28. W. Dupree and J. West, United States EnergyThrough the Year 2000 (Government PrintingOffice, Washington, D.C., 1972, updated 1975).

29. E. Cook, Study ofEnergy Futures (Environmen-tal Design Research Associates, Chapel Hill,N.C., 1975).

30. R. Knecht and C. Bullard, Direct Use ofEnergyin the U.S. Economy (Technical Memo No. 43,Center for Advanced Computation, Universityof Illinois, Urbana, 1975).

30a.Stanford Research Institute, Patterns ofEnergyConsumption in the U.S.-1968 (GovernmentPrinting Office, Washington, D.C., 1972).

30b.M. Beeler, Source Book for Energy Assessment(Brookhaven National Laboratory, Upton,N.Y., 1975).

31. Verksamheten (Activities) (Angpannefore-ningen, Stockholm, 1971-1976).

32. SOS Industri 1970-73 (Sweden's Official Statis-tics-Industry 1970-73) (Statistiska Central By-ran, Stockholm, 1972-1975).

33. Energi Program Kommitteen (EPK), Energi-forskning: vol. A. Energin och Naringslivet (En-ergy R & D: vol. A, Energy and the Economy)(Allmanna Forlaget, Stockholm, 1974).

34. The latest energy use data for Sweden are pub-lished as part of the Energy Conservation Bill inthe Parliament: Energy-Hushdllning (EnergyHusbandry) (Ministry of Industry, Stockholm,February 1975). See also Energiforsoriningen1975-1980, Rapport av Statens Industriverk(Allmanna Forlaget, Stockholm, 1976).

35. R. Herendeen and C. Bullard, Energy Syst., inpress.

36. E. Hirst, "Energy intensiveness of passengerand freight traffic, 1950-1970," Oak Ridge Natl.Lab. Rep. ORNL-NSF-EP41 (1973).

37. Federal Highway Administration, EstimatedMotor Vehicle Travel in the United States(Washington, D.C., 1970-1975).

38. Federal Highway Administration, NationwidePersonal Transportation Survey (GovernmentPrinting Office, Washington, D.C., 1970).

39. Federal Energy Administration, Project Inde-pendence: Energy Conservation: Trans-portation (Government Printing Office, Washing-ton, D.C, 1974).

40. Ministry of Communications, Regionala Traf-fikplanering (Regional Traffic Plan) (Stockholm,1972); Transporter i Sverige (Transportation inSweden) (Report Dsk 1975: 4, Stockholm, 1974);Swedish Institute, Fact Sheet ofSwedish Trans-portation (Swedish Information Service, NewYork, 1973).

41. Energi Beredskaps Utredningen (EBU), Energi:Beredskap i Kristid (Energy: Preparedness Dur-ing Crisis) (Allmanna Forlaget, Stockholm,1975).

42. Motor Vehicle Manufacturers Association, Au-

tomobile Facts and Figures (Detroit, 1975); Mo-tor Vehicles and Energy (Detroit, 1974).

43. K. Austin and K. Hellman, Passenger Car FuelEconomy (SAE Paper 730790, Society of Auto-motive Engineers, New York, 1973).

44a.Motor Traffic in Sweden (AB Bilstatistik, Stock-holm, 1974-1975).

44b.See also J. Ullen, Bilfakta (Auto Facts) (JanUllen AB, Stockholm, 1975), vol. 28.

45. Energi Program Kommitteen, Energiforskning:vol. C, Transport och Samfdrdsel (Energy R &D: vol. C, Transportation) (Almanna Forlaget,Stockholm, 1974).

46. Statens Vagverk, Personbilarnas ArligaKorlangd (Yearly Private Car Vehicle-Miles)(Stockholm, 1974).

47. K. Austin and K. Hellman, Passenger Car FuelEconomy as Influenced by Trip Length (SAEPaper 750004, Society of Automotive Engineers,New York, 1975).

48. Svenska Esso, Oljedret 1974 (Oil Year 1974)(Stockholm, 1975).

49. A. Makhijani and A. J. Lichtenberg, An Assess-ment of Residential Energy Use in the U.S.A.(ERL M-310, Univ. of California EngineeringResearch Laboratory, Berkeley, 1973).

50. J. Moyers, "The value of thermal insulation inresidential construction," Oak Ridge Natl. Lab.Rep. ORNL-NSF-EP-9 (1972).

5la.Federal Energy Administration, Residential andCommercial Energy Use Patterns, 1970-90(Government Printing Office, Washington,D.C., 1974).

5lb.See also S. Dole, Energy Use and Conservationin the Residential Sector: A Regional Analysis(R-1641-NSF, Rand Corporation, Santa Monica,Calif., 1976).

52. Indeed, the mortgage law of 1957 gave priorityto home builders or buyers who intended toinsulate beyond the building code minimums. Anewer program (1975) provides greater subsi-dies, direct grants, and local testing programs,with total outlays of over $350 million over 3years.

53. Centrala Driftledningen, Sveriges Elkonsum-tion, 1975-1985 (Sweden's Electricity Use,1975-1985) (Stockholm, 1972).

54. Energi Program Kommitteen, Energiforskning:vol. D., Uppvarmning och Lokal Komfort (Ener-gy R & D: vol. D, Heating and Space Comfort)(Almanna Forlaget, Stockholm, 1974).

55a.Swedish District Heating Association, Com-bined Heat and Power, 1975 (Varmeverks-forening, Stockholm, 1976).

55b.R. Johnson and L. Bell, Malmo's Powerplants(Industriverket, Malmo, Sweden, 1973).

56. J. Meyers et al., Energy Consumption in Manu-facturing (Ballinger, Cambridge, Mass., 1975).

57. Department of Commerce, Census ofManufac-

turers (Energy Data for 1971) (GovernmentPrinting Office, Washington, D.C., 1972).

58. W. Chern, "Electricity demand by manufactur-ing industries in the U.S.," Oak Ridge Natl.Lab. Rep. ORNL-NSF-EP-87 (1975).

59. E. Gyftopolous, L. Lazaridis, T. Widmer, Po-tential Fuel Effectiveness in Industry (Ballinger,Cambridge, Mass., 1975).

60. S. I. Kaplan, "Energy demand patterns of elev-en major industries," Oak Ridge Natl. Lab.Rep. ORNL-TM4610 (1974).

61. B. Carlsson, in IUIs Ldngtidsbed6mning 1976(Industrins Utrednings Institut, Stockholm,1976), p. 270; lecture at Industrial Energy Day,Stockholm, May 1976.

62a.Swedish Federation of Industries, Energy Con-servation in Swedish Industry (IndustriforbundetsForlag, Stockholm, 1974).

62b.A. Iveroth and B. Helmerson, Industrin ochEnergi (Industry and Energy) (Indus-triforbundets Forlag, Stockholm, 1974).

63. Statens Industriverk, Tdtorternas och denTunga Industrins Energi F6rs6rjning (EnergySupply to Industry and Built-Up Areas) (LiberForlag, Stockholm, 1976).

64. U. Norhammer et al., Samlat Energikunnande(Collected Energy Skills) (Angpanneforeningen,Stockholm, 1975).

65. Dow Chemical Co.; Environmental ResearchInstitute; Greenspann-Townsend, Inc.; Cra-vath, Swaine, and Moore, Inc., Energy Indus-trial Center (National Science Foundation,Washington, D.C., 1975).

66. In Sweden, government-owned power stations,including hydroelectric facilities, produced 50percent of the total electricity. These stationswere financed at commercial interest rates, incontrast to most government-owned power inthe United States.

67. B. Hannon, Science 189, 95 (1975).68. F. Felix, Electr. World 184 (No. 9), 64 (1975).69. J. Holdren, Bull. At. Sci. 31, 26 (December

1975).70. This research was supported in part by the Law-

rence Berkeley Laboratory, University of Cali-fornia, under ERDA contract W-7405-Eng-48.We also acknowledge the support of the Energyand Resources Group, University of California.One of us (L.S.) has also received grants fromthe Society for the Advancement of Scandinavi-an Studies and the Swedish Information Ser-vice, San Francisco. We acknowledge the help-ful suggestions of J. Holdren, R. Herendeen, C.Bullard, and B. Carlsson. Mans Lonnroth (ofthe Secretariat for Future Studies, Stockholm)provided invaluable help in coordinating ourresearch efforts in Sweden. This work is basedon report LBL-4430 from the Lawrence Berke-ley Laboratory.

Life Events, Stress, and IllnessJudith G. Rabkin and Elmer L. Struening

Studies relating social factors and lifeevents to illness appear with remarkableregularity in the major psychological,psychiatric, psychosomatic, and socio-logical journals, and to a lesser extent inthose of clinical medicine and epidemiol-ogy. While some of these publicationsderive from the cumulative efforts of in-vestigators who have worked in this fieldfor many years, concern has been ex-pressed that many recent studies repeat

3 DECEMBER 1976

both the findings and the flaws of earlierones, delaying a hierarchical growth anddevelopment of knowledge in the field.Accordingly, there is a need for criticalevaluation of this literature, taking inissues of method as well as content. Inthis article our goals are (i) to reviewselectively the research literature on therelations of life events, stress, and theonset of illness; (ii) to delineate trends inits development; (iii) to evaluate the con-

ceptual and methodological approachesemployed; (iv) to identify major vari-ables mediating the impact of stressfulevents on individuals and groups; and (v)to recommend more comprehensive ap-proaches to substantive issues.

Despite historical recognition of thepredisposing role of social factors in theonset of illness, it is only during the last40 years that scientists have attempted tostudy these phenomena systematically.In 1936 Hans Selye articulated his con-cept of stress as the "general adaptationsyndrome," a set of nonspecific physi-ological reactions to various noxious en-vironmental agents (1). This formulationwas largely responsible for popularizingthe concept of stress in the scientific

Dr. Rabkin is research scientist in and Dr. Struen-ing is director of the Epidemiology of Mental Dis-orders Research Unit, New York State Departmentof Mental Hygiene, Psychiatric Institute, 722 West168 Street, New York 10032. This article is adaptedfrom a paper presented at the annual meeting of theAmerican Association for the Advancement of Sci-ence, New York, 27 January 1975.

1013

vocabulary of medicine, and it initiatedan era of research and theoretical devel-opment conducted with accelerating en-thusiasm on an international scale in nu-merous branches of the medical and laterthe social sciences. Also in the 1930'sFranz Alexander and his psychoanalyticcolleagues in Chicago became interestedin relating personality characteristics toselected organic syndromes within theframework of psychosomatic theory. De-velopment of the stress and psycho-somatic models of illness has proceededapace, with a gradual convergence ofinterest and assumptions so that todaystress research and psychosomatic re-search are to some extent overlapping.The notion of socially induced stress

as a precipitating factor in chronic dis-eases is gaining acceptance among awide spectrum of scientists. It is becom-ing recognized that stress can be one ofthe components of any disease, not justof those designated as "psychoso-matic." As Dodge and Martin (2) haveexpressed it, "the diseases of our times,namely the chronic diseases, are etiologi-cally linked with excessive stress and inturn this stress is a product of specificsocially structured situations inherent inthe organization of modern technologicalsocieties." Even susceptibility to micro-bial infectious diseases is thought to be afunction ofenvironmental conditions cul-minating in physiological stress on theindividual, rather than simply of expo-sure to an external source of infection(3).

In the formulation of a revised etiologi-cal model, illness onset is generally asso-ciated with a number of potential factors,including the presence of stressful envi-ronmental conditions, perception by theindividual that such conditions are stress-ful, the relative ability to cope with oradapt to these conditions, genetic pre-disposition to a disease, and the pres-ence of a disease agent. In this contextthe stress concept does not only explainwhy some people are more prone to ill-ness than others. Stress, like anxiety, isa broad and general concept describingthe organism's reactions to environmen-tal demands. Its utility derives from itsrole in identifying productive lines ofresearch on the etiology of disease, en-compassing external events that influ-ence individuals and populations and al-so their appraisals and interpretations ofsuch events. Accordingly, we will turn toa consideration of the body of researchthat has focused on the correspondencebetween life changes, stress, and illnessonset.

In the following review we limit ourattention to life changes of a primarily

1014

personal nature. Changes caused bywidespread social processes also height-en the individual's vulnerability to stressand stress-related diseases, but we shallconsider only life events which are expe-rienced primarily on an individual level,such as changes in family status or occu-pation.

Definitions

We consider the following sequence ofconditions: social stressors, mediatingfactors, stress, onset of illness. In thepresent context the term "social stres-sors" refers to personal life changes,such as bereavement, marriage, or lossofjob, which alter the individual's socialsetting. A more specific definition is pro-posed by Holmes and Rahe (4), whodefine as Epial-stressors any set of cir-cumstances the advent of which signifiesor requires change in the individual'songoing life pattern. According to thisconception, exposure to social stressorsdoes not cause disease but may alter theindividual's susceptibility at a particularperiod of time and thereby serve as aprecipitating factor.

"Mediating factors" are those charac-teristics of the stressful event, of theindividual, and of his social support sys-tem that influence his perception of orsensitivity to stressors. Some are long-term predisposing factors which height-en the individual's risk of becoming ill,such as high serum cholesterol in rela-tion to myocardial infarction. Othersmay render the individual less vulnerableto stress, such as prior experience withthe stressor. In general, consideration ofmediating variables contributes to an un-derstanding of differential sensitivities tosocial stressors.

"Stress" is the organism's response tostressful conditions or stressors, con-sisting of a pattern of physiological andpsychological reactions, both immediateand delayed. "Onset of illness" is de-fined by the appearance of clinical symp-toms of disease.

"Predisposing factors" are long-standing behavior patterns, childhood ex-periences, and durable personal and so-cial characteristics that may alter thesusceptibility of the individual to illness."Precipitating factors," in contrast, in-fluence the timing of illness onset; theterm refers for the most part to more orless transient changes in current condi-tions or characteristics, and it is suchchanges that constitute our present sub-ject of inquiry. "Chronic disease" refershere very generally to syndromes whichare of long duration and are non-

infectious. It is the chronic diseases rath-er than the acute, infectious ones that areusually thought to be particularly influ-enced by the experience of stress.

Life Events Research

The role of stressful life events in theetiology of various diseases has been afield of research for the last 25 years.Derived from William B. Cannon's earlyobservations of bodily changes related toemotions and Adolph Meyer's interest inthe life chart as a tool in medical diag-nosis, the field was first given formalrecognition at the 1949 Conference onLife Stress and Bodily Disease spon-sored by the Association for Research inNervous and Mental Diseases. Sincethen several groups of investigators haveadopted this general framework in inde-pendent long-term projects.

In general, the purpose of life eventsresearch is to demonstrate a temporalassociation between the onset of illnessand a recent increase in the number ofevents that require socially adaptive re-sponses on the part of the individual.The impact of such events is presumedto be additive; more events are expectedto have greater effect. The underlyingassumption is that such events serve asprecipitating factors, influencing the tim-ing but not the type of illness episodes.Onset of psychiatric as well as physicaldisorders and accidents have been stud-ied in both retrospective and prospectivedesigns within the life events framework.Most investigators working in this

field have adopted in original or modifiedform a 43-item checklist developed byHolmes, Rahe, and their colleagues. Thechecklist items are intended to representfairly common situations arising fromfamily, personal, occupational, and fi-nancial events that require or signifychange in ongoing adjustment. Scores onthe first version, known as the Scheduleof Recent Experience (SRE), consistedof the number of items checked. Subse-quently, weights were assigned to eachitem based on ratings by a standard-ization sample ofjudges who were askedto rate the life events "as to their relativedegree of necessary readjustment ...the intensity and length of time neces-sary to accommodate to a life event" (5).On this scale, known as the Social Read-justment Rating Scale (SRRS), death ofspouse, for example, is weighted at 100(the highest point on the scale), marriageat 50, change in recreation at 19, vaca-tion at 12. This and comparable check-lists, usually covering the previous 6 to24 months, are typically used as the mea-

SCIENCE, VOL. 194

sure of stressful life events. Modified junction with medical records. In addi-forms have been developed for specific tion to military personnel, employees ofpopulations such as children, college stu- large corporations and clinic or hospitaldents, and athletes. patients have been popular subjects inThe most elaborate and extensive pro- retrospective studies because of the

gram of life events research has been availability of long-term records.conducted by Rahe, Holmes, Gun- Cn both retrospective and prospectivederson, and their colleagues. Their investigations, modest but statisticallywork, originally based largely on Ameri- significant relationships have been foundcan naval shipboard personnel, has been between mounting ife change and theextended on an international basis to occurrence or onset of sudden cardiacother naval samples and to diverse civil- death, myocardial infarctions, accidents,ian groups, and has evoked considerable athletic injuries, tuberculosis, leukemia,comment in the literature, both positive multiple sclerosis, diabetes, and the en-and negative. This brief description of tire gamut of minor medical complaintstheir overall approach and representa- (6, 9). High scores on checklists of lifetive findings is intended to illustrate the events have also been repeatedly asso-kinds of research and major issues and ciated with psychiatric symptoms andproblems in the field. disorders, and such scores have been

In their early retrospective studies, found to differ between psychiatric andRahe and his colleagues (6) asked over other samples (10, IIj2000 navy personnel to report their life It has been further noted that lifechanges and histories of illness during /events may be related to the course ofthe previous 10 years. Number of illness illness and recovery, whatever the etiol-episodes was related to scores on the ogy of the primary disease (12). In addi-SRRS; these scores are referred to as life t/ tion, periodic analyses of life events maychange units or LCU's. In general, ac- serve to monitor and help predict thecording to these investigators, those who course of illness, as illustrated in a post-recorded fewer than 150 LCU's for a hospital follow-up of mental patients bygiven year reported good health for the Michaux et al. (13).following year; of those with annuaL--LCU's between 150 and 300, about halfreported illness in the next year; and Statistical Issueswhen annual LCU scores exceeded 300,as they did for a small proportion of the Is it, then, reliably established thatrespondents, illness followed in 70 per- stressful life events commonly precedecent of the cases, and furthermore the onset of a wide variety of physicaltended to entail multiple episodes. and psychiatric disorders in populations?

In prospective studies of 2500 Ameri- As presented in the literature, the resultscan naval personnel aged 17 to 30, life are impressive. Their sheer number, theevents that occurred in the 6 months variety of populations studied, and theprior to shipboard tours of duty were range of disorders implicated togethercompared with shipboard medical rec- suggest that this is a useful and meaning-ords of the 6-month cruise. Respon- ful procedure for predicting illness and,dents were grouped into quartiles based more generally, for learning more abouton their precruise LCU scores, and vulnerability to illness. However, closermean rates of illness were computed for scrutiny of methodological and theo-each group. Those in the first quartile retical aspects of the research, as well ashad a mean of 1.4 recorded illnesses,K the actual data that have been reported,those in the fourth quartile 2.1 recorded., uncovers a host of serious issues. Theseillnesses, a statistically significant differ- are attracting increasing attention bothence. Similar results were obtained with from critics of the life events approacha sample consisting of 821 Norwegiansailors (7) and in a study of the entirecrew of 1005 men on a warship on com-bat duty off Vietnam (8).Numerous other studies by various in-

vestigators have similarly shown associa-tions between number and intensity oflife events and the probability of specificillnesses in the near future [see (5) or (6)for extensive lists of references]. Inmany of these studies data about lifeevents and illness episodes were gath-ered from large, heterogeneous samplesby means of questionnaires in con-3 DECEMBER 1976

and from investigators who use the meth-od.The most immediate issue, and one

which has received only cursory atten-tion from investigators in this field, con-cerns the size and practical sicanceof the correlation between number andnature of life events and subsequent ill-ness episodes. The vast majority of lifeevents studies have, until very recently,relied on statistical methods of the mostrudimentary nature to analyze this rela-tionship. Between-group differences areoften reported only in percentages, or

else exclusively in terms of statisticalsignificance (P levels). Given the verylarge sample sizes characteristic of lifeevents research, even very small correla-tions of no practical utility may passtests of statistical significance.

Reports of obtained correlationcoefficients are often conspicuously miss-ing. When present, they are typicallybelow .30, suggesting that life eventsmay account at best for 9 percent of thevariance in illness. In Rahe's naval data,coefficients of correlation between lifeevents and illness were consistentlyaround .12 (7), and other investigatorshave reported equally low, albeit statisti-cally significant, correlations (11, 14, 15).Similarly, when statistically significantdifferences in illness rates are reportedfor groups classified in terms of prior lifeevent scores, or groups of differinghealth status are compared with respectto number of prior life events, attentionis often focused exclusively on groupmeans. Variability of scores withingroups tends to be overlooked, evenwhen it is extreme, as observed by Wer-show and Reinhart (16). In practicalterms, then, life events scores have notbeen shown to be predictors of the proba-bility of future illnesses.

Psychometric Issues

It seems likely that stronger relation-ships between life events and illness epi-sodes might be obtained if the psycho-metric properties of the measuring in-strument were improved and the out-come criteria refined. Although fewstudies of the reliability and validity oflife events checklists have been pub-lished, available evidence suggests weak-nesses in both these respects. Rahe (7)reports correlations ranging from .26 to.90 in test-retest reliability of the SRRS.He attributes such wide variation to vari-ations in intervals between questionnaireadministration, differences in samplecharacteristics, and complexity of word-ing used in the questions. As Sarason etal. (17) have concluded, by any reason-able standard the reliability of the SRE islow.Rahe et al. (7) report that wives' inde-

pendent scores of their husbands' recentlife changes show correlations with thehusbands' self-reports ranging from .50to .75. Other questions about validityconcern respondents' errors of omissionor commission and definition of the cri-teria of illness with which checklistscores are correlated. Brown (18) hasreferred to the problem of "retrospectivecontamination" where respondents may

1015

,exaggerate past events from a need tojustify subsequent illnesses. He cites astudy of mongolism, published in 1958before chromosomal abnormalities wereassociated with the syndrome, which"demonstrated" the etiologic impor-tance of stressors of the mother duringpregnancy. On the other hand, Rahe (7)reported that in studies of patients withcoronary heart disease where recent lifechanges were gathered both by question-naire and interview, the patients rarely ifever listed life changes in the question-naires that were not substantiated in theinterviews.Another form of contamination that

may be a more significant source of erroris that a given life event and an illnessperceived or reported shortly thereaftermay be products of the same phenome-

/non, so that one cannot be said to dis-tinctly precede or precipitate the other.This problem may arise when the causeand the effect of a life event are both atleast a partial result of the actor's behav-ior, as, for example, in the case of acollege student who drops out of schooland then manifests psychiatric symp-toms. Divorce can be regarded as a lifechange contributing to depression, butdepression in some cases may be a con-tributory factor in divorce. Although theproblem of clearly differentiating be-tween life change and observed outcomehas not been ignored in the literature,satisfactory solutions have not yet beenachieved. According to Hudgens (19), 29of 43 events on the SRRS checklist areoften the symptoms or consequences ofillness, and as such are possible sourcesof contamination.

Several investigators have wonderedwhether life event checklist scores areactually associated with care-seeking be-havior rather than with the onset of ill-ness. Since care-seeking-that is, thefact of a medical record-is frequentlyused as the operational definition of ill-ness onset in college populations, navalshipboard studies, and elsewhere, theissue is not easily resolved. Cadoret (20)and Hudgens (19), in their studies of lifeevents and psychiatric depression, bothsuggest that mounting life changes pre-cipitate psychiatric hospitalization, notthe appearance of symptoms. Hudgensnoted that while causal relations havebeen found between stressful life eventsand worsening of psychiatric conditionsalready existing, and between life eventsand subsequent admission to treatmentfacilities, he has not found it con-vincingly demonstrated that ordinary lifeevents cause illness. Instead, it may bethat life changes lead people to seekmedical treatment, that they are equiva-

1016

lent, perhaps, in their etiological role to score for the year preceding hospital-the availability of medical facilities or ization was very low, and 19 percent offunds with which to pay for treatment. the sample reported absolutely no lifeMechanic (21), studying the use of out- events at all, apart from Christmas. Be-patient medical services, has also sug- fore concluding that this population in-gested that stress helps to trigger use of a deed experienced few ordinary lifemedical facility, if not the development changes, it is necessary to verify theof symptoms. This distinction between appropriateness and relevance of theillness onset and treatment-seeking be- checklist items for these particular re-havior may apply to disorders of gradual spondents. This question can be extend-onset and to those which often go un- ed to consider the appropriateness oftreated, such as colds and headaches. various life event items for members ofThe issue is, however, irrelevant to the different socioeconomic and ethnicassociation of life changes with acci- groups.dents, suicide, mortality rates, and epi- In attempting to evaluate the adequacysodes of acute, severe illness such as of item selection, Dohrenwend (25)myocardial infarctions. It is perhaps in asked samples of community residents,the realm of psychiatric disorder that the community leaders, psychiatric patients,most care is warranted in handling this and convicts to respond to an open-issue. ended question regarding "the last major

event in your life that . . . changed yourusual activities" and then to the standard

Content Validity checklists of life events for the precedingyear. He found that surprisingly few of

Investigators using the life events ap- the events reported in the checklist wereproach have differences of opinion about previously described by respondents asthe nature of the events to be included in events they considered major. Further,checklists. Though the various in- his different samples cited different kindsstruments in use have overlapping items, of events. He concluded that there reallythey vary in length, content, relative are several domains of life events, andnumber of positive and negative items, those to be sampled must depend on theand number of items over which respon- ,goals of a given study.dents have no control (such as "death Considering the same issue, Kellamof a friend," in contrast to "marriage"). (26) regards the present checklists as tooMost checklists selectively emphasize simple and conceptually deficient. Heevents of young adulthood, undesirable suggests the stratification of life eventsevents, and subiectively evaluated with items representing the following cat-events; this may make it difficult to inter- egories: age group, stage of life, locus ofpret findings when various groups are control (fate or personal responsibility),being compared. Holmes and Masuda positive versus negative events, and lev-(13), Dekker and Webb (22), and Uhlen- el of social organization involved (suchhuth et al. (23) found, for example, that as family, neighborhood, community).young adults aged 20 to 30 reported Considerable attention has been de-twice as many life changes as those over voted to the question of whether an60, and throughout the age range a signifi- event must be unfavorable to evokecant inverse relationship prevails. It is stress. In their original work, Holmesunclear, however, whether this finding is and Rahe scaled life events in terms ofdue to the character of the scale or to "the intensity and length of time neces-greater degrees of stress in early adult- sary to accommodate to a life event,hood. The former possibility is support- regardless of its desirability"; B. S. Doh-ed by data from the Midtown Manhattan renwend (27) also endorses this position,Community Survey of 1660 adults which which is supported by extensive clinicalshowed that stresses accumulated with work on normal life events such as en-advancing age (24).The "common" events represented on

life events checklists may be largely irrel-evant to certain groups, or else thosegroups experience far fewer changesthan are usually reported. For example,findings of very few life changes werereported by Wershow and Reinhart (16)in their study of 88 chronically ill, margi-nally employed men who were con-secutively admitted for medical reasonsto a southern Veterans Administrationhospital. In this study, the mean LCU

gagement and marriage (28). Gersten etal. (15), however, disagree; they regardundesirability rather than simply totalamount of change as the better definitionof stressor. On the basis of communitysurvey data about nearly 700 children,they have concluded that the number ofundesirable life events or a balancedscale (sum of undesirable events minussum of desirable events) is a better pre-dictor of behavioral impairment than isthe total number of changes.Another unresolved issue concerns

SCIENCE, VOL. 194

the scoring of life event checklists. Someinvestigators assume that there is onlyone population of events and so measurestress additively by counting number ofevents that have occurred in a specifiedtime interval. Others believe that sub-categories and weights are preferable.The most common method, noted ear-lier, is to apply weights derived fromjudge samples that showed strong con-vergence of opinions regarding appropri-ate weighting for particular items. Morerecently some investigators have askedsubjects to rate events in terms of thesubjective distress these caused and toindicate the number of times each oc-curred within the period under study.These subjective ratings were then usedas weights in arriving at a total score.Rubin et al. (8) found that weights de-rived by stepwise multiple regressionanalysis of questionnaires provided bynaval personnel enhanced the correla-tions between total scores of the lifeevents measure and counts of sub-sequent illness episodes of their navalrespondents. Finally, the factorial struc-tures of the most commonly used check-lists have not been adequately explored;it would be useful to determine empirical-ly how many dimensions are included intheir scope and whether separate factorscores may be more useful than thesingle total score currently Imployed.

Confounding Variables

Another issue in life events researchthat warrants further attention is thepossibility of interaction between lifechanges and other factors, such as avail-ability of social support systems to serveas protective buffers for the affected indi-vidual. As defined by Caplan (29), socialsupport systems consist of enduring in-terpersonal ties to a group of people whocan be relied upon to provide emotionalsustenance, assistance, and resources intimes of need, who provide feedback,and who share standards and values.Ideally, one belongs to several support-ive groups situated at home, at work, inchurch, and in a series of recreational oravocational sites. Cassel (30) has ob-served that deficiencies in support sys-tems will not in themselves contribute tosusceptibility to illness in the absence ofsocial stressors. The converse is alsoprobable: social stressors in the presenceof strong social support systems willhave only minor effects on health. Anexcellent illustration of the value of mea-suring the interaction of these sets ofvariables is provided by Nuckolls et al.(31), who studied life changes and social3 DECEMBER 1976

supports for women during pregnancy,in relation to complications of later preg-nancy and delivery. Neither the life-change score alone nor the social sup-port score alone was related to com-plications. When the two scores wereconsidered jointly, however, significantfindings emerged: 90 percent of the wom-en with high life change scores but lowsocial support scores had one or morecomplications, whereas only 33 percentof women with equally high life changescores but with high social supportscores had any complications. The socialsupport scores were irrelevant in the ab-sence of high life change scores. Theseresults clearly document the need formore analytical approaches.

Questions have been raised about thecomposition of samples in many lifeevents studies. Wershow and Reinhart(16) refer to the common failure to "dis-aggregate groups," as in the study wherepatients attending a dermatology clinicand those with coronary heart diseaseare together classified as suffering fromchronic disorders, or where protocols ofrespondents with vastly different back-grounds and life styles are combined foranalysis. In the earlier naval studies ofHolmes and Rahe, for example, the onlydistinction made between respondentswas based on the ship to which theywere assigned. Draftees, career militarymen, officers and enlisted men, new-comers and old-timers, and those withhazardous and not hazardous jobs wereall grouped together. (In recent publica-tions some of these distinctions havebeen taken into consideration.)Another design issue cpncerns the ad-

visability of controlling the variables ofsocioeconomic status and ethnicity insample selection and data analysis. Pre-liminary work by Holmes and Rahe (4)suggests that respondents grouped bysocial class or color rank life events simi-larly in terms of their perceived impactor magnitude. More direct evidence hasbeen compiled by Dohrenwend and Doh-renwend (32), who addressed the issue ofa possible relationship between class,ethnicity, and differential experience oflife events. After reviewing a wide vari-ety of published studies on class andethnic differences, they concluded thatboth class and ethnicity influence expo-sure to stressful events. They found thatlower-class members experience moresevere though not more frequent stress-ful events than do middle-class mem-bers. Within social class, stressful situa-tions are both more frequent and moresevere for blacks than for whites. Thusfar such relationships have been onlytentatively explored.

It must be noted that the extensivecritical appraisals of life events studiesare possible primarily because ofthe rela-tively large and coherent body of re-search that has been published. The factthat different groups of investigatorshave produced coordinated and cumula-tive research programs over many yearsprovides critics with an adequate pictureof how far work has progressed, whatare current deficiencies and weaknesses,and what remains to be done. The field oflife events research, like that of psycho-therapy research, seems to evoke almostas much critical commentary as empiri-cal data. In life events research, how-ever, communication channels are evi-dently effective. The quality of recentwork surpasses that of earlier studies,and many suggestions have been incorpo-rated into research programs. Increasingnumbers of studies are prospective indesign, and the researchers concernthemselves with sample selection, seekappropriate and relevant items in theirchecklists, try to refine their outcomecriteria, and use multivariate statisticalmethods in data analyses. With theseimprovements in design and methods,investigators may be able to demonstratemore accurately the nature of the rela-tionship between life events and sub-sequent illness episodes than has beendone to date.We would conclude that the life events

approach to the measurement of stressand subsequent illness offers a methodthat is attractive in its simplicity, direct-ness, ease of data collection, and com-mon sense appeal or "face validity."Much work remains to be done in apsychometric sense as well as concep-tually, to improve the reliability and va-lidity of the measuring instruments, todevelop stratified domains of life events,and to select for investigation only thoseevents that are relevant to the topic andpopulation being studied. It might beprofitable to study conditions underwhich the probability of illness is en-hanced by the occurrence of prior lifechanges in contrast to those where suchchanges have little impact. Comparisonof groups who handle life changes effec-tively with those who appear to breakdown with little apparent provocationmay also further our understanding ofthe possible role of life changes in precip-itating illness. Another helpful approachmight be the extension of the dependentvariables examined after the occurrenceof life changes.

In short, instead of trying repeatedlyto answer the question whether lifeevents play a precipitating role in illness,the next step in the progressive devel-

1017

opment of this field entails examination trof the circumstances under which such c(effects occur and do not occur. w

01

Mediating Factors atis

Some people develop chronic diseases sLand psychiatric disorders after exposure dito stressful conditions, and others do e:not. Indeed, most people do not become Tdisabled even when terrible things hap- ctpen to them, as Hudgens (19) has ob- alserved. Exposure to stressors alone is fralmost never a sufficient explanation for sIthe onset of illness in ordinary human stexperience, and other factors that influ- ibence their impact require consideration. VThese may be grouped in three broadcategories: characteristics of the stress-ful situation, individual biological andpsychological attributes, and character-istics of the social support systems avail-able to the individual that serve as buf-fers.

Before turning to a review of thesemediating factors, it is important toemphasize both their cumulative impactand the reciprocal relationship betweenthem. That is, the more rigorous andsevere the external situation, the lesssignificant are social and individual char-acteristics in determining the likelihoodand nature of response. When conditionsare sufficiently harsh, as in some war-time situations, prolonged sensory depri-vation, or concentration camps, break-down is virtually universal and individ-ual variations are reflected only in thelength of time before the reaction occursand perhaps in subsequent recoverytime. When the stressful situation is lesssevere, social supports and individualcharacteristics contribute to an under-standing of why some people become illand others do not. Finally, although itseems probable that extreme environ-mental conditions can induce disabilityeven in those who do not have social orpersonal deficits, vulnerability alone, inthe absence of stressful conditions, doesnot precipitate chronic disease or psychi-atric disorder.

Stressor Characteristics

Formal characteristics of stressfulevents that have been found to influenceillness onset include their magnitude (de-parture from baseline conditions), in-tensity (rate of change), duration, unpre-dictability, and novelty. The most widelystudied of these is magnitude, which hasbeen investigated among survivors of ex-

1018

lae)

h(v3

vz

vi

Iin

mofthPCfrthse

mat

eeme experiences such as internment inoncentration camps or as prisoners oftar. A linear correspondence has been

competence individuals have demon-strated in the past, the more likely it isthat they will cope adaptively with a

bserved repeatedly between magnitude current stressor. The more experiencef the stressor and extent of both psychi- they have had previously with a particu-tric and physical disability (33, 34). It lar stressor, the more probable that theirinow widely agreed that stressors of present responses will be effective (37).ifficient intensity and duration will in- The correspondence of personalityuce an acute stress reaction in all so type to stress reactions and to vulnerabil-xposed, regardless of predisposition. ity to disease is less clear-cut. As noted inhere has been less consensus con- the introductory section, the subject haserning long-term or permanent dis- been of major interest among those con-bilities, but recent longitudinal data cerned with the psychosomatic ap-.om concentration camp survivors have proach. Over the years, investigatorshown that profound and protracted have proposed several models to ac-tressful conditions may have irrevers- count for the impact of intrapsychic fac-le effects on all (35). / tors on bodily function, such as Adler'sSpeed of change, prolonged exposure, concept of organ inferiority (39), Alexan-tck of preparedness, and lack of prior der's idea that specific emotional con-xperience have each been found to flicts are determinants of disorderedeight the impact of stressful events function in a particular organ (40), and6-, J3. Cumulatively these findings sug- Dunbar's that personality constellationszst that the formal properties of stres- are associated with specific psycho-:rs constitute a significant source of somatic disorders (41). With the passageariation affecting their influence on indi- of time and accumulation of experience,iduals. these approaches to the understanding of

personality and illness have become lesspopular. Investigators who have contin-

idividual Characteristics as ued to work within this tradition havelediating Factors turned their attention to the delineation

of broad life styles and behavior patternsA critical factor in evaluating the im- rather than specific intrapsychic con-act of stressful events is the individual's stellations and conflicts. A major focuserception of them. Such perception de- within this framework has been on per-ends on personal characteristics deter- sonal correlates of premature coronarylining the appraisal of the significance heart disease, myocardial infarction, andf potentially harmful, challenging, or sudden death. Studies of the behavior ofireatening events. It is this cognitive individuals prone to coronary diseaserocess which differentiates a stressor have identified distinctive behavioral andom a stimulus and which determines characterological styles which may servee nature of the stress reaction and sub- as predisposing factors (42). The exten-quent coping activities (38). sive research on clustering of life eventsThe perceStion of stressful events is in association with myocardial infarctioniediated by two broad categories of vari- and sudden cardiac death does not con-:les, one consisting of personal or "in- tradict these findings, since such life

ternal" factors and the other of inter-personal or external ones, following the

*Dohrenwends' (32) conceptualization.Personal factors include, for example,biological and psychological thresholdsensitivities, intelligence, verbal skills,morale, personality type, psychologicaldefenses, past experience, and a sense ofmastery over one's fate (7, 32, 34). De-mographic characteristics such as age,education, income, and occupation mayalso contribute to the individual's evalua-tion of stressful conditions and his re-sponse to them (23).The effects of most personal variables

in mediating stressful conditions are fair-ly obvious: persons with more skills,assets, and resources and with more ver-satile defenses and broader experiencetend to fare better. In general, the more

events apparently serve as triggering orprecipitating elements influencing thetiming rather than the risk of illness on-set.

External Mediating Variables

Another broad set of contingencies, ormediating varia6les, in the stress equa-tion which may be considered social ortransactional in nature consists of thebuffers and supports accessible to theindividual in his social environment. Thesocial positions individuals or groups oc-cupy in a community can materially influ-ence their experience of stress and pre-sumably, therefore, their vulnerability toa broad range of chronic diseases. Whilethe effects of exposure to stressful

SCIENCE, VOL. 194

events may be reduced for those who are \ individual occupies two or more distincteffectively embedded in social networksor support systems (29, 43), they arecommonly exacerbated by deficienciesor impairments of such systems. Threesuch categories-social isolation, socialmarginality (minority membership), andstatus inconsistency-may be consid-ered in this context.Urban sociologists recognized years

ago that deteriorating areas of the centralcity had disproportionately high rates ofdisorders, both medical and psychiatric(44). More recently, social isolation hasbeen delineated as a major factor in in-creased risk of disease.LThere is nowconsiderable evidence to suggest thatthose who live alone and are not in-volved with people or organizations havefor this very reason a heightened vulnera-bility to a variety of chronic diseases(45-47^ A generalized "failure tothrive among institutionalized childrenis Qften associated with a lack of mean-ingful relationships to other people (48).Also in this context, it has been observedthat bereavements are a potential sourceof ill health, apparently in relation to thesocial isolation created by the loss of aspouse (49).While social isolation is perhaps the

most extreme example of impairment ofone's position in the community, margin-al social status due to membership in alow-status group or simply in one thatconstitutes a numerical minority in thearea has also been associated with in-creased health risks (47). Sheer numeri-cal size of a given group, sometimesreferred to as ethnic density, has beenfound to be inversely related to psychiat-ric hospitalization rates: as a given eth-Vnic group constitutes a smaller propor-tion of the total population in a particulararea, diagnosed rates of mental illnessincrease in comparison both to the ratesfor other ethnic groups in that area andto the rates of the same ethnic group inneighborhoods where its members con-stitute a significant proportion or major-ity. This observation has been made withrespect to Chinese in Canada (50), Frenchand English minorities in neighboringQuebec towns (51), Italians in differentBoston areas (52), and black and whiteresidents of various census tracts in Bal-timore (53){Presumably the smaller thecommunity of ethnically similar mem-bers, the less the social support availableto any one member. Equivalent fiffdingshave been noted for socially marginalgroups with res2ct to such diseases astuberculosis (45)jThe third social variable, status incon-

sistency, refers to the situation where an3 DECEMBER 1976

social statuses or roles that involve in-compatible social expectations{CFor ex-ample, mother-married-adult are threecompatible statuses, in contrast to moth-er-unmarried-adolescen'tj Other formsof inconsistency may entail lack of fitbetween education and occupation, orbetween age or sex and employment.Studies exploring the stressful effects ofstatus incongruence have been con-ducted both with individuals in surveyformat and with populations for whichaggregate data are derived from publicrecords (2).

Studies of individuals have dealt withobserved discrepancies between educa-tion and income level, or education andoccupational rank, *which were pre-sumed to generate role conflict. While afew investigators have failed to find anassociation between status incongruenceand measures of health (54), several havefound it, using different kinds of samplesand measures of health (55).

Hinkle and his colleagues at the Hu-man Ecology Study Program at Cornell(56) analyzed the medical histories of2600 semiskilled workers who workedfor the New York City telephone compa-ny continuously for 20 years. Admit-tedly, this is an atypical group in terms ofgeographical and employment stabilityand consequent lack of exposure to so-cial change. The investigators found anenormous range in the number of illnessepisodes recorded, from fewer than 5days a year of absence due to illness toan average of 50 days a year for 20 years.The very healthy workers were found tobe people whose social backgrounds, as-pirations, and interests coincided withtheir present circumstances, whose fam-ily, educational, and occupational status-es were consistent. This was not the casefor the frequently ill workers, whose edu-cational or family status was often in-appropriately high for the kind of workthey were doing. Hinkle did not invokethe concept of status inconsistency in hisconceptual analysis, but his findings lendthemselves to an understanding in thiscontext.The literature concerning variables

that mediate the impact of stressfulevents on individuals derives from somany sources that a general critical ap-praisal would be unsuitable. Some of itrepresents conventional, well-executedlaboratory studies with clearly definedindependent and dependent variables.Other studies, such as those of concen-tration camp survivors, are necessarilyretrospective in design, based on sam-ples of convenience and ad hoc measures

of change. However, most of the find-ings so briefly summarized here havebeen reported by several investigatorsworking independently, with differentpopulations. The results are thereforecumulatively persuasive, and open a vari-ety of areas for future exploration.

Summary

Athough conceptual and theoreticalorientations should play an importantpreparatory role in the design and execu-tion of empirical studies, this does notoften appear to be the case in the litera-ture reviewed on the reiation of lifeevents, stress, and illness. It is clearlyrecognized that illness onset is the out-come of multiple characteristics of theindividual interacting with a number ofinterdependent factors in the individual'ssocial context in the presence of a dis-ease agent. The conceptual model is com-prehensive, multicausal, and interactive;empirical designs should consider thiscomplexity. In spite of the repeatedlyobserved trivial relationships betweenmeasures of change in life events andillness onset (or care-seeking behavior),many investigators continue to focus onlinear relationships between independentand dependent variables without consid-eration or control of intervening and me-diating variables, some of which easilylend themselves to standard measure-ment procedures. To advance the accu-rate prediction and understanding of ill-ness onset, the design and execution ofempirical studies must take into account,as Mechanic and others have stressed,the complexity of the phenomena beingstudied.

Crucial in the measurement processare the psychometric properties of themeasures used and the methods of col-lecting data that are employed. Investiga-tors in the area of life events research arevulnerable in their operational defini-tions of both independent and dependentvariables.More emphasis should be placed on a

thorough conceptualization and sam-pling of the universe of life events, fol-lowed by multidimensional scaling ofitem samples in a variety of respondentsamples drawn from theoretically mean-ingful populations to identify commondimensions of life events. The internalconsistency and test-retest reliability ofsummary scales derived from those anal-yses should be studied across samples todetermine the true variance and stabilityof these measures over a variety of popu-lations.

1019

The use of unidimensional scales withquestionable content validity continuesto be a problem in the operational defini-tion of such complex domains as report-ed symptoms of illness or mental illness.The continued use of one measure torepresent an obviously complex domainof symptoms will frequently lead to limit-ed and erroneous conclusions. An exten-sive literature also indicates that symp-toms of mental and physical illness arenot unidimensional.

In retrospective studies importantsources of error in the measurement oflife events include selective memory, de-nial of certain events, and overreportingto justify a current illness. In prospectivestudies, the subjective evaluation of thesignificance of a life event to the respon-dent has been neglected.The data analytic procedures used in

life events research do not adequatelyinform the reader of the nature of ob-tained results. Certain procedures cru-cial to the understanding of results sel-dom have been undertiken. For ex-ample, not one instance of an estimate ofthe internal consistency reliability of alife events scale was discovered in thisreview, though such values are impor-tant in the evaluation of measures and inthe interpretation of the magnitude ofrelationships. Further, the application ofsimilar data analytic procedures to thedata of a number of studies would en-hance the comparability and communica-tion of results and the possibility of rnak-ing generalizations. It is concluded thatimprovement in data analytic proceduresremains a major challenge for life eventsinvestigators.

Refinements of method and content inthis field are to be encouraged, in theexpectation that they will contribute to abetter understanding of the disease pro-cess and also to the development of tech-niques of primary prevention of illnessand rehabilitation of the chronically ill.

References1. H. Selye, The Stress ofLife (McGraw-Hill, New

York, 1956).2. D. Dodge and W. Martin, Social Stress and

Chronic Illness (Univ. of Notre Dame Press,Notre Dame, Ind., 1970).

3. R. Dubos, Man Adapting (Yale Univ. Press,New Haven, Conn., 1965).

4. T. Holmes and R. Rahe, J. Psychosom. Res. 11,213 (1967).

5. T. Holmes and M. Masuda, in Stressful LifeEvents, B. S. Dohrenwend and B. P. Dohren-wend, Eds. (Wiley, New York, 1974), p. 49.

6. R. Rahe, Ann. Clin. Res. 4, 250 (1972).7. , in Stressful Life Events, B. S. Dohren-

wehd and B. P. Dohrenwend, Eds. (Wiley, NewYork, 1974), p. 73.

8. R. Rubin, E. Gunderson, R. Arthur, J. Psycho-som. Res. 15, 89 (1971).

9. S. Bramweil, M. Masuda, N. Wagner, T. Holmes,J. Hum. Stress 1, 6 (1975); A. Antonovsky andR. Kats, J. Health Soc. Behav. 8, 15 (1967); R.Rahe and E. Lind, J. Psychosom. Res. 15, 19(1971); R. Rahe, ibid. 8, 35 (1964); M. Selzer andA. Vinokur, Am. J. Psychiatry 131, 903 (1974);T. Theorell and R. Rahe, J. Psychosom. Res.15, 25 (1971).

10. D. Dekker and J. Webb,J. Psychosom. Res. 18,125 (1974); E. Jaco, in Social Stress, S. Levineand N. Scotch, Eds, (Aldine, Chicago, 1970), p.210; J. Myers, J. Lihdenthal, M. Pepper, D.Ostrander,J. Health Soc. Behav. 13, 398 (1972);E. Paykel, in Stressful Life Events; B. S. Doh-renwend and B. P. Dohrenwend, Eds. (Wiley-New York, 1974), p. 135; B. Prusoff,J. Myers, Arch. Gen. Psychiatry 32, 327 (1975);E. Uhlenhuth and E. Paykel, ibid. 28, 473(1973).

11. R. Markush and R. Favetb, in Stressful LifeEvents, B. S. Dohrenwend and B. P. Dohren-wend, Eds. (Wiley, New York, 1974), p. 171.

12. A. Kagan and L. Levi, Soc. Sci. Med. 8, 225(1974).

13. W. Michaux, K. Gansereit, 0. McCabe, A. Kur-land, Comm. Ment. Healh J. 3, 358 (1967).

14. L. Bieliavskas and J. Webb, J. Psychosom. Res.18, 115 (1974).

15. J. Gersten, T. Langner, J. Eisenberg, L. Orzek,in Stressful Life Events, B. S. Dohrenwend andB. P. Dohrenwend, Eds. (Wiley, New York,1974), p. 159.

16. H. Wershow and G. Reinhart, J. Psychosom.Res. 18, 393 (1974).

17. I. Sarason, C. de Monchaux, T. Hunt, in Emo-Rons: Their Parameters and Measurement, L.Levi, Ed. (Raven, New York, 1975), p. 499.

18. G. Brown, in Stressful Life Events, B. S. Doh-renwend and B. P. Dohrenwend, Eds. (Wiley,New York, 1974), p. 217.

19. R. Hudgens, inibid., p. 119.20. R. J. Cadoret, G. Winokur, J. Dorzab, M. Bak-

er, Arch. Gen. Psychiatry 26, 133 (1972).21. D. Mechanic, in Stressful Life Events, B. S.

Dohrenwend and B. P. Dohrenwend, Eds.(Wiley, New York, 1974).

22. D. Dekker and J. Webb, J. Psychosom. Res. 18,125 (1974).

23. E. Uhlenhuth, R. Lipman, M. Balter, M. Stern,Arch. Gen. Psychiatry 31, 759 (1974).

24. T. Langner and S. Michael, Life Stress andMental Health (Collier-Macmi-an, London,1963).

25. B. P. Dohrenwend, in Stressful Life Events, B.S. Dohrenwend and B. P. Dohrenwend, Eds.(Wiley, New York, 1974), p. 275.

26. S. Kellam, in ibid., p. 207.27. B. S. Dohrenwend, J. Health Soc. Behav. 14,

167 (1973).28. R. Rapoport, Fam. Process 2, 68 (1963).29. J. Caplan, Support Systems and Community

Mental Health (Behavioral Publications, NewYork, 1974).

30. J. Cassel, inHandbook ofEvaluation Research,E. Struening and M. Guttentag, Eds. (Sage,Beverly Hills, Calif., 1975), p. 537.

31. C. Nuckolls, J. Cassel, B. Kaplan,Am. J. Epide-miofl. 95, 431 (1972).

32. B. P. Dohrewend and B. S. Dohrenwend, SocialStatus and Psychological Disorder (Wiley Inter-science, New York, 1969).

33. R. Arthur, in Life Stress and Illness, E. Gun-derson and R. Rahe, Eds. (Thomas, Springfield,111., 1974), p. 195; L. Eitinger, in Society, Stress

and Disease, L. Levi, Ed. (Oxford Univ.Press, London, 1971), vol. 1; F. Hocking, Am.J. Psychother. 24, 4 (1970).

34. S. Wolf and H. Goodell, Eds., Harold G.WoOs Stress and Disease (Thomas,Springfield, III., ed. 2, 1968).

35. 4. Horowitz, Stress Response Syndromes(Aronson, New York, 1976).

36. J. Cassel, in Social Stress, S. Levine and N.Scotch, Eds. (Aldine, Chicago, 1970), p. 189; J.Coleman, Am. J. Occup. Ther. 27,169(1973); R.Lauer, Soc. Forces S2, 510 (1974).

37. N. Miller, in Psychopathology ofHuman Adap-tation, G. Serban and A. Kling, Eds. (Plenum,New York, in press).

38. J. Groen, in Society, Stress and Disease, L.Levi, Ed. (Oxford Univ. Press, London, 1971),vol. 1, p. 91; M. Jacobs, A. Spilken, M. Nor-man,Psychosom. Med. 31, 31(1969).

39. H. Ansbacher and R. Ansbacher, Eds., TheIndividual Psychology of Aifred Adler (BasicBooks, New York, 1956).

40. F. Alexander, T. French, G. Pollack, Psycho-somatic Specificity, (Univ. of Chicago Press,Chicago, 1968), vol. 1.

41. F. Dunbar, Psychosomatic Diagnosis (Hoeber,New York, 1943).

42. M. Friedman and R. Rosenman, Type A Behav-ior and Your Heart (Knopf, New York, 1974);R. Rosenman et al., J. Am. Med. Assoc.233, 872 (1975); S. Wolf, in Society, Stressand Disease, L. Levi, Ed. (Oxford Univ. Press,London, 1971), vol. 1.

43. J. Cassel, Mt. Sinai J. Med. 40, 539 (1973).44. R. Faris and H. Dunham, Mental Disorders in

Urban Areas (Univ. of Chicago Press, Chicago,1939).

45. T. H. Holmes, in Personality, Stress and Tu-berculosis, P. J. Sparer, Ed. (International Uni-versities Press, New York, 1956), p. 65.

46. L. Levy and L. Rowitz, The Ecology ofMentalDisorder (Behavioral Publications, New York,1973); J. Cassel and A. Leighton, in MentalHealth Considerations in Public Health, S. E.Goldston, Ed, (Government Printing Office,Washington, D.C., 1969), p. 67.

47. A. Linsky; Soc. Psychiatry S, 166 (1970).48. R. Spitz, Psychoanal. Study Child 2, 113 (1946);

A. Schmale, S. Meyerowitz, D. Tinling, inMod-ern Trends in Psychosomatic Medicine, 0. W.Hill, Ed. (Appleton-Century-Crofts, New York,1970), p. 1.

49. P. Clayton,Arch. Gen. Psychiatry 30, 747 (1974);Ain. J. Psychiatry 132, 133 (1975); S. Kiritz andR. Moos, Psychosom. Med. 36, 96 (1974); C.Parkes, in Modern Trends in PsychosomaticMedicine, 0. W. Hill, Ed. (Appleton-Century-Crofts, New York, 1970), vol. 2, p. 71; C.Parks, B. Benjamin, R. Fitzgerald, Br. Med. J.1, 740 (1969).

50. H. B. M. Murphy, Milbank Mem. Fund Q. 39,385 (1961).

51. I. Rootman and D. Sydiaha,Psychiatr. Q. 43,131(1969).

52. N. Mintz and D. Schwartz, Int. J. Soc. Psychia-try 10 101 (1964).

53. G. Klee, E. Spiro, A. Balm, K. Gorwitz, inPsychiatric Epidemiology and Mental HealthPlanning, R. Monroe, G. Klee, E. Brody, Eds.(American Psychiatric Association, Washing-ton, D.C., 1967), p. 107.

54. R. Meile and P. Haese, J. Health Soc. Behav.10, 237 (1969).

55. J. Abramson, Milbank Mem. Fund Q. 44, 23(1966); E. F. Jackson, Am. Sociol. Rev. 27, 469(1962); S. King and S. Cobb, J. Chronic Dis. 7,466 (1958); S. Cobb and S. Kasl, Am. J. PublicHealth 56, 167 (1966); S. King and S. Cobb, J.Chronic Dis. 7, 466 (1958).

56. L. E. Hinkle and H. G. Wolff, in Explorations inSocial Psychiatry, A. Leighton, J. Clausen, R.Wilson, Eds. (Basic Books, New York, 1957), p.105; L. Hinkle, in Stressful Life Events, B. S.Dohrewend and B. P. Dohrenwend, Eds.(Wiley, New York, 1974), p. 9.

SCIENCE, VOL. 1941020