A Global Measure of Perceived Stress Sheldon Cohen; Tom Kamarck; Robin Mermelstein Journal of Health and Social Behavior, Vol. 24, No. 4 (Dec., 1983), 385-396. Stable URL: http fink jstor.orgsici?sici=0022- 465% 281983 12% 2024%3 A4%3C385%3AAGMOPS%362.0,CO®B2-2 ournct of Health and Social Behavier is currently published by American Sociological Association, Your use of the ISTOR archive indicates your acceptance of ISTOR's Terms and Conditions of Use, available at flip: feworwjtor org/aboutterms.htmal. ISTOR's Terms and Conditions of Use provides, in par, that unless you fave obtained pcior permission, you may not dowaload an cnt isus of @ journal or multiple copies of articles, and you may use content inthe ISTOR archive only for your personal, non-commercial uss. Please contact the publisher cegarding any further use of this work. Publisher contact information may be obtained at bupsorww.jstoc.org/joumals/asa. hl. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transtnission. ISTOR is an independent not-for-profit organization dedicated to creating and preserving a digital archive of scholarly journals. For more information regarding ISTOR, please contact support @jstor.org- hup:shrwwjstor.orgy Sun May 16 22:37:22 2004A GLOBAL MEASURE OF PERCEIVED STRESS Richardson, Jean, and Julie Solis T981. "Blae of death of Hipanie cancer nants in Tas. Angeles County.” Unpublished Stoddard Sando) 1978" The Hospice Movement. Brarli Mana, New Yon Stein and Piss, Valle, Rand L Mendota ‘Wilt The Eller latino, San Diego, CA: Cam- anil Press 385 Wieging, Jery 8. 1973 ‘Personality and Prediction: Frizciplee of Personality Assessment. Reading. MA; Addison-Wesley Publishing Co. Wilder Foundation {981 Care forthe Dyins: A Study ofthe Need for Hlospice in Ramsey County, Minnesota. A ‘eport 0 the Northwest Arca Foundation {rom the Amherst H, Wilder Foundation, 4358 Washington St, St Pail, MN. A Global Measure of Perceived Stress SHELDON COHEN Carnegie Mellon University TOM KAMARCK, University of Oregon ROBIN MERMELSTEIN University af Oregon Journal of Health and Sosial Behavior 1983, Vol. 24 (December):385-396 This paper presents evidence from three samples, two of college students and one of parte aris in a community smaking cessation program, forthe reliablity and valid of a Hier insirument, the Perceived Stress Scale (PSS). designed to measwre the degree to which Situations in one's life are appraised as stressfe. The PSS showed adequate reliably ad, at Iredieied. was correlated with Ife-event scores, depressive and physical symptomatology ludicaion of health sevoices, social aniet», and smoking reduction maintenance. tall camt= Dorisons, the PSS wos a better predicir ofthe outcome in question than were lfeevent scores Wher compared 10 a depressive symplomatology sccle, the PSS was found to measure 2 Aiferent ond independently predictive construct. Additional data indicate adequate reliability ‘and vatdity of four stem version ofthe PSS fay telephone interviews. The PSS is suggested or examining the role af nonspecific appraized sores: i the eology of diteate and bohatoral disorders and as an outcome measure of experienced levels of S165. It is @ common assumption among health researchers that the impact of “objectively” sttessful events is, co some degree, determined Research reported inthis paper was supported by {#4013 trom the National Science Foundation (BNS. Fons4s3) and the Heart, Lung and Blood Testitute (HL 29847. The authors are indebted to Edward Lichtenstein and Karen Melatyce for their callab- oration on the smoking cessation project; he sal of University of Oregon dormitory: housing for their cooperation in secoring Student Sample I Pam Biel for her cooperation in obainig class participant for Sample I, and to Susan Fiske, Lew Goldberg, Ea ward Lichtenstein, Karen Matthews, Michze! Scheier, and Tom Wills for comments on eariet drafts ofthis manuscript ‘Address communications to: Shelton Cohen, De parent of Psychology, Camegie-Melion Univer Sity, Pitsburg, PA 5206 ‘by one's perceptions of their stressfulness, eg, see Lazarus (1966, 1977). Surprisingly, this theoretical perspective has nat heen ac” companied by development of psychometr cally valid measures of perceived stress. This article discusses the limitations of abjective and subjective measures of stress used in the assessment of both global and event-specific. stress levels, I¢ argues that a psychometrically sound global measure of perceived stress could provide valuable additional information about the relationship between stress and pathology. Data are presented on the psychometric properties of the Perceived Stress Scale (PSS), an instrument developed in response to these issues. The PSS measures the degree to which situations in one's life are appraised -as stressful386 Research on the role of psychosocial and cenviroamental stressors as risk factors ia both physical and psychological illness has typically employed relatively objective stressor mea sures, This work includes studies ofthe effects ‘of specific stressful events, such as unem- ployment (Cobb and Kasl, 1977; Dooley and Catalano, 1980; Gore, 1978), bereavement (Stroebe et al., 1982), and exposure to intense levels of noise (Cohen and Weinstein, 1981) and high levels of population density (Sundstrom, (978), fa addition, there is an abundant literature on the cumulative effect of objective stressful life events (Dohrenwend and Dohrenwend, 1974, 1981). [n these studies, ‘various versions of life-eVent scales (original scale developed by Holmes and Rahe, 1967) are used (o produce & cumulative stress score. These scores are usually based on either the ‘umber of events that have occurred within the specified temporal framework (in most cases, six to 12 manths) of oa a sum of event weights that are based on judges’ ratings of the diff culty of adjusting fo these events. ‘There are some clear advantages to objective measures of stressful events. First, such mes. sures permit an estimate of the increased risk for disease associated with the occurrence of ‘casily identifiable events. Second, the mea surement procedure is often simple, 8. did this event occur during the last six’ months?, ‘and in many cases, persons experiencing & articular event can be identified without the ‘aecessity of asking them about the occurrence of the event, e-g., persons living in noise- impacted communities. Third, these measure ‘ment techniques minimize the chance of vari: ‘ous subjective biases in the perceptions and reporting of events. ‘On the other band, the use of objective mea sures of stréss implies that events are, in and of themselves, the precipitating cause of pathol ogy and illness behavior. This implication is ‘counter ta the view that persons actively inter- act with their enviroaments, appraising poten- tially threatening or challenging events in the light of available coping resources (Lazarus, 1966, 1977), From this latter perspective, stressor effects are assumed to aceur only when both (a) the situation is appraised as theeatening or otherwise demanding and (b) in- sufficient resources are available to cope with the situation. The argument is that the causal JOURNAL OF HEALTH AND SOCIAL BEHAVIOR “event” is the cognitively mediated emotional response to the objective event, not the objec- tive event itself (Lazarus, 1977; Mason, 1971), ‘An important part of this view is that this re sponse is not based solely on the intensity or ‘any other inherent quality of the event, but rather is dependent on persosal and contextual factors as well ‘The assumed centrality of the cognitive ap- praisal process suggests the desirability of measuring perceived stress as opposed or in addition ta objective stress. For example, ‘comparison of the predictive validities of ob: jective and subjective measures could help to ‘clarify the role of the anpraisal pracess in the relationship between objective stressors and illness. Perceived stress scales could also he used in conjunction with objective scales in aay effort to determine whether factors such as social support (Pearlin e¢al., 1981), hardiness (Kobasa, 1979), and lacus of control (Johnson ‘and Sarason, 1979) protect people from the pathogenic effects of stressful events by alter- ing stressor appraisal or by altering the process ‘or processes by which appraised stress results in physiological or behavioral disarders (Gore, 1981). Finally, perceived stress can be viewed san outcome variahle—measuring che expert ‘enced level of stress as a function of objective stressful events, coping processes, personality factors, etc Previous work has employed a number of approaches to assess hath global and event- specific levels of perceived stress, For exam ple, several investigators have modified life- ‘event scales in an attempt (0 measure global perceived stress, The modification invalved ‘asking respondents to rate the stressfulness or ‘impact of each experienced event. In general, stressfulness are beiter predictors of health- related outcomes than are scores detived from either a simple counting of eveats (unit. weighting) ot normative adjustment ratings (Sarason et al., 1978 Vinokur and Selzer, 1975), However, the inereases in predictability Provided by diese ratings are small. It is noteworthy that any increase in predictability of a weighted event score over a simple count ‘of events is likely to be small since alternative weighting schemes yield composite scares that are substantially correlated with the event count (Lei and Skinner, (980). fa short, cal-‘A GLOBAL Measure OF PERCEIVED STRESS culating global perceived sess levels om the tris of eaciton to fdividual event assumes that perceived stress levels are very tighly coe related withthe umber of reported events Other weaknesses of global perceived stress scales that are based ona specif ist of events include an insensitivity to ehtonie ates fom ongoing life cireumstances, to stress fom ‘events occurring in the lives of close friends nd fay, fon expectations concerning fo ture events and ffom events no sted on the seal Subjective measures of response to specie stressors have also been widely used, 4. measures of perceived accipation strets Kan et, 1968), There are, however, some priccal and theoretical limitations of mea fires of specific stessrs, Practically, sit full and time-consuming to adequately de- velop and psjehometnially validate an ind itoal measure everytime 4 ew stressor Susied. Theorsialy, there a0 iste of whether measures of perceived response to 2 speifc stressor really asese a persons eval ations of that syesson These is in acl, ev dence that people often misattibute ther feelings of sress to paniculat source when fiat sees Is actually Gue fo soother source (Gochman, 1979; Keating, 1379: Worshel, 1978; Worchel and Teddie, (976), Another probiem with measures of response to specific stressors is that such measures imply the inde- pendence of that event n the precipitation of Aiaense. However, ii likely that the ines proces is affected by person's aba stress Tove, ot just by higher response Cos paricn lar event. "The above discussion indicates the de- sirablty of developing an insrument to nese sure labal level of perceived stress, This Brice presents data onthe Perceived Stress Seale, I4item measure of the degree to which situations in one's fe are appratsed ax Stressfil. PSS items sre designed fo tap the degre to which Fespndents foun tei ives unpredicable, uncontrolible, and. overload ing These thee lies have been repeatedly found to be cemcal components af the exper fence of saree (Avery 1973, Cohen, 1978 Glass and Singer, 1972; Lazarus, 1966, 1977, ‘Seligman, 1975). The scale also includes a number of dtcet queries about current levels of experienced sess. The PSS isan econom: 387 {cal scale that can be administered in only a few ‘minutes and is easy t0 score. Because levels of appraised stress should be influenced by daily hassles, major events, and changes in coping. resources, the predictive validity of the PSS is expected to fall off rapidly after four ta eight weeks. Evidence is presented from three samples—two of college students and one of a ‘more heterogeneous community. group—for the concurrent and predictive validities and the internal and test-retest reliabilities of the new scale. ‘The paper also examines the relative predictive validity of the PSS and ovo life ‘event instruments. Our premise is that the PSS should provide a hetter predictor of health out ‘comes than does a global measure of objective sttessors, such as life-event scales, This should ‘occur because a perceived stress instrument provides a more direct measure of the level of stress experienced by the respondent. Presun- ably, icis this level of appraised stress, not the objective occurrence af the events, that detes- s one's response to a stressor{s) (Lazarus, 1966, 197), Also, the new measure is more global than life-event seales. That is, it is sen sitive to chronic stress deriving from ongoing life circumstances, to stress from expectations ‘concerning future events, to stress from events not listed on a particular hfe-events scale, and to reactions to the specific events included on any scale. METHODS Validation data were collected in three sam- ples — two consisting of college students and fone consisting of a more heterogeneous group ‘enrolled in a smoking-cessation program. The samples and assessment procedures are de- seribed below. Perceived Stress Seale (PSS) ‘The 14 items of the PSS are presented in Appendix A. PSS scores are obtained by re- vversing the scores on the seven positive items, eg, 0-4; 1=3, 2=2, etc., and thea summing across all 14 items. Items 4, 5, 6, 7, 9, 10, and 13 are the positively stated items. ‘The PSS was designed for use with comm tity samples with at least a junior high schoo!388 education. The items are easy 10 understand and the response alternatives are simple to srasp. Moreover, as noted above, the ques: tions are quite genera} in nature and hence rolatively free af content specific 19 any sub- population group, ‘The data reported in this article are from somewhat restricted samples, in chat they are younger, more educated, and contain fewer ‘minority members than the general popuation. In Fight of the generality of scale content and simplicity of language and response alterna- fives, we feel that data from representative samples of the general population would aot differ significantly feom those reported below. College Student Sample I The respondents were 332 (121 male, 209 female, two with sex not specified) freshman college students living in dormitories at the University of Oregon. The mean age of the sample was 19.01 with a standard deviation of 2.75, All respondents gave written consent allowing access to their student health center records, Measures. Respondents completed five scales: one measured life events; another so- cial anxiety; a thicd depressive symp- tomatology, fourth, physical symptomatology; and finally perceived stress (the PSS). Allin: struments were completed during a one and ‘one-half hour sessian, ‘A modified version of the College Student Life-F.vent Scale (CSLES) was used as a mea: sure of stressful life events; the original scale was developed by Levine and Perkins (1980) This scale is composed of 99 items that repre- sent events that fall into 14 different categories characterizing the adjustment demands of col- lege students, ¢.g., academic affairs, male- female relationships, and family matters. Nine items dealing with health-related issues were ‘not used in calculating life-stress scores be- cause of the possibilty that these items were measuring the same thing as items in the symptom checklists. Analyses including the unused items indicated that their exclusion did not affect the cesults reported below, Respondents were asked to indicate whether each event had occurred during the last year. ‘They were asked to rate the impact of events JOURNAL OF HEALTH AND SOCIAL BEHAVIOR {extremely negative) to +3 (extremely posie tive), the format used by Sarason etal. (1978). Separate scores were generated based on sell- rated impact and unweighted events. The un- weighted score was the total number of life events checked, The score based on impact ratings was the summed impact of checked events, The impact score is nota pure measure of the occurrence of objective events, but rather takes into account the respondents’ per ceptions of the events. ‘The Cemter for Epidemiologic Studies De- pression Scale (CES-D) was employed as a measure of current level of depressive symp tomatology (Radlofl, 1977). Twenty items, each representing a state characteristic or not characteristic of a depressed person, are rated ‘on a four-point scale to indicate the frequency of their occurrence during the last week. Re- sponse options range from “rately or none of the time™ to “most ar all of the time.” Physical symptomatology Was measured by the Cohen-Hoberman Inventory of Physical ‘Symptoms (CHIPS). The CHIPS is a list of 39 ‘common physical symptoms (Cohen and Hoberman, 1983). [tems were carefully selected to exclude symptoms of an abviously psychological nature, e-g., felt nervous or de- pressed. The scale, however, is primarily made up of symptoms, such as headache, back ache, acid stomach, that have been traditionally Viewed as psychosomatic. Each item is rated for the degree to which that problem bothered ‘or distressed the individual during the past «wo ‘weeks, Items are rated on five-point scale from “not at all” to “extremely.” The CHIPS has been found to have adequate reliability and to predict use of student health services in the seven-week period following completion of the scale (¢ = -29 and -22in independent samples). ‘The Social Avoidance and Distress Seale CSADS) was ned to measure socal ansety (Watson and Friend, 1969), This. 28 tructike seale taps both the desire to avoid others (social avoidance) and the experience of distress in social interactions (social distress). Uitiization of the student health center was also monitored. All university students are re- ‘quited to pay a fee that provides for outpatient medical care. The physicians at the student health center routinely fill out 2 standardized form describing the problem(s) for which theA GLOBAL Measure oF PERCEIVED Staess student was treated at each visit. The form is hhased on the International Classification of Diseases, 9ch Revision (Commission on Profes- sional and Hospital Activities, 1980) and allows each visit to be classified as illness-related (Codes 001-779), injury- and poisoning-related (codes £00-999), or “other” (V codes), €.8., receive prophylactic vaccination. For each student, the number of visits in each of the dhree categories was recorded. Up to three symptoms or problems could be checked for any individual visit. Ifthe peob- lems were all in the same category, only one it was recorded for that category. If the problems were in different categories, a sepa- rate vist was recofded for each of the pertinent ‘categories. When total visits (collapsing over categories) were calculated, each visit was ‘counted as one, irrespective of whether there was a single problem or a number of problems in different categories, Both the number of ile ness visits and the number of total visits are analyzed in this report. ‘The number of visits was calculated for each ‘of (wo independent time blocks: the 44 days preceding the testing session and the 46 days following the testing session. The inital 44-day period was used as an indicator ofthe base-rate of visits. College Sudent Sample It Respondents in the second sample were 114 _members of a class in introductory personality psychology (S3 females, 60 males, and one with ‘eX not specified) who received class credit for participating in the study. The mean age of the sample was 20.75 with a standard deviation af 441. These students completed the same five ‘questionnaires as those in the previous study during # one and one-half hour session during the second week of the Spring Quarter. For this sample, the data on health services utiliza- tion were divided into the 90 days preceding the testing session and the 46 days following the testing session. The 90-day period was used as an indicator of the hase-rate of visit Smoking-Cessation Sample Subjects. Subjects were 27 males and 97 fe- males participating in a smoking-cessation pro- ‘gram run by the University of Oregon Smok- 389 ing-Control Program. Participants were soli ited through newspaper, television, and radio advertisements. To qualify for participation in the program, subjects either had to be marvied ‘or living with a partner. The mean age of the subjects was 38.4 years (s = 11.57). Thirty seven percent of the sample made over $25,000 peer year, and 74 percent bad formal education beyond high school. They bad been smoking for an average of 20.9 years (s = 11.82). Only three of the participants were students; all of these were graduate students, 25 to 31 years old, The mean self-monitored baseline smoking rate was 25.6 cigarets a day. Treatment. Treatment groups met for six consecutive weekly sessions lasting approx imately two hours each. ‘The target quitting date occurred on the fourth session. Interven- tion strategies included behavioral sell management techniques, nicotine-fading, and a cogoitive-behavioral relapse prevention pro: gram. Overall, 64 percent (N = 41) of the sub- jects were abstinent at the end of treatment. Measures. During a pretreatment testing session, subjects completed a life-event scale, 4 physical-symptom checklist (CHIPS), and the PSS. The life-event seale consisted of 71 ormatively negative events chosen from the Unpleasant Events Schedule (Lewinsobn and Talkington, 1979). This scale replaced the ove used with seudent samples because it provided fan item pool appropriate to the communi population. Subjects identify the events that have happened (0 them in the last six months ‘and rae the impact of each event on a seven- point scale, canging from extremely negative (0 extremely positive. As in the college student samples, both the number of events and the ‘event-impact scores were analyzed Subjects completed the PSS and the CHIPS prior to the first treatment session and at the ‘end of the six-week treatment. Roth tests were ‘administered in the same manner as with the college studeat samples, except that with the group of smokers, a ane-Week time frame was used for the CHIPS. RESULTS Means, Variance and Reliability Estimates Mean scores on the PS$ for the complete ‘samples {males and females combined) were390 23.18 and 23.67 in the student samples and 25.0 in dhe smoking-cessation sample. Standard de- viations were 7.31, 7.79, and 8.00, and ranges were 6 t0 50, 5 to 44, and 7 to 47. Mean PSS scores for females were 23.57 and 25.71 in the student samples and 25.6 in the community sample. Standard deviations were 7.58, 6.20, ‘and 8.24, Mean PSS scores for males were 22.38 and 21.73 in the suudent samples and 24.0 in the community sample. Standard deviations were 6.79, 842, and 7.80, respectively. Ab though the mean PSS score for females was slightly higher than the mean seore for males in all three samples, this difference did not ap= roach statistical significance in any sample. ‘Age was unrelated to PSS in all three sam- ples. Since the age distribution in the college student samples was severely skewed, a cor relation between the PSS and age was unlikely ‘The correlations between age and PSS were .04 and —.08 in the college samples and —.02 inthe snioking-cessation sample. Coefficient alpha reliability for the PSS was 84, 85, and. .R6 in each af the theee samples. For a stace measure, test-retest correlations should be much higher for short retest intervals tham for longer ones, For the PSS, two inter: Vals are available, two days and six weeks. The PSS was administered, an two occasions sepa- rated by two days, to 82 college students en- rolled in courses at the University of Oregon; When responding to the retest, the subjects were asked to strive for accuracy rather than for consistency across time, The. test-retest correlation in this sample was 85, whereas the cortelation was only .55 for the 68 subjects in the smoking study who were retested alter six weeks. Fvidence for Concurrent and Predictive Validity Separate correlations between the PSS and the validity criteria were calculated for males JOURNAL OF HEALTH AND SOCIAL BEHAVIOR ‘and females in each sample. The relative mag- nitudes of the correlations for males snd fe rales with each of the criteria were compared by transforming the correlations to scores (Fisher's transformation) and dividing the dif ference between the 2 scores by the standard error of the difference between the 2 coeffi- cients (ef. Guilford, 1965). None of the result- ing z scores were significantly different from 0 at the p < 05 level. Since there were no dif ferences beeween males and females, only data for the entire sample are reparted below. Separate correlations be:ween the PSS and validity eriteria were also calculated for those below and above median age in the smoking cessation sample, These analyses scemed un- rcessary in the student samples, since 98 per- cent of the students were between 16 and 25 years old. The 31 persons in the “young” ‘community group ranged from 22 t0 38. The 33 persons in the “old” group ranged from 36 10 70. Correlations for young and okd were com pared with the samme procedure described in the foregoing, Only one of these compatisons ind ‘cated a statistically significant difference at the p< 05 — actually p < 01 — level. Data on this comparison are presented in the appropri ‘ate section inthe following and addressed in the discussion section. Conelations between PSS and Life-Event Scores. Since perceived stress should gener ally increase with increases in objective ‘cumulative stress levels, the PSS should be related to the number of life events. Moreaver, these correlations sliould be higher when the Tife-event scores are based on the selfrated impact of the events, since impact scores re- flect same of the same stressor appraisal mea ‘sured by the PSS. As apparent from Table |, here is a small to moderate correlation be- tween number of life events and the PSS in all three samples, Moreover, in all but one case, ‘that correlation increases when the scale score takes into account the respondent's perception ‘TARLE 1, Coreelations hetween Life-Eyent Scores and PSS College Seudent umber of tite Events 20 Impact of Lite Events *pxal College Student “Beginning of Fed of Semple Treatment Treatment 38 3 ae 3A GLOBAL MEASURE OF PERCEIVED STRESS. TABLE 2, Correlations af Stress Measures with De- pressive Symplamatclony College Student College Student Sample! Sample 1) Number of ie Events We 4 Impact of Life Events a a Perceived ‘Stress Seale a6 as =2<.001 of the events. Hotelling ‘tests, testing the statistical significance of differences between correlations, indicate that this increase is sig ficant {p< .08) for Student Sample I and for the smoking-cessation sample at the beginning of treatment. ‘There was a difference in the correlation between PSS and number of negative events for young and old participants in the smoking cessation sample. For the young, the correla- ‘on was 65 (p <_05); for the old, it was .19. PSS Versus Life Events as a Predictor of Sympiomatology. As noted earlier, we ex- ‘pected that the PSS would be a better predictor of the various health outcomes than would stressful life-event scores. The data presented in Tables 2 and 3 support this prediction in the cease of both depressive and physical symp- tomatology. Hotelling Ctest ( < 05) provide statistical support for these differences in all ‘cases. Since these are cross-sectional correla tions, no causal inferences are implied. For ‘example, it is possible that increased symp- tomatalogy caused increased stress, rather TABLE 3. Correlatons* of Stress Measures with Physesl Symptomatology allege College Smoking. Student Student Cessation Sample 1 Samole M_“Study* Number of Tite Evens a 30 lypact of Tite Events aM si Perceived ‘tess Seale 3 6s 0 * Since ie life-events questionnaire was adminis: tered only. at Seginning of treatment, only heginning-atcreaiment data are presented so that SS and ie-event carelations are equivalent p= 001 forall eorelations. 391 than that the siress caused the symp- tomatology. In regard to establishing the validity of the PSS, itis important co note the substantial cor relations between the scale and both symp- tomatology measures. There is probably some ‘overlap between what is measured by the de- pressive symptomatology scale and what is measured by the PSS, since the perception of stress may be a symptom of depression, This may, to some degree, account for the mag- nitude of that correlation. Tn light of the very high correlation beween the PSS and the CES-D (depressive symptom scale) it is desir. able to demonstrate that these scales are not measuring the same thing. Hence, partial cor- relations were calculated; in these, depressive symptomatology was partialled out of the cor- relations between the PSS and physical symp- tomatolagy, and the PSS was partialled out of the correlation between depressive symp- tomatology and physical symptomatology. In the case of PSS and physical symptomatology, the correlation was .16, p <2 .O1, for sample I and .17,p < 7, for sample IT. In the case of the CES-D and physical symptomatology, the correlation was 31. p < .O1, in sample T and 38, p < 01, in sample Il, Hence, even with the ‘very high correlation between the PSS and CES-D, both scales still independently pre dicted physical symptomatology. PSS Versus Life Events as a Predictor of Utdization of Health Services. Table 4 pre- sents the correlations between the PSS and Utilization of health services, both before and after administration of the scale. {a Sample 1, the PSS significantly predicted utilization dur- ing the five-week periad after completing the ‘TABLE 4, Correlations of PSS and Health Center Udliatin Before and After Conpeting Seale ‘Stident_ Smemt Sample 1 Sample It Before Sale Administration ‘ysical ness Visits =. All Visits ue 05 ‘Aller Seale- Administration Physical ilness Vist 0 AA Visits snes - pel ve p<.3a scale. In Sample Il, there was a nonsignificant correlation between the PSS and all visits after administration of the scale. Correlation he- tween the PSS and physical ilness visits ater administration of the scale, with before-ad- rinistration visits partalled out, was 15, p< 07, for Sample T and — 02 for Sample IL Similar partial correlation for the PSS aad all isits after administration, with all preadmin- istration visits partialled out, was 16, p < 01, for Sample Land .13 for Sample IL. These cor- relations suggest that the PSS is predictive of changes in health center utilization. Correla- tions of fife-event scores with utilization were ‘not significant in both samples for both physi- cal illness and all visits; these correlations ranged from ~.04 to +.08. PSS Versus Life Events as a Predictor of Social Ansiety. Levels of perceived stress in college students, especially freshmen, should be related to their ability to become integrated into the university community. We would ex- ‘pect thatthe poorer the integration, the geeater the perceived stress. The social anxiety seale provides a trait measure that presumably taps difficulty in making friends and social contacts, ‘ce. the ability ro integrate into the community In both student samples, increases in social anxiety were associated with increases in per. ceived stress (37 and 48, p < .Q01 far both). Although number of life events was unrelated to social anxiety in both samples, there were small correlations between event impact and social anxiety (~.13, p < 02, ~.26, p < 01) Again, these are cross-sectional data and no ‘causal inferences are implied. PSS and Smoking-Reduction Maintenance: The Four-trem Scale. To examine the role of the PSS as a predictor of maintenance of smoking-rate reduction, perceived stress level was also assessed one and three months fol lowing treatment. Since posttreatment data Were collected by telephone interview. a short version of the scale, consisting of the four items (numbers 2, 6, 7, and 14) that were cor- related most highly with the 14 item scale, was ‘employed. The mean score on the four-item scale was 5.6 at one month and $.9 at three months. Standard deviations were 3.6 and 4.0, and the scores ranged from 0 to 15 and 0 10 14 ‘Mean PSS scores for males were 4.8 at one ‘month and 5.9 at three, while mean scores for females were 6.2 and 5.9, respectively. The JOURNAL OF HEALTH AND SOCIAL BEHAVIOR ‘coefficient alpha reliability estimate for the Four-item PSS was 72. The test-retest reliabil- ity of the fourvtem scale aver a two-month interval was 38 ‘The four-item scale at one month was corre: lated (31, p <.01) with the average number of cigarettes smoked per day at one month; the Jatier information was from a self-report can- firmed with a carbon monoxide measure. Similarly, dhe four-item scale at three months ‘was correlated (37, p < -001) with smoking rate at three months. Finally, the short PSS scale at one month after tealment predicted smoking rate chree months after treatment (.39, p< O01). In all eases, the greater the PSS score, the more cigarettes smoked. A number ‘of partial correlations were calculated to clarify the nature of the relationship between the abridged version of the PSS and smoking rate In the first partial correlation, the end-of- treatment PSS score and smoking rate were. Partalled out of the correlation between the our-item PSS and smoking rate at one month following treatment. The partial correlation ‘was .29,p < 01. Inthe second correlation, the four-item PSS and smoking rate at ane month following treatment were partialled out of the ‘correlation between the four-item scale and smoking rate at three months. The partial cor- relation was 34, p< 01. These analyses indi- ‘cate that changes'in perceived stress as mea sured by the PSS are predictive of changes in smoking rate. Another partial correlation indi cated that the abridged PSS predicted changes in smoking rate over a two-rionth period, Spe- sifically, @ correlation of .26 (p < .05) was found when smoking rate at the one-month follow-up was partalied out of the correlation between the four-item PSS at one month and smoking rate at three months, These data sug- est that the four-item scale provides & useful measure of perceived stress for use in tele- phone interviews and other situations where a very short scale is cequited, DISCUSSION ‘The PSS has adequate internal and test retest reliability and is correlated in the ex- pected manner with a range of self-report and Dehavioral criteria. Moreover, the PSS is mare closely related to a life-event impact score,AGLORAL MEASURE OF PERCEIVED STRESS which is ta some degree based on the respon dens appraisal ofthe event, than to the more objective measure of the number of evens o¢- curring within a particular timespan. The PSS fal proved to be a better predictor of health and Health-related outcomes than either of the ‘0 lifeevent scales. Finally, the PSS, al- though highly correlated with depressive symptomatology, was found to measure a dit ferent and independently predictive constrict. Teis noteworthy thatthe level of correlation hetwecn the lfeevent scales and the symip- tomatological outcomes (18 1 36 range) is ‘equivalent to, if not beer than, similar care Intions reported in the erature (ef, Rabkin and Strvening, 1976; Tausi, 1982) Hence, the superior predictability ofthe PSS is not attrib- table to psyehometeie weaknesss in the life- event seales that were employed orc idiosya- sratic aspects of the samples under study Moreover, from an absolute perspective, the SS correlations with symptomatologicl mea- ‘sures are quite high (.52 to .76). However, in ‘the case of depressive symptomatology, the correlation may be somewhat inflated by the ‘overlap in the operational deititions of per- ceived stress and of depressive symo- tomatology (ef ohrenwend tal, 1978; Gore, 1981) Tn general, the rclationships between PSS and the validity criteria were unaffected by sex or age. The one exception was the strong re- Javonship between the PSS and numberof life events forthe young and the lack of such a relationship for the ot. These data may ceflect 2 difference in the role of life events in deter- mining stess levels for these two age ous. “That i, other chronic stressors, expectations, ete. may be more important forthe older re- Spondents “The PSS differs from life-event scales in a number of ways. Fest, the PSS asks about a shorter period, one month as opposed to the Ustal six 0 12 months covered By typical life event scales This worth noting that with a subjective scale, the shorter petiod should be surficien since perceived stress during the last month should reflect any objective events that ae stil affecting respondents’ stress levels ‘A second difference between the PSS and a lie-eventseale is the period of time after ad- ministration thatthe seale provides predictions of health-related outcomes, Presumably, life- 303 ‘event scales will be predictive over fairly long periods, such as several months (o several years. We have examined the predictive ability of the PSS over four to 12 week periods after administration, These data suggest that the best predictions occur within a one- or two-month period. {cis our feeling that as chis period is lengthened, the predictive validity of the scale will fal. Afterall, perceived levels of stress should be influenced by daily hassles, major events, and changes in the availability of coping resources, all of which are quite vari- able over a short period. In fact, test-retest rolability analyses indicate that test-retests in- volving a very shart time (two days) result in fairly substantial correlations, whereas admin- istrations six weeks later produce more moder- ate test-retest correlations. ‘As mentioned earlier, the PSS can be used to determine whether “appraised” stress is an etiological (or risk) factor in behavioral dis. ‘orders or disease, It can. also be used to look more closely at the process by which various moderators of the objective stressor/pathology relationship operate. For example, we could determine whether social support protects one from the pathogenic effects of stressful events by altering the appraisal of those events or by altering the process by which appraised stress ‘causes an illness outcome. This second kind of analysis, however, is limited to the degree that the PSS reflects responses ta events outside of those measured hy the objective event instru- iment, That i, itis limited to the degree that i is more global than the objective stressor mea- sure, Finally, the PSS can be used as an out- come variable, measuring people's experi enced jevels of suress as a function of abjective stressful events, coping resources, personality factors, ete, ‘The four-item version of the PSS provides a useful fool when data must be collected over the phone. This scale makes repeated mea- sures of perceived stress in large samples feasible. It should be noted, however, that be- ceause of the limited number of items, the abridged scale suffers in internal reliability and ‘hus provides a less adequate approximation of perceived stress levels than che entire scale, ‘Although not tested in this study, the PSS. may also provide aa economical tool for as sessing chronic stress level. Either the abridged version of the PSS or the complete394 item scale could be used. Monthly adminis- trations of the scale could be summed or aver- ‘aged, providing a reliable, ie, based on more samples, measure of chronic stress, as well as a predictor chat represents a longer term than the ‘one-month period covered by the scale. Two commonly used measures of nonspecific psychological distress, the nine- scale, 54-item PERT Demoralization Measure (Dohrenwend et al., 1980) and the 28-item General Health Questionnaire (Goldberg, 1972), include a number of items that are si lar to those in the PSS. These scales, however, fare much broader in scope. They are designed as epidemiological measures of symp- tomatology, and both include a broad range of ‘ems tapping common psychiatric symptoms, such as hostility, diminished self-esteem, de- pression, and anxiety, as well as psychoso- matic complaints. Although appraised stress may be symptomatic of psychological disorder when viewed in combination with elevated scores on other psychiatric symptoms, itis our ‘contention that the perception of stress itself, as assessed by the PSS, is not a measure of psychological symptomatology. This conten- tiom is, infact, supported by the data indicating APPENDIX A: ‘eamts and Instructions for Perceived Sire Seale JOURNAL OF HEALTH AND SOCIAL BEHAVIOR independent predictive validities of the PSS and a depressive symptomatology scale Hence, the PSS can be viewed as assessing a slate that places people at risk of, ic. is an- tecedent to, clinical psychiatric disorder even though that state is also part of a diverse set of feelings and states that are characteristic of disorder. In sum, the PSS is a brief and easy-to- administer measure of the degree to which situations in one's life are appraised as stress- ful, Tehas been proven ta possess substantial reliability and validity; thus, it provides a potential too} for examining issues about the role of appraised stress levels in the etiology of disease and behavioral disorders NOTES There are ovo recently developed life-event scales chat enters events aver # one-month pe- Fiod, the Hassle Scale (Kanner etal, 1981) and the Unpleasant Events Seale (Lewinsohn and ene oat 2, Peatlinetal (1981) have poled out that theres ro single time frame that is optional fr observ- ing the effects af diverse le events. However, existing works, including tbe Peaelia et al. ‘Study, examine rather long periods ‘The questions inthis scale ask you about your feelings and thoughts during the lst month. In each case, you willbe asked ta indeste how afte you fel oF thought a certain wa. AIRHOUH sone ofthe questions are Kinilar, there ae differences between them and you should ceat each one a5 a separate question, The best Approach sta answer each question faely quickly. Thats, don cry count up the number of umes you felt ‘a'particular way, but rather indicate the alternative tha seems like # reasonable estimate For each question choose fom the fllosing altemaives almost sever fairly often very often In the last month, how often have you been upset because of something tht happened unexpectedly? Tn the last month, how often have you felt that you were unable to contra the imparane things ia your life? {othe last month, how offen have you felt nervous and “stressed”? 42 In the last month, bow often have You deat svecessully with ivatng life hassles? 54 Inthe last manth, how often have you felt that you were efectively coping with portant changes that were occuring i your life? 6+ Jn te las ar, ow fle fave you fle confident about your aby to handle your persona! problems ‘12 Tate last month, how afien have you felt that things were going your way? 8, In the last month, how often have you found that va could not cape with al he things dhat you had to-do" 9h the last month, how allen have you been able t9 canrol icsitations in your ie? 10." Tn the last month; how often have you fet that yon were om top of things?A GLOPAT. MEASURE OF PERCECVED STRESS APPPNDDY A (Continued) a5 |. Inthe ist month, how often have you been angered because of things that happened that Were outside of your cancel? 12, Inthe last month, how often have yoo found yourself thinking about things that you have to accomplish? 13. Ih the last month, how often have you been sble to control tke way ¥ou spend your time? 4. Inthe last month, how often have vou fel dificuliles were piling up so high tha you could noc overcome hers? Scored inthe reverse direction REFERENCES Averil, 1. B. 1979" “Personal contol over aversive stimuli and 42s relationship tose." Psychological Bulletin 80286-2302. ‘Cobb, Sidney, and Stanley V. Kas) 1907” Termination: "The Consequences of Sob Lass, Report No. %6-1261-Cincinati, OH: National Institute for Occupational Safety land Health, Behavioral and. Mosivational Factors Research, Coben, 8. 178 “Exviconmental load and the allocation of attention.” Pp. 1-29 in Andrew Baum, Terome E. Singer, and Start Valin eds.) ‘Advances in Enviconmental Psychology, Vol. 1. Hilludale, NI Erlbauoy Cohen. S..and H. Hoberman 1963 “Positive events and social supports as buffers of life change stress." Journal of Applied Social Psychology 13:99=125. Cohen, $,, 3a N. Weinstein 1961, “Nonauditory effets of noise on behavior and health.” Jourea) of Social Tssues 3736-7, Comaission 0 Professional and Hospital Acuivuies 1880" Irvernational Classfeation of Diseases, 2h ‘ev, Clinial Modification ICD.$ CM, Vol. 1 Ant Arbor, ME: Commission on Prof sional and Hospital Activities. Dohtenwend, Baroara 5. and Bruce P. Dobreo- wend (eds) 1974" Suessful Life Even: Their Nature and EF esis. New Yook: Wiley Stressfl Life Events and Their Comexts New York: Prodist. Doohienwend, B. S., L. Krasnoff, A. R. Askenasy, ‘and BB, Dohreowend 1978 “Beemnplifeation of a method for sealing life events: The PER] Life Events Seale Journal of Health and Social Behsvior 19208229, Dohtenwend, 6. P., P. E. Shout, G. Bart and F. 3. Mendelsaha 1980 "What psychiatric screening scales mea- sure in the general population: Part I The components of demtoralization by conuast ‘Wid otter dimensions of psychopashol- ogy." Archives of General Psychiatry SPinao36 Dooley, D. and R. Catalano 1980 "Econom change as a cause of bebavioral 188) disorder.” Bulletin 87-430-68, Glass, David C., and Jerome E. Singer 191i Urban Stress: Experiments on Noise and Social Stressors. New Yark: Academic Press Gochmnan, I 1999 “Arousal attribution and environmental stress." Bp. 67-92 in lewin G. Sarason and ‘Charles 1 Spielberger (eds), Suess and ‘Aaniety, Vol. 6 Washington: Hemisphere Goldberg, David P. 1972 The Detection of Psyehiauic ness by Questionnaire. London: Oxford Univesity Press Gore, 8. 1978 “The effect of social suppor in moderating the health consequences of unewploy feat" Journal of Health and Socal Be- Bavior 19:157-65, ‘Stresscutfering Anctions of social sup- pons. An apprasat and clacifcation of re- ‘earch models.” Pp. 202-22 in Barbara S. Dabrenwend and nice P, ‘Dahrenwend (eds. Stressful Lite Events and Their ‘Contexts. New York: Prodist. Guilford, Joy Favi 96S. Fundamental Statistics in Psychology and Education, New Yoru: McGraw-Hill, Holmes, T. Ho, and R. Hf. Rabe 1967 The Social adjustment scale.” Journal af Peychosomatle Medicine 1121318 Johnson, J and C0, Sarason 1979""Maderator vatlables in life. stress 12. seateh." Pp 151-68 in lewia C. Sarason and hanes D. Spielberger (eds), Surese and Anxiety, Vol. 6. Washington: Hemisphere Kahn, Robert L., Donald M. Wolfe, Robert P Quing, J. Diedriek Snack, and’ Robert A Rosenthal 1964" Organizational Stress: Studies in Role Cos: fet and Anibigity. New York: Wiley Kanner, A.D... C. Coyne. C. Schaefer, and & 8, stars Psychological 1981 1961 “Comparison of wo modes of stress mea surement: Daly hassles and uplits versus ‘aor life events.” Journal of Behavioral Medline 4:1-39. Keating, J 1999" “Environmental stressors misplaced em: phasis” Pp. 58-86 ip Irwin G. Sarason and Charles Di Spietberger (eds. Stress and Anxiety, Vol. 6, Washington: Hemispnere,396 Koabasa, 8. C 1979 "Stressful lite events, personality, and health: An inguty inta hardinets.” Jourgal of Personality and Social Psychology Sete Lazaras, Rechant 8 1966 "Psychological Stress and the Coping Pro- cess, New York: McGraw-Hill 1977 "Paychological stess and coping in adapta: tion and illness.” Pp. Lda26 ia 2 J Lipowski, Don R. Lipsi, and Peter C. Whybrow (eds), Psychosomatic Medicine Current Trends, New Yark: Oxfard Uni sersity Press Lei, H., and H. A. Skinner 1380"""A. psychometric study of life events and social readjustment.” Journal of Psychosomatic Research 2457-65, Levine, Mand D. V. Perkins 1980" "Tailor making fe events scale.” Pre sented a the mecting of the American Psy- chological Association, Montreal Lewinsabo, P.M., and J. Talkington 199" "Stadies on the measurement of unpleasa events and relations with depression Applied Psychological Measurement 3:83- 1 Mason, J. W. 1971! “A re-evaluation of the concept of “non: specificity in siress theory.” Soureal of Poyehiatrie Research 8:338- 33, Peaslin, LL, M. A. Lieberman, E. G. Menaghan, and fT Mullan rsa “The stress process." Fournal of Health and Social Behavior 2:157-56 Rabkin, 1 G., and E. L. Scuesing 1996" “Life events, sicess, and illness.” Science (9a:t013-20, Radlett, L, 1977 “The CES.D seale: A sellrepor: depression scale for research in the genera) pon tion.” Applied Psychosocial Measurement E3ts-401 JOURNAL OF HEALTH AND SOCIAL BEHAVIOR Sarason, .G., J. HL Johwson, and J. M. Siege 1978 “Assessing te impact of life changes: De- velopment of the life experiences survey.” ‘Fournal of Consulting and Clinical Psych oy a6:932-46 Seligman, Martin FP IMS” Helplessness: On Depression. Develop- ‘ment and Death. San’ Prancisco: W. Hi, Freeman Stuoebe, W., M.S. Strocbe, KJ. Gergen, and M. Gergen 1982 "The effects of hereavemment on mortality A social psychological analysis."" PD. Soret in J. Richard Eiser {ed}, Social Psychology and Behavior Medicine, New York: Wiley. Sundstrom, F STB "Crowding as «sequential process: Review of research on tbe effects of population ‘density on humans.” Pp. 32-116 tm Andrew Baum and Yakov Epstein (eds), Human Response to Crowding. Hillsdale, NU Ebeum, Tausig. 1982. “Measuring lie events" Journal of Health and Social Behavior 23:52-6¢ Vinokur, A.) and ML. Selzer 1975. “Desirable versus undesitabe ie events: Thee felationship to stress and mental ise tgess." Journal of Personality and Social Psychology 32:329-77 ‘Watson, D., and F. Friend 1309" Reasurement of socia-evaluative. anni ery” Journal of Consulting and Clinical Peychology 38:448-57. Worehel, 5 1978 "Reducing crowding without increasing space: Some applications ofan attributional theory of crowding.” Journal of Population 1306-30, Worchel, 8, and ©. Tedalie 1976. "The experience of crowding: &. «wo factor theory.” Journal of Personality and Soc Peychology 3420-40,