Well-Being: -~ -The Foundations ofHedonic Psychology

Daniel ICluhnemm,

Ed Dieneq and

Norbevzt Scbwmz E D I T O R S

RUSSELL SAGE FOUNDATIONNEW YORK

I

4 Reports of Subjective Well-Being: JudgmentalProcesses and Their Methodological Implications

Norbert Schwarz and Fritz Strnck

The cognitive and covnmzmrcative pv-ocesses rrvrder-lying individuals reports of happiness and satisfac-tion with their lives as a whole are reviewed in thischapter. Reports of subjective well-being (S WB) donot re$ect a stable inner state of well-bein&.Rather, they are judgments that individuals formon the spot, based on information that is chron-ical<v or temporarily accessible at that point intime, resrrltin, in pronounced context effects. Theway in which accessible infov-matiovJ about av1 indi-vidlral’s life influences the judLgment depends on

i~ow it is used. Information that is used in forminga mental representation of the individual’s life as awhole or of some extended episode results in assimi-lation ejyects, SUC~J as hicqher reports of S WB II~JLW a/~nppy mther than sad event comes to vnivid. Ivijbr-mntion that is rued in jbrminLq a standard of covn-

pnrisovi results iv) c o n t r a s t efiects. In t/Jis cmc’, the

indisidlrali lty~ looks bland by comparison to a/~nppy event. The variables that detervvhe nssivni-latiovi or covttrast efi?cts are identified. Givcvi thatthe same event can increase or decrease an individ-tlal’sjudLflvvicvit of S WB, dependivg on its itse iv1 thecovutrnal of the individunl’s life or of a stniliinrd,the rt*lntiovuhip between objectivr evcvrts and srrb-jcctillc rvalrtntiovrs is viecessarily weak. Hcnct,, S WRcannot bc predicted OVL th b a s i s o f objectir7z cir-cI~rvutavrces, tt vtless ovlc takes t/JL’ construal processesivlto ncrottvlt. Iv1 addition to ivifortnntion aborrt hiso r /JCI* owv~ pnst, pwscvrt, or fiittrrc, the ividiiGdrta1urfl-v use iviji)rmfltiovi about others’ lives in nssessivgth qllnli<v $his or hv life. Ahbqh peoph oj~~w

ji.d httCr 1dJt’N t/JCy COVklpWt’ t~JtWlSt’iVt*S t0 dJt7’S

W/JO II 1-c less 1l’L’ll ofi t/JP specijic orrtcorilc npivi h-

/MiS 011 t/Jt’ S/It’CijiC VllltllW O j ’ t/JS WlCfl tfli COP1 -

strrral. Individuals vviav simplij-i the complt*.uitizsoj’ evaIrrativg their lives by drawivrc/r on thtvir j>el-izgs at the time oj’jrid&ment as a sourct’ oj- izi-jbrtnntion. Hence, they rtvpov-t ~~&JCY S WB W~JLT ina pood rather than bad mood (and jivrdivtq a dimeis strficievit to increase temporarily one’s life sfltis-jirctiovl). Moods are vnore likely to ajyect jud’~ments

ofBenera1 SWB than jztdflments of specific lzfe do-mains. As a result, a particularly happy event indomain X may increase an individual’s satisfac-tion with his or her life as a whole but decrease sat-isfaction with the specipc domain by way of con-trast. Such divergent influences decrease therelationship between Blobal S WB and dovnain satis-faction. Public reports of SWB are often inflatedowing to self-presentation concerns. Methodologicalimplications are discussed.

hlUCH OF WHAT we know about individuals’ sub-jective Lvell-being (SWB) is based on self-reportsof happiness and life satisfaction. Since thegroundbreaking studies of Bradburn (1969), An-drews and Whithey ( 1976), and Campbell, Con-\*erse, and Rodgers (1976), hundreds of thou-sands of survey respondents around the worldhave been asked questions like, “Taking all thingstogether, how would you say things are thesedays-tvould you say that you are very happy,pretty happy, or not too happy?” or, “How satis-fied are !vou with your life as a whole these days?,-Ire you \*ery satisfied, satisfied, not very satisfied,not at all satisfied?” Questions of this type areintended to assess the subjective qualiv of life inJJI attempt to monitor the subjective side of so-&l change. These szr6$xtivr socinl indicntors sup-plement measures of the objective standard ofLiving, \\*hich have long dominated nvelfare re-search in the social sciences.

As Angus Campbell ( 1981) noted, the “use ofthese measures is based on the assumption that allthe countless experiences people go through fromd3y to dq add to . . . global feelings of ivell- being,that these feelings remain relatively constant overextended periods, and that people can describethem \iith candor and accuracy” (23). These as-sumptions have increasingly been drawn intoquestion, however, as the empirical work has pro-gressed. First, the relationship between individ-uals’ experiences and objective conditions of life

3

62 Well-Being

and their subjective sense of well-being is oftenweak and sometimes counter-intuitive. Most objec-tive life circumstances account for less than 5 per-cent of the variance in measures of SWB, and thecombination of the circumstances in a dozen do-mains of life does not account for more than 10percent (Andrews and Whithey 1976; Kammann1982; for a review, see Argyle, this volume). Sec-ond, measures of SWB have low test-retest reli-abilities, usually hovering around .40, and not ex-ceeding .60 when the same question is asked twiceduring the same one-hour interview (Andrews andWhithey 1976; Glatzer 1984). Moreover, thesemeasures are extremely sensitive to contextual in-fluences. Thus, minor events, such as finding adime (Schwarz 1987) or the outcome of soccergames (Schwarz et al. 1987), may profoundly af-fect reported satisfaction with one’s life as a whole.Most important, however, the reports are a func-tion of the research instrument and are stronglyinfluenced by the content of preceding questions,the nature of the response alternatives, and other“technical” aspects of questionnaire design (Schwarzand Strack 1991a, 1991b).

Such findings are difficult to reconcile with theassumption that subjective social indicators di-rectly reflect stable inner states of well-being(Campbell 1981) or that the reports are based oncareful assessments of one’s objective conditions inlight of one’s aspirations (Glatzer and Zapf 1984).Instead, the findings suggest that reports of SWBare better conceptualized as the result of a judg-ment process that is highly context-dependent.This chapter reviews what is known about hornpersons determine whether they are happy withtheir lives as a whole or not. Our focus is on eval-uations of one’s life as a whole or of some ex-tended episode of one’s life, rather than the eval-uation of single events (which is addressed inother chapters in this volume; see, for example,Kahneman). As will become apparent later, con-textual influences do often have an opposite im-pact on evaluations of a specific event versus eval-uations of an extended episode. A particularlydreadful event, for example, makes more moderateevents look good by comparison (Parducci 1995),yet it decreases the evaluation of the episode ofwhich it is a part (Strack, Schwarz, and Gschnei-dinger 1985). We note such discrepancies whereappropriate but primarily focus on how peopleevaluate their “life as a whole,” as survey questionsask them to do. Similarly, our review does not ad-dress how differences in personality may influencethe judgmental processes of interest here (but see

the chapters in this volume by Cantor and Sander-son; Diener and Lucas; and Higgins, Grant, andShah-). -

A PIKEvIEW

Not surprisingly, individuals may draw on a widevarieE of information when asked to assess thesubjective quality of their lives. Ross, Eyman, andKishchuck (1986) explored the range of informa-tion used by asking respondents how they arrivedat a judgment of SWB. They observed that explicitreferences to one’s momentary affective state ac-counted for 41 to 53 percent of the reasons thatvarious samples of adult Canadians provided fortheir reported well-being, followed by future ex-pectations (22 to 40 percent), past events (5 to 20percent), and social comparisons (5 to 13 per-cent). The experimental literature confirms therelevance of these different sources of information,and we address them in turn.

We first explore the impact of information aboutone’s own life, such as past events or expectationsabout the future. This review indicates that thesame event may increase as well as decrease gen-eral life satisfaction, depending on how informa-tion bearing on the event is used in forming ajudgment. Next, we address the role of‘compari-sons of one’s own lot with the lot of others. Al-though people generally tend to feel better whenthey compare themselves to others who are worseoff, the dynamics of social comparison are morecomplicated than early theorizing and commonsense \vould suggest. Following these discussionsof intra- and interindividual comparisons, we turnto the influence of temporary mood states and ad-dress how one’s momentary feelings may overridethe impact of other information relevant to one’slife. Finally, we integrate these processes in a judg-ment model of SWB before we turn to an assess-ment of the methodological implications for sur-vey research into SWB.

USING INFORMATION ABOUT ONE'S OWN

LIFE: INTRAINDNIDUAL COMPARISONS

Comparison-based evaluative judgments require amental representation of the object of judgment,commonly called a target, as well as a mental rep-resentation of a relevant standard to which thetarget can be compared. The chosen standard maybe intmindividual (for example, a previous state

IReports of Subjective Well-Being

of one’s life or one’s expectations) or interin-dividual (the situation of close others or a relevantreference group). The outcome of the comparisonprocess depends on (a) which information is usedin constructing (b) the target or (c) the standard(Schwarz and Bless 1992a). We first address whichof the many aspects of one’s life are likely to beused in forming a judgment.

Which Information Is Used?

When asked, “Taking all things together, howwould you say things are these days?” respondentsare ideally assumed to review the myriad of rele-vant aspects of their lives and to integrate theminto a mental representation of their life as awhole. In reality, however, individuals rarely re-trieve all information that may be relevant to ajudgment. Instead, they truncate the search pro-cess as soon as enough information has come tomind to form a judgment with sufficient subjectivecertainty (Bodenhausen and Wyer 1987). Hence,the judgment is based on the information that ismost accessible at that point in time. In general,the accessibility of information depends on the re-cency and frequency of its use (for a review, seeHiggins 1996). Information that has just beenused-for example, to answer a preceding ques-tion in the questionnaire-is particularly likely tocome to mind later on, although only for a limitedtime. This temporarily accessible information is thebasis of most context effects in survey measure-ment and results in variability in the judgmentwhen the same question is asked at different times(see Schwarz and Strack 1991b; Strack 1994a;Sudman, Bradburn, and Schwarz 1996, chs. 3 to5; Tourangeau and Rasinski 1988). Other infor-mation, however, may come to mind because it isused frequently-for example, because it relates tothe respondent’s current concerns (Klinger 1977)or life tasks (Cantor and Sanderson, this volume).Such chronically accessible information reflects im-portant aspects of respondents’ lives and providesfor some stability in judgments over time.

Accessibility

As an example, consider experiments on questionorder. Strack, Martin, and Schvvarz (1988) ob-served that dating frequency was unrelated to stu-dents’ life satisfaction when a general satisfactionquestion preceded a question about the respon-dent’s dating frequency, Y = - 12. Yet reversingthe question order increased the correlation to

Y = .66. Similarly, marital satisfaction correlatedwith general life satisfaction Y = .32 when thegeneral question preceded the marital one in an-other study (Schwarz, Strack, and Mai 1991). Yetreversing the question order again increased thiscorrelation to Y = .67. Findings of this type indi-cate that preceding questions may bring informa-tion to mind that respondents would otherwisenot consider. If this information is included in therepresentation that the respondent forms of his orher life, the result is an assimilation effect, as re-flected in increased correlations. Thus, we woulddraw very different inferences about the impact ofdating frequency or marital satisfaction on overallSWB, depending on the order in which the ques-tions are asked.

Theoretically, the impact of a given piece of ac-cessible information increases with its extremityand decreases with the amount and extremity ofother information that is temporarily or chron-ically accessible at the time of judgment (seeSchwarz and Bless 1992a). To test this assump-tion, Schwarz, Strack, and Mai ( 1991) asked re-spondents about their job satisfaction, leisure timesatisfaction, and marital satisfaction prior to assess-ing their general life satisfaction, thus rendering amore varied set of information accessible. In thiscase, the correlation between marital satisfactionand life satisfaction increased from Y = .32 (in thegeneral-marital satisfaction order) to Y = .46, yetthis increase was less pronounced than the Y = .67observed when marital satisfaction was the onlyspecific domain addressed.

In light of these findings, it is important tohighlight some limits for the emergence of ques-tion-order effects. First, question-order effects ofthe type discussed here are to be expected only\vhen answering a preceding question increases thetemporary accessibility of information that is notchronically accessible anyway. We would assume,for example, that respondents who are currentlyundergoing a divorce would consider their mar-riage independent of whether it was addressed ina preceding question or not. Second, the impactof information rendered accessible by precedingquestions decreases with the amount and extrem-ity of competing information. Hence, chronicallyaccessible current concerns would limit the size ofany emerging effect, and the more they do so, themore extreme the implications of these concernsare. This implies that question-order effects shouldbe relatively small for respondents who are preoc-cupied with a current concern, but rather sizablefor respondents who are not. If so, the differential

size of context effects may cloud actual differencesin SWB. Unfortunately, data bearing on these pos-sibilities are not available. Finally, information maybe rendered temporarily accessible by other fortu-itous events, such as what happens to be in thenews (Iyengar 1987). From a methodologicalpoint of view, such influences are less problematicthan the impact of question order. Most fortuitousevents affect only a small subset of the sample, inparticular when data collection extends over sev-eral weeks, as is typical for surveys. Question or-der, however, affects most members of the sample,thus introducing systematic bias.

Conversational Norms

Complicating things further, information renderedaccessible by a preceding question may not alwaysbe used. In daily conversations, speakers are sup-posed to provide information that is new to therecipient, rather than to reiterate information thatthe recipient already has (Grice 1975; for moredetailed discussions, see Schwarz 1994, 1996;Strack 1994b). Having just answered a questionabout her marriage, for example, a respondentmay therefore assume that a subsequent questionabout her life in general pertains to new aspectsof her life, much as if it were worded, “Asidefrom your marriage, how’s the rest of your life?”Whether the general question is interpreted in thisway or not depends on whether it is assigned tothe same conversational context as the more spe-cific question.

Partidy Rduvhnt @wtionr I n t h e a b o v estudies (Strack et al. 1988; Schwarz ct al. 1991),the conversational norm of nonrcdund,lncy wasevoked by a joint lead-in that informed respon-dents that they would now be asked hvo questionspertaining to their well-being. Following this lcad-in, they first answered the specific question (aboutdating frequency or marital satisfaction) and sub-sequently reported their general life satisfaction. Inthis case, the previously observed correlations of Y= .66 between dating frequency and life satisfac-tion, or y = .67 between marital satisfaction andlife satisfaction, dropped to r = -15 and .18, re-spectively. Thus, the same question order resultedin dramatically di&rent correlations, dependingon the elicitation of the conversational norm ofnonredundancy. Consistent with this interprcta-tion, a reworded version of the general question“Aside from your marriage, which you already toldus about, ho\v satisfied are you with other aspects

of your life?” resulted in a similar correlation with- --marital satisfaction, Y = .20 (Schwarz et al. 1991).Again, we would draw very different substantiveconclusions from the obtained data, depending onquestion order and the presence or absence of ajoint lead-in that assigns both questions to thesame conversational context.

The Redtrvdnncy of Hi@ly Similar Questions Inthe earlier examples, a general question was ren-dered partially redundant when preceded by amore specific one that addressed a subset of therelevant information. The same logic, however,also applies to cases in which several highly similarquestions are presented. Strack, Schwarz, andWsnke (199 1) asked respondents to report theirhappiness as well as their satisfaction with life.When both questions were introduced as the lastand first question of two different questionnaires,presented by two different researchers, both re-ports correlated r = .96. Moreover, respondents’mean happiness ratings (M = 8.0) did not differfrom their mean satisfaction ratings (A4 = 8.2),suggesting that they did not differentiate betweenthese concepts. Presumably, they assumed thathvo different researchers were asking the samething in somewhat different words. When bothquestions where presented by the same researcher,howe\-er, one at‘ter the other in the same question-naire, the correlation dropped to Y = .75, and re-spondents reported higher happiness (M = 8.2)than satisfaction (M = 7.4). Thus, assigning bothquestions to the same conversational context elic-ited a diff&cntiation because the hvo questionsw o u l d othcnvisc have been r e d u n d a n t - W h ywould the same rcscarchcr ask both questions ifthey i\*crc not supposed to tap diKerent aspects?These processes may undcrlic apparent inconsis-tencies in the relationship bctwccn reports of hap-piness and satisfaction and their respective prcdic-tors in dif+ercnt studies and may contribute to lot\*test-retest reliabilities when the same question isrcitcratcd within a short time span.

Strmnravy

Judgments arc based on the subset of potcntiall!applicable infi,rmation that is chronically or tem-porarily accessible at the time. Accessible informa-tion, however, may not be used when its repeateduse \i*ould violate conversational norms of non-redundancy. Next, we turn to the diEerent waysin Lvhich accessible information may influence ajudgment.

.i

Reports of Subjective Well-Bein- 65

MENTALCONSTRUALSOFONE'S LIFE ANDARELEVANT STANDARD: WHAT Is, WAS,WILLBE, AND MIGHT HAVE BEEN

The way in which chronically or temporarily acces-sible information about one’s life affects the judg-ment depends on how it is used (Schwarz and Bless1992a; Strack 1992). Suppose that an extremelypositive (or negative) life event comes to mind. Ifthis event is included in the temporary representa-tion of the target “my life now,” it results in a morepositLye (negative) assessment of SWB, reflecting annssimihtion effect, as observed in an increased cor-relation in the studies discussed earlier. However,the same event may also be used in constructing astandard of comparison, resulting in a contrast ef-fict: compared to an extremely positive (negative)event, one’s life in general may seem relativelybland (or pretty benign). These opposite influencesof the same event are sometimes referred to as en-dol\*ment (assimilation) and contrast effects (Tver-sky. and Griffin 1991). To understand the respectiveco~~ditions of their emergence, we need to under-stand hoiv individuals ztse accessible information.

The Lvariables that determine the use of informa-tion in constructing standards and targets can beconceptualized in terms of several broad decisionsa respondent has to make (for detailed discussions,see Schwarz and Bless 19923; Struck 1992). Themost important one is whether the information“belongs to,” or is representative of, the targetcategory (for our current purposes, “my life now”).Information that bears on a diflercnt episode ofone’s life, for example, or that seems extreme andunusual, \\.ill not be used in forming a representa-tion of the target, thus making it available for con-structing a standard.

What Is, Was, and Will Be: Is the InformationRcprescntntive of My Life?

Information about one’s life, such as specific lifee\*ents, \\ill be used in constructing a representa-tion of one’s current life only when it seems repre-sentJtive of the target. If the event is categorizedas pertaining to a dil-Ferent episode of one’s life, oras being unusual, it will serve as a standard ofcomparison, as a few examples may illustrate.

Tempovnl Distance Strack, Schwarz, and Gschnei-dingcr ( 1985, Experiment 1) asked respondents toreport either three positive or three negative re-cent life ei’ents, thus rendering these events tem-

T A B L E 4.1 Subjective Well-Being: The Impact ofValence of Event and Time Perspective

Valence of Event

Positive Neda tive

Time perspectivePresentPast

8.9 7.17.5 8.5

Category boundaryNot salientSalient

8.7 7.46.2 8.2

Sotrvce: Top panel adapted from Strack et al. (1985, Experi-ment 1). Copyright 1985 by the American PsychologicalAssociation. Bottom panel from Sch\s*arz and Hippler (un-published data).Notes: For the mean score of happiness and satisfactionquestions, the range is 1 to 11, with higher values indicat-ing reports of higher well-being.

porarily accessible. As shown in the top panel oftable 1, these respondents reported higher currentlife satisfaction after they recalled three positiverather than negative recent events. Other respon-dents, however, had to recall events that happenedat least five years before. These respondents re-ported higher current life satisfaction afier recall-ing negative rather than positi\Fe past events. Thisindicates that respondents included accessible re-cent events in the representation formed of theircurrent lives but used distant events as a standardof comparison (see also Dermer et al. 1979; Tver-sky and Griffin 1991).

These experimental results are consistent withcorrelational data (Elder 1974) indicating thatU.S. senior citizens, the “children of the GreatDepression,” are more likely to report high subjec-tive well-being the more they sufered under ad-verse economic conditions when they were adoles-cents. The accumulation of negative experiencesduring childhood and adolescence presumably es-tablished a baseline against which all subsequentevents could only be seen as an improvement. Por-tra!ing the other side of the coin, Runyan (1980)found that the upwardly mobile recollected theirchildhood as less satiseing than did the down-wardly mobile, presumably because they used theircurrent situation in e\paluating their past.

Chnkin~ the Strenm of Life: Cnttyoy BoundariesWhereas the use of life events \vas determined bytheir temporal distance in the above studies, othervariables may similarly influence how the stream oflife is chunked into discrete units. One of thesevariables is the salience of relevant transition

Well-Being-

points. For example, Schwarz and Hippler askedfirst-year students to report a positive or negativeevent that “happened nvo years ago.” As shown inthe second panel of table 1, this again resulted inan assimilation effect on current life satisfaction.Other students, however, were subtly alerted to amajor role transition, namely, their change fromhigh school to university status. Specifically, theywere asked to report a positive or negative eventthat “happened two years ago, that is, before youtame to the university.” These respondents reportedlower life satisfaction after recalling a positive ratherthan a negative event, indicating that their identi-fication of the event as a “high school” event re-sulted in its use as a standard of comparison.

Similarly, thinking about positive or negativeevents that might happen in the fixture resulted inassimilation effects on current life satisfaction(Strack, Schwarz, and Nebel unpublished data).Yet reminding the student respondents that theywould meanwhile have left the university again re-versed the pattern, resulting in a contrast effect.Hence, positive expectations about the future canincrease as well as decrease current SWB, depend-ing on their use in the judgment process.

Extremity Similarly, extreme ei’ents may seemunusual and not representative of ho~v one’s life isgoing in general. They may therefore be excludedfrom the representation formed and serve as stan-dards of comparison. If so, extreme and unusualevents are likely to result in contrast effects, atleast af%er some time has passed (thus providingtemporal distance) and their immediate emotionalimpact (to be addressed later) has waned. Al-though such exclusion processes have been ob-served in other domains of judgment (Herr, Sher-man, and Fazio 1983; Herr 1986; for a discussion,see Schwarz and Bless 1992a, 230-31), the rclc-vant studies in the domain of SWB are limited toevaluations of specific other events in people’slives, not evaluations of their lives as a whole.

Cuteguy Width: judgments of Specific Eveuts VCYSUSLife as a Whole Importantly, a highly positive(negative) event is likely to afI%ct judgments ofother specific events and judgments of one’s life asa whole in opposite directions. This suggests thatthe event can be included in the representationformed of one’s life in general (a “wide” targetcategory), resulting in assimilation effects. How-ever, it cannot be included in the representationformed of another specific event (a “narrow” tar-get category), and hence it serves as a standard of

comparison, resulting in contrast effkcts. In an ini-tial--test af the impact of category width, Schwarzand Bless (1992b) had respondents think about apolitician who was involved in a scandal (say,Richard Nixon). This decreased judgments of thetrustworthiness of politicians in general, reflectingthat the exemplar could be included in the repre-sentation formed of the group. However, it in-creased judgments of the trustworthiness of allother individual politicians assessed, reflecting thata given exemplar cannot be included in the repre-sentation of other exemplars-after all, Bill Clin-ton is not Richard Nixon, and compared to Rich-ard Nixon, Bill Clinton looks fine.

Consistent with this notion, thinking about anegative (positive) event decreased (increased) re-ported satisfaction with one’s life as a whole in theexamples reviewed earlier, unless the temporal dis-tance of the event or the salience of the categoryboundary elicited its exclusion fi-om the represen-tation formed. In contrast, Parducci (1995; seealso Smith, Diener, and Wedell 1989) observedthat an extreme negative (positive) event increased(decreased) satisfaction with subsequent modestevents (see Kahneman, this volume, for a moredetailed discussion). Thus, the occasional experi-ence of extreme negative events facilitates the en-joyment of the modest events that make up thebulk of our lives, whereas the occasional experi-ence of extreme positive events reduces this enjoy-ment (for an extensive review, see Parducci 1995).

Hence, what we conclude about the impact ofextreme events on individuals’ subjective well-being will oficn depend on the measure we use.When we draw on self-reports of satisfaction withlife as a whole (a “wide” category), we are likely toobserve assimilation eKccts bccausc the extremecvcnt will b *c i n c l u d e d i n t h e reprcscntationfijrmcd, unless its exclusion is triggcrcd by one ofthe variables discussed carlicr. Accordingly, wewould conclude that the experience of extremepositive events increases, and the experience of ex-tremc negative events decreases, overall life satis-faction. As an alternative, however, WC’ may drawon nlomellt-t(~-mo11lent measures of hedonic ex-pcricncc, as suggested by Parducci ( 1995) (andothers), who proposed that happiness is “the bal-ance of pleasure over pain” (9). This approach isbased on evaluations of specific hedonic events,which are likely to show contrast cffccts. Accord-ingly, we would conclude that individuals had bet-ter avoid extremely positive experiences becausethey reduce the overall balance of pleasure overpain, whereas occasional negative experiences en-

Repotis of Subjective Well-Bein& 67

hance this balance (for recommendations on whichexperiences we should seek or avoid, see Parducci1995). Given that both sets of findings are reliablyreplicable, self-assessments of general well-beingand measures of moment-to-moment hedonic ex-perience are likely to diverge under many condi-tions.

Summaq In combination, the reviewed researchillustrates that the same life event may affect judg-ments of SWB in opposite directions, dependingon its use in the construction of the target “my lifenow” and of a relevant standard of comparison. Ittherefore comes as no surprise that the relation-ship between life events and judgments of SWB istypically weak. Today’s disaster can become to-morrow’s standard, making it impossible to pre-dict SWB without a consideration of the mentalprocesses that determine the use of accessibleinformation. Whereas the results of our experi-mental manipulations illustrate the power of theseprocesses, we know little about how people spon-taneously parse the stream of life events into dis-crete chunks. Exploring this issue provides apromising avenue for future research at the inter-face of autobiographical memory and social judg-ment.

What Mi@t Have Been: Counterfactuals

So far, we have seen that the way people thinkabout actual outcomes may influence their judg-ments of SWB (as well as their momentary mood,to be addressed in a later section). In a differentline of research, it has been observed that themental construction of fictitiuus outcomes ma!haire similar efects.

Assume that for some trivial reason a personmisses the plane for which she had a reservation,and then learns that this very plane has fatallycrashed. Although everybody who was not on theplane has reason to be relieved, this person is morelikely to experience relief because she had almostbeen a victim. Conversely, assume a car driver hasto endure a long wait at a construction site andmisses an important business appointment. Thisdriver may experience anger and self-blame if hefocuses on the possibility that he wozild have ar-rived in time had hc not deviated from his usualroute to work. In both cases, what might have beenserves as a standard that influences the assessmentof the actual event.

The antecedents and consequences of our con-struals of “what might have been” (Roese and

Olson 1995a) have been investigated in a researchprogramti on- counterfactual thinking (Kahnemanand Miller 1986; Miller, Tumbull, and McFarland1990; Roese 1997; Roese and Olson 1995b; Wellsand Gavanski 1989). This work is based on theinsight that when outcomes deviate from norms orexpectancies, people construct the normative out-come as an alternative (Kahneman and Miller1986). The likelihood of this construction de-pends on the ease with which an actual abnormalevent can be mentally converted into the counter-factual normative outcome. If only a minor aspectof the actual outcome needs to be altered, coun-terfactual thinking is more likely than if some fi.m-damental components need to be changed. Forexample, our lucky passenger should be morelikely to see herself as a potential victim of theplane crash if the change of reservation was aspontaneous decision, made immediately beforeboarding the plane, than if it was a deliberate ac-tion taken several days earlier. As a consequence,more relief would be experienced in the formerthan in the latter case.

Counterfactual thinking can influence affect andsubjective well-being in several ways (see Roese1997; Roese and Olson 1995b). First, the mentalconstruction of the normative outcome provides astandard of compnrisorJ against which the actualoutcome can be evaluated, resulting in contrast ef-fects. This is more likely the easier it is to con-struct the counterfactual. For example, winners ofOlympic bronze medals reported being more satis-fied than silver medalists (Medvec, Madey, andGilovich 1995), presumably because for winnersof bronze medals, it is easier to imagine havingwon no medal at all (a “downward counterfac-tual”), while for winners of silver medals, it is eas-ier to imagine having won the gold medal (an“upward counterfactual”). As a further conse-quence of this comparison, bronze medalists canbe expected to experience more joy, while silvermedalists may be more likely to experience disap-pointment.

Second, counterfactual thinking may suggestspecific causal implications that influence judg-ments and affective experiences (Roese 1997). Asobserved in numerous studies (for a review, seeWeiner 1985), causal attributions determine spe-cific emotions. It is therefore not surprising thatdifferent explanations of why the abnormal out-come occurred, rather than its normative altema-tive, may elicit different reactions. For example, asurprising failure is more likely to raise hope forfUture improvement if it is attributed to transient

3

6 8 Well- Bein.

circumstances rather than to stable personality FIGURE 4.1 Asymmetries in Feature Comparisoncharacteristics (Boninger, Gleicher, and Strathman1994). Past - Present

Finally, our feelings may be influenced by thevery act of explaining the abnormal outcome. Spe-cifically, counterfactual thinkers may ruminate

A - - - - - - - -->B --w-w------)

about the cause of the abnormal event and men- C we---w--m.--)tally try to undo what has happened. Continuous 0 (---w--w-> 0thoughts of “if only . . .” and “why me . . .” may E (m------->

Edominate the person’s cognitive activity and lead F

(-----m-->F

to self-pity and depression (see Martin and Tesser(---------

G1989). Such a prevalence of counterfactual rumi-nation may result from two related aspects ofcounterfactual reasoning. On the one hand, coun-

(-----w--- I( - - - - w - w - - K

terfactual thinking is most likely when the norma-tive outcome is easily constructed; hence, thecounterfactual may intrude on people’s thinkingwhenever the abnormal outcome comes to mind.At the same time, the superficial aspects of an ab-normal event that elicit its counterfactual alterna-tive are rarely the best candidates for a causalexplanation. Thus, by directing attention to thenormative outcome, aspects of the abnormal eventdeny themselves as plausible causes.

In summary, judgments of SWB can be pro-foundly influenced by mental constructions ofwhat might have been. Hence, the impact of agiven event will be more pronounced the easier itis to imagine that things could have turned outotherwise.

Direction of Comparison

So far, we have reviewed diKerent intraindividualstandards of comparison-pertaining to what is,was, will be, or might have been-and focused onthe proccsscs that dctcrminc whether a given pieceof information is used in fi)rming a rcprcscntationof the target or of the standard. Next, we needto consider an additional, and somewhat counter-intuitive, complication. On logical grounds, weshould assume that comparing X to Y results inthe same outcome as comparing Y to X. For cx-ample, when our present situation (X) is bcttcrthan our past situation (Y), WC should bc plcascdno matter whcthcr we compare the present to thepast or the past to the prcscnt. Yet the specific in-k)rmation we actually draw on is likely to differ inthese two cases, resulting in diKerent outcomes.

This possibility is suggested by Tversky’s ( 1977;Tversky and Gati 1978) research into similarityjudgments and has recently been confirmed forcomparison processes (Dunning, Madcy, and Par-pal 1995; Wtinkc, Schwarz, and No&-Neumann

1995; Schwarz, k%.nke, and Bless 1994). Suppose,for example, that a respondent’s representation ofthe past includes features A through F, as shownin figure 4.1, whereas her representation of thepresent includes features D through K.

According to Tversky’s (1977) model of sim-ilarity judgments, a comparison of the past to thepresent would involve the respondent’s assessmentof whether features A through F are also part ofthe present. The features G through K, which arepart of the present but not of the past, are likely toreceive little attention in this case. Conversely, acomparison of the present to the past would bebased on the features D through K. However, thefeatures A through C, which characterize the pastbut not the present, \vould go largely unnoticed.As a result, the outcome of the comparison pro-cess would diKer, depending on whether we com-pared the past to the present or the present to thepast.

Such judgmental asymmctrics arc particularlypronounced when the to- be-compared targets arerepresented in diccrcntial detail (Srull and Gaelick1984; Tvcrsky 1977). For csamplc, Dunning et al.(1995) suggested that people may possess a richarray of information about the present that theymay have forgotten about the past. If so, our rcp-rcscntation of the present ivould include a largerset of unique features than our representation ofthe past. Hence, \ve should dctcct more uniquefeatures when comparing the present to the past,rather than the past to the present, and thus con-cludc that more change has occurred in the formerthan in the latter. Dunning, Madcy, and Parpal‘s(1995) results confirmed this prediction.

While such findings alert us to the impact of dif-ferenccs in question \\*ording (for a methodologi-cal discussion, see Winke et al. 1995), they alsosuggest sonic troublesome (but as yet untested)

Repovts of Subjective Well-Bein& 69

implications for the comparisons we are likely tomake spontaneously. Of course, in making ourown spontaneous assessments, we may, in princi-ple, use either direction of comparison. In mostcases, however, our spontaneous attempts to assessthe quality of life are likely to be triggered bysome current problem. If so, the current problem-atic situation is in the focus of our attention, mak-ing it likely that we compare the current situationto some previous (or counterfactual) state of af-fairs, rather than vice versa. Owing to the logic ofthe comparison process, the outcome of this en-terprise is bound to be negative: chances are thatour current problem is not a feature of our past.Other problems that we had in the past, however,are unlikely to be considered because the consid-eration of features of the past is constrained by thefeatures that make up our representation of thepresent. Accordingly, the problems of the past mayescape our attention, contributing to the impres-sion that the past was the time of the “good olddays” (for a more detailed discussion, see Schwarzet al. 1994).

The Outcome of Comparisons: The DifieventialImpact of Losses and Gains

Finally, let us turn to the outcomes of the compar-isons we make. Whichever of the above standards\ve use, the comparison may tell us that our actualsituation either falls short of the chosen standardor exceeds it. Unfortunately, the former observa-tion is likely to have a more pronounced impacton judgments of SWB than the latter, reflecting ageneral tendency to give more weight to perceivedlosses than to gains. This is particularly likely whenwe make intraindividual comparisons across time,but it has also been obsenlcd for comparisons withothers (Rrandstitter 1998). As described in Kahne-man and Tvcrsky’s (1979) prospect thcocy, the\~luc function f&- losses is steeper than the valuefimction for gains. Hence, gains and losses of anequal magnitude may not result in “zero netcl1mg.” Rather, the steeper value function forlosses implies, for example, that a $100 increase inrent, which constitutes a loss relative to the refer-cncc point of one’s previous rent, has a higher im-pact on one’s subjective sense of economic well-being than an apparently equivalent pay raise of$100, constituting a gain relative to one’s previousincome. As a result, the net efect of both changesivould not be neutral but negative. Accordingly,the gains must far exceed the losses to result in anoverall sense of improvement, and relatively large

improvements may be offset by comparativelysmaller losses.

Again,-however, the specific outcome is likely todepend on the mental representations formed. Ifthe wording of the judgment task induces respon-dents to balance their separate mental accounts(Thaler 1985) for rent and income prior to evalua-tion of the net result, they may indeed perceivezero change. Thus, the parsing of reality into dif-ferent chunks is again likely to affect the judgmen-tal outcome, as we have seen in the preceding dis-cussion (see Schwarz et al. 1994).

What Gets Lost= Duration Neglect

In combination, the discussion in the precedingsections suggests that nearly any aspect of one’slife can be used in constructing representations ofone’s “life now” or a relevant standard, resultingin many counterintuitive findings. Sometimes,however, the surprises do not result from what isused in which way, but from what is neglected.

Common sense suggests that misery that lastsfor years is worse than misery that lasts only for afew days. Hence, the evaluation of a given episodeshould depend not only on the episode’s hedonicvalence but also on its duration. Recent researchsuggests, however, that people may largely neglectthe duration of the episode, focusing instead ontwo discrete data points, namely, its most intensehedonic moment (“peak”) and its ending (Fred-rickson and Kahneman 1993; Varey and Kahne-man 1992). Hence, episodes whose worst (orbest) moments and endings are of comparable in-tensity are evaluated as equally (un)pleasant, inde-pendent of their duration (for a more detailed dis-cussion, see Kahneman, this volume).

Although the a\Glable data are restricted to epi-sodes of short duration, it is tempting to speculateabout the possible impact of duration neglect onthe evaluation of more extended episodes. If dura-tion neglect applies to extended episodes, we mayexpect, for example, that three years of economichardship may not seem much Lvorse in retrospectthan one year, provided that the peak and endvalues of both episodes are comparable. In addi-tion, ale may speculate that the level of hardship atpoints other than the peak and the end may proveirrelevant as well. By the same token, the degree ofvariation ivithin an episode should prove largely ir-relevant when the changes occur gradually and arenot marked by salient events. On the other hand,if the changes are pronounced, or are marked bysome salient event, the episode may be broken

I

70 Well- Bein.

down into a series of shorter episodes, with eachone having its own peak and end. Moreover, ret-rospective evaluations should crucially depend onthe hedonic value experienced at the end of therespective episode. Thus, a period of ten years ofscarcity may benefit from some improvement inthe final year to a much larger extent than the rel-ative durations would seem to justiQ, whereas adecline at the end may cloud longer periods of rel-ative well-being. Assuming some variation overtime, the hedonic value of the end of the episodeis likely to depend on the specific boundary cho-sen, which may be a fimction of other, rather for-tuitous events, including the context provided inthe research situation. Accordingly, not only maythe choice of category boundaries determine whatwe include in the representation of the respectiveepisode, as discussed earlier, but the chosen end ofthe episode may also determine what will be givenspecial weight in evaluating the episode as awhole. Unfortunately, the limited data available donot yet allow us to assess these possibilities.

Summary

As our selective review illustrates, judgments ofSWB are not a direct function of one’s objectiveconditions of life and the hedonic value of one’sexperiences. I&her, they crucially depend onthe information that is accessible at the time ofjudgment and how this information is used inconstructing mental representations of the to-be-evaluated episode and a relevant standard. Thisstandard may reflect previous states of aKairs (whatwas), expectations about the fUture (what will be),counterfactual alternatives (what might havebeen), or the lot of others (to be addressed later).As we have seen repeatedly, how individuals parsethe stream of life into discrete units determineswhether the event is included in the episode, re-sulting in an assimilation erect, or excluded fromthe episode, resulting in a contrast effect. More-over, the direction of comparison chosen, or sug-gested by the wording of the question, influenceswhich features are likely to be considered. Onefeature that is likely to be neglected in retrospec-tive evaluations is the duration of the episode, rc-fleeting reliance on a peak-and-end rule. Finally,the perception that one’s current situation fallsshort of the standard is likely to have a more pro-nounced impact than the perception that it ex-ceeds the standard to the same degree, reflectingthat losses loom larger than gains.

As a result of these construal processes, judg-

ments of SWB are highly malleable and difficultto predict on the basis of objective conditions._ --Hence, it-is not surprising that the relationship be-tween the objective conditions of life and theirsubjective evaluation is weak and often counterin-tuitive. Theoretically, we may expect that this rela-tionship is more pronounced, and more straight-foxward, in a person who is preoccupied with acurrent concern, such as a severe illness. This con-cern would presumably be chronically accessible inmemory and would hence come to mind indepen-dent of whether it has been addressed in precedingquestions. Moreover, it would be likely to be in-cluded in the representation formed of one’s cur-rent situation, reflecting its numerous links toother aspects of daily life. Even under these condi-tions, however, the current concern may be delib-erately disregarded, for example, when its repeatedconsideration would violate norms of conversa-tional conduct. Moreover, the evaluation wouldstill shift as a function of the standard of compari-son used, as research into social comparison illus-trates. We turn to this work next.

USING INFORMATION ABOUT OTHERS:SOCIAL COMPARJS~NS

Obviously, the range of potentially relevant stan-dards is not restricted to those aspects of one’sown life that pertain to what was, will be, or mighthave been, all of which may serve as intuaindi-vidual standards. Rather, interindividual stm-dards provided by information about others’ livesmay have similarly pronounced effects on judg-ments of SWB. In this section, we address differ-ent interindividual standards and the determinantsof their use in real life and in research situations.

Choosing Comparison Others: Downward,Upward, and Lateral Comparisons

Not surprisingly, we may feel better about ourlives when \vc compare ourselves to others whoarc less \\*ell off (a downward compc~ison) thanwhen we compare ourselves to others who are bet-tcr off (an npwnrd comparison). In fact, the morepeople assume that their own living conditions arcbetter than those of others, the more satisfactionthey report (see Campbell et al. 1976; Carp andCarp 1982), although such correlational findingsdo not unequivocally bear on the causal role ofcomparison processes. However, the causal impactof comparison processes has been well supported

Reports of Subjective Well-Being 71

in laboratory experiments that exposed respon-dents to relevant comparison standards, further il-lustrating that respondents are likely to draw onwhatever information is most accessible at thetime of judgment (for reviews, see Miller andPrentice 1996; Wills 1981; Wood 1989). For ex-ample, Strack and his colleagues (1990) observedthat the mere presence of a handicapped confeder-ate was sufficient to increase reported SWB underself-administered questionnaire conditions, pre-sumably because the confederate served as a salientstandard of comparison. Consistent with this ac-cessibility principle, numerous studies found thattemporarily accessible standards can overridechronically accessible standards (for a review, seeMiller and Prentice 1996). For example, mostpeople are presumably very familiar with societalstandards of physical attractiveness. Nevertheless,exposing research participants to photographs ofhighly attractive women has been found to de-crease women’s self-assessments of their ownphysical attractiveness (Cash, Cash, and Butters1983), as well as men’s satisfaction with the at-tractiveness of their romantic partner (Kendrickand Gutierres 1980).

However, recent naturalistic studies suggest amore complicated picture (for a review, see Taylor,Wayment, and Carrillo 1996). Under uncon-strained conditions, respondents may engage indownward, upward, or lateral comparisons; more-over, the impact of any comparison standard maychange over time and affect different dependentvariables in different ways. These complicationssuggest that self-initiated social comparisons mayserve a variety of different fimctions.

SelfAssessmeut First, social comparisons maysenre a self-assessment fimction, as initially pro-posed by Fcstinger (1954), who assumed that as-sessments of one’s own abilities and outcomes arebest sealed by comparisons with similar others(lateral comparisons). Speci@ng what exactly de-termines whether another is suficiently similar tosc~c as a relevant comparison other has been oneof the vaguest points of social comparison the-ory- a n d indeed, the accessibility principle illus-trated earlier guarantees that relatively dissimilar,but highly salient, others may often be chosen. Ingeneral, however, “given a range of possible per-sons for comparison, someone who should beclose to one’s own performance or opinion, givenhis standing on characteristics related to and pre-dictive of performance or opinion, will be chosenfor comparison” (Goethals and Darley 1977,

265). This “related attributes” hypothesis is em-pirically.-well- supported (for a review, see Millerand Prentice 1996), although it is ofien difficult tospecify a priori which attributes will be considered“relevant and predictive.”

Self-Enhancement Second, social comparisonsmay sexve a self-enhancement function, which ismost easily satisfied by downward comparisonswith someone who is less well off (Wills 1981), asseen in Strack and his colleagues’ (1990) findingthat the mere presence of a handicapped confeder-ate may increase reports of SWB. Note, however,that such downward comparisons should be com-forting only when we can assume that the other’sunfortunate state does not provide a glimpse atour own future. A person who has been diagnosedas HIV-positive, for example, may derive littlecomfort from exposure to a person with advancedAIDS. Hence, the outcome of downward compar-isons depends on the perceived mutability andcontrollability of the relevant outcome, as well asthe time frame employed and the individual’ssense of self-esteem (see Major, Testa, and Bylsma1991; Taylor et al. 1996). If the outcome is mut-able and controllable, and one’s own self-esteemsuggests one has the necessary skills, downwardcomparisons do indeed increase an individual’ssense of SWB. If the outcome is uncontrollable, orone perceives a lack of relevant skills, downwardcomparisons may be comforting only in the shortterm and in fact may elicit a sense of despair aboutthe likely fixture development. Much as we haveseen for the impact of information about one’sown life, it is not the information about theother’s situation per se that determines the out-come, but the use of this information in construct-ing representations of one’s ofvn present or futuresituation and a relevant standard.

Moreover, researchers may have overestimatedthe prevalence of downward comparisons, as Tay-lor, Wayment, and Carrillo (1996) noted. Al-though people typically report that they are betteroff than others, even under very unfortunate cir-cumstances (Taylor and Brown 1988), more de-tailed investigations suggest that these reports maybe based on comparisons with manufactured hype-theticnl others rather than on comparisons with ac-tual individuals (Taylor, Wood, and Lichtman1983), with whom contact is of?en avoided (Tay-lor and Lobe1 1989).

Se!f- Improvement As a third function, socialcomparisons may sewe self-improvement goals,

72 Well-Bein

which are best satisfied by upward comparisonswith individuals who are better off and whose suc-cess may provide relevant performance informa-tion. Early research concluded that the potentialfor long-term self-improvement comes at the priceof short-term dissatisfaction because the upwardcomparison highlights one’s own shortcomings(Morse and Gergen 1970; Salovey and Rodin1984). Confirming this conclusion, Wayment,Taylor, and Carrillo ( 1994) observed in a longi-tudinal study that college freshmen who engagedin upward comparisons felt worse over the shortterm. Four months later, however, these freshmenwere better adjusted to college life than those whodid not engage in upward comparisons, suggestinga positive long-term effect of the actual self-im-provement facilitated by upward comparisons.

Moreover, the impact of upward comparisonsdepends on how close and similar the comparisonother is and on how self-relevant one considers therespective performance dimension (Tesser 1988;for a recent review, see Tesscr and Martin 1996).If a close and similar other, such as a good friend,outperforms us on a self-relevant attribute, thecomparison results in dissatisfaction and with-drawal from the friend. If the attribute is not self-relevant, however, we may take pleasure in thefriend’s achievement. Finally, highly dissimilarothers may not be perceived as relevant compari-son standards and may hence not pose a particular-threat, independent of the self-relevance of thecrucial attribute. We conjecture that this diReren-tial impact of similar and dissimilar others rctlects,in part, the processes we discussed in the sectionon counterfactuals: the more similar the other is,the easier it is to imagine that we might have ob-tained a similar outcome, yet we didn’t.

AJZintion As a final timction, social compari-sons may serve af-liliativc needs, as initially pro-posed by Schachtcr ( 1959). Recent naturalisticstudies (for example, Helgcson and Taylor 1993;Taylor and Lobe1 1989; Ybema and Buunk 1995)suggest that “people may compare themselveswith others sharing a similar fate not only to eval-uate their own emotional experiences, but also tocreate the cxpcrience of social bonding and com-ti)rt that arise from the observation of a sharedfate” (Taylor et al. 1996, 5). These con&rts ma)mitigate the othenvise expected impact of evalua-tive comparisons.

Summnly As this discussion indicates, the im-pact of social comparison processes on SWB is

n

more comples than early research suggested. Asfar as judgments of global SWB are concerned, wecan- expect that exposure to someone who is lessLvell off will usually result in more positive-andto someone who is better off in more negative-assessments of one’s own life. However, informa-tion about the other’s situation will not always beused as a comparison standard. Rather, relevantinformation about the other’s situation may enterthe representation of one’s own future, for exam-ple, resulting in assimilation rather than contrasteffects. We therefore emphasize that knowing I&Oindividuals compare themselves to does not allowus to predict the impact of the comparison otheron individuals’ sense of SWB unless we know howthis information is used in the relevant mentalconstruals.

Stundurds Provided by the Social Environment

So far, our discussion of social comparison pro-cesses has had a distinctly individualistic and voli-tional favor, focusing on who we choose as com-parison others. This perspective needs to becomplemented by a consideration of the influenceof more stable aspects of our social environment.

First, our degree of freedom may of--ten be moreconstrained than experimental research suggests,and our immediate so&L environmevzt may forcestandards upon us that are difficult to ignore. Thishas been most consistently observed in researchthat addressed the impact of students’ standingwithin their school on their sense of self-esteem.Students with a given level of performance onstandardized tests have higher self-esteem whenthey are at a low-quality scl~ool, where many stu-dents do poorly, rather than at a high-qualityschool, \\*hcrc many students do well (Bachmanand O’hlallcy 1986; Marsh 1993; Marsh and Par-ker 1984). Although thcsc findings may in part r&tlcct that thcsc students arc likely to receive differ-cntial acknowledgment from their teachers, theyalso indicate that it is difficult to escape the normprovided by one’s environment by pursuing self-enhancement through the choice of comparisonothers \\*ho arc doing less well. In a similar vein,Morawctz ( 1977) obscmcd that citizens of a com-munity \\*ith a relatively equal income distributionreported higher well-being than citizens of a com-munity \\*ith an unequal income distribution, al-though the latter’s absolute level of income washigher. This finding at the community level is con-sistent \vith Easterlin’s ( 1974) conclusion that in-creasing levels of income within a given country

)I

Reports of Subjective Well-Being

are not related to increasing reports of life satisfac-tion. Instead, Easterlin’s findings suggested thatthe effect of income is largely relative, increasingone’s sense of well-being if one earns more thanothers (but see Diener and Suh, this volume, for areview of the contradictory evidence bearing onthis hypothesis). As a final example, Seidman andRapkin (1983) found that the usually obsemed in-crease in the prevalence of mental illness during aneconomic downturn was most pronounced in het-erogeneous communities, where the recession didnot affect everyone equally. In combination, thesefindings illustrate the power of highly accessiblestandards provided by one’s immediate environ-ment. Such standards presumably limit individuals’freedom in pursuing the comparison goals dis-cussed earlier. If so, we may be most likely to seedifferential construals of comparison standardswhen the judgment pertains to an attribute forwhich one’s environment provides a range of com-parison others with widely different standings, as istypical for the health-related research reviewed inthe preceding section. In contrast, judgments thatpertain to attributes on which one’s social environ-ment is homogeneous (as in the earlier examplesof unemployment and income) may be less opento dil++ercntial construal processes.

Second, an individual’s position in the socialstructure may influence \irhich comparison othershe or she deems relevant, as suggested by refer-c11cc group theory (Hyman and Singer 1965). Forexample, Runciman ( 1966) noted that Britishworkers strong sc11sc of social class constrainedthe range of jobs they considered relevant in mak-ing income comparisons to a larger degree thanwas the case for American \\vorkcrs, at least in the1960s. Thus, self-categorizations with regard toclass or other relatively stable social attributes ma>constrain the range of comparison others to nicm-bcrs of the same, or closely related, categories. Im-portantly, these self-categorizations are likely tochange in cases of social mobility, resulting inchanges in the comparison group deemed rele-vant. Such changes in the comparison standardmay lc~d to decrcascd satisfaction despite irn-proved objcctiirc circumstances (see Frederick andI,oc’\\.cnstciri, this voli~n~c). Several researcherssupgcstcd, for cxamplc, that objective impro\re-ments in women’s situation in the workforce didnot result in increased satisfaction because the)were accompanied by an increase in the legitimacyof comparisons with IIICII, \\*ho arc still doing bet-ter (Elstcr 1983; Walstcr, Walster, and Berschcid1978).

Finally, socially shared norms may replace spe-cific comparison groups or individuals as relevantstandards, implying, for example, that ever)l citizenis entitled to certain outcomes. Although percep-tions of entitlement are themselves a function ofsocial comparisons (see Major 1994), they mayobliterate the need for specific comparison othersonce they are formed.

In combination, these examples draw attentionto the possibility that salient comparison standardsin one’s immediate environment, as well as sociallyshared norms, may constrain the impact of fortu-itous temporary influences. At present, the inter-play of chronically and temporarily accessible stan-dards on judgments of SWB has received littleattention. The complexities that are likely to resultfrom this interplay provide a promising avenue forfuture research.

Interindividual Standards Implied by theResearch Instrument

Finally, we extend our look at the influences of theresearch instrument by addressing a frequentlyoverlooked source of temporarily accessible com-parison information. In many studies, researchersassess respondents’ experiences, their objectiveconditions of living, or the frequency with whichthey engage in a certain behavior, by asking themto check the proper ans\ver on a list of responsealternatives provided to them. As an example, ta-ble 4.2 shows different response alternatives pre-sented as part of a question about daily televisionconsumption (Schwarz et al. 1985).

As numerous studies have indicated (for a re-view, see Schwarz 1996, ch. S), respondents as-SIIIIIC that the list of response alternatives reflectsthe researcher’s knowledge of the distribution ofthe behavior: they assume that the “abverage” or“usual” behavioral frequency is represented byvalues in the middle range of the scale, and thatthe extremes of the scale correspond to the ex-tremes of the distribution. Accordingly, they usethe range of the response alternatives as a frame ofreference in estimating their own behavioral fre-qLlcIlcy, resulting in different estimates of theirown behavioral frequency, as shown in table 4.2.More important for our present purposes, theyfurther extract comparison information from theirow11 location on the scale. Checking “hvo and ahalf hours” on the low-frequency scale suggeststhat one’s own television consumption is above av-erage, whereas checking the same television con-sumption on the high-frequency scale suggests it is

I74

TABLE 4.2 Reported Daily Television Consumption and Leisure Time Satisfaction as a Functionof Response Alternatives

Low-FrequencyAlternatives (Percentage)

Reported daily television consumptionUp to half an hourHalf an hour to one hourOne hour to one and a half hoursOne and a half hours to two hoursTwo hours to two and a half hoursMore than two and a half hours

Leisure time satisfaction

11.526.926.926.9

7.70.0

9.6

Hi&&FrequencyAltema tives (Percentage)

Up to 2 1/2h 70.42 1/2h to 3h 22.23h to 3 1/2h 7.43 1/2h to 4h 0.04h to 4 1/2h 0.0More than 4 1/2h 0.0

8.2

SOWCC: Adapted from Schwarz et al. (1985, Experiment 2). Reprinted with permission from The University ofChicago Press.

below average. Hence, respondents in this studyreported lower satisfaction with the variety ofthings they do in their leisure time when the low-frequency scale suggested they watch more televi-sion than others than when the high-frequencyscale suggested they watch less-despite the factthat the former respondents reported watchingless television to begin with (see table 4.2).

Similar findings have been obtained with regardto the frequency of physical symptoms and healthsatisfaction (Schwarz and Schcuring 1992), thefrequency of sexual behaviors and marital satis-faction (Schwarz and Schcuring 1988), and var-ious consumer behaviors (Mcnon, Raghubir, andSchwarz 1995). In combination, they illustratethat response alternatives convey highly salientcomparison standards that may profoundly affectsubsequent evaluative judgments. Researchers aretherefore well advised to assess information aboutrespondents’ behaviors or objective conditions inan open-response f&nat, thus avoiding the intro-duction of comparison information that respon-dents would not draw on in the absence of theresearch instrument.

Summary

In summary, the use of intcrindividual comparisoninformation follows the principle of cognitive ac-cessibility that WC have highlighted in OLIN discus-sion of intraindividual comparisons. Individuals of-ten draw on the comparison information that isrendered temporarily accessible by the research in-strument or the social context in which they formthe judgment, although chronically acccssiblcstandards may attenuate the impact of temporarilyaccessible information. Despite this caveat, the sc-

lection of comparison standards is not solely deter-mined by relatively stable attributes of the respon-dant that may be expected to change only slowlyover time, such as reference group orientation (Hy-man and Singer 1968; Runciman 1966), adapta-tion level (Brickman and Campbell 1971), or aspi-ration level (Michalos 1985). Rather, individualsconstruct a relevant social comparison standardbased on the information that is most accessible atthe time of judgment. Moreover, these construc-tions may reflect different goals, including self-assessment, self-enhancement, self-improvement,or affiliation. Which of these goals is being pur-sued at a given point in time is likely to be itselfcontext-dependent, rendering general predictionsdifficult.

THE IMPACTOFMOOD STATES

In the preceding sections, we considered how re-spondcnts USC information about their own livesor the lives of others in comparison-based evalua-tion strategies. Howcvcr, judgments of well-beingarc a function not only of what one thinks aboutbut also of how one fir/r at the time of judgment.A wide range of experimental data confirms thisintuition. Finding a dime on a copy machine(Schwarz 1987), spending time in a pleasantrather than an unpleasant room (Schwarz ct al.1987, Expcrimcnt 2), or watching the Germansoccer team lvin rather than lose a championshipgame (Schwarz et al. 1987, Experiment 1) all re-sulted in increased reports of happiness and satis-faction with one’s life as a whole.

Two diRerent processes may account for theseobservations. On the one hand, it has been shown

Reports of Stibjective Well-Being 75

that moods may increase the accessibility ofmood-congruent information in memory (for re-views, see Blaney 1986; Bower 1981; Morris, thisvolume; Schwarz and Clore 1996). That is, indi-viduals in a happy mood are more likely to recallpositive information from memory, whereas indi-viduals in a sad mood are more likely to recall neg-ative information. Hence, thinking about one’s lifewhile in a good mood may result in a selectiveretrieval of positive aspects of one’s life, and there-fore in a more positive evaluation.

On the other hand, the impact of moods maybe more direct. People may assume that their mo-mentary well-being at the time of judgment is areasonable and parsimonious indicator of theirwell-being in general. Hence, they may base theirevaluation of their life as a whole on their feelingsat the time of judgment and may evaluate theirwell-being more favorably when they feel goodrather than bad. In doing so, laypeople follow thesame logic as psychologists who assume that one’smood serves as a “barometer of the ego” (Jacob-sen 1957) that reflects the overall state of the or-ganism (Ewert 1983) and the countless experi-ences one goes through in life (Bollnow 1956). Infact, when people are asked how they decidewhether they are happy or not, most of them arelikely to refer explicitly to their current affectivestate, saying, for example, “Well, I feel good”(Ross et al. 1986).

Experimental evidence supports this assumption.For example, Schwarz and Clore (1983, Experi-ment 2) called respondents on sunny or rainy daysand assessed reports of SWB in telephone inter-views. As expected, respondents reported being in abetter mood, and being happier and more satisfiedwith their life as a whole, on sunny rather than onrainy days. Not so, however, when respondents’attention was subtly drawn to the weather as a plau-sible cause of their current feelings. In one condi-tion, the interviewers pretended to call from out oftown and asked as a private aside, “By the way,how’s the weather down there?” Under this condi-tion, respondents interviewed on rainy days re-ported being as happy and satisfied as respondentsintenicwed on sunny days. In addition, a measureof current mood, assessed at the end of the inter-viciv, was not af5ectcd by the attention manipula-tion, indicating that the weather question did notal%ect respondents’ current mood itself but onlytheir inferences based upon it. Accordingly, themood measure was more strongly correlated withreported SWB when the weather was not men-tioned than when it was mentioned.

These and related findings (see Keltner, Locke,and Audrain 1993; Schwarz 1987; Schwarz andClore 1-983, Experiment 1) demonstrate that re-spondents use their affective state at the time ofjudgment as a parsimonious indicator of theirwell- being in general, unless the informationalvalue of their current mood is called into question.Moreover, the discounting effects (Kelley 1972)obtained in these studies rule out an alternativeexplanation based on mood-congruent retrieval.According to this hypothesis, respondents may re-call more negative information about their lifewhen in a bad mood rather than a good mood,and may therefore base their evaluation on a selec-tive sample of data. Note, however, that the im-pact of a selective database should be independentof respondents’ attributions for their currentmood. Attributing one’s current mood to theweather discredits only the informational value ofone’s current mood itself, not the evaluative impli-cations of any positive or negative events one mayrecall. Inferences based on selective recall shouldtherefore be unaffected by salient explanations forone’s current feelings. Thus, the reviewed datademonstrate that moods themselves may serve in-formative functions according to a “How do I feelabout it?” heuristic, a hypothesis that has receivedconsiderable support in different domains of judg-ment (for a review, see Schwarz and Clore 1996).

When Do People Rely on Their Mood RatherThan on Other Inforvnation?

The observation that individuals may evaluatetheir well-being either on the basis of (intra- orinterindividual) comparisons or on the basis oftheir momentary feelings raises an obvious ques-tion. Under \i*hich conditions will they rely on onerather than the other source of information?

General Life Satisfaction Versus Specific Life Do-mains On theoretical grounds, we may assumethat people are more likely to use the simplifyingstrategy of consulting their affective state the moreburdensome it would be to form a judgment onthe basis of comparison information. Note in thisregard that evaluations of general life satisfactionpose an extremely complex task that requires alarge number of comparisons along many dimen-sions with ill-defined criteria and the subsequentintegration of the results of these comparisons intoone composite judgment. Evaluations of specificlife domains, on the other hand, are often lesscomplex. In contrast to judgments of general life

Well-Being

satisfaction, comparison information is usuallyavailable for judgments of specific life domains,and criteria for evaluation are well defined. An at-tempt to compare one’s income or one’s “life as awhole” with that of colleagues aptly illustrates thedifference. For these reasons, judgments of do-main satisfaction may be more likely to be basedon inter- and intraindividual comparisons, whereasjudgments of one’s life as a whole may be basedon one’s momentary feelings. Supporting this rea-soning, the outcome of the 1982 championshipgames of the German national soccer team af-fected respondents’ general life satisfaction, butnot their satisfaction with work and income(Schwarz et al. 1987, Experiment I).

If judgments of general well-being are based onrespondents’ affective state, whereas judgments ofdomain satisfaction are based on comparison pro-cesses, it is conceivable that the same event mayinfluence evaluations of one’s life as a whole andevaluations of specific domains in opposite direc-tions. For example, an extremely positive event indomain X may induce good mood, resulting in re-ports of increased global SWB. However, the sameevent may also increase the standard of compari-son used in evaluating domain X, resulting injudgments of decreased satishction with this par-ticular domain. Again, experimental evidence sup-ports this conjecture. In one study (Schwarz ct al.1987, Experiment 2), students were tested in ci-ther a pleasant or an unpleasant room, namely, afriendly ofice or a small, dirty laboratory that wasoverheated and noisy, with flickering lights and abad smell. As expected, participants reported lowergeneral life satisfaction in the unpleasant roomthan in the pleasant room, in lint with the moodsinduced by the expcrimcntal rooms. In contrast,they rcportcd higher housing sntisfi~ction in theunpleasant than in the plcasant room, consistent

with the assumption that the rooms scnfcd as sa-lient standards of comparison.

In summary, the same event may intlucncc judg-ments of general life satisfaction and judgments ofdomain satisfaction in opposite directions, rctlcct-ing that the former judgment is based on themood elicited by the event whereas the latter isbased OJI a comparison strategy. This dilGrcntia1impact of the same objective event fl~rthcr contrib-utes to the weak relationships between global andspecific evaluations, as well as measures of objcc-tive circumstances, that we addressed earlier.

~The Relative Salience of Mood and Competing In-formation Finally, we return to the impact of rc-

called life events on judgments of SWB. In thesection on intraindividual comparison processes,we mentioned that the same event may result inassimilation as well as contrast effects, dependingon whether it is used to construct a representationof the target or a standard. These processes arefLther complicated by the degree to which the re-call task is emotionally involving. In the absence ofemotional involvement, the impact of recalledevents follows the mental construal logic describedearlier. If recalling a happy or sad life event elicits ahappy or sad mood at the time of recall, however,respondents are likely to rely on their feelingsrather than on recalled content as a source of in-formation. This overriding impact of current feel-ings is likely to result in mood-congruent reportsof SWB, independent of the mental construal vari-ables discussed earlier.

The best evidence for this assumption comesfrom experiments that manipulated the emotionalinvolvement that subjects experienced while think-ing about past life events. In one experiment(Strack et al. 1985, Experiment 2), subjects wereasked either to give a short description of only afew words or to provide a vivid account of one totwo pages in length. In another study (Strack et al.1985, Experiment 3), subjects had to explain“Why” the event occurred, or “how” the eventprocecdcd. Explaining why the event occurred orproviding a short description did not aff‘ect sub-jects’ current mood, whcrcas “how” descriptionsand vivid reports resulted in pronounced mooddiffcrcnccs bctwecn subjects who reported positiveand ncgativc cxpcricnccs.

Table 4.3 shows the results. When no pro-nounced mood state was induced, subjects re-potted higher SWH nficr recalling negative rather

TAIKE 4.3 Subjcctivc Well-being: The Impact otStyle of Thinking

Vdtwcc IQ’ Event

Positivt! Neya tiw

Dctsilcci description 9.1 7 . 9

Short description 6.8 8 . 4

“ How” ctcscription 8.2 6 . 3

“Why” description 7.8 8.9

Sunrcc: Copyright 1085 by the Amcricm Psychological Awxi-xion. AdaptcJ from Stnck ct d. ( 1985, Expwimcnts 2 and 3).Nuts: For rhc nxan score of happiness atd .sxisGction qucs-tions, rhc tmgc is 1 m 1 1, with higher vdws indic;lting rcyott.sof highcr well-lxing.

Repotis of Subjective Well-Being 77

than positive past events, thus replicating the con-trast effects discussed earlier (see table 4.1). Whenthe recall task did induce a pronounced moodstate, on the other hand, mood had an overridingeffect: in that case, subjects who had to describenegative past events reported lower well-beingthan subjects who had to describe positive pastevents, replicating the mood effects found in otherstudies. Subsequent experiments by Clark and hercolleagues (Clark and Collins 1993; Clark, Col-lins, and Henry 1994) provided conceptual repli-cations of these findings.

In combination with the research reviewed inthe section on information about one’s own life,these studies demonstrate that the impact of anevent is a joint function of its hedonic quality, vari-ables that govern the use of information in men-tal construals of the target and standard (such asthe event’s temporal distance or salient categoryboundaries), and the person’s emotional involve-ment while thinking about the event. That the re-lationship between objective events and subjectivewell-being is as weak as the subjective indicatorliterature has demonstrated is therefore not sur-prising. Knowing the hedonic quality of an eventdoes not allow a prediction of its impact on re-ported well-being in the absence of knowledgeabout other judgmental variables.

REPORTING THEJUDGMENT

Once respondents have formed a judgment, eitherbased on their mood or based on a comparisonprocess, they need to communicate it to the re-searcher. Self-presentation and social desirabilicconcerns may arise at the reporting stage, and re-spondents may At their private judgment beforethey communicate it (for a more detailed discus-sion, see Strack and Martin 1987; Sudman et al.1996, ch. 3). III general, social desirability influ-CIICCS are more pronounced in face-to-face inter-views than in telephone interviews and are of leastconcern under the confidential conditions of self-administered questionnaires (for a review, see De-Maio 1984). Consistent with this generalization,Smith ( 1979) obsemcd in a meta-analysis thathigher Lvell-being is reported in face-to-face inter-vie\\rs than in mail sun’cys.

Experimental research confirmed this finding(Strack et al. 1990) and indicated that self-presen-tation eff;cts are moderated by interviewer charac-teristics. Specifically, respondents reported higherwell-being in personal interviews than in self-

administered questionnaires. Moreover, this differ-ence was more pronounced Lvhen the interviewerwas of -the dpposite sex but was not obtainedwhen the interviewer was severely handicapped.Respondents apparently hesitated to tell someonein an unfortunate condition how great their ownlife was. In contrast, when the handicapped con-federate did not serve as an interviewer but waspresent in the room as another research participantfilling out his own questionnaire, his presence didincrease subjects’ reported SWB, presumably be-cause the handicapped confederate served as a sa-lient standard of comparison.

In summary, the available research indicates thatpublic reports of SWB may be more favorable thanrespondents’ private judgments. On the otherhand, individual differences in social desirabilityshow a weak relationship with measures of SWB(Y = .20) (see Diener 1984). In combination, thissuggests that respondents’ editing of their reportsis more strongly affected by characteristics of theinterview situation than by individual differencesbehveen respondents.

A JUDGMENTMODELOFSUBJECTIVEWELL-BEING

Figure 4.2 summarizes the processes reviewed inthis chapter. If respondents are asked to reporttheir happiness and satisfaction with their “life as awrhole,” they are likely to base their judgment ontheir current affkctive state; doing so greatly sim-plifies the judgmental task. If the informationalivalue of their affective state is discredited, or iftheir affective state is not pronounced and otherinformation is more salient, they are likely to use acomparison strategy. This is also the strategy thatis likely to be used for evaluations of less complexspecific life domains.When using a comparison strategy, individuals

draw cm the information that is chronically ortemporarily most accessible at that point in time:lvhatever comes to mind first, and is relevant tothe judgment at hand, is most likely to be used,unless the conversational context renders the useof information that has already been “given” inad-equate. Whether information that comes to mindis used in constructing a representation of the tar-get “my life now” or a representation of a relevantstandard depends on the variables that govern theuse of information in mental construal (Schwarzand Bless 1992a; Strack 1992). Information that isincluded in the representation of the target results

78 Well-Being

c

FIGURE 4.2 A Judgment Model of Subjective Well-Being

Task

Global WB 4 I + Domain -specific WB

1Affective state No Retrieveinformative? * relevant

I information

I Yes

Use mood asinformation

Constructrepresentationof target andstandard

IEvaluate

Social aspects to 4be considered?

yes 4 ReportEdit “private”judgment

J.Report

in assimilation effects, whereas information that isused in constructing a standard results in contrasteffects. Hence, the same information may influ-ence judgments in opposite directions, dependingon its use in mental construal.

If the accessibility of information is due to tcm-porary influences, such as preceding questions in aquestionnaire, the obtained judgment is unstableover time and a different judgment will be ob-tained in a different context. On the other hand, ifthe accessibility of information reflects chronic in-fluences-such as current concerns or life tasks, orstable characteristics of the social environment-the judgment is likely to be less context-dependent. The size of context-dependent assim-ilation effects increases with the amount and

extremity of the temporarily accessible informationthat is included in the representation of the target,and it decreases with the amount and extremity ofchronically accessible information. Conversely, thesize of context-dependent contrast effects in-creases with the amount and extremity of the tem-porarily accessible information used in construct-ing a standard, and it decreases with the amountand extremity of chronically accessible informationthat enters this representation.

Finally, afier having formed a judgment on thebasis of comparisons or on the basis of their affec-tive state, respondents have to report their judg-ment to the researcher. At this stage, they need toformat their answer according to the response al-ternatives provided by the researcher, and they

Reports of Subjective Well-Bein

may or may not edit their report to conform tosocial expectations, depending on the nature ofthe situation.

METHODOLOGICAL IMPLICATIONS

Our review emphasizes that reports of well-beingare subject to a number of transient influences.Like other social judgments, they are best consid-ered constructions in response to particular ques-tions posed at a particular time. Although the in-formation that respondents draw on reflects thereality in which they live, which aspects of this re-ality they consider and how they use these aspectsin forming a judgment is profoundly influenced byfeatures of the research instrument.

Implications for Survey Research

The reviewed findings have profound methodo-logical implications. First, the obtained reports ofSWB are subject to pronounced question-order ef-fects because the content of preceding questionsinfluences the temporary accessibility of relevantinformation. Moreover, questionnaire design vari-ables, like the presence or absence of a joint lead-in to related questions, determine how respon-dents use the information that comes to mind. Asa result, mean reported well-being may differwidely, as seen in many of the reviewed examples.Moreover, the correlation between an objectivecondition of life (such as dating frequency) andreported SWB can run anywhere from Y = - .l toY = .6, depending on the order in which the samequestions are asked (Strack et al. 1988), suggest-ing dramatically diKerent substantive conclusions.

Second, the impact of information that is ren-dered accessible by preceding questions is attenu-ated the more the information is chronically acces-siblc (see Schwarz and Bless 1992a). Hence, apreceding question about the respondent’s healthis likely to affect respondents with minor or nohealth problems to a larger degree than respon-dents Lvith severe health problems; the latterLvould be likely to think of their health concernsindependent of the preceding question. Accord-ingly, the same question may affect dil?erent sub-sets of a sample to diRerent degrees.

Third, the stability of reports of SWB over time(that is, their test-retest reliability) depends on thestability of the context in which they are assessed.The resulting stability or change is meaningful

when it reflects the information that respondentsspontan&sly -consider because the same, or dif-ferent, concerns are on their mind at differentpoints in time. It is potentially misleading, how-ever, when it indicates that the research instru-ment is drawing attention to the same or differentaspects of the respondent’s life. In the former case,the influence of the research instrument may cloudthe impact of actual changes in other domains ofrespondents’ lives; in the latter case, it may sug-gest changes where none have occurred by ensur-ing that respondents draw on different aspects atdifferent points in time.

Fourth, in contrast to influences of the researchinstrument, influences of respondents’ mood atthe time of judgment are less likely to result insystematic bias. The fortuitous events that affectone respondent’s mood are unlikely to affect themood of many others. An exception to this ruleare events of national importance, such as the out-come of major international sports events (Schwarzet al. 1987), which may affect a larger segment ofthe population. Even the impact of these events,however, is unlikely to last for the whole durationof data collection, which usually extends over sev-eral days, if not weeks, for large-scale surveys.Hence, mood effects are likely to introduce ran-dom variation, nrhereas instrument effects intro-duce systematic bias relative to a population thathas not been exposed to the instrument, but tokvhich the findings are to be generalized.

Fifth, as our review indicates, there is no reasonto expect strong relationships between the objec-tive conditions of life and subjective assessments ofwell-being under most circumstances. To begin\\ith, manv aspects are not considered when mak-ing a judgment, although they would have a pro-nounced impact if they were. Moreo\ver, even ifconsidered, the same information can drive thejudgment in different directions, depending onhow it is used in the construal of targets and stan-dards. As we have seen repeatedly, today’s tragedycan be tomorrolv’s standard, depending on thevariables that determine its use in mental con-strual. Our analvsis does allow us, however, to cir-cumscribe the conditions under which strong rela-tionships should be observed.

Specifically, strong positive rehtiovzships behveena given objective aspect of life and judgments ofSWB are likely to emerge when most respondentsinclude the relevant aspect in the representationthat they form of their life and do not draw onmany other aspects. This is most likely to be the

c

80 Well-Being

case when (a) the target category is wide (“my lifeas a whole”) rather than narrow (a more limitedepisode, for example); (b) the relevant aspect ishighly accessible; and (c) other information thatmay be included in the representation of the targetis relatively less accessible. These conditions weresatisfied, for example, in the Strack, Martin, andSchwarz (1988) dating frequency study, in ivhicha question about dating frequency rendered thisinformation highly accessible, resulting in a cor-relation of Y = .66 with evaluations of the respon-dent’s life as a whole. Yet, as this example illus-trates, we would not like to take the emergingcorrelation seriously when it reflects only the im-pact of the research instrument, as indicated b!the fact that the correlation was r = - .l ivhcnthe question order was reversed.

If we \\*ant to alvoid misinterpretations of method.-effects as substantii-e effects in this as well as otherareas of psychological and social research, we needto team more about the cognitive processes thatunderlie the reports that our respondents provide.Perhaps the recent collaboration of survey meth-odologists and psychologists will advance ourknowledge of these important aspects of social re-search (for rcvieLvs of the current state of this field,see Schivarz, Groves, and Schuman 1998; Sudmanet al. 1996).

Which Measures At-e We to Use?

Similarly, strong negative refatiovd%ps bch\*een agiven objective aspect of life and judgments ofSWB are likely to emerge when most respondentsuse the relevant aspect in constructing a standardof comparison and do not draw on man!’ otheraspects in forming this representation. This is mostlikely to bc the case when (a) the target catcgoy isnarrow (for example, a relatively short cpisodc ofthe respondent’s life) rather than wide (“my life asa whole”); (b) the rclcvant aspect is highly acccs-siblc; and (c) other information that may bc usedin constructing a standard is rclativcly less acccssi-blc. Thcsc conditions would bc satisfied, ti)r CS;IIII-

ple, in a follow-up study of previously ii~~ci~iplo~uiworkers who arc asked to rcpott on their prc\iousepisode of uncmploymcnt ;IIX. subscqucntl!, c\*alu-ate their current SWR.

When tlic instrument dots not guide rcspon-dents’ t h o u g h t proccsscs, howcvcr, di!-fcrcnt rc-spondcnts arc likely to draw on diKcrcnt inti)rma-tion and to use the same inti)rmation in diff;rcntconstruals, resulting in the kvcak relationships bc-twccn objcctivc conditions and subjcctivc c\.alua-tions that arc typically obtained in suw’cy rcscarchinto SWB. Thcsc weak relationships arc thcrct;)rc anatural conscqucncc of tlic complcsity of the un-derlying judgmental proccsscs-and the occa-sional observation of strong relationships is causeti)r iiictlioclological suspicion.

By now, most readers have probably concludedthat there is little to be learned from self-reports ofglobal ii-ell-being. Although these reports do re-tlect subjectively meaningfill assessments, what isbeing assessed, and how, seems too contcxt-dependent to provide reliable information about apopulation’s well- being, let alone infi>rmation thatcan guide public policy (but see Arb&, this vol-ume, for a more optimistic take). As an alternativeapproach, scvcral rcscarchcrs have returned toBentham’s ( 1789/1948) notion of happiness asthe balance of plcasurc over pain (fi)r examples,SW Kahncman, t h i s vo lume ; Parducci 1995).1Clthcr than asking rcspondcnts to provide aglobal asscssmcnt of SWB, such an approachwould rclv on nioiiicnt-to-nionicnt mcasurcs oflicdonic cspcricncc. While the h e d o n i c expcri-cnccs asscsscd by thcsc measures arc thcmsclvcsdcpcndcnt on the context provided by rcspon-dents’ other lift cspcricnccs (see Parducci 1995),reporting one’s nionicntary hedonic state poses aless ti~rmidablc task than providing an evaluationof OIIC'S lift as a \\*liolc. Such momentary reportscan bc asscsscd with cspcricncc sampling methods(Stone, SliiKmnn, DC Vrics, this volume; Csik-szcntmihalyi and 1Voiig 199 1; ti)r a mctliodologicalrcvicw, XC Hormuth 1986), such as bccpcrs thatremind rcspondcnts at randomly chosen times toreport their ciirrcnt a#cctivc state. Such mcasurcsarc unlikely to corrclatc w*cll \!*itli global reports ofSWR, as l’arducci ( 1995, l-3) noted because tlics;uiic cvcnt is likclip to ;ifGct evaluations of other

Finally, it is worth noting that the contest cf- spccifc cvcnts and cvnluations o f cstcndcd cpi-fccts rcvicwcd in this chapter limit the compara- sodcs in opposite directions, as discussed earlier.bility of results obtained in diftircnt studies. Un- To what cxtcnt measures of momentary hedonic&)rtunatcly, this comparability is a key prcrequisitc state are likely to show strong and meaningful rc-for many applied uses of subjective social indica- lationships kvith objcctivc conditions of living istors, in particular their use in monitoring the sub- difficult to dctcrminc at the present stage of af-jective side of social change over time (for esam- fairs, yet optimism SCCIIIS warranted (see Kahnc-plcs, see Campbell 198 1; Glatzcr and Zapf 1984). man, this volume). Howcvcr, experience sampling

Reports of Subjective Well-Being 81

methods are considerably more expensive than therelatively cheap option of asking respondents toprovide global assessments of their lives as awhole. Hence, considerable methodological effortneeds to be invested before the use of these mea-sures in large-scale representative studies can bejustified.

REFERENCES

Andrews, F. M., and Whithey, S. B. (1976). Social indi-cators of well-being: Americans’ perceptions of l;f qual-ity. New York: Plenum.

Bachman, J., and O’Malley, l? (1986). Self-concepts,self-esteem, and educational experiences: The frogpond revisited (again). Journal of Personality and So-cial Ps?)chology, 50, 35-46.

Bentham, J. (1948). An introduction to the principles ofmorals and legislation. New York: Hafner. (Originallypublished in 1789)

Blaney, I? H. (1986). Affect and memory: A review. Psy-chological Rtilletin, 99, 229-46.

Bodenhausen, G. V., and Wyer, R. S. (1987). Socialcognition and social reality: Information acquisitionand use in the laboratory and the real world. In H. J.Hippler, N. Schwarz, and S. Sudman (Eds.), Social in-formation processin and survey methodolofly (pp. 6-4 1). New York: Springer-Verlag.

Bollnow, 0. F. (1956). Das Wesen der Stimmtrvgen (Thenature of moods). Frankfurt: Klostermann.

Boninger, D. S., Glcicher, F., and Strathman, A. (1994).Counterfactual thinking: From what might have beento \vhat may be. Journal of Personality and Social Py-dJOh&qV, 67, 297-307.

Bo\i’cr, G. H. (1981). Mood and memory. America11l’~~vhl~gist, 36, 129-48.

Bradburn, N. M. (1969) ‘The strzrcture of psycholocqicalwell-ixinq. Chicago: Aldinc.

Brandst3tt~r, E. ( 1998). Ambivalente ZtjLiedenheit: DcrEiz@s sosialer K&eicbe. (Am bivalrn t satisfaction:The impact of social comparisons). Muenster: Was-mdnn.

Brc\vcr, M. B. ( 1988). A dual process model of impres-sion formation. In T. K. Srull and R. S. Wyer (Eds.),Advaplces iv1 Social CoLqnition, 1, l-36. Mahwah, NJ:Erlbaum.

Brickman, P., .~nci Campbell, D. T. ( 1971). Hedonic rcl-;Iti\ism and planning the good socicv. In 1M. H. Ap-plcy (hi.), Adaptatiozr-he1 theow(pp. 215-31). Nc\t,York: Academic Press.

Brickman, P., Coates, D., and JanoK-Bulman, R. ( 1978).Lotte? Lvinncrs and accident victims: Is happiness rcl-ativc? Journal of Personality and Social Psychology, 36,917-27.

Campbell, A. ( 198 1). 7he sense of well-beilk, in America.Ne\v York: McGraw-Hill.

Campbell, A., Converse, l? E., and Rodgers, W. L

(1976). The quality of American 1ijL New York:RusselLSage. Foundation.

Carp, F. M., and Carp, A. (1982). Test of a model ofdomain satisfaction and aging. Research on Aging; 4,503-22.

Cash, T. F., Cash, D. W., and Butters, J. W. (1983).“Mirror, mirror, on the wall”: Contrast effects in self-evaluation of physical attractiveness. Personality andSocial Psychology Bulletin, 9, 35 l-58.

Clark, L. F., and Collins, J. E. (1993). Rememberingold flames: How the past affects assessment of thepresent. Personality and Social Psychology Bulletin, 19,399-408.

Clark, L. F., Collins, J. E., and Henry, S. M. (1994).Biasing effects of retrospective reports on current self-assessments. In N. Schwarz and S. Sudman (Eds.),Autobiographical memory and the validity of retrospec-tive reports (pp. 291-304). New York: Springer-Verlag.

Csikszentmihalyi, M., and Wong, M. M. (1991). Thesituational and personal correlates of happiness: Across-national comparison. In F. Strack, M. Argyle,and N. SchLvarz (Eds.), Subjective well-being: An in-terdisciplinary perspective (pp. 193-2 12). Oxford: Per-gamon.

DeMaio, T. J. ( 1984). Social desirability and surveymeasurement: A review. In C. F. Turner and E. Mar-tin ( Eds.), Suweying sztbjective phenomena (vol. 2, pp.257-81). New York: Russell Sage Foundation.

Dermer, M., Cohen, S. J., Jacobsen, E., and Anderson,E. A. (1979). Evaluative judgments of aspects of lifeas a function for vicarious exposure to hedonic ex-tremes. Jownal of Personality and Social Psychology,37, 247-60.

Diener, E. (1984). Subjective well-being. PsychologicalBulletin, 235, 542-75.

Dunning, D., Madey, S. F., and Parpal, M. (1995).Frames and counterfactual thought: On comparingthe “road taken” to hypothetical alternatives and thepast. In N. J. Roese and J. M. Olson (Eds.), Whatmi&t have beevj: The social pgcho10~qy of colrnterfactrralthizzki~g. Mahwah, N. J.: Erlbaum.

Easterlin, R. A. ( 1974). Does economic growth improvethe human lot? Some empirical evidence. In l?. A.David and M. W. Redcr (Eds.), Nations and hotrse-holds ipi econovnic cqrowth (pp. 98- 125). New York:Academic Press.

Elder, G. H. ( 1974). Children of the Great Depression.Chicago: University of Chicago Press.

Elster, J. (1983). Sour &grapes. New York: CambridgeUniversity Press.

Eivert, 0. (1983). Ergcbnisse und Probleme der Emo-tionsforschung (Finding problems in emotinal re-search). In H. Thomae (Ed.), 77Jeorien und Formender Motivation. Enzykloptidie der Psychologie, Series C,Part IV, Vol. 1. Gottingen: Hogrefe.

Festinger, L. ( 1954). A theory of social comparison pro-cesses. Htrvnan Relations, 7, 117-40.

Fredrickson, B. L., and Kahneman, D. (1993). Duration

c

82 Well- Being

neglect in retrospective evaluations of affective epi-sodes. Journal of Persona&y nnd Social Psychology, 65,45-55.

Glatzer, W. (1984). Lebenszufriedenheit und alternativeMasse subjektiven Wohlbefindens (Life-satisfactionand altermative measures of SWB). In W. Glatzer andW. Zapf (Eds.), Lhensqtuzlitiit in der Barndesrepublik(Quality of life in Germany). Frankfurt: Campus.

Glatzer, W., and Zapf, W. (1984). Lebensqualidt in dcrBundesrepublik (Quality of life in Germany). In W.Glatzer and W, Zapf (Eds.), Lxbensquaiitci’t in derBuzzdesrepubfik. (Quality of life in Germany). Frank-furt: Campus.

Goethals, G. R, and Darlcy, J. M.( 1977). Social com-parison theory: An attributiona! approach. In J. M.Sulls and T. A. W11is (Eds.), Social compnrison (pp.59-278). Washington, D.C.: Halstead.

Grice, H. I? (1975). Logic and conversation. In I? Coleand J. L. Morgan (Eds.), Symx nnd smuzntics, vol. 3,Speech acts (pp. 41-58). New York: Academic Press.

Hclgcson, V. S., and Taylor, S. E. (1993). Social com-parisons and adjustment among cardiac patients. Juzcr-nal of Applied Social P@~~~ko~y, 23, 117 l-95.

Herr, I’. M. (1986). Consequences of priming: Judgc-mcnt and behavior. Journnl of Pmsonality and SocialPsycbolo~y, 51, 1106-l 115.

Higgins, E. T. ( lYY6). ffi~owledgc: Accessibility, appli-cability, and salience. In E. T. Higgins and A. Kru-glanski (Eds.), Socinl pyholo~~: Hnlldbook oj’ basicprinciples (pp. 133-68). New York: Guilford.

Hormuth, S. E. ( 1986). The sampling of experiences insitu. Jorrrnnl of Person&y, S-l, 262-93.

Hyman, H. H., and Singer, E. (Eds.). ( 1968). Rmdirgsin referencecqroxrp theory and rtwnrclt. New York: FreePress.

Iycngar, S. (1987). ‘I‘*!c cvision ric\vs and citizens’ cxpl.\-natrons ot national at!$rs. A wrricfzN I’ofitic~zl SciencrRevirw, 81, 8 I S-3 1 .

Herr, I? M., Sherman, S. J., and Fazio, R H. (1983).On the Consequences of priming: Assimilation andcontrast effects. Jownnl of Experimental Social @v-cbofoc~y, Z9, 323-40.

Jxobscn, E. ( 1957). Norm,t! and pathological moods:Their nature and function. In I<. S. Eislcr, A. F. Freud,H. Hartman, and E. Kris (hds.), 17~~’ psychwm’yticstudy of the chifd (pp. 73-I 13). SC\V York: Intcrna-tiorial University I’rcss.

Kthneman, D., and Miller, I). ‘I’. ( 1986). Norm theory:<:omparing reality to its alternatives. Psyrhfo~irfzf Re-view, 93, 136-53.

Kahneman, D., and Tversky, A. ( 1979 ). I’ros!xct thcoy:An analysis of decision under risk. kouom~triccz, 47,363-91.

h=;lmriiariri, R ( 1982). Persona! circumstances and lifeevents as poor predictors of happiness. Paper prc-sentcd at the annual convention of the American Psy-chologica! Association, ~V,~shington, D.C. (August).

Kellcy, H. H. (1972). CSlllSfd SChlWl h-l fld t/Jt Ll ttribll-

tion process. Morristown, N. J.: Genera! Learning Press.

Keltner, D., Locke, K. D., and Audrain, I? C. (1993).The influence of attributions on the relevance of neg-

.- ative feelings to satisfaction. Person&i9 nnd SocialP~ychofo~y Btrlletin, 19, 2 l-29.

Kendrick, D. T., and Gutierres, S. (1980). Contrast ef-fects and judgments of physical attractiveness: Whenbeauty becomes a problem. JownnL of Personality andSocial Psycholofi~, 34 13 l-40.

Klinger, E. (1977). Meaning and void. Minneapolis:University of Minnesota Press.

Major, B. (1994). From social inequality to persona! en-titlement: The role of social comparison, legitimacyappraisals, and group membership. In M. Zanna(Ed.), Advnnces in cxperimentnl socini psychology (vol.26, pp. 293-355). San Diego: Academic Press.

Major, B., Testa, M., and Bylsma, W. H. ( 1991). Re-sponses to upjvard and downward social comparisons:The impact of esteem-relevance and perceived con-trol. In J. Suls and T. A. Wills (Eds.), Sucinl compnri-son: Contempornq theory nnd research (pp. 237-60).Hillsdale, N. J.: Erlbaum.

Marsh, H. W. (1993). Academic self-concept: Theory,measurement, and research. In J. Suls (Ed.), Psycho-fgirnl perspectives OII the sey( vol. 4, pp. l-26). Hills-dale, N. J.: Erlbaum.

Martin, I.. I,., and Tesscr, A. ( 1989). Toward a motiva-tional and stnrctura! theory of ruminative thought. InJ. S. U1crn.m and J. A. R.wgh (Ms.), U&tended~~JUI&J~ (pp. 306-26). Nc\v York: Guilt&-d.

Mcdvcc, V. H., MJdc): S. F., and Gilovich, T. ( 1995).When less is more: Counterfactual thinking and satis-faction among Olympic medalists. Jurrrnnl of Persorr-n&v nrrd Social P~c~JcJ~~~~, 69, 603- 10.

Menon, G., Raghubir, P., and Schwarz, N . ( 1 9 9 5 ) .I~elxwiora! frequency judgments: An acccssibi!ity--diagnosticity framc\\x)rk. JUUF-IJ~ o j ’ <hmmt~ Kc-MYdJ, 22, 2 12-28.

Marsh, H. W., and Parker, J. W. (1984). Dctemlinantsof student self-concept: Is it better to be a relativelylarge tish in a small pond even if you don’t learn toswim as WC!!? Jorrrtrnl of Personnliry md Social Psychol-“,riv, 47, 2 13-31.

Mich&s, A. ( 1985). Xlcrltiple ciiscrclxwcics theory. Su-cirzl Iudicntors Rcs~*fircIJ, 16, 347-4 13.

Miller, D. T., and P rentice, D. A. ( 1996). The constnrc-tion of norms and st~ndnrds. In E. 1’. Higgins and A.W . Kruglanski (Eds.), Social p~cltulc~yv: Hnrrdbook ofbasic pritrcipfes (pp. 799-829). New York: Guilt&d.

Miller, 1). I‘., Turnhu!!, CV., and XIcF.w!and, (1. ( 1990).Counterfactua! thinking and so&l perception: Think-ing about ivh.it might have been. In M. I? Zarina(Ed.), fhIiwIlc~*s in cxp0-itricr~t~I socinl p~cl~oIoc/Tv (vol.

23, pp. 305-31). Nc\v York: Ac.rdcmic Press.Morawtz, I). ( 1977). Income distribution and sclf-

rated happiness: Some cmpiricJ1 evidence. EconomicJournnl, 87, 5 1 l-22.

Morse, H., and Gergen, K. J. (1970). Social compari-son, self-consistency, and the concept of the self. Jour-nal of Pcrsorrnlity curd Social l3-ycl~010~9~, 36, 148-56.

Reports of Subjective Well-Being 83

Parducci, A. (1995). Happiness, pleasltre, and jzrdgment:The contextlral theory and its applications. Hillsdale,N. J.: Erlbaum.

Roese, N. J. (1997). Counterfactual thinking. Psychologi-cal Blrlletin, 121, 133-48.

Roese, N. J., and Olson, J. M. (Eds.). (1995a). Whatmight have been: The social psychology of corrnterfactlralthinking. Mahwah, N.J.: Erlbaum.

Roese, N. J., and Olson, J. M. (1995b). Counterf‘actualthinking: A critical overview. In N. J. Roese and J. &I.Olson (Eds.), What might have been: The social psychol-oyy of counterfacttral thinking (pp. l-55). Mahwah,N. J.: Erlbaum.

Ross, M., Eyman, A., and Kishchuck, N. (1986). Deter-minants of subjective well-being. In J. M. Olson, C. I?Herman, and M. Zanna (Eds.), Relative deprivationand social comparison (pp. 78-103). Hillsdale, N.J.:Erlbaum.

Runciman, W. G. (1966). Relative deprivation and so-cial jzhce. London: Routledge and Kegan Paul.

Runyan, W. M (1980). The life satisfaction chart: Per-ceptions of the course of subjective experience. Inter-national Jolrrnal of Apincq and Hlrman Development,1 I, 45-64.

Salovcy, I?., and Rodin, J. (1984). Some antecedents andconsequences of social comparison jealousy. Jorrmal ofPersonality and Social Psychology, 47, 780-92.

Schachtcr, S. (1959). The psycholocqy of afilintiou. Stan-tilt-d, Calif.: Stanford University Press.

Schwarz, N. ( 1987). StimmunJ als Information: Unter-s~rch~gen ~nm EinJ3Irjj’ voll Stimmrqgen nnf die Be-wevtrrplq dcs eiqenen Lebens (~Mood as information onthe i;pact 0; moods on evaluations of one’s life).Heidelberg: Springer-Vet-lag.-. (1994). Judgment in a social context: Biases,

shortcomings, and the logic of conversation. In 1M.Zanna (Ed.), Advances in experimental social p?choloJy(vol. 26, pp. 123-62). San Diego: Academic Press.-. ( 1994). Co&tion and commtrniratiox JrrdJ-

mental biases, research vnethods, and the 1oLqic of conver-sntiozr. Hillsdale, N.J.: Erlbaum.

Schivarz, N., and Bless, H. ( 1992a). Constructing realityand its altcrnativcs: Assimilation and contrast ellects insocial judgment. In L. I,. Martin and A. Tesscr (Eds.),‘lhe construction of social j$rdJment (pp. 217-45).Hillsdale, N. J.: Erlbaum.-. ( 19921~). Scandals and the public’s trust in poli-

ticians: Assimilation and contrast efects. Persozzalityarrd Social Psycl~olo~9y Bulletin, 18, 574-79.

Schwarz, N., and Clorc, G. I,. (1983). Mood, misat-tribution, and judgments of Lvcll-being: Informativeand directive filnctions of affictive states. Jurrrnal ofI’ersunality and Social Psycho49~, 45, 5 13-23.-. ( 1996). Feelings and phenomenal experiences.

In E. T. Higgins and A. Kruglanski (Eds.), Social py-chulo~y: A handbook oj’basic principles (pp. 433-65).New York: Guilford.

Schwarz, N., Groves, R., and Schuman, H. (1998).Survey methods. In S. Fiske, D. Gilbert, and G. Lind-

zev (Eds.), Handbook of social pq&~olo~y (4th ed., vol.1, 143-79). New York: McGraw Hill.

Schwarz,.N., Hippler, H. J., Deutsch, B., and Strack, F.(1985). Response categories: Effects on behavioral re-ports and comparative judgments. Public OpinionQuaYterly, 49, 388-95.

Schwarz, N., and Scheuring, B. (1988). Judgments ofrelationship satisfaction: Inter- and intraindividualcomparisons as a &nction of questionnaire structure.Elrropean Jotrmal of Social Psychology, 18, 485-96.-. (1992). Selbstberichtete Verhaltens- und Symp-

tomhiufigkeiten: Was Befiagte aus Anwomorgabendes Fragebogens lernen (Frequency reports of psycho-somatic symptoms: What respondents learn from re-sponse alternatives.) Zeitschrzft fiir Klinische Psychol-oyie, 22, 197-208.

Schwarz, N., and Strack, F. (1991a). Evaluating one’slife: A judgment model of subjective well-being. In F.Strack, M. Argyle, and N. Schwarz (Eds.), Szsbjectivewell-beirtg: An in terdisciplinaly perspective (pp. 2 7-47). Oxford: Pergamon.-. (1991b). Context effects in attitude surveys:

Applying cognitive theory to social research. In W.Stroebe and M. Hewstone (Eds.), European Review ofSocial Psychology (vol. 2, pp. 31-50). Chichester:Wiley.

Schwarz, N., Strack, F., Kommer, D., and Wagner, D.(1987). Soccer, rooms, and the quality of your life:Mood effects on judgments of satisfaction \\ith life ingeneral and with specific life domains. Ezrropean Jour-nal of Social P~choh9y, 17, 69-79.

Schwarz, N., Strack, F., and IMai, H. I? (1991). Assimi-lation and contrast effects in part-whole questions se-quences: A conversational logic analysis. Pzrblic Opin-ion Qtrarterly, 55, 3-23.

Schwarz, N., W5nke, LM., and Bless, H. (1994). Subjec-tive assessments and evaluations of change: Some les-sons from social cognition research. In M. Hewstoneand W. Stroebe (Eds.), European Review of Social Psy-chology (vol. 5, pp. 181-2 10). Chichester: \\?ley.

Scidman, E., and Rapkin, B. (1983). Economics andpsychosocial dysfunction: Toward a conceptual frame-work and prevention strategies. In R D. Feiner, L. A.Jason, J. N. Moritsugu, and S. S. Farber (Eds.), Pve-ventive ps3/cholugy (pp. 175-98). New York: Per-g3mon.

Smith, R. H., Diener, E., and Wedell, D. H. (1989).Intrapersonal and social comparison detemlinants ofhappiness: A range-frequency analysis. Jurrnral of Per-sonality and Social Psychulu~/~_v, 56, 3 17-25.

Smith, T. W. ( 1979). Happiness. Social P~chofu~y Qtmr-t&y, 42, 18-30.

Srull, T. I(., and Gaelick, L. (1984). General principlesand individual diEerences in the self as an habitual ref-erence point: An examination of self-other judgmentsof similarity. Social Cognition, 2, 108-2 1.

Strack, F. (1992). The di&ent routes to social judg-ments: Experiential versus informational strategies. InL. L. Martin and A. Tesser (Eds.), 7&e construction of

84 Well-B&g

social jzrdLqmcnts (pp. 249-76). Hillsdalc, N.J.: Erl-baum.

. (1994a). Zur P~chologie der standardisicmnBefradunp (The psychology of standardized intcr-views). Heidelberg: Springer-Verlag.-. (1994b). Response processes in social judg-

ment. In R S. Wyer and T. K. Srull (Eds.), Handbookof social co_qnition (2nd cd., 1~01. 1, pp. 287-322).Hillsdale, N.J.: Erlbaum.

Strack, F., and h4artin, L. ( 1987). Thinking, judging,and communicating: A process account of context cf-fects in attitude sunpeys. In H. J. Hippler, N. Sch\f*arz,and S. Sudman (Eds.), Social information proccssijlc/land surwy mcthodoloLqv (pp. 123-48). Nc\tr York:Springer-Verlag.

Strack, F., Martin, L. L., and Schw-arz, N. (1988). Prim-ing and communication: Social determinants of infor-mation use in judgments of life satisfaction. EzrropcnnJournal of Social Psycholm9~, 18, 429-42.

Strack, F., Schlvarz, N., Chasscin, B., Kern, D., and \l’ag-ner, D. ( 1990). The salience of comparison standardsand the activation of social norms: Consequences forjudgments of happiness and their communication.British Jownal of Social P~cl~olog?~, 29, 303- 14.

Strack, F., Schwarz, N., and Gschncidinger, E. (1985).Happiness and reminiscing: The role of time perspec-tive, mood, and mode of thinking. Joz4~71nl of h-son-aliv and Social Ps?,cholo,/r?~, 49, 1460-69.

Strack, F., Sch\frarz, N., and Wsnkc, h4. (1991). Semanticand pragmatic aspects of context effects in social andpsychological research. Social Cogrlitiolz, 9, 111-25.

Sudman, S., Bradbum, N., and Sch\varz, N. (1996).Thin kin. a bold t answers: Z’ge application of cop itiwprocesses to survey methodology. San Francisco: Jossey-Bass.

Taylor, S. E., and Brown, J. D. (1988). Illusion andwell-being: A social psychological perspective on mcn-tal health. Pycholopical Bulletin, 103, 193-2 10.

Taylor, S. E., and Lobel, M. (1989). Social comparisonactivity under threat: Downward evaluation and up-ward contacts. Psychological Review, 96, 569-75.

Taylor, S. E., Wayment, H. A., and Carrillo, M. (1996).Social comparison, self-regulation, and motivation. InR M. Sorrentino and E. T. Higgins (Eds.), Hand17ookof motivation and cognition (vol. 3, pp. 3-27). Ne\\pYork: Guilford.

Taylor, S. E., Wood, J. V., and Lichtman, R. R. (1983).It could be worse: Selective evaluation as a responseto victimization. Joutxal of Social Issues, 39, 19-40.

Tesser, A. (1988). Toward a self-evaluation maintenancemodel of social behavior. In L. Berkowitz (Ed.), Ad-

wznces in exprrinrcn tai social pn,cl~olu~qy (\wl. 2 1 , pp.18 l-227). New York: Academic Press.

-Tesscr; A., and h4artin, I,. I,. (1996). The ps!*choloby* ofe\*aluation. In E. T. Higgins and A. \I’. Kruglanski( Eds. >, Social p~cltolo_l~?~: Ha~~dbuok of bnsic pt-inciph(pp. 400-432). Ne\if York: Guilford.

Thaler, R. H. ( 1985). hlental accounting and consumerchoice. MnrX’ctilcq Scicncr, 4, 199-2 14.

Tourangeau, R., and Rasinski, K. A. ( 1988 ). (:ogniti\.cprocesses undcrl!ing context ctlccts in l tritudc mea-surcmcnt. l?~citolo_lTica/ Brrllrtill, 103, 299-3 14.

T\*ersky, A. ( 1977). Fcaturcs of‘ similarit\: I’.~h~l~~qirnIRclGcw, 84, 327-52.

T\*ersky, A., and Gati, I. (1978). Studies of‘similarin*. InE. Rosch and 13. 13. Llo!‘d (Eds.), Cuqnitiutl nnd cats-

_qorization (pp. 81-98). Hillsdalc, N.J.: Erlb~um.T\.ersky, A., and Griffin, I). ( 1991 ). On the d!wamics of‘

hedonic cspcricncc: Endo\\*riicnt arid contrast injudgments of \\rcll-being. In E‘. Strack, hl. Arg*lc, mlti

N. Sch\\farz (kids.), Szrbjrctiw well-bcinq (1’~. 10 l-18). Osfcxd: Pcrgamon.

\‘arey, C., and Kahncman, 1). ( 1992). Expcricnccs cs-tended across time: Evaluation of moments and cpi-sodcs. Jow-nal of Hclm~~iornl Dccisiotl Mnkiyg, 5, 169-95.

li’alstcr, E., Walster, G. W., and Bcrschcid, E. (1978).Eqlrit?, hot? and t~scn~*cI~. Boston: All!x and Ihcon.

\l’snke, M . , Schwarz, N . , a n d h’ocllc-h’cuma~ln, E.

( 1995). Asking comparati\ve questions: The impact of’the direction of comparison. I’rMic Opirrion &at*-tcrlv, 59, 347-72.

Wayment, H. A., Taylor, S. E., and Carrillo, M. ( 1994).The motivational and performance implications of up-ward and downward comparisons. University of Cali-fornia at Los Angeles. Unpublished paper.

Wells, G. L., and Gavanski, I. (1989). Mental simulationof causality. Jorrrxal of Personality and Social l~~chol-03.y, 49, 1460-69.

Weiner, B. (1985). An attributional theory of achic\fc-ment motivation and emotion. Pgchohqical RrGm,89,548-73.

Wills, T. A. ( 198 1). Dowrx\*ard comparison principles insocial psychology. Psychological Bulletin, 90, 245-7 1.

Wood, J. V. (1989). Theor)l and research conccming so-cial comparisons of personal attributes. P~chologicafBulletin, 106, 2 3 l-48.

l’bema, J. F., and Buunk, B. I? (1995). The cffccts ofsocial comparison direction and social comparison di-mension upon affect and identification among dis-abled individuals. British Journal of Social Psycholu~y,34,279-92.