Content uploaded by Gerald Clore
Author content
All content in this area was uploaded by Gerald Clore on Apr 08, 2020
Content may be subject to copyright.
PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL
Simulation, Scenarios, and Emotional Appraisal:
Testing the Convergence of Real and
Imagined Reactions to Emotional Stimuli
Michael D. Robinson
North Dakota State University
Gerald L. Clore
University of Virginia
Appraisal research has relied heavily on vignette- and recall-
based methodologies in theory construction; however, the valid-
ity of these methodologies in capturing the concomitants of
online experience is unknown. To assess the convergence of
online and simulated accounts of emotion, the authors assigned
undergraduate research participants to either online or simu-
lated conditions. Those in the online condition reported on their
appraisals and emotions after viewing a series of 10 emotional
slides, whereas those in the simulated condition estimated their
likely reactions on the basis of short descriptions of the same
slides. Despite the different information available in the two con-
ditions, there was a surprising degree of correspondence in the
reports. This convergence was seen in mean levels of appraisal
and emotion but even more dramatically in the pattern of
appraisal-emotion relations across slides. It is concluded that
vignette methodologies can play a useful role in theory construc-
tion. In addition, the findings raise interesting questions about
the role(s) of implicit theory in emotion.
After nearly 20 years of research on appraisal, we are in
a better position to make specific predictions about the
unique cognitive determinants of different emotions
(e.g., Frijda, Kuipers, & ter Schure, 1989; Ortony, Clore,
& Collins, 1988; Roseman, 1991; Scherer, 1984; Smith &
Ellsworth, 1985; Smith & Lazarus, 1993; Stein, Trabasso,
& Liwag, 1993; Weiner, 1985). In characterizing progress
so far, several reviews have concluded that there is sub-
stantial agreement on the relevant appraisal dimensions
as well as on the patterns of appraisal associated with dis-
crete emotional states (Clore, Schwarz, & Conway, 1994;
Scherer, 1988; Smith, Griner, Kirby, & Scott, 1996). At
the same time, however, researchers have argued that
further work along these lines will be necessary to refine
existing accounts (Smith & Lazarus, 1993), test compet-
ing accounts (Roseman, Spindel, & Jose, 1990), and
generally improve the predictive power of the models
(Reisenzein & Hofmann, 1993; Scherer, 1993).
From one perspective, then, the story of appraisal
research is one of incremental progress, with further
refinements as part of the future agenda (Scherer, 1993;
Smith & Lazarus, 1993). Whereas such a perspective
would lead us to be optimistic about the state of our
knowledge, an emerging number of commentaries have
begun to question the basic assumptions of existing
research, offering both methodological and theoretical
reasons why we should be skeptical of the appraisal
enterprise (Berkowitz, 1993; Frijda, 1993; Lazarus, 1995;
Lewis, 1996; Parkinson, 1997; Parkinson & Manstead,
1993; J. A. Russell, 1987; Zajonc, 1984). A complete eval-
uation of these criticisms is beyond this article, but two of
the key arguments are that (a) existing results tell us
more about appraisals as correlates of emotion than as
causes of emotion (Frijda, 1993; Parkinson, 1997) and
(b) existing research presumes that appraisal judgments
are encoded in a conscious, verbalizable form, whereas
this may not be the case (Berkowitz, 1993; Lazarus,
1995).
Authors’ Note: Support is acknowledged from a National Science
Foundation (NSF) grant (SBR 98-17649) to Robinson and from an
NSF grant (SBR 96-01298) and a National Institute of Mental Health
grant (MH 50074) to Clore. The authors would like to thank Carol
Gohm and Larry Hubert for suggestions on the data analysis. Corre-
spondence concerning this article can be sent to Michael Robinson,
Department of Psychology, North Dakota State University, Fargo, ND
58105-5075; e-mail: michael.d.robinson@ndsu.nodak.edu.
PSPB, Vol. 27 No. 11, November 2001 1520-1532
© 2001 by the Society for Personality and Social Psychology, Inc.
1520
Similar to Frijda (1993) and Lazarus (1995), we are
proponents of appraisal theory but also believe that new
methodologies will be necessary to deal with these
important criticisms. The bulk of the article, however,
examines another issue that is relevant to evaluating pre-
vious research, namely, the degree of correspondence
between reports obtained in the presence versus
absence of concurrent emotional stimulation. We argue
that such results are crucial to the evaluation of current
theories but that directly relevant evidence on this
important issue is lacking.
Issues of Convergence
Although Scherer (1988) lists five methodologies that
have been used in appraisal research, two of these have
arguably been foundational to the great majority of well-
known appraisal theories. Scenario methods ask partici-
pants to imagine their reactions to descriptions of hypo-
thetical situations (e.g., Roseman, 1991; Smith & Laza-
rus, 1993; Weiner, 1985), whereas recall methods ask
participants to recall emotional episodes from their past
(Frijda et al., 1989; Scherer, Wallbott, & Summerfield,
1986; Smith & Ellsworth, 1985). By contrast, studies of
concurrent appraisal and experience (Folkman & Laza-
rus, 1985; Smith & Ellsworth, 1987) have been compara-
tively rare.
In considering the validity of recall- and scenario-
based results, Parkinson and Manstead (1992, 1993)
have raised some important considerations. Although
there are differences between simulation and recall
tasks, they are similar in the sense that a person must
report on emotions that they are not currently feeling.
Many of the contextual details that might be important if
a person were actually experiencing an emotion are pre-
sumably not available in retrospective or hypothetical
attempts at reconstruction (Lazarus, 1995). And if, as we
have claimed (Clore et al., 2001), emotional experience
cannot be stored in memory but only represented in
memory, both retrospective and simulated reports
would seem to be based on representations about emo-
tion rather than on emotion itself. This is a subtle distinc-
tion but one that could set the stage for systematic bias
and misestimation in nonconcurrent reporting. Finally,
it is also the case that both methods have been used to
target rather prototypical emotional reactions that may
be unrepresentative of more complex daily experiences
(Smith & Ellsworth, 1987). The extent to which real
emotions correspond to hypothetical, prototypical ones,
especially in the case of appraisal-emotion relations, is
thus a matter of some uncertainty.
Potentially relevant evidence was extensively reviewed
in a recent article by Robinson and Clore (2001). Their
main conclusion, in concert with Parkinson and
Manstead (1992), was that beliefs about emotion do
appear to play a larger role when people report on non-
concurrent or hypothetical emotions than when they
report on their current experiences. Several examples
should suffice. There is a good deal of evidence indicat-
ing that sex differences in emotion are more consistent
with gender stereotypes, an internalized source of
beliefs, when people are asked to report on retrospec-
tive, hypothetical, or general emotionality than when
they are asked to report on specific concurrent experi-
ences (see also Shields, 1991). Another series of studies
converge on similar conclusions regarding mood and
the menstrual cycle. These studies convincingly make
the point that the connection between phase of cycle
and negative affect is weak to nonexistent, as evidenced
in daily mood reports, but that beliefs about this connec-
tion influence retrospective and general reports about
mood states during the different phases (see McFarland,
Ross, & DeCourville, 1989, for a particularly convincing
study).
Based on such dissociations between online and non-
concurrent reports, Robinson and Clore (2001) pro-
posed a general model of emotion reporting. Online
experience, according to the model, is primarily a func-
tion of the appraisal of concurrent situational details.
These details, however, are relatively less accessible when
reporting on feelings that are not current, making it nec-
essary for people to access their generalized beliefs
about emotion. The model therefore predicts that
online versus nonconcurrent reports will tend to
diverge, with the latter being more consistent with gen-
eralized beliefs. The research reviewed in Robinson and
Clore (2001) suggests caution in accepting the validity of
retrospective or simulated reports of emotion, thus sup-
porting the concerns raised by Parkinson and Manstead
(1992, 1993). However, the review examined emotion
reports exclusively and not relations between appraisal
and emotion. It is entirely possible that retrospective and
simulated reports accurately capture the appraisal deter-
minants of emotion, even if they are sometimes biased
concerning mean emotional intensity. To put it con-
cretely, who would doubt the connection between threat
and fear, whether the reports are obtained online or on
the basis of simulated reactions? In this connection, con-
sider the following quote from Roseman (1991):
Although it is possible, there is no particular reason to
believe that subjects would in fact respond differently
from what is indicated by this data if they were actually in
the vignette situations. For example, is there any reason
to believe that subjects would not respond with joy if they
got the grade they wanted in an exam, with sadness if
they did not, or with anger if they blamed the teacher for
low grades? (pp. 196-197)
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL 1521
Roseman’s (1991) position gains some support in a
study that looked at emotional reactions to an examina-
tion. In this study, Smith and Ellsworth (1987) found
many of the same appraisal-emotion relations that were
found in an earlier study using a recall methodology
(Smith & Ellsworth, 1985). Indeed, none of the robust
relations found in 1987 contradicted earlier conclu-
sions. However, it was also true that a number of relations
found in the 1985 recall study were not found in the 1987
online study. Anger, happiness, sadness, and guilt had
fewer unique predictors in the online study, and
appraisal ratings were less able to discriminate guilt and
sadness. A direct comparison is difficult because of the
different situations involved in the two studies, but the
results generally indicate the value of the Smith and
Ellsworth (1985) framework for predicting online
emotion.
However, in another study looking at convergence
between vignette (study 1) and online (study 2) emo-
tional reactions, D. Russell and McAuley (1986)
obtained more problematic results. The ability of causal
attributions (e.g., luck, effort) and causal dimensions
(e.g., locus of causality, controllability) to predict spe-
cific emotions was substantially different in the two stud-
ies. These constructs predicted hypothetical reactions to
vignette situations much better than they predicted
online reactions to examination feedback, providing
some support for the concerns about vignette studies
raised by Parkinson and Manstead (1993). Again, how-
ever, the situations differed between the two studies, and
the online study had a relatively small sample size, mak-
ing a direct comparison between the two studies
difficult.
In summary, the validity of existing appraisal theories
would seem to depend in part on the representativeness
of results obtained from vignette- and recall-based meth-
odologies. To the extent that these methodologies tap
people’s implicit theories about emotion, and to the
extent that these implicit theories misrepresent actual
appraisal-emotion relations, existing theories may be
unrepresentative of real-world emotion (Parkinson &
Manstead, 1992, 1993). Because of the conflicting results
reported above (D. Russell & McAuley, 1986; Smith &
Ellsworth, 1987), and because no direct evidence on
these issues seemed to exist, we performed the following
study.
Appraisal Correlates of Online
Versus Simulated Emotion
To induce a variety of emotional reactions within an
hour-long session, we presented emotional slides from
Lang’s well-validated set of more than 600 pictures
(Lang, Bradley, & Cuthbert, 1999). In a series of influen-
tial studies, Lang and colleagues have found that the
slides give every evidence of producing genuine emo-
tion. Viewing the slides alters subjective experience and
physiological indices of emotion in a robust and reliable
manner (Bradley, Cuthbert, & Lang, 1996; Greenwald,
Cook, & Lang, 1989; Lang, Greenwald, Bradley, &
Hamm, 1993), both across slides and across participants.
Relevant dependent measures have included corrugator
and zygomatic electromyographic activity, electroder-
mal activity, heart rate, affect-potentiated startle, occipi-
tal and frontal lobe activation, and reaction time and
memory measures along with self-reported emotion. In
addition to being the most extensively validated method
for inducing emotion, the slides also possess several
other advantages that made them ideal for present pur-
poses. First, they permit a high degree of experimental
control, and second, by use of the slides, one can induce
a variety of discrete reactions within a short period of
time. No other induction is capable of accomplishing
the latter purpose as readily and efficiently.
A selection of these slides was shown to participants in
an online condition. After viewing each slide, partici-
pants in this condition reported on their appraisals of
the slide as well as their emotional reactions to it. Partici-
pants in a simulated condition, by contrast, were never
shown any of the slides. Their task was to estimate, based
on very short slide descriptions, how they would appraise
each slide and how they would feel if they had actually
seen it. Convergence between the two conditions was
assessed by a series of correlation coefficients. Effect size
statistics, the results of a series of analyses of variance,
also allowed us to examine the degree of convergence
and divergence between the two conditions. Because of
the conflicting prior results, as well as the lack of directly
relevant data, we retained an open mind during the
investigation. As we hope the reader will see, the results
proved to be highly informative.
METHOD
Participants and Design
Participants included 105 undergraduates from the
University of Illinois who participated in return for
credit for their introductory psychology classes. By ran-
dom assignment, 52 were assigned to the online emotion
condition and 53 to the simulated emotion condition.
The order of the dependent measures was counterbal-
anced across participants so that 54 participants com-
pleted appraisal ratings of each slide or slide description
before reporting emotions, whereas 51 reported on
their emotions before making appraisal ratings.1
1522 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Materials and Procedures
Rating norms for the slides (Lang et al., 1999) are
based on a two-dimensional space defined by valence
and arousal (J. A. Russell, 1980). In recent pilot work,
however, we have found that selected pictures also can be
used to target discrete emotions, especially negative
ones. For example, a picture of a snake tends to elicit
high ratings of fear but comparably low ratings of sad-
ness and disgust. By contrast, a picture of a filthy toilet
tends to elicit high ratings of disgust but comparably low
ratings of fear and sadness. Anger and sadness can be tar-
geted by selected slides, as can excitement and content-
ment. In the present investigation, we used these prior
rating norms to select slides that targeted a wide range of
discrete emotional states. No neutral slides were used,
although a couple of the positive slides were relatively
low in arousal.
In the online emotion condition, participants were
shown 10 slides ordered in a semirandom sequence.2In
the simulated emotion condition, participants received
the same ordering of themes, but stimuli for this group
consisted exclusively of one or two sentence descriptions
of each of the slides. The descriptions briefly conveyed
the central object of each picture as well as its most
salient details. The descriptions, along with the corre-
sponding international affective picture system (IAPS)
slide numbers, are presented in the appendix. Before
making their ratings, participants were asked to contem-
plate each slide or slide description for 5 seconds to facil-
itate imagery and experience. Participants were run in
small groups of two to five.
Instructions and Dependent Measures
Participants in the simulated emotion condition were
asked to estimate their likely reactions to each of the
slides, responding as if they had seen the slides rather
than merely read slide descriptions. By contrast, partici-
pants in the online emotion condition were asked for
their actual reactions to each of the slides. Appraisal rat-
ings were made on a 9-point scale that was only slightly
different in the online (1 = the depicted scene has none of this
quality,9=the depicted scene has a lot of this quality) and sim-
ulated (1 = the depicted scene would have none of this quality,
9=the depicted scene would have a lot of this quality) condi-
tions. Similar slight alterations, designed to acknowl-
edge the lack of experiential data in the simulated condi-
tion, characterize the difference between the 9-point
emotion scale used by participants in the online (1 = I feel
none of this in response to the slide,9=I feel a lot of this in
response to the slide) and simulated (1 = I would feel none of
this in response to the slide,9=I would feel a lot of this in
response to the slide) conditions.
The emotion scale consisted of 20 emotions selected
from Storm and Storm (1987). Their goal was to provide
a detailed map of distinctions among emotion words at
several levels of hierarchical organization. Using cluster
analyses and expert ratings, seven broad categories were
discovered, subsuming 20 intermediate-level and 61
lower-level groupings. We sampled primarily at the inter-
mediate level, but the terms are most conveniently sum-
marized at the broadest level. Our emotion scale con-
sisted of embarrassed, sad, and discouraged (cluster 1 =
terms related to shame/sadness/pain); anxious, fearful,
worried, upset, and nervous (cluster 2 = terms related to
anxiety/fear); angry, irritated, and disgusted (cluster 3 =
terms related to anger/hatred/disgust); affectionate
and caring (cluster 4 = positive terms with interpersonal
relevance); contented, happy, excited, and proud (clus-
ter 5 = positive terms with no interpersonal relevance);
bored and calm (cluster 6 = passive); and alert (cluster 7 =
active). An emotion scale was created by randomizing
the 20 emotion items.
The appraisal scale consisted of 18 appraisal judg-
ments based on four theoretical frameworks, although
the resultant terms are likely to overlap considerably
with other accounts (Clore et al., 1994; Scherer, 1988).
From the Smith and Lazarus (1990) framework, we
chose terms to represent motivational relevance (impor-
tant, significant, self-relevant), motivational congru-
ence/incongruence (dangerous, beneficial, threaten-
ing), and secondary appraisal (powerful, problematic,
and uncontrollable). From the Ortony et al. (1988)
framework, we then added terms for reactions to agents
(praiseworthy) and reactions to objects (appealing and
pleasant). From the Scherer (1984) framework, we
added terms for the novelty check (familiar, unusual,
and interesting) as well as one additional term for the
goal/need significance check (unpredictable). Finally,
from the work of Osgood, Suci, and Tannenbaum
(1957), we added two additional terms to tap the
potency dimension (complex and intense). An appraisal
scale was created by randomizing the 18 appraisal items.
RESULTS
Overview
In this research, 105 raters, divided into two condi-
tions (online vs. simulated), reported on 18 appraisals
and 20 emotions in response to 10 slides. There are
many ways to analyze the resulting ratings. Our choices
were guided by a desire to ask three questions. The basic
form of each question was the same: Would similar or dif-
ferent conclusions about emotion result from responses
obtained in the online condition versus those obtained
in the simulated condition? Specifically, would similar or
different conclusions be made with respect to (a) how
stimuli are appraised, (b) which emotions are produced
by which stimuli, and, of major interest, (c) the relation-
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL 1523
ship between appraisals and emotions. To answer the
first two questions, the same sets of analyses were per-
formed on appraisal ratings and then on emotion
ratings.
Appraisal Ratings
Do people appraise real versus imagined emotional
stimuli differently? To answer this question, we per-
formed 18 ANOVAs, one for each appraisal dimension,
as a function of the within-subject variable of slide con-
tent and the between-subjects variable of condition.
Summarizing these ANOVAs, there were no main effects
for condition, ps > .05, M2= .01. Across slides, that is,
participants in the two conditions did not differ in their
mean appraisal ratings. By contrast, all of the slide con-
tent main effects were significant, ps < .001, M2= .41.
The latter results indicate that all of the appraisal dimen-
sions differed significantly by slide content.
More important than condition main effects are
potential Condition ´Slide Content interactions. Such
interactions would indicate that participants in the two
conditions appraised the same slides differently.
Although 9 of the 18 Condition ´Slide Content interac-
tions were significant, ps < .05, the average effect size for
these significant interactions was small, M2= .03 (M2=
.02 across all 18 ANOVAs). To understand the sources of
the significant interactions, we performed one-way
ANOVAs on each appraisal rating (Slide Content ´
Appraisal Dimension) separately, with condition as a
between-subjects variable. Very few of these pairwise
comparisons were significant. However, the ones that
were significant revealed that simulated raters (vs.
online raters) thought that the gun and wolf slides would
be more novel, that the world-class runners slide would
be less predictable and more important, and that the
desert sky slide would be more pleasant and less
threatening.
Although there were some differences by condition,
they were slight in comparison to the slide content main
effects. We sought to quantify the degree of convergence
between conditions by obtaining appraisal means
(across participants) for online versus simulated condi-
tions for each Slide Content ´Appraisal Dimension
combination. For example, one set of data points
involved the mean “threatening” rating given to the gun
slide in each condition (Ms = 8.15 and 8.13 in the online
and simulated conditions, respectively). For each
appraisal dimension separately, we computed the corre-
lation between mean appraisal ratings in the two condi-
tions, with slide (n= 10) as the unit of analysis. The
median correlation across the 18 appraisal dimensions
was r= .94, all ps < .05. The size of these correlations
attests to the fact that participants in the two conditions,
to a very large extent, agreed concerning the appraisal
qualities of each slide.
Emotion Ratings
Do people imagine emotional reactions to stimuli
that are different than they would actually feel if exposed
to such stimuli? To address this question, we performed
20 Condition ´Slide Content ANOVAs, one for each
emotion separately. In contrast to the analyses involving
appraisal ratings, 8 of 20 main effects for condition were
significant, ps < .05, M2= .05 (M2= .03 across all 20
ANOVAs). In all cases, those who were simulating their
emotional reactions thought that their emotions would
be more intense than those in the online condition, par-
ticularly for negative feelings such as anxious, angry, and
discouraged. The effects for slide content, however, were
markedly larger, and all 20 of these main effects were sig-
nificant, ps < .001, M2= .48. The latter results indicate
that across the two conditions, there was substantial
agreement concerning which slides would (or did) pro-
duce more or less of a given emotion. In addition to the
condition and slide content main effects, 11 of the 20
Condition ´Slide Content interactions were significant,
ps < .05, but these interactions were rather small in mag-
nitude, M2= .04 (M2= .03 across all 20 ANOVAs).
In characterizing the significant Condition ´Slide
Content interactions, we performed pairwise compari-
sons on each emotion rating (Slide Content ´Emotion)
separately, with condition as a between-subjects variable.
Overall, most of the significant pairwise comparisons
involved the gun, wolf, and world-class runners slides.
Participants in the simulated condition (vs. the online
condition) reported that they would experience more
anxiety and fear in response to the gun and wolf slides
and that the runners slide would be more involving (e.g.,
excited). Comparably few pairwise comparisons were
significant in response to the seven remaining slides, and
the differences that did occur are difficult to characterize.
The significant effects for condition, however, must
be evaluated in light of the large effects for emotion. It
seems safe to say that participants in the simulated condi-
tion did not misrepresent their likely emotional reac-
tions. Rather, the pattern of findings pertaining to the
gun, wolf, and runners slides revealed that those
responding to the hypothetical descriptions overesti-
mated emotional responses that were consistent with the
theme of the slides (e.g., fear in response to the gun
slide). Because participants in the online condition
reported substantial amounts of the same emotions that
were overestimated by participants in the simulated con-
dition, the dominant pattern is one of parallels between
the two conditions.
This substantial convergence is again revealed by cor-
relations between mean ratings given in the two condi-
1524 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
tions. Specifically, for each emotion and each slide, we
obtained means (collapsed across participants) for those
in the online condition versus those in the simulated
condition. For each emotion separately, we computed a
correlation coefficient between means in the two condi-
tions, with slide (n= 10) as the unit of analysis. As with the
appraisal correlations, the median correlation between
the mean ratings given in the two conditions was substan-
tial, median r= .95, all ps < .005.
The results presented so far suggest that imagined
reactions to emotional stimuli are nearly identical to
online reactions. To the extent that there was diver-
gence, it was primarily due to the fact that participants in
the simulated condition overestimated their reactions to
judgment-congruent stimuli. Although one could high-
light some small but significant dissociations between
the two conditions, it is the convergence between the two
conditions that is the main story revealed by the results.
The same conclusions concerning convergence
emerged, perhaps even more strikingly, when we exam-
ined appraisal-emotion relations.
Correspondence of Appraisal-Emotion Relations
in Online and Simulated Conditions
Assessing the degree of correspondence between
appraisal-emotion relations in the two conditions was a
two-step procedure. First, for each participant sepa-
rately, we computed correlation coefficients between
each appraisal dimension and each emotion rating, with
slide (n= 10) as the unit of analysis. For example, the cor-
relation between threatening and fearful for a particular
individual would reveal the extent to which he or she
reported more fear in response to slides that he or she
appraised as more threatening. These correlations were
converted to Fisher zscores for inferential testing.
To see if imagining versus actually reacting to emo-
tional stimuli altered appraisal-emotion relations, and to
compare such divergences to agreement across the two
conditions, we performed 20 ANOVAs (one for each
emotion) on appraisal-emotion zscores. In these
ANOVAs, appraisal dimension was a within-subject vari-
able, whereas condition was a between-subjects variable.
Across the 20 ANOVAs, the mean effect size for appraisal
dimension was large, M2= .49, all ps < .001, indicating
that appraisals and emotions correlated similarly across
conditions. By contrast, the mean effect sizes for condi-
tion and the Condition ´Appraisal Dimension interac-
tion were 2= .01 and 2= .02, respectively.
To more directly assess convergence between the two
conditions, the mean z-transformed correlations were
themselves correlated. First, for each emotion and each
condition separately, we obtained 18 mean z-trans-
formed correlations, one for each appraisal-emotion
relation. Second, for each emotion, we correlated these
mean zscores across the two conditions, with appraisal
dimension as the unit of analysis (n= 18). Appraisal-emo-
tion relations were nearly identical in the online and sim-
ulated conditions, as the median correlation of zscores
was r= .99!
The Structure of Emotion
Vignette, simulated, and recalled emotion reports
often have been used to make conclusions about the
structure of the emotion space as a whole. For example,
the similarity between two emotions can be assessed by
determining the extent to which these two emotions are
similar in terms of their relations with different apprais-
als. A final question that we examined, thus, was how the
overall structure of the appraisal space (one set of analy-
ses) and the emotion space (a second set of analyses) dif-
fered by whether reports were obtained on the basis of
online experience versus imagined emotional reactions.
The analyses reported below also will allow us to evaluate
the present results in light of previous theories concern-
ing the structure of appraisal-emotion relations.
To compare relations among the different appraisal
dimensions, and to assess the convergence of this struc-
ture across online versus simulated conditions, we per-
formed cluster analyses on appraisal relations, separately
for each condition. The data for the cluster analyses were
correlations of each appraisal with the 20 emotions. Cor-
relations rather than zscores were used, and the cen-
troid linkage method was used. To compare relations
among the different emotions, we then performed par-
allel cluster analyses on emotion ratings. In this case, the
data consisted of correlations of each emotion with the
18 appraisal dimensions. Figure 1 displays cluster solu-
tions for the appraisal dimensions, whereas Figure 2 dis-
plays cluster solutions for the emotions.
Turning first to the cluster analyses performed on
appraisal dimensions, three broad clusters were
obtained, and they were identical in the two conditions.
Elements of primary and secondary appraisal (Smith &
Lazarus, 1990) formed a threat-related cluster (terms:
threatening, dangerous, uncontrollable, unpredictable,
intense, and powerful). This cluster provides indirect
support for Robinson’s (1998) contention that the
appraisal of threat may be more unitary than has been
proposed by some theories. Specifically, Robinson pro-
posed that the appraisal of threat combines elements of
motivational relevance, motivational incongruence, and
coping potential but is best described as a single
appraisal judgment rather than a combination of multi-
ple ones (see Öhman & Soares, 1994, for evidence that
only minimal processing is necessary to trigger this
appraisal).
A second cluster also consisted of elements of both
motivational relevance and secondary appraisal, as well
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL 1525
as a term related to novelty (terms: complex, problem-
atic, unusual). This cluster (along with significant and
important) then joined the threat cluster and generally
appeared to be more characteristic of negative than of
positive emotions. Particularly interesting in this regard
was that appraisal dimensions related to novelty
(unusual) and motivational relevance (significant and
important) clustered together with perceptions of threat
and problematic situations. According to Scherer
(1984), the novelty check precedes the pleasantness
check; thus, perceptions of novelty should be initially
neutral in valence. Similarly, Smith and Lazarus (1990)
propose that motivational relevance is orthogonal to
perceptions of motivational congruence or incongru-
ence. The present results, however, suggest that these
appraisals (i.e., novelty and motivational relevance) may
not be neutral in valence but rather may tend to be more
characteristic of negative emotional reactions.
A third broad cluster consisted of a core of positivity
(terms: appealing, pleasant, and beneficial) in combina-
tion with terms related to novelty (terms: familiarity,
interesting), motivational relevance (term: self-relevant),
and secondary appraisal (term: praiseworthy). Con-
sidering the two broadest clusters, then, appraisals
related to motivational incongruence (e.g., threaten-
ing) or congruence (e.g., beneficial) played a central
role. These negative and positive cores attracted other
appraisal elements so that the overarching themes per-
tained to positive versus negative qualities. Consistent
with Zajonc’s (1980) claims, appraisals related to novelty
were not affectively neutral, because “unusual” joined
the negative cluster, whereas “familiar” joined the posi-
tive cluster. Appraisals related to motivational relevance
also split into those apparently carrying negative (signifi-
cant and important) or positive (self-relevant) implications.
Turning to the cluster analyses performed on emo-
tions, a particularly interesting finding was that upset,
angry, and discouraged were the first to join, followed by
other reactions pertaining to both anger as well as other
negative emotions (terms: angry, upset, discouraged,
irritated, worried, disgusted, and sad). These results con-
verge with those of other investigators who have found
that anger- and sadness-related emotions may be more
similar than an attributional analysis would suggest
(Polivy, 1981; Smith & Ellsworth, 1987).
A separate negative cluster, which later merged with
the anger-related cluster, was composed of terms related
to fear (terms: fearful, nervous, anxious, and alert).
1526 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Figure 1 The structure of appraisal in online (top) versus simulated
(bottom) conditions, as revealed by cluster analyses.
Figure 2 The structure of emotion in online (top) versus simulated
(bottom) conditions, as revealed by cluster analyses.
Interesting in this regard was that worried joined the
anger-related cluster before it joined the fear-related
cluster, suggesting that worry and fear may differ in some
important respects (Robinson & Johnson, 2000). At a
broader level, the finding that anger- and fear-related
emotions can be discriminated on the basis of their
appraisals suggests that circumplex models of emotion
(e.g., J. A. Russell, 1980) may fail to capture the full com-
plexity of emotional reactions, because anger and fear
are both negative/high arousal states.
A third emotion cluster was anchored by the positive
emotions of calm, contented, and happy. No additional
subclusters were found, but five other terms eventually
joined (terms: calm, contented, happy, affectionate,
proud, excited, bored, and caring). Most of these terms
are positive in valence but are characterized by addi-
tional qualities such as arousal (excited), attributional
specificity (proud), or interpersonal relevance (affec-
tionate and caring). That calm and contented were
included in the core of the cluster, together with the fact
that bored joined this cluster rather than the negative
emotion cluster, suggests that the cluster was generally
characterized by lower intensity reactions (with the
exception of excited). This may be a function of the
slides, which seem to be better suited for inducing
intense negative reactions than intense positive ones.
In characterizing the degree of convergence between
the structures obtained in online and simulated condi-
tions, we note that all of the above findings were found to
be parallel in the two conditions. Although there were
some minor differences between the structures obtained
in the two conditions (e.g., appealing first joined pleas-
ant in the online condition but familiar in the simulated
condition), these difference pale in comparison to the
degree of convergence between the two conditions.
DISCUSSION
Convergence Between Conditions
This research was motivated by Parkinson and
Manstead’s critiques of recall (1992) and vignette
(1993) methodologies in appraisal research. Because
such methodologies ask participants to characterize
their emotions in the absence of concurrent experience,
without the rich contextual details that surround an
emotional episode, it seems reasonable to propose that
such reports would be vulnerable to bias due to implicit
beliefs about emotion. Indeed, in terms of mean-level
emotion ratings, this does appear to be the case. As Rob-
inson and Clore (2001) detail, it is generally true that
emotion reports obtained in the absence of concurrent
experience (i.e., those targeting retrospective, prospec-
tive, hypothetical, or aggregated time frames) tend to
correspond to generalized beliefs to a greater extent
than do online reports.
Some evidence for such dissociations was found in
this study. Participants in the simulated condition esti-
mated that they would experience more fear and anxiety
in response to dangerous slides than participants in the
online condition actually felt, and they also estimated
that they would feel more excitement in response to a
slide depicting world-class runners. Beyond this, partici-
pants in the simulated condition estimated that they
would experience 8 of the 20 emotions more intensely
across slides. These results are generally consistent with
research reviewed in Robinson and Clore (2001) in the
sense that both retrospective (Thomas & Diener, 1990)
and prospective (Wilson, Wheatley, Meyers, Gilbert, &
Axsom, 2000) reports are often exaggerated in intensity.
Such dissociations presumably arise because emotion-
relevant details acquire disproportional salience when
other, less-emotional details of the situation either fade
from memory (retrospective reports) or are not avail-
able in principle (simulated reports).
However, it is also worth emphasizing that partici-
pants in the simulated condition were far from being off
the mark, even when significant dissociations were
found. They reported more fear and anxiety only in
response to slides that actually engendered fear and anx-
iety and more excitement only in response to a slide that
actually engendered some excitement. Thus, although
significant differences were sometimes found in the
reports given by participants in online and simulated
conditions, patterns of mean appraisal and emotion
were highly consistent in the two conditions. Effect size
statistics support this point, as do correlations computed
on mean ratings given in the two conditions. For exam-
ple, with slide as the unit of analysis, mean appraisal rat-
ings were highly convergent in the two conditions, Mdn r =
.94, as were mean emotion ratings, Mdn r = .95.
Even more impressive convergence was obtained
when we looked at appraisal-emotion relations, for
which we reported a median correlation of r= .99
between the relations obtained in the two conditions.
These correlations are so high that one would generally
assume that they were obtained on the basis of a random
split of participants within the same condition. In point
of fact, however, the two conditions were quite different,
including differences in both the presence versus
absence of concurrent emotional experience and the
presence versus absence of emotion-relevant visual
details. Participants in the simulated condition were
asked to estimate their feelings in response to one- or
two-sentence slide descriptions, but their reports, espe-
cially with regard to appraisal-emotion relations, were
virtually identical to those obtained online.
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL 1527
We regard the current results as evidence for the valid-
ity of scenario methods in developing appraisal theories
of emotion. The issue is important to the validity of exist-
ing theories because scenario- and recall-based methods
have been the dominant source of evidence for such the-
ories (Parkinson & Manstead, 1992, 1993; Scherer,
1988). However, no directly relevant evidence seemed to
previously exist as to whether these accounts are
veridical when compared to online ones. Instead, the
few studies that have sought to validate their theories by
comparing online results to recall or vignette results
(e.g., D. Russell & McAuley, 1986; Smith & Ellsworth,
1987) have compared appraisal correlates across differ-
ent situations, and the results themselves have been
somewhat inconsistent. By holding situations constant,
we were able to obtain results relevant to method factors—
online reactions versus simulated estimations—rather
than content ones. The central conclusion that emerges
from this study is the pronounced correspondence in
appraisal-emotion relations obtained across these two
conditions. These results raise some interesting ques-
tions about the role of implicit theories in emotion.
Implicit Theories and Emotional Experience
Parkinson and Manstead (1992, 1993) have raised the
possibility that nonconcurrent reports about emotion
may tell us more about people’s implicit theories of emo-
tion than about the actual correlates of emotional expe-
rience. Our results, because they indicate such a substan-
tial degree of correspondence between the two
conditions, may therefore be consistent with one of
three possibilities. One possibility is that participants in
the simulated condition did not rely exclusively on their
implicit theories but rather were able to simulate their
reactions to the point that they could actually feel some-
thing in response to the slide descriptions. Another pos-
sibility is that implicit theories accurately represent, or
are constitutive of, actual emotional reactions. Finally, it
is possible that both of these processes were occurring
and that the obtained convergence reflects this fact. It is
this third possibility that we favor.
Although many of the obtained relations were consis-
tent with commonsense beliefs, others were not; there-
fore, an exclusive reliance on common sense in the simu-
lated condition would have produced more discrepancies
between the two conditions than occurred. Consistent
with common sense were the strong positive correlations
between threat and fear and between pleasantness and
happiness. However, other findings were more surpris-
ing. In cluster analyses of the appraisal dimensions,
terms related to novelty and motivational relevance split
up, joining positive and negative cores rather than each
other. Unusual and familiar, although not obviously
valence-relevant, were in fact core predictors of negative
versus positive emotions, respectively. In cluster analyses
of the emotion ratings, sadness- and anger-related emo-
tions revealed remarkably similar correlates across the
appraisal dimensions, despite clear differences in the
intuitive theories associated with these emotions (Fehr &
Russell, 1984; Shaver, Schwartz, Kirson, & O’Connor,
1987). Finally, the emotion “bored” joined a cluster
related to positive emotionality, whereas the emotion
“alert” joined a cluster related to negative emotionality.
The deviations from intuitive theory, found in both of
the conditions, suggest that participants in the simulated
condition had access to another source of information
in addition to common sense. Consistent with this point,
studies of simulated emotion suggest that emotional
imagery is capable of producing mild feelings, as
reflected in physiological responses, and that these feel-
ings are often congruent with those that would be
obtained if one were actually in the situation (Cuthbert,
Vrana, & Bradley, 1991). For example, Smith (1989)
manipulated scenario content in a guided-imagery task.
During scenarios indicative of high effort, he found
increased heart rate, and perceived obstacles also were
related to increased brow tension. An extensive body of
evidence from the Lang lab reinforces these points. In a
series of studies, manipulations of scenario content have
influenced heart rate, skin conductance, respiration,
and affect-modulated startle in scenario-congruent ways
(Lang, Kozak, Miller, Levin, & McLean, 1980; Lang,
Levin, Miller, & Kozak, 1983; Miller et al., 1987). For an
intuitive appreciation of the effectiveness of simulated
emotion, we ask the reader to contemplate the “filthy toi-
let” and “baby with tumor” slide descriptions (see the
appendix). Even a brief sustained image of either of
these objects may be sufficient to create a momentary
aversive state that could aid in appraisal and emotion
estimations.
On the other hand, it also may be the case that
implicit beliefs about emotion are either representative
of, or constitutive of, online emotional reactions (Robin-
son & Clore, 2001). The first possibility assumes that
implicit beliefs are in part built from numerous online
experiences, such that the representations accurately
map covariations in experience (see Roseman, 1991;
Smith & Lazarus, 1993, for arguments along these lines).
By this account, connections between importance and
sadness, pleasantness and happiness, threat and fear,
intensity and excitement, and so forth, are more a func-
tion of beliefs derived from personal experience than of
culturally shared theories entirely divorced from daily
experience. The second, more subtle, possibility is that
implicit beliefs about emotion influence the appraisal
process, in effect causing belief-consistent experiences
(Gross, 1998; Lazarus & Alfert, 1964; Robinson & Clore,
2000; Wilson, Lisle, Kraft, & Wetzel, 1989; Winter &
1528 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Kuiper, 1997). For example, if one expects to feel anx-
ious in a given situation, this expectation may in fact cre-
ate its own reality. Whether implicit beliefs about emo-
tion are derived from experience or influence
experience, or both, the basic point is that there are the-
oretical reasons for expecting the type of convergence
found here.
From the standpoint of cybernetic models of motiva-
tion and emotion (Carver & Scheier, 1990; Clore & Rob-
inson, 2000; Higgins, 1997), it would be difficult to over-
state the importance of people’s ability to accurately
simulate their emotions. Specifically, it would be highly
functional to pursue outcomes that are likely to produce
positive hedonic benefits and avoid outcomes that are
likely to produce negative hedonic benefits. Because the
present results suggest that emotion simulations possess
a high degree of accuracy, they also suggest that self-
regulatory efforts are likely to be properly directed more
often than they are likely to be fundamentally misdirected.
Structural Implications
Any particular method of inducing emotion (e.g.,
recalling emotions from the past) may yield structural
results that are particular to the method of induction,
and our structural conclusions must be offered along
with some caveats. In particular, emotional slides seem
better suited for inducing high-intensity negative emo-
tions than high-intensity positive ones, and our struc-
tural results are likely influenced by this method factor.
In addition, our primary motivation in conducting this
research was to test the convergence issue rather than to
assess any particular appraisal account. Despite these
caveats, the results do possess relevance for structural
accounts of emotion, and some of the more important
findings will be discussed here.
Appraisal theories differ with respect to how they han-
dle the valence dimension of experience. Some
accounts, such as Ortony et al.’s (1988), propose that
valence is the predominant factor in organizing experi-
ence, whereas others treat valence as one of a number of
dimensions that can, to some extent, be orthogonally
crossed (e.g., Smith & Ellsworth, 1985). The same theo-
retical tension can be seen in dimensional (e.g., J. A. Rus-
sell, 1980) versus discrete (e.g., Ekman, 1992) views of
emotion, because the former, but not the latter, accords
emotional valence a primary role. The present results
provide support for the centrality of valence in organiz-
ing both the emotion space and the appraisal space. Spe-
cifically, in every cluster analysis, it was the case that the
two broadest clusters could be interpreted as positive
and negative in valence. Ultimately, we believe that a suc-
cessful structural theory must both recognize the hege-
mony of valence as well as make distinctions that allow
one to discriminate discrete reactions. Any structural
account that does not accomplish both of these goals will
necessarily be incomplete.
At the level of discrete emotions, one notable finding
was our inability to readily distinguish distress-related
emotions (e.g., upset) from anger-related ones (e.g.,
angry). According to most appraisal accounts, such emo-
tions should be distinguishable in terms of either attribu-
tions of responsibility or blame (Smith & Ellsworth,
1985; Weiner, 1985), opportunities for problem-focused
coping (Lazarus, 1991; Roseman, 1991), or both (Smith &
Lazarus, 1990). Although we note that our method of
emotion induction, which involved passive exposure to
stimuli that were not extremely personal, may have lim-
ited our ability to distinguish these classes of emotion, we
also note that similar inabilities to distinguish anger and
distress have been reported in other online induction
studies before (Polivy, 1981). Because feelings of distress
often co-occur with feelings of anger (Berkowitz, 1993;
Weiner, 1985), these feelings may share more appraisal
features than discriminate them, at least when highly
prototypic elicitors of distress (“imagine your relative
dies”) and anger (“imagine someone criticizes you inap-
propriately”) are not used. We suggest that in many real-
life circumstances, distress and anger may be more
closely related than recognized by many appraisal theo-
ries (see Berkowitz, 1993).
Although our results did not allow us to distinguish
anger and distress, they did allow us to make another dis-
tinction that has perhaps received less attention in the
appraisal literature. Specifically, fear-related emotions
such as fearful and nervous formed a cluster that was dis-
tinct from the anger/distress cluster. Particularly inter-
esting in this regard was that worried joined the anger/
distress cluster before it joined the fear cluster, suggest-
ing that fear and worry are dissociable. Because similar
results were found in a previous appraisal study of ours
(Robinson & Johnson, 2000), we gain more confidence
in their validity. Robinson and Johnson (2000) found
that feelings of fear, in comparison to feelings of stress
and worry, are associated with more imminent, physi-
cally palpable threats. Feelings of stress and worry, by
contrast, are more diffuse, more prolonged across time,
and more closely associated with anticipated effortful
coping.
Furthermore, such a distinction between fear and
worry has received support in the clinical literature.
Barlow (1991), for example, makes a distinction
between anxious arousal, an intense, phasic disturbance
linked to imminent threat, and anxious apprehension, a
more diffuse experience centered on worrying about
possible, but not imminent, future threats (see also Bor-
kovec, Ray, & Stöber, 1998; Heller, Nitschke, Etienne, &
Miller, 1997). In terms of specific disorders, anxious
arousal is a central element of panic disorder, whereas
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL 1529
anxious apprehension is a central characteristic of obses-
sive-compulsive and generalized anxiety disorders. The
present results reinforce this distinction and further
demonstrate that fear versus worry have distinct
appraisal correlates.
Turning to the cluster analyses performed on
appraisal terms, we make two related observations. First,
recall that we included three terms to assess motivational
relevance (Smith & Lazarus, 1990)—important, signifi-
cant, and self-relevant—and that we generally expected
these appraisal terms to be highly related. Instead,
important and significant joined a negative cluster
(along with threatening and dangerous), whereas self-
relevant joined a positive cluster (along with interesting,
appealing, etc.). Second, recall that we included two pri-
mary terms to assess novelty (Scherer, 1984)—unusual
and familiar—that were expected to be valence inde-
pendent. Instead, unusual first joined problematic,
whereas familiar joined a cluster with appealing, benefi-
cial, and pleasant. The present results, then, suggest that
certain appraisals that are held to be valence independ-
ent may in fact not be. Furthermore, items that would
seem to be closely related may in fact not measure the
same appraisal construct.
Cluster analyses of emotion terms are relatively com-
mon (e.g., Shaver et al., 1987; Storm & Storm, 1987),
whereas cluster analyses of appraisal terms are not. This
is somewhat surprising given widespread interest in the
structure of appraisal. We suggest that a potentially gen-
erative approach for appraisal theory would be to
employ cluster analyses of appraisal terms, as we have
done here. Such a method, perhaps more directly than
those used previously, can reveal how people use apprais-
als to differentiate among their emotional reactions.
Cautions and Limitations
The motivation for this research was to provide initial
evidence concerning the appraisal correlates of real ver-
sus imagined emotional reactions. Prior results were
ambiguous with regard to the degree of correspondence
that we might find, and there were theoretical reasons
for expecting systematic dissociations (Parkinson &
Manstead, 1992, 1993). In particular, a good deal of evi-
dence indicates that simulated reports of emotion, simi-
lar to other reports that are removed from online experi-
ence, are often vulnerable to belief-consistent biases
(Robinson & Clore, 2001). The current results, however,
suggest that such biases may be more relevant to reports
of mean levels of emotionality than to appraisal-emotion
relations. Further research would be necessary to con-
firm, as well as explain, these initial conclusions. In this
connection, the findings raise a number of intriguing
questions about the relation between beliefs and experi-
ence, but the validity of nonconcurrent accounts of emo-
tion should be the topic of active investigation rather
than of uncritical acceptance.
Thus, we would not advocate abandoning the study of
real-world emotion in the study of appraisal. There are
simply too many issues—such as the process of emotion
elicitation—that can only be examined by inducing gen-
uine emotional reactions. In addition, these conclusions
regarding convergence must be regarded as prelimi-
nary. Nevertheless, we do think that these results provide
some support for extant appraisal theories, which have
been validated primarily by studies using recall or simu-
lation methods (Scherer, 1988).
Although we found a strong convergence between
imagined versus real responses to emotional slides, the
data do not indicate that such an induction method will
always produce the same degree of correspondence. For
example, the pictures used in this study were specifically
selected to elicit (a range of) emotions in a reliable fash-
ion. A random sample of pictures without regard to
agreement about their evocative powers would not nec-
essarily produce the same results. To the extent that
responses depended on the idiosyncratic interpreta-
tions of the perceiver, then evidence of convergence
would not be found even within the same presentation
mode (e.g., online or simulated), let alone across modes.
The selected nature of the stimuli was therefore critical
to testing the hypothesis that simulation would intro-
duce systematic error.
As a final limitation, we note that these results were
based on correlational analyses and therefore should
not be taken as evidence for the causal role of appraisal.
Designs such as ours are therefore most relevant to struc-
tural accounts of appraisal. These accounts seek to iden-
tify the specific cognitive eliciting conditions for differ-
ent emotions (see Clore et al., 1994; Scherer, 1988, for
reviews), but they are relatively silent concerning the
processes involved in emotion elicitation (Smith et al.,
1996).
Conclusions
Our findings address a crucial issue for previous
appraisal research, namely, the representativeness of
results based on vignette or recall methods. Supporting
the arguments of Roseman (1991) and Smith and Laza-
rus (1993), the results provide strong support for conver-
gence across online versus simulated conditions.
Although these conditions differed markedly in several
aspects (i.e., the presence or absence of visual details, the
presence or absence of moderate to strong emotional
reactions), appraisal-emotion relations were, for all
practical purposes, identical. The results therefore sup-
port the validity of previous investigations.
To gain precise control over the content of the emo-
tional situations, we chose to use emotional slides.
1530 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
Although there are several advantages of this methodol-
ogy, there are also several disadvantages that are worth
mentioning—they are static images, and the content of
the images is largely divorced from people’s daily lives.
Thus, the relevance of our findings for more protracted
emotional episodes (Frijda, 1993; Lewis, 1996), particu-
larly those occurring in daily experience (Parkinson &
Manstead, 1993), remains an open question. We do not
want to argue that our findings generalize to any and all
emotional transactions. Indeed, no single set of data
could support such claims. We appreciate arguments for
more naturalistic research on appraisal and emotion
(Lazarus, 1995; Parkinson, 1997) but also believe that a
number of issues, such as the present one, are particu-
larly amenable to laboratory-based research.
APPENDIX
Slide Descriptions Used in the Simulated Emotion Condition
IAPS
Slide Description Number
Slide 1: This scene shows two bunny rabbits sharing
a stalk of grass together. 1750
Slide 2: This scene shows a gun pointed straight at
you. It appears to be inches away from
your face. 6230
Slide 3: This scene shows a filthy toilet and an empty
bottle of liquor in a room that hasn’t been
cleaned in a very long time. The toilet is
full of feces and other waste products. 9300
Slide 4: This scene shows a wild wolf-like animal
directly in front of you. It has its jaws wide
open and is apparently ready to attack. 1300
Slide 5: This scene shows world-class runners
sprinting for the finish line in what looks
like a very important race. 8220
Slide 6: This scene shows a roller coaster diving down
a particularly steep section of the track.
The ride is full of people with their arms
upraised and their faces animated. 8490
Slide 7: This scene shows a woman who looks like
she’s been punched in the face. The
damage to her eye looks serious and
there are abrasions around her eye as well. 3180
Slide 8: This scene shows two young lovers who are
kissing each other passionately. They are
attractive and, from what you can see, naked. 4660
Slide 9: This scene shows a baby with a facial tumor.
The tumor, which is protruding through
an eye socket, is also causing one side of
the infant’s head to bulge out. 3170
Slide 10: This scene shows a desert with gently rolling
sand dunes, a setting sun, and a deep blue
sky. The desert is very still. 7580
NOTE: IAPS = international affective picture system.
NOTES
1. Although the order variable was initially included in all analyses,
its effects were inconsistent and did not modify any of our substantive
claims. For this reason, this variable will not be discussed further.
2. The order was only semirandom because we thought it desirable
to begin and end the set with positive pictures.
REFERENCES
Barlow, D. H. (1991). Disorders of emotion. Psychological Inquiry,2,
58-71.
Berkowitz, L. (1993). Towards a general theory of anger and emo-
tional aggression: Implications of the cognitive-neoassociationistic
perspective for the analysis of anger and other emotions. In R. S.
Wyer & T. K. Srull (Eds.), Perspectives on anger and emotion: Advances
in social cognition (Vol. 6, pp. 1-46). Hillsdale, NJ: Lawrence
Erlbaum.
Borkovec, T. D., Ray, W. J., & Stöber, J. (1998). Worry: A cognitive phe-
nomenon intimately linked to affective, cognitive, and interper-
sonal behavioral processes. Cognitive Therapy and Research,22, 561-
576.
Bradley, M. M., Cuthbert, B. N., & Lang, P. J. (1996). Picture media
and emotion: Effects of a sustained affective context.
Psychophysiology,33, 662-270.
Carver, C. S., & Scheier, M. F. (1990). Origins of positive and negative
affect: A control-process view. Psychological Review,97, 19-35.
Clore, G. L., & Robinson, M. D. (2000). What is emotion regulation?:
In search of a phenomenon. Psychological Inquiry, 11, 163-166.
Clore, G. L., Schwarz, N., & Conway, M. (1994). Emotion and infor-
mation processing. In R. S. Wyer & T. K. Srull (Eds.), Handbook of
social cognition (2nd ed., pp. 323-417). Hillsdale, NJ: Lawrence
Erlbaum.
Clore, G. L., Wyer, R. S., Dienes, B., Gasper, K., Gohm, C., & Isbell, L.
(2001). Affective feelings as feedback: Some cognitive conse-
quences. In L. L. Martin & G. L. Clore (Eds.), Theories of mood and
cognition: A user’s guidebook. Mahwah, NJ: Lawrence Erlbaum.
Cuthbert, B. N., Vrana, S. R., & Bradley, M. M. (1991). Imagery: Func-
tion and physiology. In J. R. Jennings, P. K. Ackles, & M.G.H. Coles
(Eds.), Advances in psychophysiology (Vol. 4, pp. 1-42). London:
Jessica Kingsley.
Ekman, P. (1992). An argument for basic emotions. Cognition and
Emotion,6, 169-200.
Fehr, B., & Russell, J. A. (1984). Concept of emotion viewed from a
prototype perspective. Journal of Experimental Psychology: General,
113, 464-486.
Folkman, S., & Lazarus, R. S. (1985). If it changes it must be a process:
Study of emotion and coping during three stages of a college
examination. Journal of Personality and Social Psychology,48, 150-
170.
Frijda, N. H. (1993). The place of appraisal in emotion. Cognition and
Emotion,7, 357-387.
Frijda, N. H., Kuipers, P., & ter Schure, L. (1989). Relations between
emotion, appraisal and emotional action readiness. Journal of Per-
sonality and Social Psychology,57, 212-228.
Greenwald, M. K., Cook, E. W., III, & Lang, P. J. (1989). Affective judg-
ment and psychophysical response: Dimensional covariation in the
evaluation of pictorial stimuli. Journal of Psychophysiology,3, 51-64.
Gross, J. J. (1998). Antecedent- and response-focused emotion regu-
lation: Divergent consequences for experience, expression, and
physiology. Journal of Personality and Social Psychology,74, 224-237.
Heller, W., Nitschke, J., Etienne, M., & Miller, G. (1997). Patterns of
regional brain activity differentiate types of anxiety. Journal of
Abnormal Psychology,106, 376-385.
Higgins, E. T. (1997). Beyond pleasure and pain. American Psycholo-
gist,52, 1280-1300.
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1999). International affec-
tive picture system (IAPS): Technical manual and affective ratings.
Gainesville: The Center for Research in Psychophysiology, Uni-
versity of Florida.
Robinson, Clore / SIMULATION, SCENARIOS, AND APPRAISAL 1531
Lang, P. J., Greenwald, M. K., Bradley, M. M., & Hamm, A. O. (1993).
Looking at pictures: Affective, facial, visceral, and behavioral reac-
tions. Psychophysiology,30, 261-273.
Lang, P. J., Kozak, M. J., Miller, G. A., Levin, D. N., & McLean, A.
(1980). Emotional imagery: Conceptual structure and pattern of
somato-visceral response. Psychophysiology,17, 179-192.
Lang, P. J., Levin, D. N., Miller, G. A., & Kozak, M. J. (1983). Fear
behavior, fear imagery and the psychophysiology of emotion: The
problem of affective response integration. Journal of Abnormal Psy-
chology,92, 276-306.
Lazarus, R. S. (1991). Progress on a cognitive-motivational-relational
theory of emotion. American Psychologist,46, 819-834.
Lazarus, R. S. (1995). Vexing research problems inherent in cogni-
tive-mediational theories of emotion—and some solutions. Psycho-
logical Inquiry,6, 183-196.
Lazarus, R. S., & Alfert, E. (1964). Short circuiting of threat by experi-
mentally altering cognitive appraisal. Journal of Abnormal and
Social Psychology,69, 195-205.
Lewis, M. D. (1996). Self-organizing cognitive appraisals. Cognition
and Emotion,10, 1-25.
McFarland, C., Ross, M., & DeCourville, N. (1989). Women’s theories
of menstruation and biases in the recall of menstrual symptoms.
Journal of Personality and Social Psychology,57, 522-531.
Miller, G. A., Levin, D. N., Kozak, M. J., Cook, E. W., McLean, A., &
Lang, P. J. (1987). Individual differences in emotional imagery.
Cognition and Emotion,1, 367-390.
Öhman, A., & Soares, J.J.F. (1994). Unconscious anxiety: Phobic re-
sponses to masked stimuli. Journal of Abnormal Psychology,103, 231-240.
Ortony, A., Clore, G. L., & Collins, A. (1988). The cognitive structure of
emotions. New York: Cambridge University Press.
Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The measure-
ment of meaning. Urbana: University of Illinois Press.
Parkinson, B. (1997). Untangling the appraisal-emotion connection.
Personality and Social Psychology Review,1, 62-79.
Parkinson, B., & Manstead, A.S.R. (1992). Appraisal as a cause of
emotion. In M. S. Clark (Ed.), Review of personality and social psychol-
ogy (Vol. 13, pp. 122-149). New York: Russell Sage.
Parkinson, B., & Manstead, A.S.R. (1993). Making sense of emotion
in stories and social life. Cognition and Emotion,7, 295-323.
Polivy, J. (1981). On the induction of emotion in the laboratory: Dis-
crete moods or multiple affective states. Journal of Personality and
Social Psychology,41, 803-817.
Reisenzein, R., & Hofmann, T. (1993). Discriminating emotions from
appraisal-relevant situational information: Baseline data for struc-
tural models of cognitive appraisals. Cognition and Emotion,7, 271-
293.
Robinson, M. D. (1998). Running from William James’ bear: A review
of preattentive mechanisms and their contributions to emotional
experience. Cognition and Emotion,12, 667-696.
Robinson, M. D., & Clore, G. L. (2001). Belief and feeling: Evidence for an
accessibility model of emotional self-report. Manuscript submitted for
publication.
Robinson, M. D., & Johnson, J. T. (2000). Temporal characteristics and
anticipated effort: A framework for differentiating between stress and other
negative states. Manuscript submitted for publication.
Roseman, I. J. (1991). Appraisal determinants of discrete emotions.
Cognition and Emotion,5, 161-200.
Roseman, I. J., Spindel, M. S., & Jose, P. E. (1990). Appraisals of emo-
tion-eliciting events: Testing a theor y of discrete emotions. Journal
of Personality and Social Psychology,59, 899-915.
Russell, D., & McAuley, E. (1986). Causal attributions, causal dimen-
sions, and affective reactions to success and failure. Journal of Per-
sonality and Social Psychology,50, 1174-1185.
Russell, J. A. (1980). A circumplex model of affect. Journal of Personal-
ity and Social Psychology,39, 1161-1178.
Russell, J. A. (1987). Comments on articles by Frijda and by Conway
and Bekerian. Cognition and Emotion,1, 193-197.
Scherer, K. R. (1984). Emotion as a multicomponent process: A
model and some cross-cultural data. In P. Shaver (Ed.), Review of
personality and social psychology (Vol. 5, pp. 37-63). Beverly Hills, CA:
Sage.
Scherer, K. R. (1988). Criteria for emotion-antecedent appraisal: A
review. In V. Hamilton, G. H. Bower, & N. H. Frijda (Eds.), Cogni-
tive perspectives on emotion and motivation (pp. 89-126). Dordrecht,
the Netherlands: Kluwer.
Scherer, K. R. (1993). Studying the emotion-antecedent appraisal
process: An expert system approach. Cognition and Emotion,7, 325-
355.
Scherer, K. R., Wallbott, H. G., & Summerfield, A. B. (Eds.). (1986).
Experiencing emotion: A cross-cultural study. Cambridge, UK: Cam-
bridge University Press.
Shaver, P., Schwartz, J., Kirson, D., & O’Connor, C. (1987). Emotion
knowledge: Further exploration of a prototype approach. Journal
of Personality and Social Psychology,52, 1061-1086.
Shields, S. A. (1991). Gender in the psychology of emotion: A selec-
tive research review. In K. T. Strongman (Ed.), International review
of studies on emotion (Vol. 1, pp. 227-245). New York: John Wiley.
Smith, C. A. (1989). Dimensions of appraisal and physiological
response in emotion. Journal of Personality and Social Psychology,56,
339-353.
Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal
in emotion. Journal of Personality and Social Psychology,48, 813-838.
Smith, C. A., & Ellsworth, P. C. (1987). Patterns of appraisal and emo-
tion related to taking an exam. Journal of Personality and Social Psy-
chology,52, 475-488.
Smith, C. A., Griner, L. A., Kirby, L. D., & Scott, H. S. (1996, August).
Toward a process model of appraisal in emotion. Paper presented at the
ninth meeting of the International Society for Research on Emo-
tions, Toronto, Canada.
Smith, C. A., & Lazarus, R. S. (1990). Emotion and adaptation. In
L. A. Pervin (Ed.), Handbook of personality: Theory and research
(pp. 609-637). New York: Guilford.
Smith, C. A., & Lazarus, R. S. (1993). Appraisal components, core re-
lational themes, and the emotions. Cognition and Emotion,7, 233-269.
Stein, N. L., Trabasso,T., & Liwag, M. (1993). The representation and
organization of emotional experience: Unfolding the emotion
episode. In M. Lewis & J. M. Haviland (Eds.), Handbook of emotions
(pp. 279-300). New York: Guilford.
Storm, C., & Storm, T. (1987). A taxonomic study of the vocabular y of
emotions. Journal of Personality and Social Psychology,53, 805-816.
Thomas, D. L., & Diener, E. (1990). Memory accuracy in the recall of
emotions. Journal of Personality and Social Psychology,59, 500-507.
Weiner, B. (1985). An attributional theory of achievement motivation
and emotion. Psychological Review,92, 548-573.
Wilson, T. D., Lisle, D. J., Kraft, D., & Wetzel, C. G. (1989). Preferences
as expectation-driven inferences: Effects of affective expectations
on affective experience. Journal of Personality and Social Psychology,
56, 519-530.
Wilson, T. D., Wheatley, T., Meyers, J. M., Gilbert, D. T., & Axsom, D.
(2000). Focalism: A source of durability bias in affective forecast-
ing. Journal of Personality and Social Psychology, 78, 821-836.
Winter, K. A., & Kuiper, N. A. (1997). Individual differences in the
experience of emotions. Clinical Psychology Review,17, 791-821.
Zajonc, R. B. (1980). Thinking and feeling: Preferences need no
inferences. American Psychologist,35, 151-175.
Zajonc, R. B. (1984). On the primacy of emotion. American Psycholo-
gist,39, 117-123.
Received July 19, 2000
Revision accepted October 5, 2000
1532 PERSONALITY AND SOCIAL PSYCHOLOGY BULLETIN
