chologists print out, “ness is intrinsic to the fife cycle. Every culture hhas had its institutes, its expert and lay roles, and tsi uals forthe entry to and exit from life and for the man. tnent of ines episodes” (p. 247). Sacal psychology is the disiptine that explicates the impact of institutions, ees ‘nd rituals on the commencement and termination of exis fence, and, more typically on daily health-rolevan cogni- tion and behavior within a culture, and comparatively eros cultures. Healtny People 2000 (U.S, Department of Health and Human Services, 199T) notes that health promotion and disease prevention objectives "depend heavily on chanzes in human behavior” (p. 8). Its change in behav the discovery of new antigens or anibodies—that is likely to be the most efficient way of reducing disease morbisty and premature mortality. Ifthe entre population quit re would be a 25 percent reduction in cancer deaths and 350,000 fewer fatal heart attacks each year A 10 percent weightloss through dietary modification and physical exercise would produce a 20 percent decrease in Coronary artery disease as well as lower incidences of stroke, some cancer, diabetes, and other kinds of heat is: ease in the population (Taylor, 1990; see also Matarazzo 1984). Moreover, we could expect dramatic reductions in disease morbidity i individuals engaged reliably in preven tion behaviors suchas regular physical exercise and sereen- ing behaviors such as mammography and Pap testing. Pro- toting such behaviors and preventing relapse once the behaviors are modified are outcomes that social psychol- ogy shouldbe especially well equipped to addres. “The response of socal psychology tothe challenge of ‘investigating social behavior relevant to health and illness has involved empirical research guided by a view of people as active processors of information eapeble of self-eilec- tion and self-regulation rater than as responding passively to environmental contingencies (Bandura, 1986: Maddux 1993). Social psychological work on health behavior re- Aects three different strategies: (1) the application of for- toal theory tothe health domain,(Gthe development of rounded theory generated specifically fr te health field, and (2) problem-focused research addressing specific so- ial Sad personality processes relevant to health behsviors (S. E. Taylor, 1984). We have organized this chapter ‘round the formal and grounded theoretical perspectives that have guided research inthis Feld, rather than by spe- cifie diseases (e.g. AIDS, cancer) or types of behavior (ce. screening, primary prevention) By formal theory. we ‘mean the investigation of health behavior from the guiding we ofa broader theory that was not developed to explain health behavior per se. We include in tis category ‘theories linking atitudes and behavior, Social Cognitive Theory. Attribution Theory, and Social Comparison The- cory. Grounded theories are conceptual frameworks that smoking, th ary Persp ional onepnng of evra Rha 30 and disease pathophysiology, or the buffering of ae Stet fate ele aloueh hse psn heh arly ace chi teas oo py cingy trope osmosis it hv en we ewe chenee ep Aer thew. >: Ane soe Amal 1995 Coen, 198; Helgaon& coe 36: ane, 90, Repel & Se 159; Gt, Carpe Risk Gas 95) FORMAL SOCIAL PSYCHOLOGICAL THEORY AND HEALTH ‘We begin with theories from social psychology, broadl speaking, that have been applied to understanding healil behavior. This section covers (1) theories linking anituded and behavior such as Subjective Expected Utility Theofj the Theory of Reasoned Action, and the Theory of Planned Behavior; (2) Attribution Theory; (3) Social Cognitive ‘Theory; and (4) Social Comparison Theory. These theoreti cal traditions were not developed originally to expl health behavior, but they have become prominent a proaches to understanding decisions to engage in healt protective and illness-reducing activities. ‘Theories Linking Attitudes and Behavior “The assumption that people make behavioral decisions the bass oftheir beliefs sa featre of many ay and sci tf theories about health promotion. Analyses of the tions between health beliefs and health behavior have £9 cused on denying (1) the kinds of beliefs that should measured, (2) the rules by which beliefs are integrated, combined in inking them to bebav tions under which we should expect strong or weak havior ofexpe ing the behavie Alt more ai conditi tity ¢ have so 1985), control of SEU distingw ‘ot pere ‘o be atWalisto nstein, 1993). In contrast to frame works such as the Health Belief Model that specify the pre dictive value of particular health beliefs (c.g. perceived vulnerability, baiviers to action), the TRA provides the mework in which the relevance pends on the issue and the situation. The theory focuses on the operationalization and measurement of variables, pro- viding specific combinatorial rules that are to be used when analyzing the relationship between health attitudes and behavior, ‘of specific heliets de- Operationalization and Measurement Ajzen and Fish- bein (1980) have provided specific instructions revarding the operationalization of each of the theory's dimensions. ‘The basic assumption underlying the TRA is that behavior dicecily reflects an individual's intention to take action. Be havioral intention is in turn a weighted linear function of an individual's atiude toward performing the behavior and the perceived subjective norms concerning whether or not to take action. The relative influence of subjective norms and atitudes is thought to vary across situations and issues (but see Evans, 1991, regarding difficulties in mak ing this comparison). Attitudes toward the action are in tum the product of salient beliefs about the consequences of the action weighted by the evaluation of those conse quences. Perceived norms are based on the degree to witich relevant individuals and groups support taking action, ‘weighted by the actor's motivation to comply withthe ex pectations of others. Any additional factors are understood to influence behavior indirectly through their effect on be= liefs about the behavior of beliefs about social norms. How researchers have operationalizedjnientiom has proved important in identifying the relation berween atti tudes and behavior (Maddux & DuCharme, in press). War- shaw and Davis (1985) have distinguished between behav- ioral_inte ‘and behavioral expectation. Infention ibdletat plan fo pecforva Bebuvior whet expecaion reflects the perceived likelihood of performing the behav= ior in the future. People have been shown to sely on their beliefs about the behavior to formulate a behavioral nen tion, whereas their thoughts about prior behavior and = presence or absence of situational constraints determine behavioral expectations (Gordon, 1990). Compared to in- tention, expectation is thought to be a better predictor of behavior. Recently, Gibbons and Gerrard (1995; Gibbons, Gerrard, & Boney-MeCoy, 1995) have noted that although people may not intend to perform an unhealthy oF risky be- havior (eg. driving while intoxicated, they are willing to do so under certain conditions, irrespective oftheir inten- tions to behave appropriately. Their willingness is a reac~ tion to features of the immediate situation rather then the implementation of deliberate planning. Thus, measuring, the degree to which, under certain conditions, people are willing to perform a behavior may be helpful in predicting’ certain types of behaviors. bs In certain situations, assessing the beliefs that undettig bbchaviors ean become quite complicated. To the extent that: the behavior in question is thought to produce an outcome’ (eg. reducing one's risk of developing skin cancer), then perceptions of that outcome need to be either measured ex. plicitly in terms of change (i.e. if you use sunscreen, your risk of developing skin cancer will decrease), or measured: separately for the current behavior (i... the tisk of devel ‘oping skin cancer without using sunscreen) and forthe tar geted behavior (i., the risk of developing skin cancer ‘when using sunscreen). Intentions to adopt the precaution. ary behavior should be a function of the perceived advan- tage of the precautionary behavior compared to the current behavior (Weinstein, 1993). Correspondence berween Attitudes and Behaviors The! observed relationship between attitudes and behavior isa funetion ofboth the match in sp ‘and_behaiar andthe stability of and intentions. Ajzen'and Fishbein (197) discuss fou tensions of correspondence between atitudes and be ior: (action (e.., using contraceptives): )tarset (4 condoms); G) time (e.g., using condoms over the next} twelve montis); and @jcontext (e.g, using condoms di ‘ng sexual intercourse with a primary partnet) Greater co respondence between attitude and behavior allows fora more accurate prediction of behavior (sce Kraus, 1995, ‘a meta-analysis), The observation that predict: tentions better than intentions predict behavior may ref Sette ase ore a ln stow what greater correspondence. Unfortunately, litle tent s directed atthe correspondence between the measured liefs and the assessed behavior (but see Ajzen & Tim 1ORG, Fishbein, Mialestade& Hicheock, 1994 Lis ee Brena Geer dnd a fee fist) The recent asf cond lized questions suchas those advocated by Weinstein an ‘Nicolich (1983) and Ronis (1992) should work to ale some ofthese problems, For example, measuring percep tions of risk that are conditonslized on a set pater af havior (¢g., your risk of developing skin cancer in the net five years if you do not use sunsereen routinely) sho provide a greater correspondence between the belief the subsequently assessed behavior Although the{context}in which a behavior is performed ‘can shape the salience and relevance of personal bei context is often left unspecified. Stasson and Fisbbel (1990) have illustrated how perceiving a diving scenario pert betw fect shan bbeha intent Be may a intent ciliate Davide ‘mation fect we edge al by inac duce th Weak r reflect ¢ sion, th salue of of takin nears, Finba home can provide benetits to one’s family, Raats, Shep. herd, and Sparks (1995) assessed the degree to which peo- ple felt obligated to care for the health of thee family. Per- ceived obligation signiticantly predicted intentions to use ower-fat milk after coatrolling for the influence of att tudes, subjective norms, and perceived eonttol. This type ‘of approach may effectively help predict any behavior that influences the health status of othets (e.g. using a condos (o protect a partner's health, taking medication or being ‘quarantined to prevent the spread of tuberculosis) Research on health behavior focuses primarily on a sin ie behavioral response to the presence of health-relevant ‘information. Yet, people frequently can respond to a health concer in a variety of ways. Studies that assess a wider range of behavioral alternatives may provide a more ace fe picture of how people are responding to & health issue Research on safer sex behavior is one area in whieh inves tigators are examining a range of behavioral responses. For example, Terry, Galligan, and Conway (1993) assessed three types of safer sex practices (avoiding casual sex being in an exclusive relationship, and asking partners about their sexual and IV drug use history). Altematvely. Cochran et al. (1992) differentiated between safer sex be- hhaviors that focused on partners (eg., reducing the number of sexual pariners) and those that focused specifically on exposure to HIV (e.g. insisting on using condoms). In i ther case, measuring a range of behavioral responses en- ables investigators to examine the relative ability of ati- tudes and norms to predict the behavioral outcome. Furthermore, Raats, Shepherd, and Sparks (1995) found that intentions were more accurately predicted when a be havioral alternative was compared with two other alterna tives than when each intention was analyzed individually ‘This mode of analysis is further supported by the finding that a within-subject analysis across behavioral alternatives provides a more accurate prediction of behavior than a tra- ditional between-subjects analysis-of a single behavioral alternative (Davidson & Morrison, 1983; see also Pagel & Davidson, 1984), Final Thoughts The Theory of Reasoned Action and the Theory of Planned Behavior describe how to assess opti mally the relationship between people's atitudes and their behaviors. Specifically, behavioral intentions are believed to reflect a combination of attitudes toward the act, subjec- tive norms, and perceived behavioral control, However, the additive nature of these theories renders them unable to predict a prior the relative influence of each of these di- ‘mensions on intentions (Bagozzi, 1992) In addition, these theories do not address the factors that facilitate the i ‘mentation of one’s atitudes nor do they address why peo- ple take action, other than to assume that they are moti- vated to be consistent with their beliefs (Sutton, 1997). By ‘way of comparison§Protection Motivation Theory (Rogers, © mote gh vated 10 fake action in order to protect themselves, age Kendzierski (1990) has advocated for distinguishing be. 4p. 9 ‘sven decision misking (ie. the transition from beliefs to_ intentions and decision implementation (ie. the Wansition 1983), co be described later, proposed that people from inention t behaviovy, suggesting Wat expectancy yp, value theories such as the TRA and the TPB are beter a: suited for explaining intentions than behavior (cf. Kuhl, [O86)- Discontinuties between intentions and behavior may reflect the limitations of verbal reports of mental events and the conscious eontrol of one's behavior (Nisbett & Wikon, 1977), and the difficulty of formulating opera- |tionaldtinitons for concepts like volition and intention | Baum & Heath 1992), Attribution Theory ‘The study of auibutional processes dates tothe thinking of Heider (1958), who argued that the social perceiver is much like a scientist who devotes considerable effort to discovering the causes of various outcomes in the World. ‘What we now know as atribution “theory” encompasses a series of rules that individuals use to draw logical cause- aand-ffect inferences about themselves and their social en- vironment (Jones & Davis. 1965; Kelley, 1967). These rules focus On the consistency of antecedents and out comes, their distinctiveness, and consensus information about the inferences drawn by others (see Gilbert, 1998, in this Handbook). Causal attributions are especially likely to be stimulated by negative, unexpected, or uncertain oute comes and, as a result, seem particularly relevant to health and illness cognition. Attribution Theory focuses attention inthe health ares on beliefs about causal with Michela & Wood, 1986). We describe atbutiona processes in two contexts:())attributions about major life traumas sich as eis ines and Q) the ole of ati ional processes in health promotion, We do not review the, literature on attributions (and misattributions) about thei causes of arousal stafes and physical sensations, since ‘much ofthis literature was covered in the third edition of @f the Handbook (Rodin, 1985), and recent work i this tai tion has not focused especially on physical health (c Schwarz, 1990) ee ‘control over aversive events, they are more likely to adjt to them successfully (Thompson, 1981). When individuals; ee ee’ acs crocs a a ne ae selves losses search & Wor scribec ment ( studies Victim studies quence 1987); Hildebr flicting « fined a: toone’s Janoft-|ever, the latter group was superior in maintaining absti nonce over time. The superior maintenance of the self lation group could be traced directly to a lower incidence ‘ol external attributions about their success in quitting ventions designed to encourage Similar results for in ‘women over forty years old to obtain regular screening ‘naninograms have also been reported (Rothman, Saloses Tues et al, 1993). Rothman and his colleagues manip tuted ateibutions regarding the reasons to have a mani” in through a videotape intervention that emphasized thor a woman's responsibility or the esponsibility of th medical system for protecting her health, Twelve months alge the intervention, women who had viewed the inter nally oriented video were significantly more likely to have cbtsined a sereening mammogram than were women inthe extcinal condition or in an “information only” contol sroup, Internal attibutions for behavior change have also ben associated with the adoption of fluoride mouth rins- ing (Lund & Kegeles, 1984) and participation in seree for hypertension (King, 1982), Although internally o: ented atrbutions effectively promote screening behaviors. they may also stimulate inappropriate sl-blame (and con: siderable guilt) when aa illness is detected. These poten- tally unwanted consequences of motivating individuals ro Lake responsibility for their health eare need to be evalu- ated, especialy in light of the sowing emphasis onthe pa- tients responsibilty for maintaining his or her wellness Many of these findings are consistent withthe frame work proposed by Brickman et a. (1982), who suggested that for an intervention to be successful, the causal atribu- tions promoted by the program should correspond to in Viduals' causal aributions about the problem and about is alleviation. Furthermore, Brickman etal. emphasized that in addition to matching the atributions of individuals and those inspired by the intervention itself, getting people 10 accept responsibly for thei el beso anes re Produce Tong-tenm change. Social Cognitive Theory: Self-efficacy “The chapter in the thied edition of the Handbook that cov- ered applied social psychology (Rodin, 1985) placed great emphasis on the importance of beliefS about predictabilcy and competence and, in particular, control. Beliefs in per- sonal control over one's behavior and over outcomes are generally adaptive in the health domain, as long as they do rot produce distress in the face of extreme uncontrollable circumstances (Taylor et al., 1991). These kinds of beliets play a particularly prominest role in Social Cognitive Ths ory (Bandura, 1986), the name given to the revised version of Social Leaming Theory (Bandura, 1977b). Social Coz: nitive Theory (Bandura, 1986) argues that individuals are driven neither by inner forces nor automatically by envi- ronmental contingencies. Human functioning and motiva- soni est thought of asthe tipo in uence of con fon. he quizypmnla god botuvionceath affecting hs Ae Thao ves that ae epscally porn nis syn are ality ows symbols and an through ncrivon oetough, self-egulaion, ad sl-refe om Cena among these cognive processes i peeve ietfieacy (andra, 197, 1983, 1997), Sella ‘tinal at “people's odpmests ofthe capa to pani nd cnet Gone of econ eied tain dey: ante types of perfomance, Wis concerned nt wih te icone has but with pment of what oe ean do with ‘shaver skills one possesses" (Bandura, 1986, p. 331), Functional selreteacy ithe bli ht ne an ary tar requted action. I is operationalized typically i tunmtof one's self-confidence tn being able to aecomplh designed behavior Selatcay i construct from Social Cognitive ‘tery that hasbeen particularly ental to predicting heath betavor: Indies do wo nate Debris ati thy do oo belee try ca pefor, cvs, sustaling behavior nthe face of bles and dificulies i asoct 2 Med wih privet selcefieny, However sltelieaey is og tor state Beliefs about one's caebilies are often ae feted diccly by health campaighs inthe medi and by tppor groups for poplecopag with various nese. tn terete designs o alec expectations of sel-cacy monsvat thatthe change in these expectations may be more important than hei Orginal baseline levee im ot 4 {ating new health behaviors (Matbach, Flos, & Nas 1991) and raining them once they arin place (Cay & Carey, 193) i Selfeffeacy is concen fence in one's bly to eat low-fat foods snot neceiay 4 related to confidence that one ean avid smoking cigt tees, for instance Like most associations between cos tion and acon selfefcacy predicts behavior best when i is mentored inthe sume domain and at he same level of. batation asthe behavior tell (Hier, Salis, Hov- J CHL 1990) Ie wantfo Know whether someone i key to ‘ara seat bet when siding inthe Back sea of ca, the Telfeticacy expecaion that is most highly related tot concerns one's confidence that one could buckle up wh “ge ek of ie ny i seleatcaey ie eased is both the eit of view (it can produce empirical segs ‘more impressive # ‘than the usual weak associations | stract attitudes and specifi behaviors) and is weakness obvious @ tha speci belts would be related to specific behavior) 4 Some investigators have atempred to broaden the tl es Consuct into a generalized expectation about | an near cd =_—_ overall perceived competence (Smith, Dobbins, & Wall: ian TOBT), but the empiiea erature supporting hse Nremains limited. Approaches that measure self- confidence that one can behave appropriately sufficient to provide a sense of control? Or must control also include Lf-efficacy seems to matter most beliefs about outcomes? § in situations in which individuals believe Tey fave Consid s that control moderates the influence of sll sacy—é though some investigators have argued that Selfeficacy and perceived contro ae essentially the same construct (Ajzen & Madden, 1986). ~~ Consider the distinction provided by Skinner (1992) in her model of perceived contol. She differentiates bevween beliefs about the causes of suecess and failure, which she calle Goi Pelz beliefs about one’s abilities 10 enact those strategies, which she calls €4paeity belies, ‘would seem difficult to discriminate these concepts from expectations of outcome efficacy and self-efficacy, respec- tively. However, Skinner (1992) argues forthe value of considering a third construct, which she termscontrol be- liefs,)generalized expectancies about the extent to which ‘OWE éan produce the desired outcomes irrespective of the ‘means involved. Control beliefs differ from self-efficacy expectations in that they do not refer to specific actions. They are very much like intemal aueibutions forthe causes of outcomes. In this sense, Skinner links Social Cognitive Theory with Ausibution Theory an integration that may prove helpful. Whether control beliefs increase the vari ance that can be accounted for in health behaviors and ill ness endpoints by expectations of outcome efficacy and self-efficacy remains to be demonstrated, although atleast as operationalized by health locus-of-control scales. the redictive validity of control beliefs has been relatively smited (Walston, 1992). The relationship between percep- tions of contol and beliefs about self-efficacy and outcome ‘fcacy remains a conceptual and empirical challenge in this area, and Rodin’s (1985) conclusions in the previous edition ofthe Handbook appear sil to hold Most important, in operationalizing the construct of trol in the real world, itis necessary to distinguish whether people believe that they have the opportunity (0, exercise control in a given situation and whether they perceive that they have the skills necessary to exercise the control effectively. On the one hand the opportunity ‘may exist to exercise control in a given situation, but the person may feel that she or he does not have the abil Bence COD) ‘covering that oe Sees We = age ides paterne tee td Ammomecnan > take advantage of that opportunity. On the other hand person may fee! that she or he has the skills to exer ise control in a given situation, but the situation does ‘ot allow for anyone to exercise control. Thus feelings. of self-efficacy may be one of the factors that determine — ‘whether or not greater conteo is desirable, (p. 851) Fi ‘ Comparten Theory] cgay, Festinger (1954 propoed Socal Compan Tey vay to unde a papreneemencsras ay a eget Hed ar ep asl Bre Bhs pclae, tnd Jb cir inforatn abut el aie Bea. Inthe dhanco of obj enudnds(utich poole peer when, thy ar seal) Fetnacr postulated tat poole simply compare hersces oor and tha sir hrs re the most dentable sure of sol comparison informs. tion Scho (199) fen proposed that desi fo aia Sneth oa angus pees tne ey 8 gee inermato isms experiment showing that ni sat cain a ane Ls cea Fac ie rtd bat ic of eh oncaig epoca Tie of veda clas Santina ol corcoee Todal Coste Tong bas ean Yoian itcatons and Gabor avr the year, ts emphasis shifting to anderandng he selon ander whch People ARIE con corona a eet MI focdoeck. cod BOOED ars stil flee lfbreton (for rviws see Suan. 1994 Colins, 1996. Taylor & Lobel, 198; Wood, 1989; nd tated cellecions by Buon & Gibbons, 1057 and Sul Wilk, 1991) Inthe heh ae, Social Comparison Theory Ts sa eased css oop ale ‘aos aca ice oe ox WOE C T AN eli Was To appa Reah Informacion, nd ube For tinger’s (1954) exclusive concen with’ comp: isons to people performing slightly better than oneself} Considerable atendon has beet devoted compara wth wore off oer, so-called dma corparoad (Wills 1961, especialy in easy cums Al though Festinger orga formula of Seal Com ee Sato of any intel wort soma such ay Beal Silvas ka lel i clizana oliiog=tin tore ‘olesae econ Taylor Buon & Aspnwal 1950) Coping with Illness Social comparison information cal} serve Various functions for people coping with an illness For example, they can be reassured by learning that situation could be worse, or they can be inspired on dis situation could improve. The frequet “_ ‘in tates tine tt from ants ual me preter rare ova Dow aatherng bons 6 de hey Second Taylor is "epi se (Ginbows fnaings in ward and ‘hich es Yooma Does dow oil adjust exclusive a concisons Bl of niney ‘hoatheir positive ive results for this enzyme, given that they know ¥ out the test or the disorder for sa supposed mark In one set of studies, the proportion of acquaintances, same health problem is manipulated (Jem- mou, Ditto, & Croyle, 1986). In the enzyme paradigm, subjects are told either that one in five individuals tested four in five tested positively. Subjecis in the later condition rated the severity of the health threat &s lower, and in a separate study, subjects who received this kind of prevalence information were also less likely to re aquest additional information of a follow-up eamination (Ditto & Jemmott, 1989). Subjects similarly minimized the health threat when a confederate did likewise. and they were less likely to indicate that they would change their health habits (Croyle & Hunt, 1991). Recently, these find hhave been replicated within the context of actual hhealth concerns, such as high cholesterol (Croyle, Sun, & Louie, 1993) and gum disease (McCaul, Thiesse-Dulfy, & Wilson, 1992). All of these studies demonstrate that sub- {jects use other people—either statistical information about ‘thers or the modeling of another—to determine whether the information received is threatening (as we will discuss in the section on optimism, however, such comparisons with others can also lead to biased judgments about one’s health status; e.g., Klein & Kunda, 1993; Klein & Wein- stein, 1997). Even when the comparison “sample” is non- representative, individuals seem willing to use it to make judgments about the severity of their “disorder” In fact, people who have had a history of a health problem such as ‘migraine headaches—compared to those without any expe rience—provided higher estimates of the problem's preva- lence and lower estimates of its severity (Jemmott, Croyle, & Ditto, 1988, Experiment 1; see also Suls, Wan, & Sanders, 1988). In short, people make use of a heuristic suggesting that unusual conditions are considered more se~ rious (Croyle, 1992). There are other means by which comparison informa- tion regularly shapes the meanings that people atribute to ‘both their own and other people's actions, Research on sur- vey methodology (see Schwarz, Groves, & Schuman, 1998, in this Handbook) has revealed that the response al- temnatives embedded in questions—such as those used to assess physical and mental health—can infiuence how these questions ae interpreted. When told that a patient ex- perienced a symptom once a month, physicians atributed ‘greater severity to the symptom when “once a month” was framed as a high-frequency rather than a low-frequency re- sponse alternative (Schwarz et al., 1991). The range of re- sponse alternatives provided also influences perceptions of | one’s sexual behavior (Rothman, Haddock, & Schwarz, 1996). Participants were asked how many sexual partaers they had had in their lifetime and were provided with re- sponse scales constructed to shift whether their responses fell toward the high or low end of the distribution, The seales did not influence the nurnber of partners that people reported (ie. approximately three), but rather whether they thought their response was above cr below average and their subsequent concerns about HIV vulnerability Recall that Festinger (1954) predicted that individuals, ‘would prefer social comparison information only when ob- yf jective information was unavailable. Yet individuals seem, ready to make judgments about thei health on the basis of| input from their friends and families even when a health care provider ean offer more objective data. The field of medical sociology has studied the impact of such “lay re- ferral structures” and their somewhat negative conse- {quences for the health care system (e.g.. Apple, 1960; Friedson, 1961). In a clever role-playing experiment, Sanders (1951) asked subjects to imagine that they had de- veloped a symptom and needed to decide whether to see a physician, He then manipulated whether there was an ob- Jective basis for believing that the symptom required med- ical attention (a positive result on a test described in a text- book that the subject tries) and whether the subject’s friends encouraged him or her to go to the doctor. The re- sults supported Festinger’s predictions. When subjects had ‘objective information indicating that they should go to the doctor, the advice of friends had no influence. In the ab- sence of this kind of objective information, however, sub- jects were stongly influenced by others. But there is one ‘caveat: when subjects had objective information indicating that they should nor go to the doctor, others’ opinions that they should still influenced their decision. So there is an asymmetry here. Objective information overrides social ‘comparison information in motivating individuals to seek health care, But social comparison information can over ride objective information when the objective information is a “noneveat—that is, when it indicates that the person is not ill, We will return to the difficulties that individuals hhave in processing “negative” information indicating health when we discuss Cioffi's work in the section on selawarges ate in gehts rm Summary So far we have discussed the application of some classic social psychological theories to the understanding of health-related behaviors. We now take a moment to reflect fon (wo interrelated questions: (1) how have these theories been useful in understanding health behavior and (2) how ‘has application in the health arena provided an opportunity to evaluate and modify the theories themselves? Much of the research on social science and health has been problem-focused. Investigators have conducted stud- ies in order to understand, say, psychological aspects of a particular disease or to gain a descriptive understanding of the correlates ofa particular health behavior. Although this oft This Rea whic beh texte nit liets behe not F ory: with Se ity of first is chs50. Part Seven buterd Interdiscip reduction framework, people were originally thought 10 be ‘motivated by fear to eliminate threats 0 their health, Yet a systematic, motivating effect of fear on behavior has been uifficult to demonstrate (Rogers, 1983; but see Schwarz, Servay, & Krumpl, 1985: Sutton, 1982), Protection Moti- vation Theory (PMT; Maddux & Rogers, 1983; Rogers, 1975, 1983) incorporates traditional assumptions about the role of fear into 2 social cognitive framework of health- related di ms to také ac- Z are elie See ree people are ‘motivated to protect Themselves, Protection motivation is ‘Considered “an intervening Variable that has the typical characteristics of a motive: i arouses, sustains, and directs activity™ (Rogers, 1983, . 158). Following Leventhals (1970) distinction between the goal of controlling fear and the goal of controlling danger, PMT proposes the specific cognitive mediators that under” lie the influence ofa fear appeal on subsequent health be- havior. Exposure to health-relevant information initiates two appraisal processes: threat appraisal and coping ap- praisal (of Lazarus & Launier, 1978). A threat appraisal as Sess the danger posed by the health issue, comparing the relative benefits of one's current (maladaptive) behavior to the severity of the danger and one’s vulnerability 10 it Maladaptive behavior is inhibited to the extent that the threat (a combination of severity and vulnerability) out ‘weighs the perceived benefis of continuing the curtent be- havior. A coping appraisal compares the costs of periorm- ing the recommended behavior tothe ability o take ection (Ge, self-efficacy beliefs) and the perceived effectiveness of the action (ie., response efficacy beliefs). The relative ‘outcome of these parallel processes determines the degree of protection motivation, which is generally measured terms of behavioral intentions Feelings of fear indirectly motivate behavior by aug- menting perceptions of vulnerability and severity. PMT. originally specified a mulipicatve relationship between threat and efficacy (Rogers, 1975), such that high levels of fear motivated behavior only when accompanied by srong feelings of efficacy. Fear inthe absence of strong eficacy expectations should undermine health behavior. However, there has been litle empirical support forthe fear-by-effi- cacy interaction (Rogers, 1983; Sutton, 1982). la its cur- rent formulation, the motivation to protect oneself is pre- dicted to be an additive function of these two kinds of beliefs (Rogers, 1983). ‘Testing the Model There have been numerous experi- mental tests of Protection Motivation Theory (e.g.. Brouw- ers & Sorrentino, 1993; Fruin, Prat, & Owen, 1991; Mad- ‘dux & Rogers, 1983; Rippeioe & Rogers, 1987; Rogers & Mewborn, 1976; Stanley & Maddux, 1986; van der Velde & van der Pligt, 1991; Wie, 1991-1992; Wurtele, 1988; ‘Wortele & Maddux, 1987). Nearly all tests of the theory ceptions of response efficacy, self-effi have focused on ps cacy, vulnerability, and severity, paying almost no attention to the perceived benetits of the maladaptive behavior or the peeved cous ofthe apie ein Pence Dunn 2 Rogers, 1986; of Feuin, Pratt, & Owen, 1991) ‘Selletficacy or response efficacy beliefs have been the strongest predictors of behavioral intentions (e.g, Fruin, Pratt, & Oven, 1991; Maddux & Rogers, 1983; Stanley & Maddux, 1986; Wurtele & Maddux, 1987). The effects of perceived vulnerability appear to be independent of per- ceived efficacy, although occasional interactions have been reported (Maddux & Rogers, 1983; Witte, 1991-1992), There has been litle empirical support for an independent effect of the severity or noxiousness of a health issue on behavioral intentions, suggesting that this factor may play 4 limited tole in the formation of decisions regarding health behavior. However, severity may be the least infor- mative of the PMT dimensions; there is often litle variabil- ity in subjects” ratings. Similar results have been obtained in nonexperimental analyses ofthe effects of beliefs about efficacy, vulnerabil- ity, and severity on health-protective behavior (c.g, Aspin- wall et al, 1991; Eppright, Tanner, & Hunt, 1994; Sutton g & Hallet, 1988; van der Velde & van der Pligt, 1991). For example, van der Velde and van der Pligt (1991) found that, beliefs concerning response efficacy, self-efficacy, and vul, nerability independently explained intentions t0 use con, doms among heterosexual men and women. These three variables, as well as severity, were associated with condom, use in a sample of gay men. Because perceptions of risk and efficacy are shaped by past behavioral experiences, the variance in fealth behav, ‘ors that can be explained by PMT will be clarified by ac- ‘counting for a person's prior behavior. For example, Witte (1991-1992) obtained effects of threat and efficacy on bee hhaviors and intentions related to safer sex only after con; ‘wolling for previous sexual activity. Alternatively, invest ‘gators can test PMT among individuals who have never, before engaged in the target behavior (e-g., Rippetoe &, Rogers, 1987; Wiitele & Maddux, 1987). Because ofits roots in research on fear appeals, PMT hhas addressed primarily situations in which people need t0j protect themselves from a threat. However, this approach] ccan be adapted to characterize a broader range of health concerns (Beck, 1984; Maddux, 1993). For example, some, studies have examined atitudes toward exercise and inten tions to exercise (Fruin, Pratt, & Owen, 1991; Stanley & Maddux, 1986; Wurtele & Maddux, 1987). In this context people are motivated to obtain the potential benefit of a bey havior rather than to avoid a potential threat. ‘The decision to adopt an advocated behavior has been, the primary outcome measure in neatly al ests of PMT However, forming a behavioral intention is but one ‘many potential coping responses to a health-oriented apy Tike the tho pro. the per und ib: ade 199) 1996 that « bility How. cond) did How:which to communicate to other people about somatic expe- riences and can lead people 10 identify symptoms even when a health problem is believed 10 be asymptomatic. For nd Leventhal (1985) found that indi Viduals believe that they can estimate their own blood pres sure even though actual blood pressure, physical sensa tions, and emotional responses correlate only to the extent that they reflect a specific eliciting event (cf. Pennebaker etal, 19% Lay theories about hypertension not only shape how people interpret somatic experiences, but also influence their compliance with medical recommendations (Bau- ‘mann & Leventhal, 1985; Meyer, Leventhal, & Gutmann, 1985; Pennebaker, 1984), In one study, people adhered to & pharmacological treatment to the extent that they believed it was influencing the symptoms they thought were associ ated with hypertension. Compliance was also a function of the perceived duration of the disease. People dropped out of treatment if they perceived hypertension to be acute, but ‘maintained treatment when it was perceived to be a chronic condition (Meyer, Leventhal, & Gutmann, 1985), Further ‘more, contextual cues (e.g, being ina stressful situation) as well as theories about an illness (e.g., the duration of hhigh blood pressure) shape how people respond to infor~ ‘mation about their blood pressure (Baumann et al., 1989). ‘Symptom sets were perceived to reflect an illness when presented in the absence of a stress-related context such as, the day before an exam (Baumann eta, 1989, Experiment 2), Beliefs about an iliness, as well as actual health status, can also influence how side effects from treatments such as chemotherapy are interpreted (Leventhal, Diefenbach, & Leventhal, 1992; Leventhal, Nerenz, & Steele, 1984). Fi nally, a person's age can alter whether a symptom is ati uted to illness of to the normal consequences of aging, at- tributions that in turn influence requests for medical assistance (Prohaska etal. 1987). Lau and Hartman (1983; Lau, Bernard, & Hartman, 1988) have identified the same set of dimensions in peo: presentations of common illnesses (e. colds, fu, attacks). However, people were less likely to em: © both the duration and the consequences of suc ill- nesses, perhaps because they are less relevant in the con- text of 2 common health problem, Moreover, people's mental models included the dimension of cure, because these conditions are curable. Lau, Bernard, and Hartman (2989) replicated their findings in a longitudinal analysis ‘that provided additional evidence for stability in the di- ‘mensions that people use to describe illnesses. However, people do not necessarily use al five dimensions in a sin- gle description of a recent illness (Bishop ct al., 1987; Tusk, Rudy, & Salovey, 1986). ‘Although there is’ substantial evidence that people meaningfully organize their understanding of an illness along the five dimensions outlined, thee has been litle di- example, Baumann Seven / Interdisciplinary Perspect = * vary in prototypicalty, the of how this information is structured in 4 memory. Bishop and his colleagues (Bishop, 1987; Bishop & Converse, 1986; Bishop os. 1987) have proposed thet health information is organized around prototypes (ef, Rosch, 1978; Smith & Mein, 1981). The identification of a health threat is thought co reece the disease prototype memory that best matches the current set of symptoms, § When people are presented with sets of symptoms that are better able 0 classify and. remember those that are highly prototypic (Bishop & Con- verse, 1986). As the provotypicality of a set of symptoms people are faster and more accurate in their iden- lification of the disease associated with the symptom set (Bishop et al, 1987). Although these findings are consis- tent with a prototype model. they can also be accounted for using altemative conceptions of memory such as a schema- based (Fiske & Taylor, 1991) or exemplar model (Hintz- oe that | 1987 In on he Loss-f BSE.1 fain by man 980 ‘ wet eereeaieinceermintniinbetstte psf sat ie «tras or hc petes vica. ‘en Maire duty etavns (Gastone b Gere 99 rence Gibbs, Gara & Boney McCoy, 1998). For sane peters adolescent smokers image the protic cgay (Meyen Stoke acl ndepenen nd suc and his eal mesg Tepecntaon ry scout npr fore appeal fel phy (a tener (Chai ls 198i Leventhal & Clery 1580) al Further, changes over time in college students’ risk behav toting operon alpriegelrrngg rere peters poting Radi piabete tod mors Peuman coed beta ple eeceie cee Sees ear ed Presence mnore pone, heeds he prototypes held by tore whe aelke smoked less over time became more negative (Gibbons & teetion'b Geran, 1995, The favorit of prasype sd the — sammie teeter dada uate eo ee deision POS GISETST Ip poesia jon fetetion SE Pete saves poviont Gott witageer bs ese bee preted seven conealing for einen to Ears Binh conta tions, Gerard & Bony Mecay. 1995) eel M1 orl Salone, fom ena cast) (evened by Rathas & Svey, o}igamy PME p This Line of research is based on the demonstration example, u ‘Tversky and Kahaeman (1981) chat gain- and loss-fr ere ston alternatives do not have a symmetrical effect on dei Dabperetve ‘making, even when these different construals afford th sees same expected utility. Specifically, people have bee fisky choice shown to avoid risky or uncertain options when consider. peers ing them in tems of gains, but to prefer risks and tolerate associated losses or costs (Tversky & Kahneman, 1981)3 i Neary all health-clated decisions involve the considera actually unc framed averse, incr (versky &Seven J Intenisl Precaution Adoption Process Weinstein and Sandman (1992) divide precaution adoption into a series of seven stages, Initially, people are unaware of the issue (Stage 1), Many theories fail to distinguish between people who know nothing about a hazard and people who have consi: ed the issue and concluded that there is no theeat. The Precaution Adoption Process, however, makes this distinc som. When people have learned something about an ssc, they are inthe awareness stage, but are not necessarily en ssaged with or thinking about the issue 10 any great extent (Stage >» In Stage 3, people become engaged art be thinking about how to respond, This decision-making stage can result in one of two outcomes. If the decision is made sot 10 fake any action, then the Precaution Adoption Process ends (Stage 4). People who form an intention to sot have reached Stage 5, Stage 6 isthe ination of action, sod Stage 7. appropriate, represents that the behavior has been maintained overtime A strength of this mode is that it identifies the factors thal promote transitions across stages. Acquiring knowl edge about an issue or knowing someone who has had a health problem predicts whether people move from Stage | to Stage 2. Perceptions of personal vulnerability predict ‘whether people decide to take precautionary action (Stage 3), However, deciding to take action (Stage 5) is hypothe- sized to occur only when beliefs about susceptibility, severity, and efficacy each reach a particular threshold (Weinsitin, 1988). Finally, moving from intention to action (Stage 6) is especially sensitive to the presence of situa tional obstacles and constraints "" Unlizing data from cross-sectional and prospective studies of public reactions to household radon and radon testing, Weinstein and Sandman (1992; Sandman & Wein- stein, 1993) have gathered support for several of the hy- potheses underlying the model. For example, radon testn ‘ver one year was almost exclusively limited to people who had originally stated they were planning to test (Stage 5). Consistent with the Precaution Adoption Process, per- ceptions of vulnerability, personal concern, and severity cistinguished people who decided to test from those Who exther decided not o test or remained undecided. Although perceptions of vulaerabilty and personal concem led to differences in testing intentions, they were not associated with actual radon testing. This observation is consistent with the assumption that situational factors play an impor- tant role in determining whether intentions are translated Transtheoretical Model The Transtheoretical Model fo- es on the process by which a person decides to take ac- tion, Separating it into five stages (Prochaska, DiClemente, ‘& Norcross, 1992). Given the model’s emphasis on behav- Jor change, each stage is specified by a person’s past be- hhavior and plans for future action. The ability to change sion indicates the intention to take action within the next behavior is thought to be function ofthe stage attained. In: the stages, consider smoking cessation, 6 c. precomemplation, identifies individuals sho express no intention to stop smoking in the near fu ture, typically operationalized as the next six months. Peo pple who indicate that they are thinking about quitting sometime in the next six months but have not cormmitted to the action have reached the contemplation stage. Prepara- The inital sta month and at least one unsuccessful attempt to change the bbchavior in the past vear. Action involves successfully al- tering behavior for one day to six months. When behavior has continued for longer than six months, a person is said to have reached maintenance. Although progression through each stage is primarily linear, relapse to an earlier stage can occur, re n a spiral-like progression through the behavior change process (Prochaska, Di- Clemente, & Noreross, 1992) ; Tis model has generated a considerable amount of em= pirical work (For a summary, see Prochaska, Redding, et al, 1994). Although initialy formulated 10 address at- {empis to change addictive behaviors such as smoking, it has been applied successfully in several other domains, in- cluding weight control, mammography utilization, exercise adoption, and safer sex practices. People are assigned to a stage on the basis of their responses to either a discrete cat- cegorical measure (.g., DiClemente et al., 1991) or a series fof scales that measure a person’s standing with regard to a each stage (e.g., MeConnaughy, Prochaska, & Velicer, 1983), Research on addictive behaviors has suggested that the vast majority of people are neither initiating change nor ready to change their behavior. Prochaska, DiClemente, 4 ‘and Norcross (1992) estimate that only 10 to 15 percent of smokers have reached the preparation stage. However, the distribution of stages appears to be sensitive to the issue 4 under consideration, as well as to the time frame used to ‘operationalize each stage. For example, when considering pregnancy prevention, people were classified predomiz nantly in the action or,maintenance stage (71.6 percent) but for STD prevention the same people were largely inthe precontemplation or contemplation stage (48.6 percent) (Grimley etal, 1993), ‘There has been some empirical support for the valiit of these classifications. For example, the classification of, smokers into either a precontemplative, contemplative, of preparative stage accurately reflected their current smoking history, such as the aumber of cigarettes smoked each day4 and their reported level of addiction based on the Fagersf sicom scale (DiClemente et al., 1991; Pallonen et al. 1992). Similarly, stage of dietary fat reduction was consis: i teat withthe reported percent of calories of fat in one’s diet, (Curry, Kristal, & Bowen, 1992). Longitudinal data have revealed that people assigned to the preparation stage are subsequently more likely to attempt to quit smoking than, il ee ee eee 198 proc Prox platc oh wechn tion ( 1985) report stages Th about marke al, 19: Janis a ing, ce Althou positive leagues of pros stages, across ¢ crease ( al, 198: tained { Rakows Clement use of 1 1993; Re although using eor Evaluati Model an descriptic behaviors distinct Who are n ity of the ory. How:blanket set of variables that can consistently predict behav ior change. An intervention designed to change behavior is predicted to be successful only co the extent that it comple- iments a person’s current stage. Prochaska etal. (1993) re cently re successful stoking-cessation interven- tion based on this approach (but see Sutton, 1996). Similaely, a 104 patient's stage has successfully encouras. ieraphy utilization among women of low socioeconomic status (Skinnér, Strecher, & Hospers, 1994), The underlying assumption in| these models—that factors such as perceived vulnerability, efficacy beliefs, or estimated costs will be more effective predictors when examined within the context of a person's stage—parallels recent developments in continuum models such as Protection Motivation Theory or the Theory of Planned Behavior. In those cases, the predictive validity of 4 factor is maximized when close attention is paid to per- sonal and situational moderators. However, stage models are distinct in that they specify a sequence in which partic- lular issues must be addressed. joring a doctor's messa 1 satin Summary We have considered both formal and grounded social psy- chological theories of health behavior. One way in which theories differ is their relative emphasis on content or process—that is, whether they focus on specific health-rel- ‘evant cognitions or on the underlying mechanisms of be havior change. The grounded theories provide a rich de- scription of the beliefs that are seen as relevant to health ‘The Health Belief Model is probably the prototypic exam- ple of theorizing at this level. The challenge, however. for ontent-oriented theories concerns construct fidelity and discriminant validity. By construct fidelity we mean, do the Gifferent operationalizations of a construct across studies ‘measure the same latent variable? It is clear that litle aren tion has been paid to the fidelity of measures designed to identify a variety of health beliefs (Harrison, Mullen, & Green, 1992). In addition, investigators often develop mea- sures in the same contexts in which they tes their associa tions with behavior; there is litte independent validation (Holden, 1991), ‘Turning to issues of discriminant validity, comparing theoretical perspectives raises the question of whether or ‘ot specified constructs are focusing on the same cognitive activities (albeit with different labels). Weinstein (1993), for instance, has pointed out the conceptual similarity of the variables instantiated by the Health Belief Model. the ‘Theory of Reasoned Action, Subjective Expected Usiity ‘Theory, and Protection Motivation Theory. Consider self efficacy, a variable delineated in both Social Cognitive ‘Theory and Protection Motivation Theory, perceived barri- ers as described in the Health Belief Model, and perceived behavioral control as described in the Theory of Planned Behavior. When the specific operaivonalizations ofthe ical constructs are revealed, they sometimes appear lek differentiated than in the original theoretical model, & ceived barrier to get a might be that a4 woman believes that she cannot obain one at & LOW eos On a self-efficacy scale, she would also respond that she ig the next month trol, she would indicate that obtaining @ mammogram would be difficult. As operationalized in this example, all theee of these constructs appear to reflect the same ex pectancy coatngeat on cach tet such hat cap al odepencil is unlikely. For example, Bernstein and Keith (1991) rex ported a high positive correla benefits of using contraception and the perceived serious ness of an unplanned preznancy. The benefits of contrad y revealed (0 individuals fot hhave the most serious conse ception are most strikin ba Whom a pregnancy woul quences. In this sense, the #0 cognitive constructs are in: Ope terdependent. To the extent that the benefit of a behav disy such as contraception, is defined as reducing the lik asi G1 an unwanted outcome, benefits, outcome severity, and faww outcome likelihood cannot be meaningfully distinguished aye (Suton, 1982). cont Given the apparent difficulty in distinguishing among com some of the constructs specified by the theories we have reviewed, pethaps the field should be moving toward Disp development of integrative models that take advantage off agpn seemingly related families of constructs and complement of of tary sets of variables (Maddux, 1993; Weinstein, 1993)} of be Grounded theories lke the Health Belief Model and formal Carw theories like the Theory of Reasoned Action and the Th ‘loba ory of Planned Behavior place considerable emphasis on sus bi attitudes and beliefs about outcome (e-., de Vries, Dijs Inset stra, & Kuhlman, 1988), which is probably why apy dices proaches emphasizing an integration of such theories wit the bit Social Cognitive Theory have produced encouraging tong : sults (eg. Ajzen & Madden, 1986; MeCaul, O'Neill, (Shei Glasgow, 1988; Rodgers & Brawley, 1993). Contextual using ing these new theories within basic research on mental taveb ‘models of health and illness would allow investigators to (esh adress the specific ways in which beliefs in a domain de surger Yelop and change asa function of intrveations. Finally, as with p structural equation modeling becomes the data analytic unsuec technique of choice for theory testing, multiple opera being tionalizatons of latent constructs will be necessary fo 1993), properly specifying measurement modes. a distress 4 illness PERSONALITY, EMOTION, AND HEALTH * cal ‘Some substantial lines of research in social psychology and dis concerning health behavior are not easily ied o an explicit] measucoptimism and subjective well-being, as well as some sense that optimism may be a factor in disease progression and recovery, Attributional Style Optimism has also been operation ized as an attributional style (Peterson & Bossio, 1991; Pe terson & Seligman, 1984, 1987; Peterson, Seligman, & Vaillant, 1988). Using a self-report instrument catled the Autributional Styles Questionnaire (ASQ) or a content-cod: ing procedure called the Content Analysis of Verbatim Es planation (CAVE), investigators identify the tendency — called pessimism—to make internal (it’s my fault), global (its all-consuming). and stable (it will never go away) at twibutions for stressful and challenging events in an indi viduals life, Optimists are individuals who make the oppo site attributions: external, specific, and unstable. Although these alternative attributional styles were devised originally to identify a risk factor for depression (Peterson & Sel man, 1984), recent work has attempted to link them to physical health outcomes. In perhaps the most impressive study, Peterson, Selig- ‘man, and Vaillant (1988) coded open-ended interview tran seripts from ninety-nine male Harvard students in the asses of 1942 to 1945, The students were twenty-five years old at the time of the initial interviews, and the habit- uual ways in which they explained negative events (often having to do with their wartime experiences) were coded as pessimistic or optimistic using the CAVE technique. Find- ings from periodic medical examinations on these individ- uals over the next forty years were also available to the in- vestigators, and an index of physical health quality was created on the basis of the judgment of an internist reading the medical records. Overall, the men who explained nega- tive events pessimistically—using internal, global, and sta- ble attributions—at age twenty-five were in significantly poorer health at ages forty-five, fifty, fifty-five, and sixty than those with an optimistic explanatory style, even con: ‘rolling for initial physical and mental health status. In {quent illness reporting. For example, among college stu dents, a global and stable attributional style assessed with the ASQ early in the semester was associated with more frequent reporting of colds, sore throats, and flu later in the semester, controlling for initial levels of illness and de- pression (Peterson & Seligman, 1987), ‘The study of individual differences in optimism and their association with health and illness faces at least 1wo challenges. One is that insufficient attention has been paid to issues of construct validity (but see Smith et al., 1989). Do the LOT, ASQ, and CAVE (let alone “home-grown' ‘optimism scales) measure the same latent construct? Cor- rolations between the LOT and ASQ are surprisingly low (Scheier & Carver, 1992). Are transparent measures such autoeld as these valid or are they overly reactive? Is optimism re ally unidimensional (see Marshal tal, 1992. who suggest © that its not)? A more important challenge concerns mech- anism: ean we specify a physiological or behavioral path. ‘way linking optanism to health? Psychoneuroinimanologi- cal data are emerging that suggest that optimism may be associated with enhanced immune functioning (Kamen. Stogel eta, 1991). Among a small sample of elderly indi- + viduals, optimism was measured using the CAVE proce- dure applied to audiotapes of nine interviews concerning sireses, hassles. wortis, mood, socal contacts, activites, snd health, Kamen-Sigel etal. focused primarily on ater ations in Tell levels in response to a mitogen challenge measured some time after the interviews. A combined index ofthe internal, global, stable explanatory syle (i, pessimism) was associated with a lower ratio of helper to Suppressor Tells and to poorer cel proliferation when challenged by the mitogen, especially at lower levels of tnitogen exposure, contlling for depression (because this explanatory style. of course, is associated with if), current health, mediation, weight change, sleep, and alcohol use, ‘This work suggests that the influence of optimism on health may be mediated by the immune system, but the specification of a precise mechanism accounting for these connections remains a mater of speculation, and other studies have not produced as promising results (Cohen et 5 al, 1989; Sieber etal, 1952) Optimistic Beets Whether optimism s viewed as ads. § Psion ox at an explana spe the erature reviewed thas fr has focused on whether individual ferences in pis actu for wrance in subjective well being and prysical heal Another perspective views optimism aa felfserving basta we default to unrealistically opi mise jadgoens about te licelitood of future event In | parca, people belee tit positive outcomes ate moe Ttely to happen o thom hart other people, whereas neg: ative outcomes are les likely. Taylor and bec colleagues have clasifed hs pervasive bias as one ofthe general “positive illusions” and have argued that these illusions (GramoW san and phyla weliogffoyloe & Brown, ‘i983, 1994; but see Block & Colvin, 1994; Colvin & Blok, 1994; Seder, Mayman, & Manis, 1993) ‘Weinstein (1980, 1989) has focused on this bis inthe domain of rik perceptions, introducing the term “unreal tc optimism’ fo describe the phenomenon whereby the overwhelming majority of people believe that their ow Tsk of negative ture cucomes is below average fru | tiew of even lteratre, see Weinstein & Kiki, 1996) | Becase perceptions of ulerabiy to harm ae essential explanatory vetables in 19 many models of change in tealrlasd bebo, wdecetnating such vilnerbiy tiny lend a flere bo adopt proces tat wold ro tee health Pe, 1983; Weinstein, 1980, 1983, 1989) adopt Perlot all mc tion if diftice Bushn who f levels it, alt (Wein: hat de sessm« stein & The liefs bh 1994; an atte cilemn compe can eit risk be orientaand the Intepersonal Hostility Assesment Technigue—ae not highly intercoreated (Muse eta 1989), Pact of the problem seems to lie in reaching con sensual definition of hostility. Barefoot (1992) suggests that a multidimensional approach. inching a cognitive component (cynicism, hostile stributions), emotions Conger, resentment, disgusb, and bebasors (verbal a8 sion, antagonistic behaviors), hs utility. Evsting messes tsualy represent only one or two of these facts of ost ity, For example, the widely used Ho scale. which ix de rived from the Minnesota Muliphasic Personality Inven tory (MMPD, assesses eynicism more than angry fel orhostle behaviors (Barefoot ctl. 1989: Smith & Feo, 185), In general, self-report measures of host quite sensitive to Social desirability biases, At the some time, the interview based measures, which cover the som ponents of hostility more evenhanded, tend to require Considerable time to administer and extensive training to score reliably. ‘What ae the potential finks between hostil Aisease? The most commonly endorsed mechanism for un derstanding the impact of hostility om health stat iis re Tated to inereases in blood pressure, heart cate, and stress hormone release, especially during the frequent anger out butts to which hostile individuals ae prone (Smith, 1992). Because hostile individuals operate in a state of Vi with respect to the environment, these cardiovascular changes ean become chronic. Exaggerated reactivity is thought 0 initiate and hasten the development of coronary artery disease leading 10 symptoms of heart disease (Williams, Barefoot, & Shekele, 1985). Evidence for this hypothesized mechanism has been mixed ovcrll but itis ‘more encouraging if one focuses on studies where hostility is induced in response to social ater than cogitveses- sors (Hardy & Smith, 1988; Smith & Brow, 1991; Suaez & Wiliams, 1989). A related mechanism is based on the idea that hostile individuals experience the world as moce stressful and receive considerably les social suppor from it Hardy & Smith 1988; Houston & Kelly, 1989). This d- ditional stress and lack of social support may increase the vulnerability of hostile individuals toa variety of diseases “These two mechanisms have been combined into a tansac- tional model suggesting that hostile individuals create through maladaptive thoughts and behaviors more fe- quent, severe, and enduring contacts with stessrs, and as 4 resule experience considerable conflict in their daily lives. This style results in the physiological changes de- seribed above and in reduced social support (Smith & Pope, 1990; Suls & Sanders, 1989). Finally, hostile iad vidusls may lead unhealthy lifestyles. Hoste individuals engage in lower levels of physial exercise, report higher alcohol intake, engage in fewer self-care behaviors, have cater body mass, and are more ikely to smoke cigarettes than Tess hostile individuals (Houston & Vavak, 1991: Koskenvuo etal, 1988; Shekel ta, 1983) and heart ininary Perspectives The literature on hostility and heart disease ate considerable consternation. Stable links reported in one study are not replicated in the next. only 10 resurfacg broader Type a ret al. (1991) thought, 5 0 obtain null effects in thi shortly thereafter In a meta-analysis of th and heart disease literature, Mi fully point out why itis s0& area of study even if there isa relatively sirong underlying Tink between the two, They point to thee problems: (1) thea selection of onl high-risk or diseased subjects, 2) the ue Uv selerepon meatucs and (3) the Wentfeaon of nyocardal infarction as the defining outcome criterion, Miler and his colleagues argue that when researcher en, roll only subjects with heart disease or suspected hear di case, as is characteristic of many studies inthis iterate, they restrict the range of disease severity, reducing the power of tests of associations between Type A and disease. 5 Second, self-report measures of Type A and hostility may ger not be especially valid in comparison to those based onthe Type A Structured Interview or other kinds of behavioral observation (Booth-Kewley & Friedman. 1987). Even with the Structured Interview, the threshold for defining some- one a5 Type A or as hostile may have sified over the yea Such that larger mimbers of subjects have been assigned these isk factors in recent studies (Schweritz, 1989). In Miller et a's meta-analysis folly 70 percent of all middle aged men with heart disease were clasified as Type A, Third, Type A and hostility may not be associated wit ll possible coronary events (Brackett & Powell, 1989). Fo example, because Type A's develop heart disease ata Younger age than people not classified as Type A (ie. Type B's), they may be more likely 1 survive a heat attack. In fact, Miller tl found that Type A was not related to fat ‘myocardial infarction. Finally, when associations between, hostility and cardiovasculc reactivity do emerge, they may be due in prt tothe anger one experiences on being beled a cardiae patient, hypertensive, or the like (Sul, Wen, & Costa 1995). Neuroticism and Negative Affectivity Many stuies in health psychology ty to link the exper cence of stressors, challenges, relationships, and the like tog health outcomes. Often such studies rely on subjective ‘measures of both the independent variable (e.g., stress) and. the dependent vriable (health oilnes) Ina typical study, individuals who report that their mariages are stressful might also repor more physical health problems fnd evaluate their overall health as worse than individuals reporting happy marriages. During the past ten years oF, the value ofthese Kinds of studies has been seriously ques- tioned. Costa and MeCise (1985, 1987) believe that an in- dividual-ditference consruct thatthe termed newrotcism (defined asthe tendency to experience negative, distressing mations) serves asa thied variable linking reported di tress to subjective measures of health and spuriousy ia- der pre