attitudes and social cognition patterns of...

ATTITUDES AND SOCIAL COGNITION

Patterns of Cognitive Appraisal in EmotionCraig A. Smith and Phoebe C. Ellsworth

Stanford University

There has long been interest in describing emotional experience in terms ofunderlying dimensions, but traditionally only two dimensions, pleasantness andarousal, have been reliably found. The reasons for these findings are reviewed,and integrating this review with two recent theories of emotions (Roseman, 1984;Scherer, 1982), we propose eight cognitive appraisal dimensions to differentiateemotional experience. In an investigation of this model, subjects recalled pastexperiences associated with each of 15 emotions, and rated them along theproposed dimensions. Six orthogonal dimensions, pleasantness, anticipated effort,certainty, attentional activity, self-other responsibility/control, and situationalcontrol, were recovered, and the emotions varied systematically along each ofthese dimensions, indicating a strong relation between the appraisal of one'scircumstances and one's emotional state. The patterns of appraisal for the differentemotions, and the role of each of the dimensions in differentiating emotionalexperience are discussed.

Most people think of emotions in categor-ical terms: "I was scared," or "I was sad," or"I was frustrated." In complicated situationsthey may say, "I felt sad and frustrated." Theidea that there is a small set of fundamentallydifferent emotions, has a long and illustrioushistory in science as well, dating back at leastto Aristotle and reemerging in the theory ofthe four humors, in the works of eighteenth-century philosophers, and in Darwin (1872/1965). In recent years the categorical ap-proach to the study of emotions has becomeprominent in psychology, stimulated by themonumental work of Sylvan Tomkins (1962,1963, 1982; Ekman & Friesen, 1971; hard,1971, 1972, 1977; Izard & Buechler, 1980).This view of emotional experience admirablycaptures our intuition that happiness, anger,and fear are basic feeling-states, easily rec-ognizable, and fundamentally different fromeach other.

This research was supported in part by a StanfordUniversity graduate fellowship and in part by a NationalInstitute of Mental Health training grant to the firstauthor.

We would like to thank Laura R. Novick for herhelpful comments on an earlier version of the manuscript.

Requests for reprints should be sent to Craig A. Smith,Department of Psychology, Stanford University, Building420, Jordan Hall, Stanford, California 94305.

However, most categorical theories presentemotions as an unstructured collection ofdistinct entities and thus fail to capture ourintuitions about the similarities and differ-ences among emotions. Some emotions arecommonly viewed as opposites, such as joyand sorrow, or fear and anger. Once we startconsidering the similarities and differencesamong emotions, more specific questions in-evitably demand our attention. Are joy andsorrow opposite in the same way as fear andanger? Clearly not, because joy is pleasant,but fear and anger are both unpleasant. Thenin what ways can emotions be "opposite?"The very idea of opposition implies a dimen-sion or set of dimensions along which emo-tions can be arranged, so that we can beginto make more specific statements of howemotions are similar or dissimilar to eachother.'

' A dimensional view of emotions is not incompatiblewith a categorical view. Even in a categorical system,dimensions may be useful for describing the similaritiesand differences among the categories. In a largely dimen-sional system, common emotion categories may representregions of the dimensional space that are frequentlyexperienced or subjectively salient. Also, it is possible fora categorical theory to include hypotheses about therelations among the discrete emotions (cf., Plutchik,1980).

Journal of Pcrsoralily and Social Psychology, 1985, Vol. 48, No. 4, 813-838Copyright 1985 by the American Psychological Association, Inc. 0022-3514/85/SOO.75

813

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

814 CRAIG A. SMITH AND PHOEBE C. ELLSWORTH

Search for Dimensions of Emotion

Questions about the interrelations amongemotions are not new. In 1896 Wundt (I896/1897) proposed a three dimensional structureof emotional experience, and in 1941 Schlos-berg began a series of investigations into thestructural interrelations among emotions(Engen, Levy, & Schlosberg, 1958; Schlosberg,1941, 1952, 1954) that marked the start of aresearch tradition that has continued up tothe present (e.g., Berglund, Berglund, & En-gen, 1982; Daly, Lancee, & Polivy, 1983;McHugo, Smith, & Lanzetta, 1982; Russell,1983). However, this tradition has remainedfairly isolated from other research on emo-tions. Progress has been slight and heuristicinsights rather sparse.

One reason for this is that the empiricalresults have been disappointing. Pleasantnessand level of activation (or arousal) are theonly dimensions that have been found con-sistently across studies. For anyone interestedin understanding the rich diversity and subtlenuance of emotional life, a simple character-ization of emotions as pleasant or unpleasant,strong or weak, seems both theoretically sterileand experientially implausible. And in fact,most of the dimensional studies have foundmore than these two dimensions. The prob-lem is that there is little consistency amongthese richer characterizations: Beyond pleas-antness and level of activation, the dimensionsthat emerge are often difficult to interpret(e.g., Block, 1957; McHugo et al., 1982), andwhen they are interpretable, they vary fromstudy to study. The failure to arrive at asolution that is both interesting and robustcan be understood by examining both themethods and the theoretical underpinningsof the previous studies.

Methodological Problems With theEarlier Research

Twentieth-century research on the dimen-sions of emotion has evolved along two dis-tinct lines. The original line, initiated bySchlosberg (1941), sought to discover thedimensional structure communicated throughposed emotional facial expressions (Abelson& Sermat, 1962; Berglund et al., 1982; Engenet al., 1958; Frijda, 1969; Frijda & Philip-szoon, 1963; Osgood, 1955, 1966; Schlosberg,

1952). The second, more recent line of re-search has examined the structure underlyingthe subjective feeling states associated withvarious emotions (e.g., Averill, 1975; Block,1957; Bottenberg, 1975; Bush, 1973; Davitz,1969;Izard, 1972; Russell, 1978, 1979, 1980;Russell & Mehrabian, 1977). There is noreason to expect that the same dimensionsshould be revealed by these two approaches.Some salient aspects of an emotional experi-ence may not show on a person's face, andsome salient aspects of facial movement maynot correspond to changes in a person's feel-ings. Furthermore, in studies of facial expres-sion, subjects are asked to interpret otherpeople's emotions, whereas in many studiesof subjective feelings subjects are asked toreport on their own. What matters to us inunderstanding someone else's emotion maybe quite different from what matters to us inunderstanding our own emotions.

Aside from possible problems created bythe confounding of domains, there have beenmethodological weaknesses within each ofthe two lines of research. First, as Ekman,Friesen, and Ellsworth (1982) point out, therange of stimuli sampled will determine thedimensions that can be found. No possibilityof discovering a valid dimension exists if thestimuli do not vary along that dimension.Several of the early studies were weak in thisrespect. For example, Abelson and Sermat(1962), who found only two dimensions, useda sample of faces that underrepresented angryand surprised expressions.

Second, the questions asked of the subjectcan restrict the dimensions it is possible tofind. A common method for discovering theways in which people differentiate amongemotions is to have them rate the stimuli ona variety of scales. A major strength of thisapproach is that the recovered dimensionsare usually interpretable. Both scale loadingsand the arrangement of the stimuli along thedimensions can be used to guide interpreta-tion. A further strength is that by havingsubjects rate each stimulus along a variety ofscales, it is possible to uncover subtle, butpsychologically important, dimensions. Themajor limitation of this method is that theselection of scales determines which dimen-sions can possibly be found. Although ameaningless scale will produce unsystematic

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

COGNITIVE APPRAISAL IN EMOTION 815

ratings, and thus can not "create" a nonex-istent dimension, if a meaningful dimensionis not represented in the scales it will not befound in the data. Therefore, studies usingdifferent sets of scales will often find differentdimensions.

Critics of this technique have argued thatthe investigator's choice of scales constrainsthe subject to use those dimensions and makesit impossible to discover unexpected dimen-sions or to find out what dimensions arenaturally salient. The usual alternative, de-signed to avoid these problems, has been toask subjects to rate the similarity of allpossible stimulus pairs (Abelson & Sermat,1962; Berglund et al., 1982; Bush, 1973;Shepard, 1962; Russell, 1978, 1980). Thisallows the subjects to choose the dimensionsthey use in their judgments. A major limi-tation of this method is that investigators nolonger have the scale loadings to help themdecide what the dimensions mean, but mustrely solely on the arrangement of the stimuli.Further, subjects may use inconsistent criteriato judge different pairs of stimuli, or maybase their ratings on only the most salientdimensions. For both reasons, this methodmay often fail to reveal more than two orthree dimensions.

Thus, both the direct rating and the simi-larity rating methods run the risk of missingdistinctions that are meaningful in people'sperceptions of emotion. One of our aims inthe present research was to create a set ofscales that would allow subjects to differentiatetheir emotional experiences meaningfully inways that go beyond the usual two or threedimensions. Therefore, in reviewing the lit-erature, we regarded a dimension as poten-tially valid if it was found in one study, andwe were not particularly disturbed by otherstudies that did not find it.2

Dimensions of Facial Expression

Table 1 summarizes the results of bothtraditional lines of research on the dimensionsof emotion. Looking first at the studies offacial expression, we see that there is consis-tent evidence for three dimensions: pleasant-ness, level of activation, and attentional activ-ity. In 1955, Osgood found evidence thatfacial expressions also differ from each other

along a dimension of control. Although heand others (Frijda, 1969; Triandis & Lambert,1958) felt that this dimension was somehowrelated to attentional activity, the two dimen-sions were clearly not identical (Frijda &Philipszoon, 1963). In a reanalysis of hisdata, Osgood (1966) found two dimensionsin the place of the old control dimension: AnInterest factor, corresponding more directlyto attentional activity, and a new Controlfactor. Frijda and Philipszoon (1963) alsofound a fourth factor, defined by the semanticdifferential scales mild-derisive, natural-arti-ficial, submissive-authoritarian, and admira-tion-despise. Frijda (1969) later interpretedthis dimension as being closely related toOsgood's control dimension, with "natural,"submissive expressions corresponding to "un-controlled" expressions, and "artificial," au-thoritarian ones being "controlled."

Dimensions of Subjective Feeling Slates

Although attentional activity emerged as aconsistent third dimension in the research onfacial expression, none of the studies of sub-jective experience has found any evidence forit. Apparently people are likely to register theattentiveness of their companions' faces, butare unlikely to discriminate among their ownfeelings in terms of attention. On the otherhand, several studies indicate control is im-portant in interpreting subjective experience,just as it is in perceiving faces (see Russell,1978, for a discussion of the convergent va-lidity of this dimension within the domain ofsubjective feeling-states).

Finally, several studies of subjective feeling-states have found evidence that people differ-entiate among their own emotions in termsof depth of experience. Daly et al. (1983)

2 In this review, the dimensions are grouped together

on the basis of the original investigators' interpretationsof them. Obviously, it is very difficult to formally comparedimensions found in different studies, particularly whenthe studies employ different methods and stimuli. There-fore, we do not intend our groupings to imply that all ofthe dimensions sharing a common label are necessarilyidentical. Instead we take the similarity of interpretationsas an indication that the original investigators believedthat they were examining conceptually related phenomena.We feel that these phenomena, and the concepts they arebelieved to represent, are worthy of further consideration.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


describe this dimension as the intensity ofthe felt emotion, ranging from "no report ofan emotion to an extremely emotional ex-perience" (p. 445), and Russell (1978) de-scribes it as contrasting "serious, profoundemotions such as spiritual and loving withmore shallow emotions such as giddy, peevish,or coy" (pp. 1152-1153). This dimension isnot typically found in facial expression stud-ies, although it bears some resemblance toan unnamed dimension found by Frijda(1969), denned by the scales simple-compli-cated and shallow-deep.

Theoretical Problems With theEarlier Research

The main theoretical problem with theearlier research is that there was very little

theory. In general, the investigators did notchoose their dimensions in order to exploretheir ideas about emotions. Some researcherswere frankly atheoretical, waiting for ideas toemerge from the data. Others (e.g., Osgood,1966) were trying to see whether dimensionsdiscovered in other domains would extend tothe experience or expression of emotion. Veryfew had a theory that included much morethan the obvious considerations of arousaland valence. In part, this theoretical sterilitywas due to a failure to consider what it isabout emotion that creates its variety. In thisregard, the categorical theorists were moresuccessful. Working from a list of differentemotions, they could ask "what is uniqueabout the feeling of anger" and "how is angerdifferent from fear?" (cf. Izard, 1977). How-ever, the categorical approach has not led to

Table 1

Empirically Derived Dimensions of Emotional Experience Found in Two or More Studies of EitherFacial Expression or Subjective Feeling States

Dimensions

Studies PleasantnessLevel of

activationAttentional

activity

1 Reanalysis of Abelson & Sermat (1962) data.b Not interpreted as such by original authors; see text for explanation.' Reanalysis of Osgood (1955) data.•* Reanalysis of Frijda & Philipszoon (1963) data.

ControlDepth of

experience

Facia] expression

Schlosberg(1952) X XSchlosberg(1954) X X XOsgood (1955) X X XEngen, Levy, & Schlosberg (1958) X X XTriandis& Lambert (1958) X X XAbelson & Sermat (1962) X XShepard(l962)' X XFrijda & Philipszoon (1963) X X X (X)"Osgood (1966)' X X X XFrijda (1969^ X X X X (X)6

Berglund, Berglund, & Engen(1982) X X X

Subjective feeling states

Block (1957)Bush (1973)Averill(1975)Bottenberg(1975)Russell & Mehrabian (1977)Russell (1978)Russell (1980)Daly, Lancee, & Polivy (1983)Russell (1983)

XX

XXXXXXX

XX

XXXXXXX

XX

XX

X

X

X

X

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Table 2Proposed Cognitive Dimensions and Their Relations to Previously l-'ounil Dimenxionx*

Proposed cognitive dimensions

Pleasantness

Motivational state"Situational state"

Anticipated effort/ Activation'Attentional activity/NoveltyControl Agency/Coping potentialDepth of experience

CertaintyGoal-path obstacle/Goal relevance

LegitimacyNorm/self-concept

compatibilityResponsibility

Previouslyfound

dimensions

X

XXXX

Roseman(1979)

XX

X

X

X

Scherer(1982)

X

XX

X

X

Smith &

Ellsworth(1985)

X

XXX

XXX

X

• Previously found dimensions are documented in Table 1.b These two dimensions combine to form a Pleasantness dimension; see text for explanation.c See text for explanation of the proposed relationship between Activation and Anticipated effort.

a general framework within which a widerange of emotions can be evaluated in thesame terms.

Dimensions of Cognitive Appraisal

A third group of theorists, identified withneither the categorical nor the dimensionalapproach, have argued that emotional differ-ences must inescapably involve differences inthe way the organism appraises its environ-ment (Arnold, 1960; James, 1890/1950; La-zarus, 1968; Lazarus, Kanner, & Folkman,1980; Mandler, 1975; Plutchik, 1962, 1980;Roseman, 1984; Schachter & Singer, 1962;Scherer, 1982). In line with these theorists,we believe that the experience of emotion isclosely associated with the organism's ap-praisal of its environment along several cog-nitive dimensions, and that a close study ofthese dimensions will help us to understandnot only the nature of distinct emotionalstates, but also their interrelations. A fewrecent theorists, most notably Roseman(1984) and Scherer (1982) have independentlyproposed dimensional models. Their models,and ours, are shown in Table 1?

Attention. Most of the research on dimen-sions of facial expression has found strongevidence that level of attention is an importantaspect of emotion. It may even be that thefirst appraisal we make is whether to attend

to a stimulus, ignore it, or avoid it (Scherer,1982). On a city sidewalk, a man with a gunin his hand is riveting, a sleeping verminousdrunk is repellent, and the rest of the massof humanity may not attract our attention atall. That this dimension is obvious in thefacial expression studies but absent in thestudies of subjective experience may be be-cause attention is so preliminary, and subse-quent appraisals seem much more salient inour descriptions. If we have barely noticed astimulus, we are unlikely to mention it in anexperiment.

Scherer's novelty dimension is related tothe attentional activity dimension. Like anorienting response, it is an appraisal of theextent to which events violate or meet expec-tations and therefore demand attention orcan be ignored. We have chosen to study this

3 Several other recent investigators have also proposedan important role for cognitive appraisals in emotion,including DeRivera (1977), Weiner, Russel, and Lerman(1979), Clore and Ortony (1984), Abelson (1983), andStein and Jewett (1984). For various reasons none ofthese are easily summarized in Table 2; however, we seeall of them as compatible with our general point of viewthat emotional life is much more than movement backand forth along a continuum from wry positive to very

negative and that in order to understand the variety ofemotions people experience we must consider their inter-pretations of their present environment and themselveswithin it.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


dimension as a bipolar one, ranging from apowerful motivation to examine the stimulusmore closely to an equally powerful motiva-tion to turn away from the stimulus. Likemost of our decisions, this one is tentative,and mainly based on the empirical evidenceof bipolarity (e.g., Schlosberg, 1952, and fol-low-up research). A state of strong interest orfascination would represent one pole of thisdimension, while disgust would usually fallat the opposite extreme.

Certainty. Roseman (1984) does not pro-pose an attention dimension; in his schemethe most similar dimension is certainty. Whileit is likely that both attention and certaintyare correlated in many situations, in that anunpredictable situation demands more of ourattention than a predictable one, we feel thatthe cases in which the correlation fails areparticularly interesting, and so we chose toinclude attention and certainty as separatedimensions. Hope and fear are the emotionsfor which uncertainty is paramount and at-tention is inescapably drawn to the source,but there are also situations in which we areuncertain about what is going on and do notparticularly care to find out, and others, suchas watching a lover or baby asleep, in whichwe feel no noticeable uncertainty, but ourattention is devout.

Control. The control dimension has ap-peared with some consistency in the tradi-tional literature, and it also appears in thecognitive models of both Roseman (1984)and Scherer. Scherer (1982) proposes that theorganism must quickly evaluate its ability tocope with the situation and that this evalua-tion will contribute to its subjective feeling.We feel, as does Roseman, that this dimensionis more complicated. We may be unable tocontrol a situation because no one can controlit, as in an earthquake, or we may be unableto control it because some other person con-trols it. Weiner, Russell, and Lerman (1979)have found support for a similar distinctionin their research on emotional responses inachievement-related contexts. They found thatdifferent emotional responses were associatedwith both success and failure depending onwhether the cause of the outcome was attrib-uted to oneself, to others, or to "luck." Notwanting to obscure this potentially importantdistinction, we chose to retain the more

complex formulation, and in assessing per-ceptions of control we examined whether theevents were controlled by the person, anotherperson, or impersonal circumstances.

Pleasantness. Pleasantness has alwaysemerged as an important dimension and hasaccounted for the lion's share of the variance.In some recent formulations (Zajonc, 1980),pleasantness is seen as such an immediate,automatic evaluation that it cannot be calleda cognitive evaluation at all, but rather pre-cedes cognition. Interestingly, none of thethree new cognitive models is content with asimple pleasantness dimension. Roseman(1984) argues that pleasantness is a functionof two appraisals—appraisals of what onewants in relation to what one has. That is,happiness can result from attaining a positivegoal or from avoiding (not having) an evil,and the two states are different; likewise, theunpleasantness of a noxious event is differentfrom the unhappiness of loss. The departureof a beloved friend is very different from thearrival of an enemy (cf. Abelson, 1983; Mow-rer, 1960). Scherer (1982) proposes that stim-uli are intrinsically pleasant or unpleasant,but that our evaluation of them will alsodepend on their relevance to our currentpurposes: "Even intrinsically pleasant stimulican interrupt ongoing plans and thus beevaluated negatively by the organism in termsof goal attainment" (Scherer, 1982, p. 559).Because of sunny days and music, we kept asimple, intrinsic pleasantness dimension.

Perceived obstacle. However, like Scherer,we are well aware that even our favoritesymphony or person can be annoying if wehave to complete a job by an imminentdeadline. We chose to call our additionalcomplicating dimension perceived obstacle, aterm both broader and less vague than goalrelevance. It is broader in that it allows afeeling of pleasure on encountering an obsta-cle, as in sports, or poetry, or intellectualpuzzles. It is less vague in that it avoids thevague term relevance in favor of a perceptionthat is usually quite concrete: the perceptionof something standing in the way, even if thegoal was previously unconscious.

Responsibility and legitimacy. A personmay be responsible for initiating a situation,but then cease to control it. A thoughtlessremark or a kind deed can set in motion

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


events that seem to go far beyond theircauses. So control may be different fromresponsibility. When we think of emotionslike anger and guilt, it seems that somedimension more human and "cognitive" thanany we have yet considered must be involved.If a small child gets knocked down and hurtin a confrontation with a large dog, we areusually not angry, but if the confrontation isbetween a small child and an adult, we oftenare. Likewise, we feel greater guilt the moreresponsible we feel for the ill effects of ourown actions.

The examples are familiar, but the concep-tual issues are difficult. Roseman (1984) pro-poses a legitimacy dimension that representsan appraisal of whether the (good or bad)outcome is deserved or undeserved. Scherer(1982) approaches the same problem by pro-posing a norm/self-concept compatibility di-mension by which people evaluate an actionin terms of general social norms or personalstandards. There is clearly a dimension herethat is important in the adult human expe-rience of anger and guilt, and that distin-guishes these emotions from, for example,frustration and embarrassment. We chose toexamine two related but potentially separabledimensions in the hope of further clarifyingthe conceptually elusive, but intuitively com-pelling set of distinctions: responsibility, theextent to which oneself, or someone or some-thing else, is responsible for bringing aboutthe event that arouses emotion, and the legit-imacy or fairness of the outcome.

Anticipated effort. By now it should beobvious that we have failed to mention levelof activation, one of the universally founddimensions. We do not question that emo-tional experiences vary in intensity; however,we do not feel that intensity is very importantin discriminating among different emotionalstates. It is worth noting that neither Rosemannor Scherer include a separate intensity di-mension either. Of course our machine, likea TV, has a volume control, but it does notaffect the content of the channels. There isanother dimension with a long and respectablehistory, that is very reminiscent of a globalactivation dimension, but potentially moreuseful in discriminating among emotions,and that is Cannon's (1929) concept of fightor flight. In fight or flight situations the

organism anticipates having to do something,to expend effort; in other situations it mayrelax, enjoy, or withdraw quietly. Because theanticipated effort is so important physiologi-cally, we included it as a dimension in theplace of level of activation.

A New Dimensional Approach

Obviously, we do not feel that the dimen-sional approach has yet been given a fairchance. Few investigators have asked questionsexplicitly designed to examine more than twoor three dimensions. The studies employingglobal similarity ratings have been disap-pointing; apparently subjects do not takeadvantage of the theoretical openness of thetechnique. Rather than revealing dimensionsundreamed of by the experimenter, they seemto sort the stimuli as crudely and easily asthey can, and seldom reveal more than agood-bad, strong-weak categorization. Ouraim was to discover whether people wouldconsistently use a larger set (eight) of a prioridimensions if given the chance. The eightdimensions were, pleasantness, attention,control, certainty, perceived obstacle, legiti-macy, responsibility, and anticipated effort,selected primarily because each of them hasbeen suggested by at least one other researcher.

These are all dimensions of cognitive ap-praisal. We use the term cognitive broadly, asLazarus (1984) does, rather than restrictively,as Zajonc (1984) does. We believe that peoplemust answer certain fundamental questionsabout the changing sensations that impingeupon them not only so as to know what todo, but also so as to know what they feel.Their answers may be incorrect, in whichcase observers might say that their emotionsare inappropriate. Their answers may be in-complete, in which case they might be un-certain of how they felt, or might vacillatebetween different emotions. The emotion,however, is in part the resultant of a series ofestimates about the present environment. Fi-nally, we are not proposing that emotion ismerely the product of cognitions; instead, weare trying to explore the cognitive aspects ofemotion.

Our approach was to ask people to remem-ber past emotional experiences, one at a time,and to describe them to us in as much detail

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


as possible. We then asked them about theirexperiences, using questions that were sys-tematically designed to tap the eight proposeddimensions. We predicted that we would findan interpretable multidimensional cognitivestructure, representing appraisal dimensionsalong which people evaluate their relationwith the environment during emotional ex-periences. And we predicted that basic cate-gories of emotions, such as fear, sorrow, andanger, would be represented by different pat-terns of appraisal outcomes along the dimen-sions, and therefore, that examination of theappraisal patterns would provide considerableinsight into the experience of the basic emo-tions.

Method

Experimental Overview and Design

In a fully within-subjects design, subjects were askedto recall past experiences associated with each of 15different emotions. For each emotional experience subjectsfirst responded out loud to a series of questions designedto encourage them to describe the experience in as muchdetail as possible, then posed the facial expression thatwould show someone else how they felt during the

experience, and finally rated the experience along thehypothesized cognitive dimensions.4 Because pretesting

indicated that the procedure lasted approximately 10min for each emotional experience, the experiment was

run in two I Vi hr sessions held between one day and oneweek apart. Each subject described between six and nine

emotional experiences during the first session and theremainder in the second session. The order in which the

experiences were described was random across subjects.

Subjects

Subjects were 16 Stanford University undergraduateswho participated in the study as part of an introductory

psychology course requirement. Students who had indi-cated on a preselection questionnaire that they were both

emotionally expressive and good at posing emotionalfacial expressions were selected to participate in the

study. These criteria were used both to facilitate theposing phase of the experiment, and because pretestingsuggested that people who say they are emotionallyexpressive give more complete descriptions of their emo-tional experiences than less expressive people.

Stimulus Emotions

The 15 emotions we asked subjects to describe werehappiness, sadness, fear, anger, boredom, challenge, in-

terest, hope, frustration, contempt, disgust, surprise, pride,

shame, and guilt. We felt that 15 emotions was themaximum we could adequately observe using a within-subjects design. The selection of these particular emotions

was based on two considerations. First, we wanted toinclude the fundamental emotions (i.e., happiness, sadness,fear, anger, interest, contempt, disgust, surprise, shame,and guilt) proposed by discrete emotions theories (e.g.,Ekman, Friesen, & Ellsworth, 1982; Izard, 1977; Tomkins,1982). Second, we included emotions that we believed

would represent the dimensional space as extensively aspossible. This consideration led us to add frustration and

boredom, as well as emotions that we expected to revealdifferentiation among pleasant or positive emotional states

(i.e., challenge, pride, and hope)—an area which hasoften been overlooked in previous studies (Lazarus et al.,1980).

Procedure

The subjects were solicited by telephone to participatein an individual 3-hr, 2-session experiment concerning

"subjective emotional experience." At the first session,the subjects were told that unlike past studies, which"had shied away from asking people about their actualexperiences of emotion," the purpose of "today's exper-iment is to find out what emotional experience is likefor you." They were told that they would be asked toanswer a number of questions about their past emotionalexperiences out loud and that what they said would betape-recorded. In order to encourage detailed descriptions

we asked subjects to answer the questions as though "youare talking to a Vulcan, like Mr. Spock from Star Trek."

That is, they were told to describe their experiences as ifthey were talking to a brilliant being, very familiar withhuman culture, but unable to experience emotions firsthand, and thus eager to learn as much about emotional

behavior as possible. The subjects were also told theywould be asked to pose the facial expression correspondingto the experience and that the poses would be videotaped.

The final task for each experience would be to fill out aquestionnaire that asked more specific questions about

what it was like while they were "actually experiencing

the emotion, not what it was like either just before orjust after the emotional experience." (These instructionswere reiterated when the subject was given the question-naire.) The experimenter answered any questions aboutthe procedure and then left the room, in order tominimize both the subjects' self-consciousness and anypotential experimenter demands. Subjects were continu-ously observed from behind a one-way mirror, andcommunicated with the experimenter through a two-wayintercom system.

The same procedure was followed for each emotionalexperience. First the subjects were asked to recall a past

4 The instructions to show the emotion on the facewere designed to study issues beyond the scope of thepresent discussion. Also, 17 additional subjects were notasked to recall a particular emotional experience, butinstead to describe what each of the 15 emotions wasusually or typically like. The results of this emotionalconcept study were generally similar to the results of theemotional experience study reported here, and will befully described in another article specifically focussed onthe comparison between general and specific representa-tions.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


experience or event in which they had felt a particularemotion. Then they were asked a number of general

questions about that particular experience. For instance,when the subjects were to recall a past happy experience

they were told:

I want you to think of a past situation or event whereyou felt happy. Picture this situation in your mind.Try and remember as vividly as you can what this pasthappy situation was like: Think of what happened tomake you feel happy, and what it felt like to be happy

in this particular situation. Tell me when you are readyand have this happy situation in your mind, and I'llask you questions about it. Remember you will betelling a Vulcan, who has never had a happy experience,

what one was like.

When the subjects said they were ready they were askedthe following series of questions about the experience.

1. Please describe this past happy situation to me.What was it like to be happy in this situation?

2. What happened in this situation to make you feelhappy?

3. Why did these things make you feel happy?

4. How did you know that you were happy in thissituation?

5. What did it feel like for you to be happy in thissituation?

6. What did you do in this situation where you werehappy?

In addition to providing a wealth of open-ended datafor exploratory analysis, the main purpose of these

questions was to encourage the subjects to recall theirexperiences as fully as possible before completing the

ratings. We felt that this prior reminiscence would max-imize the accuracy of the ratings. The instructions andquestions seemed to be engaging, and to encourage thesubjects to recollect their past experiences in detail. The

answers to the questions tended to be long and involved,averaging approximately 350 words for each experience.

Also, the experiences subjects chose to describe wereusually important and emotionally involving. For example,for their sad experiences 9 subjects described their reac-tions to the death of someone very close to them. Fortheir fearful experiences, subjects described compellingsituations like being trailed by a car while walking aloneafter dark, seeing a prowler near an unlocked door whilehome alone at night, being chased by a guard dog, andalmost being arrested for drug possession at a rockconcert. Finally, during debriefing several subjects spon-taneously commented that while making their descriptionsthey often felt themselves beginning to reexperience thefeelings associated with the past event.

After the intervening posing task the subjects rated theexperience on the dimensional ratings questionnaire.

Stimulus Ratings

The subjects rated each of their emotional experienceson the items presented in Table 3 (using happiness as asample emotion). Each of the proposed dimensions wasmeasured by two items, with two exceptions. In order totap all three of the types of control we wanted toinvestigate (i.e., situational, self, other), three items were

constructed to measure the control dimension. In additionthree items were constructed to measure the certaintydimension, because pretesting indicated that this might

be a particularly difficult dimension to tap.All items were rated on II-point scales. Pleasantness

and attention were rated on bipolar scales ranging fromunpleasant, divert attention (-5) to pleasant, devote at-

tention (5). The remaining dimensions were rated onmonopolar scales ranging from not at all ( I ) toextremely (11). The items appeared on the ques-tionnaire in the same order for all emotions across allsubjects. The order was random with the constraint thatno two items designed to measure the same dimension

were adjacent to one another.The three-part procedure was repeated for between six

and nine emotions, until an hour and a half had elapsed,at which time the first session ended. On reporting forthe second session the subject was briefly reminded ofthe procedure for each emotion. Then the experimenterleft to observe the subject from behind the one-waymirror, and the procedure was completed for the emotionsnot covered during the first session. Following the lastemotion, the subject was thoroughly debriefed and dis-

missed.

Results

Overview

The analyses were designed to address twomajor issues. First, we needed to identify thedimensions subjects used in rating their ex-periences. Then we sought to examine thehypothesis that the pattern of appraisals alongthe dimensions differed across emotions, andif they did, we wanted to describe the inter-relations of the emotions within the dimen-sional framework.

The first task was to identify the dimen-sional structure of our subjects' ratings. Someprevious researchers employing a priori ratingscales have assumed that each scale repre-sented a separate dimension (e.g., Engen etal., 1958; Russell & Mehrabian, 1977; Schlos-berg, 1952, 1954; Triandis& Lambert, 1958),but this approach does not take into accountthe problem of intercorrelations among thescales, resulting in highly redundant dimen-sions. Other researchers have attempted torecover truly orthogonal dimensions in twodifferent ways. Some have used a correlationalapproach, such as factor or principle com-ponents analysis (e.g., Block, 1957; Frijda,1969; Frijda & Philipszoon, 1963; Osgood,1966), while others have used multidimen-sional scaling techniques, a distance-fittingapproach (e.g., Abelson & Sermat, 1962;Bush, 1973; Osgood, 1955; Shepard, 1962;

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Russell, 1978, 1980, 1983). Neither approach tation and Symmetric Individual Differencesprovides strict guidelines for determining the Multidimensional Scaling (SINDSCAL, Pru-dimensionality of the solution, and the re- zansky, 1975). Because the two techniquessearcher must subjectively decide how many examine the data in quite different ways, weof the dimensions are "real" by considering felt we could be confident in the stability andtheir stability and interpretability (Chatfield validity of similar dimensions revealed by& Collins, 1980; Kruskal & Wish, 1978). both.Having designed the measures to tap eight The solutions produced by both techniquesindependent dimensions, an unusually large were examined in order to determine hownumber, we were concerned that we might and to what extent the emotions varied alongbe tempted to retain and interpret small, the recovered dimensions. Then, in order tomeaningless dimensions that reflected mainly learn about the relative importance of theerror variance. To guard against this possi- various dimensions in differentiating thebility we decided to analyze the data using emotions, we also performed discriminantboth approaches; we used both principal analyses on the component scores producedcomponents analysis (PCA) with varimax ro- by the PCA.

Table 3

Items on the Dimensional Ratings Questionnaire

Pleasantness

(Pleasant) How pleasant or unpleasant was it to be in this situation?(Enjoy) How enjoyable or unenjoyable was it to be in this situation?

Attentional activity

(Consider) Think about what was causing you to feel happy in this situation. While you were feeling happy, towhat extent did you try to consider this thing further, or to what extent did you try to shut it out?

(Attend) Think about what was causing you to feel happy in this situation. When you were feeling happy, towhat extent did you try to devote your attention to this thing, or divert your attention from it.

Control

(Situational-Control) When you were feeling happy, to what extent did you feel that circumstances beyondanyone's control were controlling what was happening in this situation?

(Self-Control) When you were feeling happy, to what extent did you feel that you had the ability to influencewhat was happening in this situation?

(Other-Control) When you were feeling happy, to what extent did you feel that someone other than yourself wascontrolling what was happening in this situation?

Certainty

(Understand) When you were feeling happy, how well did you understand what was happening around you inthis situation?

(Uncertain) When you were feeling happy, how uncertain were you about what was happening in this situation?(Predict) When you were feeling happy, how well could you predict what was going to happen in this situation?

Goal-path obstacle

(Problem) Think about what you wanted when you felt happy in this situation. While you were feeling happy, towhat extent did you feel there were problems that had to be solved before you could get what you wanted?

(Obstacle) Think about what you wanted when you felt happy in this situation. When you were feeling happy, to

what extent did you feel there were obstacles standing in the path between you and getting what you wanted?Legitimacy

(Fair) Think about what was causing you to feel happy in this situation. When you were feeling happy, how fairdid you think this thing was?

(Cheated) When you were feeling happy in this situation, to what extent did you feel cheated or wronged?Responsibility

(Self-Responsibility) When you were feeling happy, how responsible did you feel for having brought about theevents that were making you feel happy in this situation?

(Other-Responsibility) When you were feeling happy, how responsible did you think someone or something otherthan yourself was for having brought about the events that were making you feel happy in this situation?

Anticipated effort(Exert) When you were feeling happy, to what extent did you feel that you needed to exert yourself to deal with

this situation?(Effort) When you were feeling happy, how much effort (mental or physical) did you feel this situation required

you to expend?

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Data Estimation

Because both PCA and SINDSCAL requirecomplete data sets, we estimated values formissing data points. Of the 4320 (18 items X15 emotions X 16 subjects) possible datapoints 67, or 1.55% were coded as missing,54 of them due to the failure of three subjectsto recall past experiences associated withparticular emotions. One subject claimednever to have experienced shame, and 2subjects did not what the word contemptmeant. An additional 9 data points werecoded as missing because one subject indi-cated that she alternated between both polesof the two attentional activity dimension itemsduring her recalled sad, contemptuous,shameful, and guilty experiences, and betweenboth poles of just the consider item duringthe challenging experience. The remaining 4missing points appear to be random failuresto respond to particular items.

The missing data points were estimated usinga two-step regression procedure (BMDPAM,Dixon, 1981). Each item for every emotionwas entered into the analyses as a separatevariable (e.g., subjects' ratings on the enjoyitem for happiness and sadness were separatevariables). In order to stay within analysissize limitations we ran 7 subanalyses, eachof which included all of the variables designedto measure a single dimension (the variablesfor the anticipated effort and perceived obsta-cle dimensions were entered into a singleanalysis). The missing values were estimatedby using regression on up to two predictorvariables. An f-to-enter criterion of 4.00 wasemployed to ensure that only variables sig-nificantly correlated with the missing variableentered the regression. In 54 cases two vari-ables met this criterion and entered theregression; in 9 cases a single variable metthe criterion and was used; and in 4 cases novariables met the criterion so the variablemean was entered for the missing value. Theentire data set, including the 67 estimatedvalues, was employed in all subsequent anal-yses.

Dimensional Structure

As indicated above, both PCA with varimaxrotation and SINDSCAL were employed toevaluate the dimensional structure of the

emotional experiences. In the PCA each emo-tional experience was treated as a separateobservation (N = 240).5 In order to performthe SINDSCAL analysis, the raw ratings hadto be transformed into distance estimates.For these estimates we employed profile dis-tance measures instead of correlations, be-cause profile distances preserve informationabout means and standard deviations, whichis lost in correlational data (Kruskal & Wish,1978). In particular, we employed scale profiledistances; summing across subjects, a separatematrix containing distance estimates betweenall possible pairs of emotions was computedfor each scale (Equation B, Kruskal & Wish,1978, p. 72). Scale matrices were used insteadof the more commonly employed subjectmatrices for two reasons: First, the scalematrices tend to produce clearer, more stableresults (Kruskal & Wish, 1978); and second,with scale matrices, the dimension weightsproduced by SINDSCAL are similar to scalefactor loadings, and are useful in interpretingthe recovered dimensions.

The factor patterns and dimension weightsfound in the six-dimensional solutions of thePCA and the SINDSCAL analyses are presentedin Tables 4 and 5, respectively.6 The six-

s Due to the within-subjects design, individual differ-

ences in scale usage might inflate the intercorrelationsamong the variables, which in turn, could affect the

solution. To check against this possibility, in addition tothe raw-score analyses reported here, the PCA was also

performed on deviation scores (computed by subtractingeach subject's mean for a given variable from his or her

scores for that variable). The results of the two analyseswere virtually identical. Across the two solutions, boththe rotated component loading of the original variables,and the mean location estimates of the 15 emotions,were correlated above .99 along all six of the recovered

dimensions.6 For purposes of clarity the dimensions in Table 5

have been reordered to correspond to the analogousdimensions in Table 4. In both tables the order in whichthe dimensions emerged are indicated by the numbers atthe top of the columns. The presence of negative loadingsin the PCA solution and their general absence in theSINDSCAU solution reflect the very different algorithmsthe two techniques employ. PCA is a correlational ap-proach, and the negative loadings reflect the fact that theratings for different scales can covary in opposite direc-tions. In contrast SINDSCAL is a Euclidean distance fittingalgorithm, and the dimension weights reflect the relativecontribution of each dimension in determining the ob-served distances in each of the scale or subject matrices.Because negative distance has no meaning in Euclideanspace, dimensions can not make negative contributions

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Table 4

Sorted Rotated Component Loadings (Pattern) of the Principal Components Analysis (6-d Solution)

ComponentCorn-

Scale 1 2 3 4 5 6 munalily

Pleasant -0.91 _ _ _ _ _ 0.90Enjoy -0.91 _ _ _ _ _ 0.89Obstacle 0.83 — — — 0.25 — 0.80Problem 0.82 _ _ _ _ _ 0.77

Cheated 0.75 -0.38 — — — — 0.74Fair -0.74 0.47 — — — — 0.81Other-responsibility — -0.89 — — — — 0.83Self-responsibility — 0.85 — — — — 0.75Other-control — -0.79 0.28 — — — 0.72Self-control — 0.60 — — 0.32 -0.41 0.69Uncertain — — 0.80 — — — 0.72Understand — — -0.78 — — — 0.66Predict — — -0.75 — — — 0.63Consider — — — 0.94 — — 0.90Attention -0.27 — — 0.88 — — 0.88Effort — — — — 0.90 — 0.89Exert 0.39 — — — 0.83 — 0.86Situational-control _ _ _ _ _ Q.89 0.88

Eigenvalue 4.52 2.96 2.07 1.84 1.81 1.11Variance1 25.1 16.5 11.5 10.2 10.0 6.2

Note. For clarity of presentation the above component loading matrix has been rearranged so that the columns appearin decreasing order of variance explained by components. The rows have been rearranged so that for each successivecomponent, loadings greater than 0.50 appear first. Loadings less than 0.25 have been omitted.' Percentage of variance explained by the component.

dimensional solutions were chosen as optimal contrary to predictions, judgments of pleas-because the six dimensions emerging from antness and legitimacy may not be distinctboth analyses were remarkably similar and appraisals. Similarly, the second PCA dimen-highly interpretable. The sixth dimension was sion and the third SINDSCAL dimension arerelatively small in both solutions, but because made up of items designed to assess bothit clearly had the same interpretation across responsibility and control, suggesting people'ssolutions it was judged to be stable and appraisal of agency may be a combination ofretained. these two estimates. This responsibility/con-

Examination of Tables 4 and 5 indicate trol dimension reflects attributions of one'sthat the recovered dimensions by and large own responsibility/control versus that of othercorrespond to the predicted dimensions. The people. The situational control item does notfirst dimension in both analyses is easily load significantly on this dimension in eitherinterpretable as a pleasantness dimension, analysis. The third PCA and fifth SINDSCALHowever, the high loadings of the fair and dimensions represent the predicted certaintycheated items on this dimension suggest that, dimension; the fourth dimension in both

analyses reflects attentional activity; and thefifth PCA and second SINDSCAL dimensions

to the distances. In theory, therefore, the minimum are interpretable as anticipated effortpossible weight is zero. In practice, negative weights are _, Y dj • £ b th analvses issometimes observed. When small these negative weights . 1 ne slxtn Dimension in DOtn analyses ISmerely reflect statistical fluctuation, or error variance, interesting in that It was not predicted to beHowever, when these negative weights are large they a separate dimension. This dimension, definedreflect major discrepancies between the SINDSCAL model by the situational control item, reflects attri-and the data, often because the dimensionality is too ulrt:nn, ah,,..* tr,e patent tn whirh the affectivelarge for the data (Kruskal & Wish, 1978). In the present "™ons at)out ̂ extent to wmch the attectivecase (Table 5) there is only one negative weight larger Situation IS controlled by Circumstances (orthan .25, and this weight (-0.28) is small. fate) versus the extent to which it is controlled

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


by any human agent (self or other). Thisdimension, considered with the above respon-sibility/control dimension, supports Rose-man's (1984) tripartite conception of agency.When evaluating agency in an emotionalsituation, people not only distinguish betweenself- versus other-responsibility/control, butthey also evaluate the extent to which thesituation is caused by circumstances beyondanyone's control (cf. also Weiner et al., 1979).

The last of our predicted dimensions wasthe perception of an obstacle. The loadingsof the items intended to measure this dimen-sion differ greatly in the two solutions. In thePCA solution they load highly on the pleas-antness dimension, whereas in the SINDSCALsolution they load moderately on both thepleasantness and anticipated eifort dimen-sions. This difference is the one major dis-crepancy between the two solutions. It reflectsthe fact that because PCA and SINDSCAL aredifferent data reduction techniques, onemethod will often find regularities in the datathat the other method treats as random vari-

ation. This discrepancy also demonstratesthat the two analysis techniques will notnecessarily lead to the same conclusions, andtherefore increases our confidence in the con-clusions suggested by both techniques. Takentogether the two solutions suggest that theperception of an obstacle is related both topleasantness and to anticipated effort.

How Do Appraisal Outcomes DifferAcross Emotions?

The results presented thus far indicate thatpeople can reliably evaluate emotional expe-riences along at least six cognitive dimensions.However, by themselves, these results do notshow us how, or the extent to which, theemotions differ along these dimensions. Thelocation estimates of the 15 emotions alongeach of the six dimensions in the SINDSCALsolution are presented in Figures 1, 2, and 3.In addition component scores were computedfor the PCA solution. Table 6 lists the meanscores for each emotion along the six PCAdimensions. The correlations between the

Table 5Sorted Dimension Weights (Pattern) of the SINDSCAL Analysis (6-d Solution)

Dimension

Scale

Goodness

of fit"

Pleasant

EnjoyObstacle

Problem

Cheated

FairOther-responsibility

Self-responsibilityOther-control

Self-control

Uncertain

Understand

Predict

Consider

Attention

Eifort

ExertSituational-control

Variance11

0.88

0.880.420.370.900.92

0.62

0.66

0.280.880.920.700.54

0.730.720.35

-0.28

21.1 15.2

0.860.94

10.8

0.40

1.040.92

17.50.753.7

0.96

0.960.930.950.910.810.830.940.72

0.690.800.790.59

0.850.94

0.920.930.89

Note. For clarity of presentation the above dimension loading matrix has been rearranged so that the columns and

rows correspond to the analogous columns and rows in Table 4. Weights less than 0.25 have been omitted.

" Goodness of fit is the correlation between the predicted and observed distances between the objects for each scale.

" Percentage of variance explained by the dimension.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


High Effort

* ChallengeShame* * Fear

* Guilt

* Frustration

* Anger

* Sadness

* Contempt

PleasantDisgust

Unpleasant

Hope

* Interest

* Boredom

* Pride

* Surprise

HappinessLow Effort

Figure 1. SINDSCAL location estimates for the 15 emotions plotted along the Pleasantness (X-axis) andAnticipated effort (Y-axis) dimensions.

SINDSCAL and PCA location estimates for allsix dimensions are very high (pleasantness:r = .98; responsibility/control: r = .98; cer-tainty: r = .89; attentional activity: r = .89;anticipated effort: r = .76; situational control:r = .89; all ps < .001) indicating that the rel-ative locations of the emotions along each ofthe dimensions are highly similar across thetwo solutions. The lowest correlation, for theanticipated effort dimension, is easily attrib-utable to the fact that the perceived obstacleitems load moderately on the anticipatedeffort dimension in the SINDSCAL solution

but do not load on this dimension in thePCA solution. The reliable ordering of theemotions along the dimensions recovered byboth techniques indicates that the dimensionalappraisals are in some way related to theemotions, and do not reflect ratings of inci-dental properties of the situations unrelatedto emotion.7

7 Because SINDSCAL primarily fits the distances betweenmean objects (in this case the emotions), it will onlyreveal dimensions along which the objects systematicallydiffer. However, because PCA derives factors from the

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.

COGNITIVE APPRAISAL IN EMOTION

Uncertain

* Fear

827

Low Attention

* Sadness

Shame

* Surprise

* Hope* Frustration

High Attention

DisgustGuilt

ContemptAnger

* Boredom

Interest

* Challenge

Happiness* Pride

Certain

Figure 2. SINDSCAL location estimates for the 15 emotions plotted along the Certainty (X-axis) andAttentional activity (Y-axis) dimensions.

Table 6 and Figures 1,2, and 3 demonstrate

the ways in which the appraisals differ for the

correlational structure of the individual observations, itcan reveal dimensions along which the individual obser-vations vary, but the mean objects do not differ. ThatPCA revealed the same dimensions as SINDSCAL makesit is highly likely that the emotions vary systematicallyalong these dimensions. Nevertheless, in order to confirmthis fact, the PCA component scores were entered intoseparate within-subjects. one-way ANOVAS for each of thesix dimensions. As expected, the main effect for emotionswas highly significant for all six dimensions, all Fs(l4,210) > 3.20, all pi< .001.

various emotions.8 Happiness, pride, surprise,

interest, hope, and challenge are all pleasant

emotions, while shame, guilt, disgust, fear,

anger, contempt, frustration, boredom, and

sadness are all unpleasant states. Challenge,

fear, anger, and frustration are associated

* Although the PCA and SINDSCAL solutions are highlysimilar, the relative locations of some of the emotionsalong some of the dimensions vary slightly across the twosolutions. Therefore, all statements in this article con-cerning the dimensional properties of the emotions arebased on consideration of both solutions.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


with particularly high levels of anticipatedeffort, whereas boredom and surprise areextremely low effort emotions. A strong senseof other-responsibility/control is associatedwith surprise and anger, whereas strong attri-butions of self-responsibility/control are as-sociated with pride, shame, and guilt. Angerand pride are linked to strong attributions ofhuman agency, but sadness and fear aredistinctively high on situational control. Thesubjects wanted to pay particular attentionwhen they were challenged, frustrated, or

interested, but to shut out what was happen-ing when they were disgusted or bored. Fi-nally, the subjects felt least certain when theywere surprised, afraid, or hopeful, but feltparticularly certain about the situation whenthey were bored, happy, or proud. Interest-ingly, three unpleasant states (boredom,shame, and guilt) are highly differentiatedfrom the other unpleasant emotions. Figure1 shows that boredom is the only negativestate that does not involve anticipated effort.Similarly, Figure 2 reveals that shame and

Situational Control

* Sadness

* Fear

Boredom *

Other-Resp/Con * Frustr tion

* Hope

Interest

Self-Resp/Con

* Challenge

Happiness

* Surprise

* Contempt

* Disgust

* Anger

* Guilt

* Shame

* Pride

Human Control

Figure 3. SINDSCAL location estimates for the 15 emotions plotted along the Self/Other-respdnsibility/control (X-axis) and Situational control (Y-axis) dimensions.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Table 6Locations of Emotion Means Along the PCA Components

Component

Emotion

Happiness

SadnessAngerBoredomChallengeHopeFear

InterestContempt

DisgustFrustration

SurprisePride

ShameGuilt

Pleasant'

-1.460.870.850.34

-0.37-0.50

0.44-1.05

0.890.38

0.88-1.35

-1.250.730.60

Responsibility/

Control"

0.09-0.36-0.94

-0.190.44

0.15-0.17

-0.13-0.50-0.50-0.37

-0.940.81

1.311.31

Certain0

-0.460.00

-0.29-0.35-0.01

0.46

0.73-0.07

-0.12-0.39-0.08

0.73-0.32

0.21

-0.15

Attention"

0.15-0.21

0.12-1.27

0.520.310.030.70

0.80-0.96

0.600.400.02

-0.11-0.36

Effort'

-0.33-0.14

0.53-1.19

1.19

-0.180.63 •

-0.07-0.07

0.060.48

-0.66-0.31

0.07

0.00

Situational-Control'

-0.211.15

-0.960.12

-0.200.350.59

0.41-0.63-0.19

0.22

0.15-0.46-0.07-0.29

Note. Scores are standardized.* Pleasantness: high scores indicate increased unpleasantness.* Responsibility/Control: high scores indicate increased self-responsibility/control.0 Certainty: high scores indicate increased uncertainty." Attentional activity: high scores indicate increased attentional activity.

' Effort: high scores indicate increased anticipated effort.' Situational control: high scores indicate increased situational control.

guilt differ from the other unpleasant statesin attributions of responsibility/control. Bothshame and guilt are associated with high self-responsibility/control, whereas the other un-pleasant emotions are associated with other-responsibility/control.

In order to investigate the relative impor-tance of the dimensions in differentiatingamong the emotions, a series of four stepwisediscriminant analyses (BMDP7M, Dixon,1981) were performed on the PCA componentscores. (The results of these analyses shouldbe viewed as exploratory rather than infer-ential, because they employ a between-sub-jects model unlike the within-subjects designemployed in the study, and the practicalconsequences of violating the between-subjectsassumption are not well understood.) Thefirst analysis attempted to discriminate be-tween all 15 emotional states. Follow-up anal-yses examined the pleasant and unpleasantemotions separately. Finally, because bore-dom, shame, and guilt are in some ways quitedistinct from the other unpleasant states, afinal analysis was run on the unpleasantstates, excluding these three emotions. The

results of these analyses are presented inTable 7.9 As can be seen from the table, inall four analyses the emotions were correctlyclassified at rates that were much better thancould be expected by chance alone, onceagain illustrating the presence of systematicrelations between cognitive appraisals andemotional experiences.

For each analysis Table 7 also presents thethree dimensions that best discriminateamong the emotions when each dimension isentered into the discriminant analysis byitself. The values in parentheses are the per-centages of experiences correctly classifiedusing each of the dimensions. This portion

5 Although not depicted in Table 7, the discriminant

analyses also produced F values corresponding to between-subject one-way ANOVAS along each of the dimensions.The F values produced in the analysis including all of

the emotions are the between-subject analogs of the Fvalues of within-subject one-way ANOVAS, which as men-tioned above were also performed. Comparison of thetwo sets of F values revealed that they were highly similarfor every dimension (r = .999), suggesting that the con-sequences of using an inappropriate model in the dis-criminant analyses are probably minimal in this instance.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Table 7Discriminant Analyses Using PCA Component Scores

Emotions analysed

Percentage All Pleasant' Unpleasant-lb Unpleasant-2c

Percentage classified"Percentage expected'

3 best discriminators1

23

42.16.7

Pleasant (16.7)'

Effort (15.4)

Situational-control (15.0)

52.116.7

Responsibility/Control (36.5)

Effort (32.3)Pleasant (29.2)

39.611.1

Situational-

control (27.8)Effort (18.1)Attention (17.4)

41.716.7

Situation al-

control (40.6)Attention (25.0)

Certain (24.0)

Note. Each experience was classified using a jackknifed classification function. That is, the classification function for

any given experience was computed using all the data except that experience." Includes: Happiness, Challenge, Hope, Interest, Surprise, and Pride.b Includes: Sadness, Anger, Boredom, Fear, Contempt, Disgust, Frustration, Shame, and Guilt.c Includes: Sadness, Anger, Fear, Contempt, Disgust, and Frustration.

" Percentage of experiences classified correctly with all of the components entering the analysis.' Percentage of experiences expected to be classified correctly by chance.' The labels list the three dimensions for each analysis which correctly classify the most experiences when entered intothe discriminant function by themselves. The values in parentheses represent the percentage of experiences correctly

classified by using the single dimension.

of the table reveals that the importance of adimension in discriminating between two ormore emotional states depends heavily on thestates considered. When all of the emotionsare considered simultaneously, the pleasant-ness dimension is the single most importantdimension for distinguishing between them.This finding is in line with the virtuallyuniversal finding of this dimension in previousstudies. However, when the pleasant and un-pleasant states are examined separately theimportance of pleasantness diminishes greatly,and in its stead evaluations of agency pre-dominate.

Given that the situation is basically pleas-ant, the consideration that best discriminateswhat we actually feel is our sense of howmuch control we ourselves (rather than some-one else) had in initiating and maintainingthe situation. Given a negative situation, ourappraisal of the power of impersonal circum-stances is most important in determining ouremotion. By itself, the dimension of situa-tional control can correctly classify over 40%of the experiences associated with the sixemotions of sadness, anger, fear, disgust, con-tempt, and frustration, which is as well asthey can be classified when all of the dimen-sions are considered. Clearly, considerations

of human versus circumstantial agency areimportant in discriminating among negativeemotional states, much less so in differentiat-ing positive emotions.

Does the Dimensional Structure Differ forSubsets of Emotions?

The finding that the discriminatory powerof a dimension depends upon the set ofemotions considered suggests that the localdimensional structure obtained by consideringa set of similar (e.g., all pleasant or allunpleasant) emotions may differ from theglobal structure obtained by considering abroad range of emotional experience. Cog-nitive distinctions that are relatively uncom-mon or unimportant when considering avariety of emotions may be very importantfor distinguishing among a set of closelyrelated states, whereas distinctions that areglobally important may not discriminate wellbetween similar emotions.

In order to explore possible differencesbetween local and global dimensional struc-tures we performed separate PCAs on thepleasant (happiness, challenge, hope, interest,surprise, and pride; N = 96) and unpleasant(sadness, anger, boredom, fear, contempt, dis-

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


gust, frustration, shame, and guilt; N = 144)experiences.10 The analysis of the pleasantexperiences yielded a six-dimensional solutionthat accounted for 78% of the variance. Theloading patterns of the recovered dimensionswere almost identical to those of the globalsolution, and the six dimensions in bothsolutions clearly had the same interpretations.The only notable difference between the twosolutions was in the loadings of the perceivedobstacle items. In the pleasant solution theobstacle and problem items loaded moderatelyon both the pleasantness dimension (-0.68and -0.64, respectively) and the anticipatedeffort dimension (0.47 and 0.43, respectively),and therefore resembled the global SINDSCALsolution more closely than the global PCAsolution. In general, however, the strong sim-ilarities between the two solutions providedlittle evidence of differences between thepleasant and global dimensional structures.

When we examined the dimensional struc-ture of the unpleasant emotions a strikingdifference was found. The analysis of theunpleasant experiences revealed a six-dimen-sional structure that accounted for 75% ofthe variance, which was in most respectshighly similar to both the global and pleasantsolutions; however the loadings of the legiti-macy items were very different. The fair andcheated items no longer loaded highly on thepleasantness dimension as they did in boththe global and pleasant solutions (loadings <0.25), instead they loaded highly on the re-sponsibility/control dimension (0.83 and—0.70, respectively). This finding suggeststhat the appraisals of legitimacy and pleas-antness may be distinct after all. Although ingeneral, unpleasant experiences are viewed asbeing less fair than pleasant ones (as theglobal solution indicates), the present findingindicates that for unpleasant experiences theperceived fairness of the situation is notrelated to the degree of unpleasantness. In-stead, for these experiences, fairness is im-portant for appraisals of responsibility/con-trol; if someone else is responsible for a badsituation, it is seen as less fair than if youyourself are responsible. This distinction doesnot seem to be made for pleasant experiences;the more pleasant the situation is, the fairerit is seen as being.

Discussion

Our results show that people's emotionsare intimately related to their cognitive ap-praisals of their circumstances. The subjectsrated past emotional experiences along sixappraisal dimensions, and different patternsof appraisals were strongly associated withdifferent emotions. The strength of this rela-tion was demonstrated in the discriminantanalyses, in which 15 different emotions werecorrectly predicted over 40% of the time onthe basis of the corresponding patterns ofcognitive appraisal. This rate is more thansix times what would be expected by chance.If we know how a person sees his or herrelation to the environment, we are betterable to identify that person's emotional state;conversely, if we know what a person isfeeling, we can deduce much about how thatperson is interpreting his or her circumstances.

What sets this study apart from its prede-cessors is not the discovery of new and here-tofore unimagined dimensions of emotionalexperience, but the clear demonstration ofthe six dimensions simultaneously. Becausedimensions beyond pleasantness and level ofactivation have seldom been found concur-rently in the past, there has been muchuncertainty about how any additional dimen-sions should be interpreted, and whether theywere in fact distinct dimensions or merelydifferent interpretations of a single third di-mension. The simultaneous demonstration ofsix highly interpretable dimensions allows usto be confident that these dimensions aredistinct entities. Further, we can use thesedimensions to characterize the different emo-tions in much richer detail than was previ-ously possible.

Properties of the Emotions

Happiness and pride. Subjects' experiencesof happiness and pride were very similar

10 SINDSCAL analyses were not performed because thedistance fitting algorithm can always fit N objects arbi-trarily into N - 1 dimensions. Therefore the small num-bers of objects to be scaled in these analyses (pleasant,N = 6 emotions; unpleasant, N - 9 emotions) precludesthe finding of a meaningful SINDSCAL solution with morethan three or four dimensions.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


along most of the dimensions. Both emotionswere described as extremely pleasant statesinvolving very little effort, a high level ofcertainty about the situation, and a strongdesire to pay attention. Both were associatedwith attributions of human control and spe-cifically with one's own sense of responsibilityand control, but these attributions were con-siderably stronger for pride. These differenceswere readily apparent in the experiences sub-jects described. For pride, the subjects in-variably recalled a personal achievement, suchas receiving high school graduation awards,being admitted to Stanford University, receiv-ing a part in a play, and winning an election.This sense of personal achievement was notemphasized in the happy experiences. Mostsubjects (11) remembered enjoyable timesspent with other people—being reunited withclose friends or relatives after a long absence,being at a party with friends, or going outon a pleasant date.

Interest and hope. Subjects' experiencesof interest and hope were quite similar alongall the dimensions except pleasantness andcertainty. Subjects described interest as ahighly pleasant state, whereas hope was con-siderably less so. Subjects remembered antic-ipating somewhat more effort when they feltinterested or hopeful, and being less certainabout their situation than they did for hap-piness or pride. However, for interest thesubjects were still fairly certain, whereas forhope they were quite uncertain about whatwas happening. Both emotions were accom-panied by a desire, particularly strong forinterest, to attend to the situation. In boththe hopeful and interesting experiences eventswere seen as largely controlled by the situa-tion, and responsibility/control was not con-sistently attributed to either the self or toothers.

Devout attention is a distinguishing featureof interest. Subjects consistently describednovel experiences in which they wanted tolearn something about the situation, such asvisiting a foreign country, viewing surgery,and going to the opera for the first time. Incontrast, moderate uncertainty about the oc-currence of a desired outcome was salient inhope. Subjects described situations in whichthey felt that it was possible, but not certain,that they would get what they wanted. For

instance, one subject was hopeful when hewas called back for an audition for a part ina play, and another felt hopeful when shelearned that a political candidate she wasworking for was leading in the polls. Further,a number of these situations were ones inwhich the subjects could do little to affectthe outcome, such as waiting for collegeacceptances, and waiting to see if a friendor relative would recovery from surgery orcancer.

Challenge. Like hope, challenge was de-scribed as less pleasant than the other positiveemotions, and it was further distinguished byan extremely high level of anticipated effort.However, challenge resembled interest in thatsubjects felt fairly certain about the situationand strongly desired to attend to it. It wassimilar to happiness in that it was associatedwith moderately strong attributions of humanagency and self-responsibility/control. Fortheir challenging experiences subjects de-scribed situations in which they felt fairlyconfident that they could achieve a desiredgoal but knew that it would require a greatdeal of effort. Six subjects described how theyfelt immediately prior to and during athleticevents, such as when competing against anequally matched or slightly better opponent,or when skiing an advanced slope. Othersubjects described more intellectual challengessuch as taking an exam, and attempting tosucceed at a difficult job.

Surprise. Surprise was described as beingextremely pleasant, like happiness, pride, andinterest. It was also similar to happiness andpride in that the subjects anticipated expend-ing very little effort. However, subjects feltmore uncertain in surprising situations thanthey did in any others. This uncertainty wasaccompanied by a strong desire to attend tothe situation. Surprise was associated withattributions of human agency, and, unlikeother pleasant emotions, it was seen as causedby other people. Subjects always described anunexpected event, such as receiving an A ina course for which a C was expected, gettinga ticket to Hawaii as a birthday present, andgetting unexpected flowers from a boyfriend.

Boredom. Subjects described boredom asan unpleasant state quite different from theother negative emotions. As with happinessand pride, subjects anticipated expending very

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


little effort and were very certain about theirsituations. Boredom was distinguished fromevery other emotion by an extremely strongdesire to shut out or ignore what was hap-pening. It was also seen as caused by thesituation or by other people. Low effort andlow attentional activity appear to be centralto the experience of boredom. Subjects alwaysdescribed experiences in which they werecompelled to remain in a situation wherethere was littl? or nothing to keep their mindsoccupied. Ten subjects described how theyfelt during classes in which the professor wasdull or incomprehensible. Other subjects de-scribed how they felt while flying across thecountry, and being bed-ridden for three daysafter minor surgery.

Shame and guilt. The cognitive appraisalsassociated with shame and guilt were virtuallyindistinguishable from each other. Both emo-tions were described as unpleasant states re-quiring a relatively large amount of effort.Subjects remembered feeling moderately cer-tain about the situation, although less sowhen ashamed than when guilty. They alsoremembered wanting to avoid thinking aboutthe situation. In shame and guilt, unlike theother negative emotions, a sense of self-blameis central. For both emotions subjects alwaysdescribed situations in which they had donesomething that they regretted such as stealingsomething, lying, spreading gossip about afriend, or failing to meet an obligation.

Disgust. In many respects, the appraisaloutcomes associated with disgust were similarto those for shame and guilt; it is an unpleas-ant state involving considerable effort, fairlyhigh levels of certainty, and a strong unwill-ingness to attend to the situation (even stron-ger than for shame or guilt). Disgust was alsoassociated with attributions of human control.However, whereas in shame and guilt the selfwas held responsible, in disgust someone elsewas seen as responsible. Many subjects de-scribed situations in which someone else didsomething physically repulsive that theywanted to shut out and get away from. Forinstance, six subjects described experiencesin which someone was drunk and obnoxiousand in four of the cases eventually got sick.Other subjects described situations in whichthey were exposed to something they foundcognitively repulsive. Three subjects described

being disgusted by someone espousing whatthey felt were unacceptable, sexist, or racistattitudes.

Anger and contempt. Aside from a slightdifference along the anticipated effort dimen-sion, the appraisal outcomes associated withanger and contempt were much alike. Bothwere described as extremely unpleasant states,and both were associated with a considerableamount of anticipated effort, particularly an-ger. Like disgust, guilt, and interest, bothemotions were associated with a fair degreeof certainty about the situation, but neitherwas consistently associated with a strong desireto either attend to or ignore the situation.Both emotions were associated with strongattributions of human agency and other-re-sponsibility/control. The experiences de-scribed for these two emotions were verysimilar. Typically they involved someone un-fairly doing something that had negative con-sequences for the subject. Interestingly, theexperiences associated with contempt tendedto involve interpersonal relationships. Sevensubjects described situations in which a personhad backed out of a date with them, brokenup a relationship with them, or had starteddating someone with whom they had previ-ously been involved. In contrast, anger wasassociated with a broader range of experi-ences, including receiving an unfair grade,having personal property destroyed, and hav-ing wine poured over one's head in a restau-rant.

Frustration. Frustration was similar toanger in that it was described as a veryunpleasant emotion in which the subjectsanticipated expending extreme effort. It wasaccompanied by a fair degree of uncertaintyabout the situation and by a stronger desireto attend to the situation than for any othernegative emotion. Frustration was associatedwith moderately strong appraisals of bothsituational control and other-responsibility/control. An important characteristic of thisemotion seems to be failure at something forwhich success is expected. The failure is oftenaccompanied by uncertainty about its causes.Frustrating expeiences included doing poorlyon an exam for which the subject felt shewas well-prepared, being unable to get acomputer program to run, and trying to writea paper in which the subject knew what she

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


wanted to say, but "couldn't write it downcorrectly."

Sadness. Subjects described sadness asbeing extremely unpleasant. Like contempt,it was associated with a high degree of antic-ipated effort, though less so than anger orfrustration. As with frustration, subjects weresomewhat uncertain about the situation, but,as with shame and guilt, subjects indicated amoderate desire to avoid thinking about it.Sadness was distinguished from all otheremotions by extreme appraisals of situationalcontrol. These appraisals were accompaniedby moderately strong appraisals of other-responsibility/control. Apparently, a beliefthat the unpleasant situation is controlled byimpersonal circumstances and that nothingcan be done to set it right is crucial tosadness. Several subjects described their sad-ness at the death of someone they loved. Inother situations, in which perhaps the subjector someone else objectively had some control,subjects still reported that they felt that thesituation was beyond anyone's control. Forinstance, one subject described being sadwhen she learned that her parents were gettinga divorce "because I didn't want it, I knewit had to be, and yet I didn't want it."Another subject became sad when a numberof problems—difficulties with school workand problems with her family and boy-friend—all converged on her at once. Shedescribed her sadness by saying, "I felt likethe whole world was against me, and every-thing was wrong, and I—I had lost control."

Fear. Fear was described as an unpleasantstate demanding extreme amounts of effort.Like surprise, fear was associated with max-imal uncertainty about the situation. However,this uncertainty was not consistently accom-panied by a strong desire to either attend toor shut out the situation. Fear was furtherassociated with strong attributions of situa-tional control, though not as strong as withsadness, and tended to be associated withappraisals of other-responsibility/control. Fearseems to be characterized by uncertaintyabout whether or not one will be able toescape or avoid an unpleasant outcome. Inaddition to the fearful experiences alreadydescribed, subjects told of their feelings whilebeing held up at knife-point, while being lostin the mountains, and after losing control

both while driving in a blizzard and whileriding a bicycle down a very steep hill.

Properties of the Dimensions

The descriptions of the individual emotionsdemonstrate that each emotion (with theexceptions of shame and guilt, and anger andcontempt) is characterized by a unique pat-tern of cognitive appraisals. It is importantto consider the role that each of the cognitiveappraisal dimensions plays in distinguishingamong the emotions.

Pleasantness. As in every study examiningthe dimensions of emotional experience, werecovered a strong and clear pleasantnessdimension. This dimension subsumed theproposed dimensions of legitimacy and per-ceived obstacle, and accounted for a largeproportion of the variance (20-25%). Clearly,emotional experiences are readily classifiedas pleasant or unpleasant. Beyond this initialclassification, however, pleasantness is of lim-ited usefulness in differentiating among theemotions. In classifying the emotions, pleas-antness is the single most important dimen-sion when all 15 emotions are considered,but it does not help us much in makingfurther differentiations when only the pleasantor only the unpleasant emotions are consid-ered. Among the pleasant states, hope andchallenge tend to be less pleasant than theothers, but the unpleasant emotions all areabout equally unpleasant. Thus, appraisaloutcomes along the other dimensions need tobe taken into account.

Anticipated effort. Amount of anticipatedeffort is particularly important in differen-tiating the pleasant emotions. Challenge isassociated with very high anticipated effort,interest and hope with moderate effort, andhappiness, pride, and surprise with very loweffort. Among the unpleasant emotions, bore-dom is distinguished from the others by itsextremely low levels of anticipated effort; allthe rest tend to be associated with high levelsof anticipated effort. When the situation isbad we try to ameliorate it, and that requireseffort.

Certainty. Fear, hope and surprise aredistinguished from the other emotions byuncertainty. It should be noted that of thesix recovered dimensions, the solution for

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


certainty is the least satisfactory. Althoughthis dimension is clearly defined by the scalesdesigned to measure certainty, the commu-nalities or goodness of fit of these items areonly moderate, suggesting that the subjectsfound this dimension especially difficult torate consistently. Examination of the experi-ences subjects described suggests that therecovered certainty dimension may be com-bining two distinct appraisals: the violationof one's past expectations and uncertaintyabout future developments. For surprise, sub-jects invariably described situations in whichsomething unexpected had happened. In con-trast, for hope and, to some extent, fear,subjects described situations in which theywere unsure about what was going to hap-pen—whether they would be hurt, arrested,or accepted by a good college.

Attentional activity. As the early studiesof the dimensions of emotional facial expres-sion suggest, the motive to keep seeking orto look away is important in differentiatingemotional states, particularly for the unpleas-ant emotions. Frustration signifies a need toincrease attention, whereas boredom and dis-gust are distinguished by rejection. All of thepleasant emotions are associated with a strongdesire to pay attention, particularly interestand challenge. Although attention is distinctfrom certainty, the two dimensions are related.Certainty can be associated with a desireeither to attend to the situation or to shut itout, but strong uncertainty is never associatedwith inattention.

Responsibility and control. Consistent withthe ideas of Roseman (1984), we find thattwo dimensions are useful in understandingagency, and both of these dimensions providesubstantial differentiation among the emo-tions. Surprise differs from all other pleasantemotions in its attributions of other-respon-sibility/control, whereas shame and guilt differfrom all other unpleasant emotions in attri-butions of self-responsibility/control. Al-though the situational control dimension ac-counts for a small proportion of the overallvariance (4-6%), it is essential for under-standing the unpleasant emotions. Sadnessand anger are poles apart—when we areangry we believe a person is responsible, butwhen we are sad we believe that we arevictims of circumstance. In discriminating

among the fundamental unpleasant states ofsadness, anger, fear, disgust, contempt, andfrustration, the situational control dimensionis almost as crucial as all of the dimensionscombined (see Table 7).

Although the two dimensions involvingresponsibility and control are orthogonal, theappraisal outcomes along these dimensionsare clearly related. When we attribute anoutcome to ourself, we attribute it to ourselfalone, but strong attributions of both situa-tional control and other-responsibility/controlcan co-occur, as happens in sadness. It maybe that people first appraise the extent towhich they are in control of the situationand then, if they do not feel in control, seekother human agents or uncontrollable cir-cumstances.

In sum, each of the cognitive dimensionsis important in differentiating one or moreof the emotions from the others. The pleas-antness dimension chunks emotions into goodones and bad ones, but does not get us muchfurther. Sorrow and anger are the two mostunpleasant emotions, but they are very dif-ferent on other dimensions. The relative im-portance of a dimension in differentiatingamong a set of emotions depends on theuniverse of emotions to be differentiated.

Future Directions

We have demonstrated a strong relationbetween the interpretation of an event andthe emotional reaction to it. We have alsoidentified several of the important dimensionsof this interpretation and shown how thesedimensions combine to make each of theemotions we have studied unique. We hopethat we have helped to make the next ques-tions a little clearer.

Are there further dimensions of emotionalexperience? Direct rating methods are unableto reveal properties of a phenomenon thatare not specifically measured. Therefore, wecannot claim that our six dimensions captureall of emotional experience. Becaure we de-rived our dimensions from a comprehensivereview of previous theory and empirical find-ings, we suspect that they are real. Neverthe-less, there may be more. We have alreadysuggested that our certainty dimension mayconfound two distinct appraisals. There is

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


also some reason to believe that the legitimacyand perceived obstacle dimensions, eventhough they appeared to be redundant withpleasantness in this study, should be furtherexplored. When the unpleasant emotions areconsidered alone, what is seen as legitimateis related to who is in control, not to overallpleasantness. Thus appraisals of legitimacymay be distinct from appraisals of eitherpleasantness or responsibility/control. Simi-larly, the items intended to measure perceivedobstacle sometimes load moderately on antic-ipated effort, suggesting that this appraisalmay be distinct from the appraisals of eitherpleasantness or eifort. Our results clearlyindicate that typically, in college students,pleasantness is associated with fairness andfew obstacles. However, by asking for oneexample of each basic emotion, we wereprobably encouraging reports of fairly pro-totypical emotional experiences. Thus, wecannot tell whether experiences that are unfairbut pleasant, or free of obstacles but unpleas-ant, are impossible, or whether they are un-usual but meaningful emotional experiences.

A further limitation of this study is that itemployed expressive subjects in a within-subjects design. This design may have en-couraged subjects to exaggerate relatively mi-nor differences between the emotions, andexpressive subjects may have reported onexperiences that were biased in some manner.It is possible that had a between-subjectsdesign been employed with randomly selectedsubjects, somewhat different relationships be-tween the emotions might have been found.

We are currently conducting a series ofstudies that addresses these limitations. In abetween-subjects design, we are asking sub-jects (not selected for expressivity) to describepast emotional situations in which particularappraisal outcomes or combinations of out-comes applied. By avoiding the use of com-mon emotional labels to elicit the experiences,and by studying unusual combinations ofappraisal outcomes, we can avoid studyingonly prototypical emotional experiences andcan begin to determine which cognitive ap-praisal outcomes are central for particularemotions, and which are more optional. Forinstance, in the present study subjects consis-tently associated pleasantness with surprise,possibly because the first surprising experi-

ences they thought of were pleasant. Life isfull of unpleasant surprises, however, and weexpect that we did not hear about thembecause the surprise was brief and the un-pleasant emotion more lasting, and so moredefinitive. In further research we will askabout such plausible but (in the present study)unusual responses as unpleasant surprise.

How are cognitions and emotions related?The cognitive mapping of emotional experi-ence revealed by our subjects suggests thatcognitive appraisals have an important rolein the experience of emotion. The nature ofthis role needs to be further examined. Inthis study, only dimensions originally thought,and later found, to be closely related toemotional experience were recovered and ex-amined. However, we do not believe that alldimensions of cognitive appraisal are relatedto emotion, nor do we believe that the di-mensions we examined are always related toemotion. We suspect, for example, that theappraisal of a person's eye color seldom sys-tematically affects one's emotional state. Sim-ilarly, we suspect that it is possible to blamesomeone else for an unpleasant event thatdoes not affect us personally without becom-ing angry.

Along with other contemporary emotionstheorists (e.g., Arnold, 1960; Izard, 1977;Lazarus et al., 1980; Plutchik, 1980; Rose-man, 1984; Scherer, 1982), we believe thatemotions represent adaptive responses to thedemands of the environment. Interpretationof the perceived situation in terms of pre-vailing goals creates an emotional experiencethat allows the organism to respond adaptively(Scherer, 1982). This view implies that cog-nitive appraisals will lead to an emotionalresponse primarily when they are perceivedas having adaptive significance for the organ-ism's well-being. We believe that the dimen-sions revealed in the present study are animportant step in helping us understand thekinds of perceptions that trigger emotions.However, it is still necessary to examine theconditions under which such perceptions will,and will not, result in emotional experience.

We also agree with Scherer (1982) thatemotional behavior is an ongoing process—that the organism is constantly evaluatingand responding to its environment, andtherefore, that emotions are dynamic pro-

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


cesses rather than steady states. We havestarted with a method that elicits a fairlystatic view of the emotions, demonstratinghow people saw the world at a particular timewhen they experienced a given emotion. Thisapproach allowed us to identify relevant ap-praisal dimensions to be further explored inmore dynamic contexts.

For instance, challenge and frustration weresimilar to each other in that both occurredwhen subjects wanted to achieve somethingand thought that they should be able tosucceed. The major difference was that whenchallenged, subjects felt good and were fairlyconfident of success, but when frustrated,they felt bad and thought that they werefailing for reasons they did not quite under-stand. Thinking dynamically, we would pre-dict that when faced with a difficult goal,people often start out feeling challenged, butthat the feeling of challenge will evolve intoone of frustration if the goal eludes them.Further, we would predict that if the personperceived that someone else was preventingthem from reaching the goal the frustrationwould turn to anger, but that if the impedi-ment were due to uncontrollable circum-stances the frustration would become sadness.

The appraisal process itself is unquestion-ably dynamic. Outcomes along one dimensioninfluence outcomes along other dimensions.For instance, both pleasantness and uncer-tainty involve attention, unpleasantness in-volves anticipated effort. However, the self isonly responsible when the situation is not.Why do these dependencies exist? Is it becausewe failed to consider large enough regions ofemotional experience, or is the appraisal pro-cess structured so that some decisions pre-clude other decisions, and perhaps even ruleout appraisals along whole dimensions? Hav-ing demonstrated that examination of cog-nitive appraisals can reveal much about thesimilarities and differences among emotionalstates, it becomes important to examine howthese appraisals affect the ongoing experienceof emotion.

References

Abelson, R. P. (1983). Whatever became of consistencytheory? Personality and Social Psychology Bulletin. 9,37-54.

Abelson, R. P., & Sermat, V. (1962). Multidimensional

scaling of facial expressions. Journal of Experimental

Psychology, 63, 546-554.Arnold, M. B. (1960). Emotion and personality (2 Vols.).

New York: Columbia University Press.Averill, J. R. (1975). A semantic atlas of emotional

concepts. JSAS Catalogue of Selected Documents in

Psychology, 5. 330 (Ms. No. 421).Berglund, B., Berglund, U., & Engen, T. (1982). Multi-

dimensional analysis of emotions expressed in photo-

graphs and words (Report No. 583). Stockholm: De-partment of Psychology, University of Stockholm.

Block, J. (1957). Studies in the phenomenology of emo-tions. Journal of Abnormal and Social Psychology, 54,

358-363.Bottenberg, E. H. (1975). Phenomenological and opera-

tional characterization of factor-analytically deriveddimensions of emotion. Psychological Reports, 37,

1253-1254.Bush, L. E. (1973). Individual differences multidimen-

sional scaling of adjectives denoting feelings. Journal

of Personality and Social Psychology, 25, 50-57.Cannon, W. B. (1929). Bodily changes in pain, hunger,

fear, and rage (2nd ed.). New York: Appleton-Century.Chatfield, C, & Collins, A. J. (1980). Introduction to

multivariate analysis. New York: Chapman & Hall.Clore. G. L.. & Ortony, A. (1984). Some issues for a

cognitive theory of emotion. Cahiers de PsychologicCognitive. 4, 53-57.

Daly, E. M., Lancee, J. W., & Polivy, J. (1983). A conicalmodel for the taxonomy of emotional experience.

Journal of Personality and Social Psychology, 45, 443-457.

Darwin, C. (1965). The expression of the emotions in

man and animals. Chicago: University of ChicagoPress. (Original work published 1872)

Davitz, J. R. (1969). The language of emotions. NewYork: McGraw-Hill.

DeRivera, J. (1977). A structural theory of the emotions.Psychological Issues, 10 (4. Monograph No. 40).

Dixon, W. J. (Ed.). (1981). BMDP statistical software.

Berkeley, CA: University of California Press.Ekman, P., & Friesen, W. V. (1971). Constants across

cultures in the face and emotion. Journal of Personality

and Social Psychology, 17, 124-129.

Ekman, P., Friesen, W. V., & Ellsworth, P. (1982).

Research foundations. In P. Ekman (Ed.), Emotion in

the human face (2nd ed., pp. 1-143). New York:Cambridge University Press.

Engen, X, Levy, N., & Schlosberg, H. (1958). Thedimensional analysis of a new series of facial expres-

sions. Journal of Experimental Psychology, 55, 454-458.

Frijda, N. H. (1969). Recognition of emotion. In L.Berkowitz (Ed.), Advances in experimental social psy-chology (Vol. 4, pp. 167-223). New York: AcademicPress.

Frijda, N. H., & Philipszoon, E. (1963). Dimensions ofrecognition of expression. Journal of Abnormal andSocial Psychology, 66, 45-51.

Izard, C. E. (1971). The face of emotion. New York:Appleton-Century-Crofts.

Izard, C. E. (1972). Patterns of emotions. New York:Academic Press.

Izard, C. E. (1977). Human emotions. New York: Plenum.

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.


Izard, C. E., & Buechler, S. (1980). Aspects of conscious-

ness and personality in terms of differential emotionstheory. In R. Plutchik & H. Kellerman (Eds.), Emotion:

Theory, research, and experience: Vol. 1. Theories of

emotion (pp. 165-187). New York: Academic Press.

James, W. (1950). The principles of psychology (Vol. 2).New York: Dover Publications. (Original work publishedin 1890)

Kruskal, J. B., & Wish, M. (1978). Multidimensional

scaling. Sage University paper series on quantitativeapplications in the social sciences. (Series No. 07-011). Beverly Hills: Sage.

Lazarus, R. S. (1968). Emotions and adaptation: Con-ceptual and empirical relations. In W. J. Arnold (Ed.),Nebraska symposium on motivation, 1968 (pp. 175-266). Lincoln: University of Nebraska Press.

Lazarus, R. S. (1984). On the primacy of cognition.American Psychologist, 39, 124-129.

Lazarus, R. S., Kanner, A. D., & Folkman, S. (1980).Emotions: A cognitive-phenomenological analysis. InR. Plutchik & H. Kellerman (Eds.), Emotion: Theory,

research, and experience: Vol. I. Theories of emotion

(pp. 189-217). New York: Academic Press.Mandler, G. (1975). Mind and emotion. New York: Wiley.McHugo, G. J., Smith, C. A., & Lanzetta, J. T. (1982).

The structure of self-reports of emotional responses tofilm segments. Motivation and Emotion, 6, 365-385.

Mowrer, O. H. (1960). Learning theory and behavior.

New York: Wiley.

Osgood, C. E. (1955). Fidelity and reliability. In H.Quasller (Ed.), Information theory in psychology: Prob-

lems and methods, (pp. 374-386). Glencoe, IL: FreePress.

Osgood, C. E. (1966). Dimensionality of the semanticspace for communication via facial expressions. Scan-

dinavian Journal of Psychology, 7, 1-30.Plutchik, R. (1962). The emotions: Facts, theories, and

a new model. New York: Random House.Plutchik, R. (1980). Emotion: A psychoevolutionary syn-

thesis. New York: Harper & Row.

Pruzansky, S. (1975). SINDSCAL: A computer program for

individual differences in multidimensional scaling. Un-published manuscript, Bell Laboratories, Murray Hill,NJ.

Roseman, I. (1984). Cognitive determinants of emotions:A structural theory. In P. Shaver (Ed.), Review of

personality and social psychology: Vol. 5. Emotions,

relationships, and health (pp. 11-36). Beverly Hills:

Sage.Russell, J. A. (1978). Evidence of convergent validity on

the dimensions of affect. Journal of Personality and

Social Psychology, 36, 1152-1168.Russell, J. A. (1979). Affective space is bipolar. Journal

of Personality and Social Psychology, 37, 345-356.Russell, J. A. (1980). A circumplex model of affect.

Journal of Personality and Social Psychology, 39, 1161-

1178.

Russell, J. A. (1983). Pancultural aspects of the humanconceptual organization of emotions. Journal of Per-

sonality and Social Psychology, 45, 1281-1288.Russell, J. A., & Mehrabian, A. (1977). Evidence for a

three-factor theory of emotions. Journal of Research

in Personality, 11, 273-294.Schachter, S., & Singer, J. E. (1962). Cognitive, social,

and physiological determinants of emotional state.Psychological Review, 69, 379-399.

Scherer, K. R. (1982). Emotion as process: Function,origin and regulation. Social Science Information, 21,

555-570.Schlosberg, H. (1941). A scale of the judgment of facial

expressions. Journal of Experimental Psychology, 29,497-510.

Schlosberg, H. (1952). The description of facial expressionsin terms of two dimensions. Journal of Experimental

Psychology, 44, 229-337.Schlosberg, H. (1954). Three dimensions of emotion.

Psychological Review, 61, 81-88.

Shepard, R. N. (1962). The analysis of proximities:Multidimensional scaling with an unknown distancefunction. II. Psychometrika, 27, 219-246.

Stein, N. L., & Jewett, J. L. (1984). A conceptual analysis

of the meaning of negative emotions: Implications for

a theory of development. Unpublished manuscript,University of Chicago and Northwestern University,Chicago, IL.

Tomkins, S. S. (1962). Affect, imagery, consciousness:Vol. 1. The positive affects. New York: Springer.

Tomkins, S. S. (1963). Affect, imagery, consciousness:Vol. 2. The negative affects. New York: Springer.

Tomkins, S. S. (1982). Affect Theory. In P. Ekman (Ed.),Emotion in the human face (2nd ed., pp. 353-395).New York: Cambridge University Press.

Triandis, H. C., & Lambert, W. W. (1958). A restatementand test of Schlosberg's theory of emotion with twokinds of subjects from Greece. Journal of Abnormal

and Social Psychology, 56, 321-328.

Weiner, B.. Russell, D., & Lerman, D. (1979). Thecognition-emotion process in achievement-related con-

texts. Journal of Personality and Social Psychology, 37.1211-1220.

Wundt, W. (1897). Outlines of psychology (C. H. Judd,Trans.). Leipzig: Wilhelm Englemann. (Original workpublished 1896)

Zajonc, R. B. (1980). Feeling and thinking: Preferencesneed no inferences. American Psychologist, 35, 151-

175.

Zajonc, R. B. (1984). On the primacy of affect. American

Psychologist, 39, 117-123.

Received July 19, 1984Revision received November 16, 1984

This

doc

umen

t is c

opyr

ight

ed b

y th

e A

mer

ican

Psy

chol

ogic

al A

ssoc

iatio

n or

one

of i

ts a

llied

pub

lishe

rs.

This

arti

cle

is in

tend

ed so

lely

for t

he p

erso

nal u

se o

f the

indi

vidu

al u

ser a

nd is

not

to b

e di

ssem

inat

ed b

road

ly.