1,006.89.89-.15Past YearSMU undergraduates 964.87 inpatients data for Positive Affect are less clear, butone can anticipate that many patients will obtain unusuallylow scores on this scale.Internal consistency reliabilities (Cronbach's coefficientalpha) for the two higher order scales, and also theirintercorrelations, are presented in Table 4. The alphareliabilities for both scales are influenced by the rated time frame. Thus, the higher orderscales maintain their quasi-independence regardless ofwhether state (shorter term) or trait (longer term) affect isassessed.Aggregated within-subject data. The data we have presentedthus far have been derived from single, between-subjectsassessments of large subject samples. Some investigators,however, may wish to use the higher order PANAS-X scalesin designs that necessitate repeated within-subject groups of subjects who completed thehigher order PANAS-X scales (using Today instructions) on adaily basis over a period of several weeks. All subjectscompleted a minimum of 30 daily mood assessments; all ofthe assessments were completed in the evening, so that theratings would provide a reasonable estimate of the subjects'moods over the course of the day (for more details regardingthis type of design, see Kennedy-Moore, Greenberg, 15.35.6Note. CFS = Chronic Fatigue Syndrome. See Few Days1,002-.15.92.93-.10Past Few Weeks586-.10.92.92-.18Past Year649-.17.89.93-.09General663-.08.94.93-.12Note. This table is adapted from Watson et al. (1988, Table 4). See text for details.____________________________________________________________________________________________________C. Construct ValidityFactorial validity. An important step in each solutionwere then rotated using varimax. Each of the final sample was Correlations Between the General Positive Affect and Negative Affect Scales and Regression-Based Scores on theFirst Two Varimax Factors in Ten Samples Assessed with the 60 PANAS-X Mood Descriptors____________________________________________________________________________________________________Positive AffectNegative AffectScale CorrelationsScale Correlations________________________________________Rated Time Frame NFactor 1Factor 2Factor 1Factor 2____________________________________________________________________________________________________SMU undergraduates Moment1,027-.02.94.94-.05 .93-.08 35 moodassessments.The mood ratings in each sample were standardized on awithin-subject basis; that is, each subject's responses wereconverted to standard scores (M = 0, SD metropolitan area. To be included in the study, the couplehad to have been dating each other for at least one month (Mtime of dating = 21.8 months). All subjects rated boththemselves and their partners on the full PANAS-X.Before presenting these results, we should note that one can Factors Emerging in & Tellegen, 1982), differing only in thetime frames used to generate the mood ratings (Moment,Today, Past Few Days, Past Few Weeks, Past Year,General).Each of these data sets was factored separately usingprincipal factor analysis (squared multiple correlations in thediagonal) and varimax rotation. To determine the finalnumber of factors in each data set, a range of solutions--starting at two factors--was examined, Past Few Days (8), Past Few Weeks (8),Past Year (10), and General (6). In each case, thisrepresents the maximum number of interpretable factors thatcould be identified in the data.________________________________________________Table 9 Convergence Between Self- and Peer-ratings on theTwo Higher Order PANAS- .74happy.70 .39.36** scared ____________________________________________________________________________________________________Factor_______________________________________________________________Descriptor12345678____________________________________________________________________________________________________** alone.75** lonely.71** sad.62** blue.31 .39.32.49** dissatisfied with self.74** angry at self.73** disgusted with self.68** guilty.54** blameworthy.51** ashamed.51** sleepy.76** tired.74 .62** surprised.60** bashful.67** clearly in at least four of the six solutions. Seven of thesedimensions--Fear, Sadness, Guilt, Hostility, Shyness,Fatigue, and Surprise-- represented specific emotional statesand therefore were used as the basis for scale development.The terms that most strongly and consistently defined thesefactors across all of the solutions were selected as thecomponent scale items, and are indicated in Table 10. Theeighth replicable factor was the higher order Positive Affectdimension; we describe our further efforts to identifyspecific positive emotional factors in than those with of Positive Affect to split into more specificpositive emotional states. Separate positive affect factorsemerged only in the Past Year solution. In these data,Positive Affect split into Joy/Sociability andInterest/Energy. To solution (the initial Past Weeksample) is shown in Table 11.Four specific positive affect scales--Joviality, Self-Assurance, Attentiveness, and Serenity--were constructedfrom the clearest and most consistent markers of each ofthese factors. The descriptors comprising these 17.14.8.83Attentiveness13.13.1.79Serenity8.92.5.74Surprise6.62.7.80Past Week (328 Dallas-area adults)Fear9.64.1.88Sadness9.54.4.88Guilt10.14.9.91Hostility11.14.6.88Shyness6.32.7.83Fatigue9.03.8.89Joviality25.06.8.93Self-Assurance16.54.6.81Attentiveness12.73.0.78Serenity8.72.4.75Surprise5.42.3.75Past Few Weeks (678 SMU undergraduates)Fear12.34.9.86Sadness11.74.8.87Guilt12.05.2.86Hostility12.95.0.85Shyness7.73.1.81Fatigue12.73.9.88Joviality26.86.6.93Self-Assurance17.74.7.81Attentiveness13.52.9.75Serenity8.92.6.79Surprise6.82.8.80Past Month (1,006 SMU undergraduates)Fear11.34.6.86Sadness10.64.6.88Guilt11.0 18.54.8.83Attentiveness13.83.0.78Serenity9.4 19.04.4.80Attentiveness14.02.8.78Serenity9.22.5.74Surprise7.52.3.74General (1,657 SMU undergraduates)Fear11.33.8.83Sadness10.13.7.83Guilt10.8 19.14.3.81Attentiveness14.22.6.76Serenity9.82.3.73Surprise6.72.3.76General (107 mixed inpatients/outpatients)Fear15.19.0.92Sadness14.85.3.88Guilt17.7 15.84.9.80Attentiveness13.13.0.70Serenity7.92.8.83Surprise6.42.5.72________________________________________________these scales had median internal consistency values of .78,.76, and .77, respectively. Nevertheless, these medianvalues reflect mean inter-item correlations of .45 orgreater, indicating that the scales are appropriatelyhomogeneous; thus, their reliability small.Generally speaking, these analyses revealed few consistentgender-related differences in affective experience. Fivescales--general Negative Affect, general Positive Affect,Attentiveness, Surprise and Sadness--showed virtually nosignificant gender-related effects. General Positive Affectand Attentiveness both yielded significant but inconsistentgroup differences in two samples: In each case men scoredhigher in one sample and women scored higher in the other.Sadness (women scored higher in the Past Week/Adultsample) and Surprise (men scored higher in the Past Monthsample) each produced one significant group difference.Finally, the general Negative Affect scale showed no Month, General).The three remaining scales showed more consistent gendereffects. Specifically, men reported significantly higherlevels of Self-Assurance in all 10 samples; in addition, theyscored significantly Week/SMU, Past Few points higher on Self-Assurance, onepoint higher on Hostility, and 10.24.7Past Few Daysab12.75.211.05.4Past Few Daysc12.64.512.24.9Past Weekab12.44.811.04.4 2.79.32.7Past Few Daysab9.62.58.92.6Past ab10.02.3 was administered to 563 SMU undergraduates using PastFew Weeks instructions.These five PANAS-X scales measure the same basic affects ascorresponding scales from the POMS. Specifically, Fearcan be identified with POMS Tension-Anxiety, Hostilitywith PANAS-X scales is strongly related to its POMScounterpart, with convergent correlations ranging from .85to .91. These very high coefficients partly reflect itemoverlap, as the corresponding scales have one to three itemsin common. Nevertheless, the important point is that thePANAS-X scales assess affective states that are broadlysimilar to those measured in existing multi-affectinventories such as the POMS.Table 15 also demonstrates that the PANAS-X scales offer animportant advantage over their POMS counterparts: Theytend to be less highly correlated with one another, and thusshow better discriminant validity. The mean correlationamong the PANAS-X Fear, Hostility, Sadness and Fatiguescales -X scales further, we againwill consider evidence from two studies in which self-ratingson these scales were correlated with corresponding judgmentsmade by well-acquainted peers. First, in the dormitory studythat was discussed previously (see Table 9 and the Fiske, 1959), five of the scales (Fear, Hostility, Sadness,Shyness, and Fatigue) showed acceptable discriminantvalidity. (For a more detailed discussion of these data, seeWatson & Clark, 1991).____________________________________________________________________________________________________Table 15 Correlations among the PANAS-X Scales and Corresponding Scales from the Profile of Mood States (POMS;McNair, Lorr & Droppleman, 1971)____________________________________________________________________________________________________PANAS-X ScalesPOMS Scales_____________________________________________________________Instrument/Scale123456789____________________________________________________________________________________________________PANAS - Sadness.61.49 --- 4.Fatiguea.40.31.27 --- 5.Positive Affect.02 in boldface. Allcorrelations greater than I.10I are significant at p .01, two-tailed.aData are based on a preliminary, 3-item version of * together with the .30 .32 *.09.08Fatigue.01-.01.08-.06.05.02-.01.10.27* =Attentiveness; Seren = Serenity; Sad = Sadness; Host = Hostility; Shy = Shyness; Fat = Fatigue; Surp = Surprise.*p ScalesA. Testing the Hierarchical Arrangement of the PANAS-XScalesAs was discussed previously, the PANAS-X was designedexplicitly to reflect the hierarchical structure of self-ratedaffect. That is, self-rated affect is characterized by twobroad higher order dimensions (Negative Affect and PositiveAffect), each of which is composed of several correlated,yet ultimately distinguishable specific affect states. ThePANAS-X includes scales assessing both of these structurallevels--that is, the two higher order dimensions and 11specific lower order states.How accurately does the PANAS-X capture this hierarchicalstructure? To answer this question, we subjected the 11lower order PANAS-X scales to second-order principal then correlated these computed factor scores with the generalNegative Affect and Positive Affect scales. These resultsare displayed in Table 19, and they again confirm thefactorial validity of the higher order scales. Specifically,scores on Factor 1 correlated .92 and .91 with generalNegative Affect in the Moment and General solutions,respectively; conversely, scores on Factor 2 correlated .94and .91 with general Positive Affect -.13.86.85Self-Assurance-.04-.11.80.78Attentiveness Basic Negative Emotion Scales. As the data in Table factors: Fatigue is a marker ofhigh Negative, low Positive undergraduates (see their Tables 1 and 2). For instance,scores on the PANAS-X Sadness scale (assessed using PastFew Weeks time instructions) correlated .59 with the BeckDepression Inventory (BDI; Beck, Ward, Mendelson,Mock, & Erbaugh, 1961), .69 with the CES-D; Radloff, 1977). Similarly, the PANAS-X Fearscale correlated .74 with the HSCL Anxiety scale and .56 evidence of significant discriminant validity (see Watson &Clark, 1992a, Table 2).Similarly, Watson et al. (1988) present correlations interaction and exercise, but was notaffected by test stress; conversely, Negative Affect wasincreased significantly by the stressful examination, butwas not influenced by social activity or exercise.In the fourth study (Watson et al., 1992, Study 2), 127SMU undergraduates completed the two higher order PANAS-X scales each evening for 5-7 weeks using Todayinstructions (M = 42.7 assessments per subject); a subset of column of Table the retest interval ranged from 56 to 99 months (Mretest interval = 72.4 months). It also should be noted thatall subjects graduated from the university during thisintervening time period. Thus, all of the respondentsexperienced a major life change over the course of the study.Stability correlations from this study are presented in Table21. The most important aspect of these data is that both ofthe higher order scales demonstrated significant, moderatestability over this extended time interval, with retestcorrelations of .43 and .39 for Negative Affect and PositiveAffect, respectively. Note also that the discriminantcorrelations were substantially lower, again demonstratingthe discriminant validity of these scales (for more detailsregarding this study, see Watson & Walker, 1996).________________________________________________Table 21 Long-term Retest during the first week of class. Subjectsthen were reassessed on these scales (using Past Weekinstructions) once a week over the next 13 weeks. Allsubjects completed a -.10Negative Affect -.02.48*Mean Daily Ratings (N = 410)Positive Affect .64*-.03Negative Affect -.06.53**p .05, two-tailed._______________________________________________A Attentiveness (.58), Sadness (.54), Self-Assurance (.53), and Shyness (.51). Four additional scaleshad convergent coefficients in the .40 to .50 range: Serenity(.47), Fear (.46), Hostility (.46), and Fatigue (.41). It isnoteworthy that Surprise produced the lowest convergentcorrelation (.36), again challenging Negative and Positive Temperamentscales are true-false measures designed to assess traitNegative and Positive Affectivity, respectively. it is noteworthy FFI Neuroticism.82-.21-.16.04-.03 .13.27.00GTS Positive Temperament-.19.78.28.02.10PANAS-X Self-Assurance-.35.59.25-.36.16Goldberg Conscientiousness-.06.06.89.04.02NEO-FFI Conscientiousness-.11.10 are shown in but also had a summarize theextensive data presented here and elsewhere: trait scoreson the PANAS-X scales (a) are stable over time, (b) showsignificant convergent and discriminant validity whencorrelated with peer-judgments, (c) are highly correlatedwith corresponding measures of aggregated state affect,and (d) are strongly and systematically related to measuresof personality and emotionality. These data clearlydemonstrate that --with the possible exception of Surprise--the PANAS-X scales can be used validly to assess long-term individual differences in affect.V. ReferencesAlbright, L., Kenny, D. A., & Malloy, T. E. (1988).Consensus in personality judgments at zeroacquaintance. Journal of Personality and SocialPsychology, 55, 387-395.Almagor, M., & Ben-Porath, Y. (1989). The two- C. H., Mendelson, M., Mock, J.,& Erbaugh, J. (1961). Survey. Unpublished manuscript,Southern Methodist University, variation in the Assessment Resources. M., &Stone, A. A. The relationship between daily events Press.Spielberger, C. D., Jacobs, G. A., Russell, S., pp. 52-76). Hillsdale, NJ: Erlbaum.Tellegen, A. (1985). Structures of mood andpersonality and their relevance to assessing anxiety,with an emphasis on self-report. In A. H. Tuma andJ. D. Maser (Eds.), Anxiety and the anxiety disorders(pp. 681-706). Hillsdale, NJ: Erlbaum.Watson, D. (1988a). Intraindividual and interindividualanalyses of Positive and Negative Affect: Their relationto health complaints, perceived stress, and dailyactivities. Journal of Personality and SocialPsychology, 54, 1020-1030.Watson, D. (1988b). The vicissitudes of moodmeasurement: Effects of varying descriptors, timeframes, and response formats on measures of Positiveand Negative Affect. Journal of Personality andSocial Psychology, 55, 128-141.Watson, D. (1989). Strangers' ratings Affect, personality, and socialactivity. Journal of Personality and SocialPsychology, 63, 1011-1025.