Sessão3 Marcus1988

The Structure of Emotional Response: 1984 Presidential Candidates
Author(s): George E. Marcus

Source: The American Political Science Review, Vol. 82, No. 3 (Sep., 1988), pp. 737-761
Published by: American Political Science Association
Stable URL: http://www.jstor.org/stable/1962488 .
Accessed: 26/01/2015 19:39
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
American Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to
The American Political Science Review.
http://www.jstor.org
This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

All use subject to JSTOR Terms and Conditions
THE STRUCTUREOF
EMOTIONALRESPONSE:
1984 PRESIDENTIAL
CANDIDATES
GEORGE E. MARCUS
WilliamsCollege
Over the past two decadespsychologicalmodels of affect

have changedfrom valence (one-dimensional)models to multiple-dimensionalmodels.
The most recentmodels, circumplexmodels, are two-dimensional.Feelingthermometer
measures,which derivetheirtheoreticallogic from earlier(valence)modelsof emotional
appraisal,are shown to be confounded. Underlyingthe variationobtainedusingfeeling
thermometermeasuresare two dimensions of emotional response, mastery (positive
emotionality)and threat(negativeemotionality).Analysis of the 1984 NESsurvey sug-
gests that positive emotional response is twice as influential as negative emotional
responsein predictingpresidentialcandidatevote disposition to the presidentialcandi-
dates. Reliance on emotional response is shown to be uniformly influential across
various strata of the electorate.
Policy considerationshave little direct influenceon vote disposition, though policy
considerationsare indirectlyrelatedto vote disposition throughthe influenceof issues
on the degreeof feelings of threatevoked by the candidates.
It has been 15 that they add explanatorypower to our

years since it was first recognized that understandingof a variety of political
asking people about their feelings toward phenomena.'
presidentialcandidateswas equivalentto While the studies cited above have
ascertainingtheirvote disposition(Brody established the value of studying emo-
and Page 1973). Only recentlyhave feel- tions and in redressing the imbalance
ings-emotions-begun to be the subject resultingfrom the dominanceof the cog-
of serious study by political scientists. nitive approach to human behavior, it
The recent studies on emotions and poli- must also be noted that thesestudieshave
tics have shown that feelings have strong failed to provide a sound theoretical
predictive effects on candidate appraisal focus. The recent attention to affect has
(Abelson et al. 1982), on issue appraisal been markedby contradictoryand anom-
(Conoverand Feldman1986), on political alous findings;so that even when the cor-
participation(Marcus1985) and on polit- rect structureof emotional response has
ical communication(Roseman, Abelson, been identified, its significanceand rele-
and Ewing1986). The primaryconcernof vance have not been properlyunderstood
the initial studies has been to establish (e.g., Abelson et al. 1982). Contradictory
that emotions are worthy of study and findings have been reported in different
AMERICAN POLITICAL SCIENCE REVIEW

VOL. 82 NO. 3 SEPTEMBER1988

American Political Science Review Vol. 82
studiesas to the numberof dimensionsof pretations and that variations in these

emotionality.In a study of political com- strategicresponseshave importantpolit-
munications, Roseman, Abelson, and ical implications.
Ewing(1986)reportfour dimensions.In a Beforeturningto circumplexmodels of
study of economic issues, Conover and affect, it will be useful briefly to review
Feldman(1986) report three dimensions. the theoretical rationale of valence and
In a study of candidateappraisal,Abelson discretemodels of affect.
and his colleagues (1982) report two
dimensions. And a recent study of the
vote decision, Rahn and colleaguesn.d., Valence Models of Emotion
relies on one dimension of affect eval-
uation. In this approachto the study of beliefs
The anomalousresultperplexinginves- and attitudesaffect is definedas the eval-
tigators has been the consistent finding uative dimensionof an attitude,hencethe
that when individualsregard a situation term valence. Attitudes and beliefs are
or event (suchas a candidate,issue, or cir- held about some object (e.g., person or
cumstance) they often simultaneously issue). Attitudes are composed of three
reportgood and bad feelings (Dienerand dimensions: affect, behavioral disposi-
Emmons1984; Warr, Barter,and Brown- tion, and cognition.Affect and evaluation
bridge 1983). Abelson and his colleagues are treatedas synonymous (Fishbeinand
(1982)find, afterexaminingthe emotional Ajzen 1975). Eachattitudesubjectis eval-
responses to various candidates for na- uated positively or negatively or neutral-
tional office, that "perhapsour most psy- ly. The assumptionunderlyingthis class
chologicallyprovocativeresultis the near of models is that feelings are unidimen-
independenceof negative and positive af- sional and effectively bipolar (i.e., posi-
fect"(i.e., the frequentconcurrentreports tive feelings at one end, negativefeelings
of both good and bad feelings).All of the at the other, and neutral feelings in the
studies that use multiple dimensions of middle). That some may evaluate an ob-
emotionality report similar "anomalous" ject positively while others are neutralor
findings.Until and unless the structureof negativetowardsthe objectis to be under-
emotional responseis clarified,how feel- stood as a matterof individualdifferences
ings interactwith thinking and influence -an empiricalmatter to be studied. The
political behavior will remainobscured. tripartiteconception of attitudesis grad-
Threeclassesof models of the structure ually being abandonedby psychologists,
of affect are generally used: valence because emotion and cognition are com-
models, discrete models and circumplex ing to be recognizedas separate mental
models. Valence models are the basis of processes. It retains appeal to some
the familiar "thermometerscales" in the (Breckler1984).
National Election Studies surveys. Dis- In political science, feeling thermome-
crete models are the basis of the feeling ters are the most frequentapplicationof
measuresin the NES. Circumplexmodels valencemodels of affect. The feelingther-
have been a recent developmentand will mometershave been a popularmeasureof
be unfamiliarto most political scientists. assessing feelings elicited by candidates
Circumplex models offer a theoretical and by groups, parties, organizations,
approach to variation in feelings that and economic classes. The 1984 NES lists
resolves the conflicting and anomalous 50 feeling thermometers,includingthe 18
findings.More importantly,the theory of feeling thermometerson political figures.
circumplexmodels of affect suggeststhat The validity of findings that rely on feel-
feelings provide stable strategic inter- ing thermometersis dependentupon the
738

Emotional Response
assumptionthat affect has a bipolar one- (e.g., Storm and Storm 1987; Wiener,
dimensionalstructure.As this study will Russell,and Lerman1978, vol. 2). In par-
demonstrate,that assumptionis invalid. ticular, discretemodels posit that evalua-
tion (emotionalappraisal)is to be under-
stood in more strategicterms by positing
DiscreteModels of Emotion a link between specific adaptive benefit
Discretemodels of emotions share cer- (Pluthik 1980). Further, by emphasiz-
tain presumptions. First, different emo- ing the emotional content of the repre-
tions are identified as "alternativeap- sentationof sensory stimuli and its influ-
praisals of events" (Roseman1984). The ence on thought and action, discrete
number of fundamentalemotions differs models have begun the transition from
from one model to another.Davies (1980) the orthodox position that cognitive
identifies seven principal affects: moral processes can provide a comprehensive
indignation, anger or hatred, envy or account of thinking, awareness, and
resentment,fear or suspicion, powerless- behavior.2
ness or disillusion,sympathyor pity, and However, proponents of discrete
loyalty or pride. Izard (1971, 1977) pro- models disagreeat the theoreticallevel as
posed a 10-fundamental-emotions model. to the numberof discreteemotions (rang-
Roseman has proposed 13 fundamental ing from 7 to 14); and at the empirical
emotionsthat characterizethe structureof level, studies of politics using discrete
emotions (Roseman 1979). His most re- models report varying dimensions of
cent list of the fundamental emotions emotionalappraisal(rangingfrom2 to 4).
includes 14 fundamentalemotions: joy, And, of greatest importance, empirical
relief, hope, fear, frustration,liking, dis- studies consistentlyreport that individu-
liking, anger, guilt, pride, and regret als reportboth positive and negativefeel-
(Roseman1984). Different combinations ings so frequently that discrete models
of motivationalstates and situationalfac- must concludeeitherthat this is an anom-
tors present different circumstances, alous findingor that most of the timepeo-
which then, in turn, elicit differentinter- ple are in conflictedmotivationalstates.
pretations (cognition) that produce an
appropriate emotion (cognitively trig- Modelsof Emotion
Circumplex
gered affect). Perceptions trigger emo-
tional responsesdependingon the motiva- More recent developmentshave iden-
tional circumstance(for example,whether tified the structure of affect as a two-
the event is perceived as rewarding or dimensional circumplex (Plutchik 1980;
punishing)and on the situationalcontext Russell 1980; Zevon and Tellegen 1982).
(for example,whetherthe event is seen as The two dimensionsare generallyidenti-
personally invested or personally neu- fied as positive emotionalityand negative
tral). A situation sensed as punitive will emotionalityeffectivelyyunipolaraxes at
be experiencedas a negativeemotion:fear orthogonal locations). Each dimension
or distress if control of the situation is representsa specific arousalsystem. One
attributedto oneself, anger or hatred if arousal system characterizes certain
control is attributedto someone else. motor-sensory signals as positive emo-
Discretemodelshave made some useful tionality, rangingfrom dull and depressed
contributions,principallyin beginningto when arousalis low and to elated enthu-
pursueparsimonyin the structureof emo- siastic when arousal is high. The second
tional appraisal, particularly important arousal system characterizes certain
since the numberof distinguishableemo- motor-sensory signals as negative emo-
tional terms is measuredin the hundreds tionality, ranging from calm and placid
739

Figure1. EmotionalResponse to son and Tellegen 1985; see also Masters

Mondale on Seven Mood Terms and Sullivan 1986).
The circumplexmodel resolveswhy dif-
1.0
ferent studies have reported more than
0.8 two dimensions. Studies that include af-
0.6d /s usued afraid fect termsthat are heavily clusteredin dif-
0.4-
ferent regions of the circumplexand rely
on oblique solutions will indeed report
0.2
three or more dimensions(e.g., Conover
.0.01 sympathetico and Feldman 1986; Roseman, Abelson,
o .0.2 ~~~~~~~~~~~~hopmeful
and Ewing1986). However, reanalysisof
when proud
studies with multiple oblique factors
0~ -064 reports results that unambiguouslysup-
.0.2
0 port the two-dimensional circumplex
-0. model (Watson and Tellegen 1985). The
-1.0
Abelson and colleagues 1982 study of
-1.0-0.8-0.6-0.4-0.2.0.0 0.2 0.4 0.6 0.8 1.0 emotional appraisal of candidates, de-
RotatedFactor1 (PositiveAffect) signed to demonstrate the relevance of
emotions using one discrete model of
emotions (Roseman's structural model;
when arousal is low to nervous and see Roseman1979, 1984), reports results
anxiouswhen arousalis high. The circum- that validate the circumplex model
plex model locates all affect termswithin (thoughthey do not recognizeit as such).
the circumplex,the location varying as a Thus studies designed from a discrete
resultof the differentialdegreesof arousal model perspective, using measures in-
on each dimension. Studies of facial tended to demonstratethe distinct influ-
expression, tones of voice, and affect ences of each primary emotion, have
words-in Englishand in other languages found, to the contrary,two dimensionsof
-all report the same basic two-dimen- affect-positive and negative.
sional structure(studiesreviewedin Wat-
Figure2. EmotionalResponse to PreliminaryEmpiricalTest

Reagan on Seven Mood Terms
The 1984 NES data afford an oppor-
1.0 tunity to compare the circumplexmodel
- 0.8 disgustedu neasy of emotion against valence and discrete
0.6 / afaid gr models as it appliesto emotionalresponse
to presidentialcandidates.Replicatingthe
0.4 /
methodology used in previous NES stud-
? 0.2 sympathetic ies, respondents in the 1984 NES were
'O.0- ~~~~~~~~~~rud asked to regard the presidential candi-
0 -0.2- hopeful6
dates and asked if "somethingabout that
.0.4
person, or something he has done has
\
made you have certainfeelingslike anger
] -0.6 \/ or pride."Subjectsthen reportedwhether
-0.86 they have ever felt that emotion, using a
.1.0 list of seven discrete emotions selected
-1.0 -0.8 -0.6 -0.4 -0.2 -0.0 0.2 0.4 0.6 0.8 1.0 from Roseman's discrete model of the
Rotated Factor 1 (Positive Affect) structure of emotions (angry, hopeful,
740

Emotional Response
Figure 3. Gray-Tellegen Two-Factor

Model of Emotional Appraisal
Inputs outputs
Behavioral
Activation
Signalsof Reward Beha--ora (approach
learning;
mastery)
Signalsof Nonpunishment Activation

Activation Increment
in Arousal
(Enthusiasm,
excitement)
Signalsof Mastery System
FocusedSensoryAnalysisof
reward/non-punishment
stimuli
TheBASSystem-Appraisal
andEvaluation
of TaskPerformance
(Scanningfor Indicationsof Success/Failureof SubjectEngagedActions)
Inputs Output
Signalsof Punishment
l ~Behavioral Behavioral
Inhibition
Signalsof Nonreward
Inhibition N Increment
in Arousal
Novel Stimuli (Anxiety)
. --
InnateFearStimuli Increased
Attention
The BIS System-Appraisal and Evaluationof IntrusiveInformation

(Scanningfor ExternalDangeror Threat)
afraid of him, proud, disgusted, sympa- sional. These resultscontradictthe expec-

thetictowardshim, uneasy).This analysis tations of the valencemodel that affective
should provide a robust test since the evaluation is one-dimensionaland vali-
selection of the affect termswas intended dates the circumplexmodel's expectation
to supportthe discretemodel of emotions that affect is two-dimensional.The only
(Abelson et al. 1982; Roseman,Abelson, departurefrom the circumplexmodel is
and Ewing1986). Discretemodels of emo- the somewhat oblique relationship be-
tions shouldfind a multidimensionalsolu- tween the two dimensions(which will be
tion, probably oblique in structure.Fig- addressedmore fully in the later covari-
ures 1 and 2 display plots of the factor ance analysis). These results are not
loadingsfrom factoranalysesof the seven specificto the 1984 NESdata; comparable
emotional measuresfor Mondale and for findingsare reportedin Abelson and col-
Reagan. leagues 1982 and Sullivan and Masters
Only two dimensionsemergefrom the 1988.
factor analysis.3The plotted factor load- The discrete model, as well as a lay
ings replicate the findings reported in grasp of the labels positive and negative,
Abelson et alji 1982; emotional response would suggestthat people feel eitherposi-
to presidentialcandidates is two-dimen- tive or negative feelings, but not both
741

(i.e., affect dimensionswould be bipolar scene-reflect the kinds of stimuli that

ratherthan unipolar).The terms positive may elicit negative affects.
and negative seem mutually exclusive; The two systemsefficientlybalancethe
muchlike blackor white, up or down. Yet assignment of mental and physical re-
this is clearlynot the case. To resolve this sourcesbetweenself-absorption(full allo-
dilemma we need to turn to a more cation to the task, or social activity, at
specificcircumplexmodel of affect. hand, which is adaptive and efficient if
The personalitymodels of Gray (1970, anxiety is low-i.e., a safe environment)
1973, 1981) and Tellegen (1985) identify and the claims of attentionwarrantedby
emotionality as the manifestationof two externalcircumstanceswhen confronting
primary systems of emotional arousal.4 a threat (Simon 1967). Further,the two
Gray calls the first system the behavioral systems of emotionalityare well adapted
activation system (BAS). Tellegencalls it to maintainingsocial integrationthrough
positive emotionality.The second system emotional display (Mastersand Sullivan
in Gray'smodel is the behavioralinhibi- 1986; Sullivan and Masters 1988). The
tion system (BIS), which Tellegen calls developmentof models of affect has been
negative emotionality. Figure 3 depicts a steady trendmoving from a recognition
the Gray-Tellegenmodel of affect. that affect is engaged in evaluation
The primarycontributionof the Gray- (valence models) to the further recogni-
Tellegen model of emotion is that each tion that affect is also linked to motiva-
system of emotionalresponseis identified tion (discretemodels)and, more recently,
as having a distinct strategic function. to the recognitionthat affectsupportstwo
One system is oriented towards apprais- specific strategicevaluations (circumplex
ing the implementation of subject- models).
engaged action (positive emotionality). The identification of two principal
The strategicfocus of the positive dimen- sourcesof behavior, self-initiatingor self-
sion of emotionalityis self-regardingand engaged action on the one hand and
task management.3This system monitors action initiated in response to external
the ongoingmasteryand controlachieved causeson the otherhand, is a familiarone
by the individualas a resultof the recruit- in political science. Forexample,political
ment and allocation of mentaland physi- efficacy is currently defined as two-
cal resources.This systemproducesvaria- dimensional(Craig and Maggiotto 1982;
tion in the arousal of positive emotion, Shingles 1986). One dimension concerns
affects of enthusiasm, pleasure, and ex- attitudes toward self-initiating political
citement. Such affects occur when the action (internalpoliticalefficacy)and one
task at hand is going well (andconversely, dimension concerns attitudes assessing
when things are going badly, affects of actions by others (externalpolitical effi-
depressionare likely). The secondsystem, cacy). In psychology, a number of per-
negative emotionality, monitors the en- sonality theories depend on this distinc-
vironment for signs of evident danger. tion. For example, recent work on self-
This system is environment-regarding; it monitoring (Snyder and Ickes 1985) and
is concerned with scanning for threats on field dependence also focus on the
that may intrude and disrupt ongoing location of mastery(self versus situation)
subject-engagedaction. Some examplesof and the degreeof mastery (high or low).
such stimuli-loud or sudden noises or Those describedas high self-monitorsrely
the appearanceof a dangerin the form of on situational information to regulate
sudden bad weather during an airplane their behavior. Those described as low
flight or an unexpectedly violent indi- self-monitors rely on information from
vidual or hostile person appearingon the relevant inner states to regulate their
742

EmotionalResponse
behavior. The locus of control literature become so suddenlyor unexpectedly,the

(Phares1978; Rotter 1966) also identifies greaterthe anxiety and distress.
two personalitytypes on the basis of the The strategic significanceof these ap-
issue of mastery: contrasting between praisalssuggestswhy emotionalresponses
those who believe that outcomes are the to candidates,and otherpoliticalsubjects
result of their own behavior internalss) and events, are likely to have such empir-
and those who believe that outcomes are ical predictivevalidity. Helplessnessand
essentially contingent on luck, fate, or depression are inextricably intertwined
powerful others (externals). (Seligman 1975) and concerns for mas-
The Gray-Tellegenmodel offers some tery, usually couched as autonomy in
advantages over these conceptions, political thought, have been central to
though it has much that it shares with liberal democratic theory (Bay 1965;
them. First, it distinguishesbetween the Fromm 1941; Lane 1978; Mill 1956). It
two dimensions,threatand mastery, and would be surprisingindeedif information
the degree of arousal (high or low) on about candidatesas to the enhancement
each dimension. Second, it differentiates or diminishingof masteryor the increase
between mood and temperament(again or decreaseof threatdid not play a central
enabling situationaland dispositionalin- role in influencing support for political
fluencesto contend).6Third, it establishes leaders.
connectionsbetween emotional response
and motivation (somethingthat informa-
tion-processingmodels of cognitionfail to Data Analysis
do [Hastie 1986]). Finally, while it is
almost certain that the Gray-Tellegen Fitting a confirmatoryfactor analysis
model is incomplete in important ways (LISREL)model (Joreskog 1969, 1970,
(Derryberryand Rothbart1984), its iden- 1973) to the 1984 NES data is anieffective
tification of two fundamentalsuperordi- way of assessingwhether the circumplex
nate dimensions is parsimonious and model can provide a superioralternative
capable of emendationin light of further to discreteand valence model interpreta-
work. tions of the candidate feeling measures.
With respectto emotional responsesto LISRELrequiresthat a model be specified
politicians, the Gray-Tellegenmodel also on theoreticalgrounds. The LISRELpro-
has specific advantages.The two dimen- gramevaluatesthe goodnessof fit, identi-
sions of emotional response suggest that fying wherethe model and data do and do
political candidatesare appraisedon two not fit. The data consist of the covariance
fundamentalcriteria. First, Is this candi- structuredescribingthe relationshipsbe-
date capable of being in personal com- tween the various feeling measures and
mand and able to make the environment the feeling thermometers.The variables
more tractable, predictable,and bounti- are treatedas manifestindicatorsof latent
ful, assuming that such resources are constructs.The Gray-Tellegencircumplex
being used to achievepolicies and actions model identifies the proper structureof
supported by the individual (mastery)? emotionsas two dimensions,masteryand
The more these featuresare apparent,the threat, each effectively unipolar, and the
greater the enthusiasm. Second, Is the relationshipbetween the two dimensions
candidate likely to pursue policies and as orthogonal. Valence models would
actionsnot supportedby the individualor posit a single effectively bipolar dimen-
give evidence of lack of statureor moral sion and discrete models would posit a
weakness (threat)?The more these fea- multidimensionalsolution of 10, 13, or 14
tures are apparent, especially if they oblique dimensions (limited in this in-
743

Figure 4. Two-Factor Model of Emotional Response: Mondale

-.46
(Environment
) Tsk-
5 .68
~~.67 .81
.78 .63 .85 .76
angry afri uneasy disgusted feeling hope proud syptetic

thermomfeter
MONDALE.0
.10 .28 -.08

Source:1984 NES.
stance to 7 by the number of measures for ULS estimation; however, goodness-

availablein the NES). of-fit indexes and, more importantly,the
The latent constructmasteryis defined root mean squareresidualvalue are pro-
by the indicators hopeful, proud, and vided. For Mondale, the goodness-of-fit
sympathetic.The latentconstructthreatis index is .999, the adjustedgoodness of fit
defined by the indicators angry, afraid, is also .999, and the root mean square
disgusted, and uneasy. The feeling ther- residual is .013. For Reagan, the good-
mometers for Reagan and Mondale are ness-of-fit index is .999, the adjusted
identified as confounded measures; that goodness-of-fitis .998, and the root mean
is, the feelingthermometersare indicators squareresidualis .014. These values dem-
that are causally dependent upon both onstratethat each model fits the data very
latent constructs. well.
The model is estimated using poly- The variancein feeling thermometersis
choric and polyserial correlations, cal- essentiallydefinedby the two latent con-
culated by LISREL,because the feelings structsof emotional response, threat and
measuresreportedfor each candidateare mastery. The model accountsfor much of
dichotomous. As a result the models are the varianceof the feeling thermometers;
estimatedusing unweightedleast squares R2 = .78 for Mondale's feeling ther-
(ULS).7Figures4 and 5 presentthis model mometerand R2 = .74 for Reagan'sfeel-
specification and the standardizedsolu- ing thermometer.These R-squaredvalues
tions for Mondale and Reagan. are as high as are generallyreportedfor
The chi-squarevalue is not available single items in measurementmodels (the
744

EmotionalResponse
Figure 5. Two-Factor Model of Emotional Response: Reagan
-.54
((Environ ent r (Task-
.83 .2
87 865
.70.5
angry afraid disgusted uneasy feeling hope prou sympathetic
Source: 1984 NES.
remainingvarianceis assumedto be error the two latentfactors,threatand mastery,

variance;few single items are completely may be due to either the choice of affect
reliableindicatorsof a latent construct). terms (see Appendix) or perhaps to the
There are some correlated errors in propensity for respondentsto positively
each model, none from one domainto the respondto politicalcandidates(Lippmann
other domain of great magnitude (the 1922).
highestbeing r = .10 from afraidto sym- These results provide the preliminary
patheticin the Reaganmodel). The factor evidencethat using the feelingthermome-
analysis, reportedin Figures1 and 2, sug- ter measures, even as a summary, "on-
gestedthat sympathyis more problematic balance" measure of affect evaluation,
as a measureof mastery (positive affect). obscures more than it reveals. Mastery
The relatively low epistemiccorrelations plays a more influentialrole than threatin
(reliabilities)in each model and the epis- defining the feelings. towards Mondale
temic relationship from threat to sym- and Reagan. The epistemic correlations
pathy in the Mondalemodel confirmthat suggest that mastery may be as much as
assessment.Tomkins(1981)has suggested twice as influentialas threat(an interpre-
that some affective terms are combina- tation consistentwith the reportedfactor
tions of biological primary affects and analysis).9 This suggests that identifica-
cognitivereferences.The LISRELand fac- tion with the candidate(positiveaffect)is
tor analysis of sympathy suggests that it the more potent considerationinfluencing
may be one of these combinations.8The the disposition to vote for a candidate-
substantialnegative correlationsbetween though threatis also a powerfulinfluence
745

(not surprisingin the light of the frequent found that respondentsutilizea variety of
use of "negativecampaigning"). considerations,the variousconsiderations
But before turning to the question of can be divided into two domains. The
how feelings may influence voting, can first domain contains appraisalof moral
more evidence be brought forward to leadership (integrity, empathy, honesty,
validate the task management-environ- and the like). This domain identifies
ment regarding distinction? Theories of features of a candidatethat are likely to
democratic leadership identify at least elicit feelings of respect and trust. The
two majorideal types of leadershiproles, second domain identifies appraisal of
trustee and delegate (Pitkin 1967). The command: Does the performanceof the
trustee role, as an ideal type, suggests candidate yield, whether from luck or
leadershipprincipallyevaluatedon bonds skill, positive outcomes? This second
of trust and identification(good trustees domain contains appraisals of perform-
would displaypositive qualitiesof compe- ance in meetingthe crisesand demandsof
tence and rectitude).The delegaterole, as the world at large. The failureof a leader
an ideal type, suggests leadershipprinci- to meet the confronting challenges suc-
pally evaluated on common policy ac- cessfully should, according to the task
tions and issue agendas. management-environmentregardingdis-
While as normative ideals these types tinction elicit more feelings on the threat
are drawnas distinct,we can assumethat, dimensionthan on the masterydimension
empirically, elements of each may have of emotional appraisal. Though, if the
some influenceon leadershipappraisal.1o failure is perceivedto be due to a failure
Of course, specificcandidatesarelikely to of character, depression rather than
display features relevant to each ideal anxietymay be a dominantfeelingamong
type of representativeand thereforeelicit followers. The 1984 NES contains varia-
assessmentsby the public on standards bles that can be used to construct mea-
derivedfrom each type. If we can identify sures of both moral leadershipresources
characteristicfeatures of candidatesthat (the first domain) and leadershipcompe-
are apparentlylinked to leadership,such tence (the second domain).
as trustand identification,and featuresof The 1984NEScontainsa numberof dif-
candidatesthat are apparentlylinked to ferent assessmentsof the two presidential
policy, then it would be possible to test candidates:their personal characteristics
whetherthe formerare likely to elicit feel- (15 measures)and Reagan'sjob approval
ings of confidence and enthusiasm and ratings (4 measures). In addition, ques-
whetherthe latter are likely to elicit feel- tions ascertaining issue positions were
ings of threat. We can hypothesize that used to construct measuresof the close-
personal characteristicsof a candidate, ness of the respondentto the two candi-
featuresthat suggestpersonalcontrol and dates on leadingforeignpolicy issues and
probity, are more likely linked to feelings domesticpolicy issues. ForReaganwe can
of mastery and that issue positions, de- assess leadershipcompetenceby combin-
scribing-plansto shapethe environmentin ing the measuresof Reagan'sjob approval
particularways, are more likely linked to (a simple additive scale combining the
feelings of threat. four measuresof Reagan'sjob approval,
Studies of candidate appraisal have V258, V260, V262, and V264. Cronbach's
found that a variety of considerations a = .88).11A comparableset of measures
may play a role in assessing candidates for Mondale is not available.
(Kinder 1986; Miller, Wattenburg, and For both candidatesthe personal char-
Oskana 1986). While these studies have acteristicsof the candidatecan be assessed
746

Emotional Response
by using the 16 assessmentsfor Reagan It would seem plausible to argue that

(V319-34) and Mondale (V335-50). Fac- perception of a candidate taking a dif-
tor analyses of these 16 assessments ferent policy position would most likely
("How well does the word de- generate feelings of threat (since, if
scribe 7") shows a single factor elected, that candidatewould attempt to
accountingfor approximately60% of the move the society in directionsnot favored
common variance for both candidates by the respondent).Thuswe can hypothe-
(the second factor in each case accounts size that the closer the respondent per-
for less than 10% of the common vari- ceives the candidateto his or her position,
ance). Other studies have found more then the less threatening;and that this ap-
dimensions (Miller, Wattenburg, and praisal should have little impact on feel-
Oskana 1986). A factor analysis of the ings of mastery.On the other hand, inso-
1984 NES measuresfinds that single fac- far as the relationshipbetween political
tors best describeReagan'scharacteristics leader and supporteris one of identifica-
and Mondale'scharacteristics.Five terms tion (as almost all varietiesof theoriesof
loaded high on both first factors: hard- political leadership attest), then percep-
working, decent, compassionate, moral, tions of the candidateregardingthe effec-
and kind. The common content of this tive handlingof the tasks at hand should
factorcan best be describedas morallead- be more strongly linked to the elicitation
ership(i.e., mattersof rectitudeand lead- of positive emotionalitythan to the elici-
ership resources). Simple additive scales tation of negative emotionality.
for Reagan and Mondale were created In the two regression equations that
from thesefive items (Cronbach'sa = .88 follow, a number of candidate charac-
for Reagan and Cronbach'sca = .85 for teristics, policy appraisals, and one
Mondale). dimension of emotional response are en-
Issue appraisalmeasureswere created tered as independentvariablespredicting
using the simple absolute distances be- to the other of the two dimensions of
tween respondent(V122) and candidates emotionality. The purpose of the regres-
on the liberal-conservativescale (V126for sion equationsis to identify which candi-
Reagan and V130 for Mondale) and on date features are particularlyrelated to
the four domestic policy issues on which each dimension of emotionality. The re-
the respondent'sposition is ascertained gression equations should not be inter-
(V375, V382, V401 and V414: govern- preted as claims of causality. The recip-
ment spending, minority aid, economic rocal influences of emotional and cog-
status of women, and governmentguar- nitive appraisal remain insufficiently
anteed living-guaranteedjob). The relia- understoodto be able to specify with any
bilities for the respondent-Reaganprox- confidence the causal interplay of the
imity scale on domestic issues (Cron- various interpretiveprocesses. There is a
bach'sa = .81) and respondent-Mondale lively debatein psychologyon this subject
proximityscale on domesticissues (Cron- and little consensus(Fiskeand Pavelchak
bach's ax = .76) are acceptable. Similar 1985;Lazarus1982, 1984;Tassinaryet al.
proximity scales were created on three 1984; Tomkins 1981; Tsal 1985; Zajonc
foreign policy issues (V395, V388, and 1980, 1982). There is little doubt that
V408: defense spending, involvement in emotional processes influence memory
Central America, and cooperation with (Bower1981) and that emotionaland cog-
Russia) and acceptable reliabilitieswere itive processes are mutually influential,
obtained;Cronbach'sca= .75 for respon- but precisely how is currently not well
dent-Reaganand Cronbach'sca = .67 for understood (but see Millar and Tesser
respondent-Mondale. 1986). The primary purpose of these
747

equations, then, is to gain a betterunder- Given the tenor and rhetoricof the 1984
standingof the two distinctdimensionsof campaign,the respectivevulnerabilitiesof
feelings. the two candidates match closely these
Table1 presentsthe relationshipof can- findings; Reagan vulnerable on foreign
didate characteristics and policy ap- policy and Mondalevulnerableon domes-
praisals to the scale of threat (a simple, tic policy (Pomperet al. 1985). For Mon-
summated,equally weighted scale of the dale only, perceptionof distancefrom the
four negative affect terms, angry, afraid, voter on ideology has the consequenceof
disgusted, and uneasy; Cronbach'sa = evoking threat.This confirmsthe conven-
.74 for Reagan and Cronbach'sa = .75 tional wisdom that Reaganwas generally
for Mondale). Table 2 presents the rela- not perceivedin ideological terms. Table
tionship of candidate characteristicsand 2 shows that only on domesticpolicy and
policy appraisalsto the scaleof mastery(a only for Mondale is closeness between
simple, summated,equallyweightedscale candidateand respondentrelatedto posi-
of the threepositive affectterms, hopeful, tive emotionalresponse.
proud, and sympathetic;Cronbach'sa = Party identificationplays a more influ-
.68 for Reaganand Cronbach'sa = .68 ential role in determiningMondale'sap-
for Mondale). In order to obtain a mini- praisalthan it does Reagan's(thoughthis
mally acceptable scale, in terms of reli- is probablyinfluencedby the lack of a job
ability, the term sympatheticis retained performance assessment for Mondale;
for use in the scales for Reagan and for party identificationmay "pick up" that
Mondale. variance). Interestingly, the small but
The findings generally support the negative values for party identification
threat-masterydistinction as based on and feelings of mastery associated with
self-initiatingand environment-regarding Reagan suggest that the most partisan
appraisals.12 ForReaganand for Mondale Republicanshad somewhatmorereserved
moral leadership is strongly related to feelings about Reagan than less partisan
both threatand masterybut more strong- supporters (perhaps as a result of con-
ly associatedwith feelingsof masterythan tinued conservative complaint that the
it is with feelings of threat. For Reagan, Reagan performance in office had not
leadershipcompetence-his job approval done enough for the conservative agen-
-is more associatedwith threatthan it is da). More importantly,these small13'sfor
with mastery:the ratio of the fi for leader- party identificationshow that emotional
ship competenceto the fi for moralleader- response to Reagan was broadly shared
ship is about two to one.'3 and not very differentwhatever the par-
Of the six possible issue appraisals, tisan identificationof the respondent.By
closenessof the respondentto Reaganand comparison, Mondale evokes different
to Mondaleon ideology and domesticand feelings among Democrats than among
foreign policy issues, three have a sub- Republicans.This suggeststhatMondale's
stantial association with threat. The far- appeal (or lack thereof)was, much more
therReaganis perceivedfrom the respon- than Reagan's, limited to the pool of
dent on foreign policy issues, the more Democraticparty identifiers.
threateninghe is to the respondent(a rela- The equation for Mondale is less
tionship stronger than the influence of powerfulthan that for Reagan(the multi-
party identification). For Mondale, the ple R-squaredfor Mondale is some 20
crucialappraisalis perceivedclosenesson points lower). Most of this differenceis
domestic issues. The farther Mondale is apparentlydue to lack of informationon
perceived from the respondent on Mondale's competence. The importance
domestic issues, the more threatening. of job approvalin definingthe emotional
748

Emotional Response
Table 1. Regressionof Feelings of Threaton CandidateEvaluations

Reagan Mondale
CandidateEvaluation b t b 3 t
Candidatecharacteristics
Leadershipcompetence(Reaganjob
approval) -.47 -.44 -12.7
Moral leadership -.45 -.21 -6.8 -.56 -.20 -7.7
Party identification(Republican) -.06 -.08 -3.1 .13 .21 6.9
Policy appraisals
Domesticissues (closeness) -.03 -.02 - .8* -.18 -.15 -4.9
Foreignissues (closeness) -.13 -.13 -4.7 -.09 -.05 -1.6*
Ideologicalcloseness -.06 -.06 -2.2* -.10 -.10 -3.2
Emotionalresponse
Mastery .13 .13 4.9 -.04 -.05 -1.6*
R2 .52 .29
AdjustedR2 .52 .28
F 208.90 85.71
df 7, 1369 6, 1273
Source:1984 NES.
*Not significantat p .01.
Table 2. Regressionof Feelings of Masteryon CandidateEvaluations
Reagan Mondale
CandidateEvaluation b t b 3 t
Leadershipcompetence(Reaganjob
approval) .39 .36 10.2
Moral leadership .74 .33 11.5 .85 .28 11.5
Party identification(Republican) -.06 -.06 -3.1 -.22 -.31 -11.3
Policy appraisals
Domesticissues (closeness) .05 .04 1.4* .13 .10 3.4
Foreignissues (closeness) .05 .05 1.9* .06 .05 1.8*
Ideologicalcloseness .06 .05 2.2* .05 .04 1.5*
Emotionalresponse
Threat .13 .13 4.9 -.05 -.04 -1.6*
R2 .53 .34
Adjusted R2 .53 .34
F 224.8 111.5
df 7, 1369 6, 1273
Source:1984 NES.
*Not significantat p = .01.
749

characterization of President Reagan evaluationsmay be subjectto persuasion

might, as many have concluded, suggest and projection, thus making differentiat-
the benefit of incumbencyto candidates ing causalinfluencesamongthe proximity
for reelection. One handicap facing the judgments difficult. Further, a national
challengeris that demonstratingability to cross-sectionalsurvey is not a research
performin the office must be made with- design best adapted to the unravelingof
out the electoratehaving the opportunity the interplay between various sequential
to observe the challengerin the role of and simultaneousrelationships.Nonethe-
office holder (unless the challengerpre- less, we can begin to assess the relative
viously held the office). Of course, some influenceof the various assessmentscon-
incumbents find their history not so sideredon the propensityto vote for pres-
valued an asset (Fordand the pardon of identialcandidateby examiningthe asso-
Nixon; Carter and the perceived eco- ciations between the different influences
nomic record), and in this instanceMon- and vote intention.
dale has been Carter's vice-president. Brody and Page (1973), Kelley and
There is no formal evaluation of Mirer(1974)and Rahnand her colleagues
Mondale's expected job performance (n.d.) argue that people compare candi-
availablein the 1984 NESdata, so the two date images, summarizingjudgmentsand
analyses are not precisely equivalent. feelings so as to form a voting decision
Apart from this difference,the patternof based on which candidatefares better in
findings is remarkablysimilar. the comparativeprocess. Rahn and col-
On balance,the patternof findingssug- legues go on to show that using com-
gests that feelingsof threatare evoked by parativemeasuresconsistentlyproducesa
perceptionof job performance,by percep- more successfulpredictivemodel than do
tions of moral leadership,and by policy models using the respondents'positions
appraisals.The findings also suggest that on the issues and the respondent'sassess-
feelings of mastery are evoked equally ments of the candidates'characteristics.
stronglyby appraisalsof moralleadership The comparativeperceptionsapproach
rectitudeand leadershipcompetenceand is replicatedhere for most of the candi-
more weakly by party identificationand date characteristics.A measure for each
perhaps some policy appraisals. Finding respondent was created by subtracting
that each dimensionof affect covaries in Mondale's leadership scale score from
different ways and to different degrees Reagan'sleadershipscale score. The pro-
with policy appraisals and leadership cedure yields a positive value when the
characteristicsis consistentwith the theo- comparison advantages Reagan, zero
reticalassertionthat each system of emo- when the comparisonis seen as tied and
tional response has a different strategic negative when the comparison advan-
orientation. tages Mondale. The scores describe the
extent to which a respondent perceives
Reagan or Mondale to be the better
Vote Disposition leader, in the sense of more hardworking,
decent, compassionate,moral, and kind.
Evaluating how much influence cog- The more positive the scores, the more
nitive and emotional appraisals have, likely the respondentwill decide to vote
especially with respect to causality, is for Reagan;the more negativethe scores,
problematic, given the current state of the more likely the respondentwill decide
knowledge. Brody and Page (1972) argue to vote for Mondale (insofaras that com-
that distinguishingcausality is difficult parisonis influential).Similarlycompara-
because issue positions, perceptions,and tive judgmentsare measuredon ideology
750

Emotional Response
Table 3. Regression of Vote Disposition on Comparative Candidate and

Policy Evaluations
CandidateEvaluation b (3 t
Leadershipcompetence(Reaganjob approval) .27 .30 10.7
Moral leadership .04 .03 1.2*
Party identification(Republican) .09 .15 6.9
Policy appraisals
Domesticissues .06 .08 3.3
Foreignissuesa .03 .04 1.9*
Ideologicala .00 .00 .2*
Emotionalresponse
Masterya .14 .27 10.3
Threat -.08 -.14 -5.7
R2 .79
AdjustedR2 .79
F 489.3
df 8, 1016
Source:1984 NES.
aComparativecloseness.
(the respondent'sdistance from Mondale tion to vote for Reagan and a negative
subtractedfrom the respondent'sdistance value predicts to a greater intention to
from Reagan), domestic policy issues, vote for Mondale.15
foreignpolicy issues, mastery(subtracting A positive assessmentof Reagan'scom-
the mastery scale for Mondale from the petenceand the comparisonof feelingsof
masteryscale for Reaganfor each respon- masteryelicitedby the two candidatesare
dent), and threat. the two most importantfactorsin predict-
Leadership competence is available ing a vote disposition favoring Reagan
only for Reagan, as previously noted, so (j3'sof .30 and .27 respectively).Next in
this is the only measurethat is not com- importanceare party identification(3 =
parative. I assume, with Rahn and her .15) and feelings of threat (,3 = - .14).
colleagues,thatpartyidentificationincor- Consistent with the LISRELanalysis of
porates a comparativejudgment.Unlike the feeling measuresand the feeling ther-
Rahn and her colleagues, I treat affect as mometersfor Mondaleand Reagan,posi-
two-dimensional, with candidates being tive feelingsare more weighty than nega-
compared on the degree of feelings of tive feelings (a ratio of 2 to 1).16 As a
mastery and on the degree of feelings of group the policy appraisal measures do
threat.Table 3, on vote disposition(using badly in predicting the vote. Only the
the four-point scale for vote disposi- comparisonof perceptionof closenesson
tion),14 is regressed on candidate char- domestic issues reachesstatisticalsignifi-
acteristics, party identification, policy cance although its influence (,3 = .08) is
appraisals, and the two factors of emo- weak compared to the assessment of
tional appraisal, threat and mastery. As Reagan'scompetence, the two emotional
the vote dispositionscale is coded so that appraisalsand party identification. The
intention to vote for Reagan is high, a overallstrengthof the model is quitehigh,
positive value predictsto a greaterinten- the adjustedR2 = .79,17
751

It is strikingthat issue appraisalshave to define and measureelectoratecompe-

little apparentdirect influenceon voting. tency, among them differentindicatorsof
Feelingsabout the candidates,ratherthan cognitive complexity, ideological sophis-
thoughtful assessmentsregardingpublic tication, and constraint (Neuman 1986).
policies, appearto be centralto the voters' Therestill remainslittle consensuson how
choices. But perhaps it is prematureto best to measurecognitive ability in poli-
conclude that issues have little direct tics (but see Lushkin1987). We need not
influence in determiningthe outcome of select such a measure,as we are here con-
elections.After all, few candidatespublic- cernedwith the searchfor segmentsof the
ly assert that issues do not matter.A fre- electorate that eschew emotional re-
quent conclusionof voting studies is that sponsesto the candidatesand rely instead
the electoratedoes not engage in careful upon the policy positions and ideological
policy analysis and comparison of the perspectives of the candidates. Perhaps
party platformsofferedby the majorpar- the single measureon which most atten-
ties of candidates.This conclusion is fre- tion has centeredis education.Whileedu-
quentlyheld to demonstratethat the elec- cation may not measurecognitive ability
torate does not meet the standards re- in politicsdirectlyand may contributeless
quired by "classicaldemocratictheory." to its development than expected, most
The obvious references are to Schum- democratictheoriesexpectthat education
peter, Berelson,and the Michiganschool. preparespeople for citizenship(see, e.g.,
Perhaps, as is frequently asserted, the Thompson 1971). Further, educational
public at largedoes not performto expec- attainment has often been used as one
tations, but there is a competentpolitical measureof sophistication,as it measures
stratathat does meet thesestandards.?8 If, the ability and willingnessto acquirecog-
so the argumentgoes, we divide the elec- nitive skills as well as to learn citizenship
torate into levels of competency,we can (Neuman1986). If lack of sophisticationis
discerna more active and influentialseg- the reason for so few using policy con-
ment that does have the cognitive skill siderationsto decide which candidateto
and experience to utilize the necessary support, then we can expect to find that
sophisticatedskills required of a demo- stratifyingthe electorateon educationwill
cratic electorate.19 reveal a difference between those who
In the preliminaryanalyses that follow rely upon feelings about the candidate
three cautionarypoints should be kept in and those who rely upon policy con-
mind. First,the combinationof stratifica- siderations.
tion of the sampleand listwise deletionof Table 4 presentsthe same analysisas in
missingvalues reducesthe size of the sub- Table3, the regressionof vote disposition
samples, in some cases substantially. on candidate characteristicsand policy
Second, the measurement of cognitive evaluations, but with the data stratified
ability and its effects may be better into those with less than a high school
attempted in experimentalresearch de- diploma, those with at least a high school
signsthanin secondaryanalysisof a cross- diploma,and those with some collegeand
sectionalnational sample. And third, the postsecondaryeducation. Only unstand-
1984 NES data are collected quite late in ardized regression coefficients are pro-
the impressionformationprocessand as a vided (bs), as the relevant comparisons
result may be less helpful in illuminating are between educationstrata.
how impressionsare formedand evaluat- The only suggestion that education
ing what role cognitive ability and emo- divides the adult population into strata,
tional appraisal,respectively,play. each with its own mode of reasoning, is
Various measureshave been put forth that assessment of Reagan's leadership
752

Emotional Response
Table 4. Regression of Vote Disposition on Comparative Candidate and

Policy Evaluations: Stratified by Education
Low Medium High

Education Education Education
CandidateEvaluation b t b t b t
Leadershipcompetence
(Reaganjob approval) .12 1.8* .38 7.7 .28 8.5
Moralleadershipa .23 2.0* .03 4* .01 3*
Partyidentification(Republican) .22 5.8 .05 2.0* .08 4.6
Policy appraisals
Domesticissuesa .06 1.2* .04 1.5* .07 3.1
Foreignissuesa .25 1.2* .13 1.1* .16 1.6*
Ideologicala .02 .6* .00 .0* .03 1.6*
Emotionalresponse
Masterya .13 3.0 .14 5.9 .12 7.1
Threat -.06 -1.3* -.09 -3.3* -.08 -5.1
R2 .83 .80 .80
AdjustedR2 .82 .78 .79
F 60.5 145.2 264.9
df 8, 100 8, 310 8, 588
Source: 1984 NES.

competenceand moral leadershiphas lit- dent on foreignpolicy issues, or closer to

tle influence on vote disposition among the respondent on domestic issues does
the less educated;in this stratapartyiden- not seem to play a greaterrole in the vote
tification has a substantial association decision for the more educated than for
with vote disposition.This might suggest the less. Moreover, comparativefeelings
that the less educated, compared to the about the candidatesseem, if anything,
more educated, rely more on partisan more influentialamong the more than the
cues and less on candidatecharacteristics. less educated (and the ratio of the influ-
Studiesby Lau(1986)and Miller,Watten- ence of feelingsof masteryto the influence
burg, and Oskana (1986) find that the of feelings of threat remainsrather con-
more educatedhave more political infor- stant at two to one, suggestingthat across
mation than the less educated. the differenteducationalstrata,feelingsof
Only the domesticissue closenesscom- threathave about half the impactof feel-
parisonis significantlyrelatedto the vote ings of mastery on vote decisions).
disposition in the more educated strata Stratifying the sample by interest in
(which suggests that issues may have politics, anotherfactor commonlyheld to
some influence on voters actions, but lead to political sophistication(Neuman
most likely this influence is an indirect 1986), finds even less evidence of greater
effect through emotional response). But reliance on cognitive cues among the
apart from these differences,the percep- higher strata. All of the nine policy ap-
tion that one candidate is ideologically praisals, three within each stratum,have
closerto the voter, is closerto the respon- weak coefficientsand none are statistical-
753

AmericanPolitical Science Review Vol. 82
iy significant(Marcus1987). The coeffi- sistent reliance on emotional response in

cients for the two dimensions of emo- the sampleas a whole and within various
tional response are weaker in the least subsamples. While these findings must
interestedstratum(.10 and -.06 for mas- await confirmationfrom more appropri-
tery and threat)and strongerin the mod- ate researchdesigns before being treated
erate and high intereststrata and display as more than suggestive,the basic finding
the same two-to-one ratio found in the of pervasive and uniform reliance on
analyses above (.17 and .15 for mastery emotional response is consistent with a
and -.07 and -.09 for threat). These numberof studies in politics (Kinderand
resultssuggestthe not-surprisingrelation- Sears1981; Kuklinskiet al. 1987; Searset
ship that interest in politics and pas- al. 1980).
sionate engagement toward or against Notwithstanding these findings, it
candidatesis linked. would be incorrectto concludethat issues
One explanationfor why issuesplay so do not play an influentialrole in voters
little influence in candidate choice is choices between presidentialcandidates.
offered by Page (1976, 1978). Page's Table 1 presentsresults that suggest that
analysisof party platformsand candidate issues and policy considerationsdo influ-
positions concludesthat candidatesavoid ence voters by acting upon emotions and
controversialissues and attempt to pre- particularlyon the environment-regard-
sent ambiguouspositions that can appeal ing dimension of emotion, feelings of
to the broadestrangeof likely voters. Ob- threat. Emotionsare rarely, in the West-
taining definitive, or even moderately ern tradition,takenas rationalor of adap-
accurate,informationon the positions of tive benefit, althoughDavid Hume is one
candidateson even the most controversial notable exception(Lloyd1984). But emo-
issues of the day may be quite difficult tions may providepeoplewith an efficient
and thereforeprevent many voters from means of comprehendingtheir circum-
issue-voting even if they are so inclined stances. Reliance on emotional response
(Page and Brody 1972). Yet for those to characterizecandidates on leadership
issues that are important to the voter, ability, competence,and issuesand policy
issue comparisonsmay play an important positions provides the electorate with a
role. To evaluatethis possibility,the sam- speedyand efficientmeansof determining
ple was stratifiedinto thirdsby the degree candidate suitability on criteria that are
of importanceof threedomesticissues on consistentwith the requirementsof repub-
which the 1984 NES ascertained how lican forms of government.
important the issue is to respondents
(government spending, economic status
of women, and government guaranteed
living-guaranteedjob; V380, V407, and Conclusion
V420 respectively).20The results essen-
tially mirrorthose in Table4. None of the The circumplexmodel providesan eco-
bs for the nine policy appraisals,three in nomical approach to the study of emo-
each stratumof importance,has a signifi- tional response. More importantly, by
cant influenceon vote disposition. As in demonstratingthat emotionality consists
the previous tables, the influence of the of two concurrent appraisal processes,
two dimensionsof feeling is stable across one strategicallydedicatedto self-initiat-
the three strata (mastery has bs of .15, ing action and one strategicallydedicated
.13, and .15 and threat has bs of - .05, to evaluating threat, the results reported
-.08, and -.07 across the three strata). here resolve the anomalous findings of
The resultsare noteworthyfor the con- previous studies that positive and nega-
754

EmotionalResponse
tive emotions are often concurrently picting the candidate as uncertain and
reported. likely to take unpredictablepositions. It
The relationshipbetween the structure would seem that the most successfulcam-
of emotions and candidateappraisalcan paign strategywould try to build support
also be useful in gaining a better under- throughdepictinga candidateas compe-
standingof the likely success of different tent to producegood resultsin the job and
strategies of political campaigning.The of good probity and to use issues only
consistentfinding that positive emotional defensively (to sustain calm feelings
response to the candidatesis more influ- ratherthan build enthusiasticsupport).21
ential than negative emotional response Position taking, if misguided, can hurt.
suggeststhat electionsturnmore on moral But unless necessary to respond to chal-
leadershipand leadershipcompetenceand lenges raised by the opposition, position
less on issues. Issue appraisals do not taking is not likely greatly to help a can-
seem to have much, if any, influenceon didate build enthusiasm.
the emotionalenthusiasmcreatedby can- Finally, these results suggest, as has
didates. Yet issues are shown to have an been long noted, that public debate is
influenceon the threatdimensionof emo- often accompaniedby heat and passion.
tional response. This suggests that run- Madisonidentifiedan importantrelation-
ning a campaignbased on matching the ship between emotions, deliberation,and
issue positions of the voters has limited the attentiongiven to publicmeasures:"It
prospectsfor strengtheninga candidate's is a measure of misfortune, inseparable
image of leadership and competence. from human affairs,that public measures
Indeed,as GaryHart'sdeclinein the sum- are rarely investigatedwith that spirit of
mer of 1987 suggested, running a cam- moderation which is essential to a just
paign solely "on the issues"may createa estimateof theirreal tendencyto advance
disquietingsense that somethingis being or obstructthe public good; and that this
hidden or that there is somethinglacking spirit is more apt to be diminishedthan
in the characterof the candidate.This in prompted by those occasions which re-
turn is likely to provoke press inquiries quirean unusualexerciseof it" (Federalist
into the characterof the candidate. Papers, no. 37; italics added). Madison's
Insofaras issuepositionsof a candidate insight is that passion (or to use our ter-
are seen to be discrepantfrom the issue minology, emotional response) is what
positions of the voters, candidates risk draws individualsto politics. Madison is
arousing feelings of anxiety and threat properlyskepticalof the influenceof emo-
among voters. Thus candidatesmay best tions, but he recognizesthat emotionsare
use issues to forestall others from using an unavoidableand dynamicpart of poli-
issues against them, as in negative cam- tics.22Emotions do not just form empty
paigning, or from taking positions feelings,are not merelysubjective;rather,
throughmiscalculationthat place them in they constitutevital strategicunderstand-
opposition to their supporters. ings (Marcusand Rahn n.d.). While the
It is importantto add that threat feel- process by which these emotional re-
ings are elicited by the prospect of an sponses are arrived at is hidden (Tsal
unpredictableenvironmentas well as by 1985) and is likely to be fallible, emotion-
that of a hostileenvironment(Gray1981). ality nonethelessdoes serve instrumental
This would suggestwhy candidatesare so purposes. It is perhaps appropriate to
quick to paint their opponents as "flip- reexaminethe receivedwisdom that emo-
flopping on the issues." Successfulcam- tions constitute "a regrettableimperfec-
paign rhetoricof this genre can make the tion in an otherwise perfect cognitive
opposition candidate vulnerable by de- machine"(Scherer1982).
755

Table A-1. Regression of Vote Disposition on Comparative Candidate and

Policy Evaluations with Feeling Thermometers To Assess Affect
CandidateEvaluation b j3 t
Leadershipcompetence(Reaganjob approval) .27 .30 10.6
Moralleadership .01 .01 A*
Party identification(Republican) .09 .15 6.8
Policy appraisals
Domesticissues .05 .07 3.3
Foreignissues .02 .04 1.8*
Ideological -.01 -.02 -.9*
Emotionalresponse
Feelingthermometers .01 .42 12.0
R2 .79
Adjusted R2 .79
F 555.3
df 7, 1026
Source: 1984 NES.
*Not significant at p = .01.
Appendix: The Measurement be gained by using the feeling thermome-

of Emotional Response ters. However, this conclusionwould be a
substantial mistake. The feeling ther-
Thereis a furthertest availableto assess mometers mask two quite different ap-
the claim that the feeling thermometers praisalprocesses.
are confoundedmeasures.The regression The thermometerscale, as a measureof
presentedin Table 3 can be reestimated emotionalresponse,is a confoundedmea-
with the differencebetween Reagan and surementdevice. Substantively,responses
Mondale feeling thermometersreplacing to feeling thermometerswill be an indis-
the two scales of mastery and threat. If tinguishablemixture of the intensity of
relationshipsare differentwhen the feel- threatand the intensityof masteryevoked
ing thermometersare used, then it would by the objectof appraisal.Continuedreli-
suggest that the feeling thermometers ance on thermometer scales to assess
covary with these independentvariables global feelings regarding an object of
in ways not capturedby the two scales of appraisalwould be ill advised.
mastery and threat. Table A-1 presents The feelingsmeasuresdevelopedfor the
this regressionanalysis. National ElectionStudiesare an improve-
The results,exceptfor the measurement ment over the feelingthermometers.They
of emotional response, are almost iden- provide multiple measures of the two
tical to those reportedin Table 3, with no dimensions and these indicators enable
differencesgreaterthan .02 for standard- the two dimensions to be discriminated
ized or unstandardizedregressioncoeffi- from each other. However, they suffer
cients, R-squaredor adjusted R-squared three liabilities. First, one of the positive
values. This might suggest that the two- mood terms, sympathetic, is not a good
dimensionalmodel adds little to what can measureof mastery.It is less reliablethan
756

Emotional Response
is generally desired. And its location in director of studies in 1984. The data were collected
the circumplex,Figures1 and 2, and the under a grant from the National Science Foundation.
I thank Kimberly Jordan for help in conducting the
two LISRELmodels, suggest that sym- statistical analyses, and Andy Crider, George
pathy is a compound mood term and Goethals, Saul Kassin, Donald T. Campbell, W.
should be dropped as a measure of the Lance Bennett, Tim Cook, Roger Masters, Lyn Rags-
mastery dimension. Adding one or two dale, Wendy Rahn, and John L. Sullivan for their
additional mastery emotion indicators useful comments and suggestions.
1. In the data used here, the correlations be-
would seem advisable (see Watson and tween feeling thermometers for Mondale and
Tellegen1985, for numerousalternatives). Reagan and vote intention (V425) in the 1984 NES
Second, the feeling measuresare dichoto- are .50 and .53 for Mondale and Reagan respective-
mous, thus sacrificing the range of rely. Further, if we use the strength of voting prefer-
ence (V426) and voting intention to create a four-
sponses offeredby the feeling thermome- point scale, the correlation between Reagan's feeling
ters. Third, current emotional response thermometer score and the vote intention scale in-
measuresdo not identify the nonaffective creases to .78, and the correlation between Mon-
pole of each dimension. An alternative dale's feeling thermometer and the vote intention
scale increases to .71 (the correlation between the
procedurewould be to use semantic dif- vote intention scale and V425 is .94; the increases in
ferential formats with the appropriate correlation between feeling thermometers and vote
mood term anchoring the low-arousal intention scale result from using a scale with more
pole scale and the appropriatemood term categories [see Bollen and Barb 19811).
anchoring the high-arousal pole of the 2. It is necessary to differentiate between con-
scious and nonconscious cognitive processes, each of
scale. For example, mastery mood terms which can influence behavior, as well as between the
and their low-arousal polar opposites interdependent influences of emotional and cogni-
such as the following could be used: indif- tive processes (Lewicki 1986; Millar and Tesser 1986;
ferent, curious,enthusiastic,unenthusias- Tassinary et al. 1984).
tic; interested,withdrawn;and depressed, 3. For Mondale, the first factor has an eigen-
value of 2.62 and the second factor has an eigen-
happy. For threat mood terms and their value of 1.54; together the two factors account for
low-arousalpolar oppositesthe following 60% of the variance. For Reagan, the first factor has
could be used: tense, calm; upset, com- an eigenvalue of 2.74 and the second factor has an
fortable; worried, contented; and ner- eigenvalue of 1.37; together the two factors account
for 59 % of the variance. In each case the third eigen-
vous, relaxed (Marcus 1985). Using this value is well below 1.00 (.77 and .78 respectively).
measurement approach each measure 4. Fonberg (1986) reviews the neurophysio-
would provide a wider range of scores logical basis for the two-system paradigm.
than the currentmeasures.Further,using 5. While the mastery dimension is hypothesized
this proposed measurement approach to be self-orientated, the capacity for identification
with others (e.g., empathy and group identification)
would provide for multiplemeasuresand extend appraisals of mastery to collective action and
thus ensure ample opportunity for item to the actions of others.
analysis (reliability and validity assess- 6. The full Gray-Tellegen model accounts for
ment) and enable scale buildingand mul- mood and temperament (stable emotional differ-
ences). Temperament can be defined as reactivity
tiple indicatormethods of analysis. levels. Temperament is the disposition, across situa-
tional states, to interpret any environment as, for
example, threatening. Because the NES data, used
Notes here, does not contain any acceptable measures of
temperament, this aspect of the Gray-Tellegen
The data utilized in this analysis were made model can not be pursued.
available by the Inter-UniversityConsortiumfor 7. The correlation matrices are positive definite.
Politicaland Social Research.The data for the 1984 Reestimating with maximum likelihood gives essen-
AmericanNational ElectionStudy were originally tially the same solutions. Also, defining the model
collectedby the Universityof Michigan,Centerfor with three latent concepts, threat, mastery, and feel-
PoliticalStudiesof the Institutefor Social Research ing thermometer-the latter defined by one indica-
for the National ElectionStudiesunder the overall tor-provides the same solution, differing only in
directionof WarrenE. Miller. Santa Traugottwas that rather than epistemic correlations running from
757

mastery and threat to the feeling thermometer cients increase marginally by about the same small
indicator, the solution has structuralparameters amounts. The R-squared drops by an average of .05.
runningfrom masteryand threatto the latent con- More importantly, policy appraisals remain uni-
cept, feelingthermometer.The values are the same. formly small and emotional responses become more
8. It may also be the case that there is a third influential. Alternative tables, with leadership com-
dimension of emotion, empathy, which has been petence removed, are available from the author.
shown to have a cognitiveand affectivedimensions 16. The proportion may well change from candi-
(Davis 1983;Chlopanet al. 1985). The collectionof date to candidate and from campaign to campaign,
feeling measuresin the NES series does not enable depending on the particular strengths and liabilities
this possibilityto be pursued.In any case, this and of specific candidates. More studies will be needed to
other studies (Mastersand Sullivan 1986) suggest ascertain how labile these appraisals are and pre-
that the two dimensions of circumplex model cisely what kinds of information have most conse-
account for the predominantvariance in feelings. quence for emotive appraisal.
Seen. 3. 17. Consistent with the findings reported in Rahn
9. See n. 4. and colleagues n.d., the comparative model does
10. Sullivanand Masters(1988)have shown us- better than models that predict vote choice based on
ing experimentalproceduresthat emotionaldisplays ideological and issues closeness and level of threat
by presidentialcandidatesevoke two dimensionsof and mastery reported for each candidate (i.e., a non-
emotional appraisal in observers, hedonic and comparative model). For the model predicting vote
agonic, which are equivalentto the two dimensions for Reagan, using distance measures between
conceptualizedhere as masteryand threat. Reagan and the respondent, the adjusted R-squared
11. Scale scoresreportedfor these and all subse- is .75. For Mondale, the adjusted R-squared is .76
quentscalesare the averagevalues of the sum of all when using Mondale's characteristics and closeness
valid responseswhen the numberof valid responses to the respondent (including Reagan's job perform-
is equal to, or one less than, the maximumnumber ance in the Mondale model).
of itemsin the scale. Forexample,a scale with four 18. This tradition has flourished in areas other
itemswill returna scalescoreif respondentsprovide than voting, for example in the analysis of who com-
valid responsesto any three,or all four, scaleitems. prehends and subscribes to the rules of the game and
This preserves more cases for subsequentmulti- political tolerance (Mueller n.d.).
variateanalysis(which, using listwisedeletion,can 19. The studies that find such strata have been
sharply reduce the number of cases available for countered by studies that dispute the methodology
analysis). of the studies and the findings themselves (Sullivan,
12. With samples as large as those in the NES Piereson, and Marcus 1982). But on balance the
series, statisticalsignificanceis readilyachievedfor more accepted position seems to be that put forth by
all but the smallestrelationships.Substantiveassess- Lippmann, Schumpeter, and others in that tradition
mentsmust be made, ratherthanhavinginterpreta- (for a review, see Krouse and Marcus 1984).
tion rest on the issue of inferentialrisk. 20. In order to identify the strata, those who iden-
13. Since I am interested, in this analysis, in tified an issue as extremely important were assigned
comparingthe relative influenceof differentinde- a score of three on that issue; those who identified
pendentvariablesand as the scales are "scale-free" the issue as very important were assigned a score of
(thatis, the rangeof the scalescoresis a resultof the two on that issue; those who identified the issue as
availablenumberof itemsmakingup eachscale),the somewhat important were assigned a score of one on
relevantstatisticis the j3,the standardizedregression that issue; and all others were assigned a score of
coefficient. zero. The scores were aggregated to form an index.
14. See n. 1. All those with a total of three or less are the low-
15. It has beensuggestedthatbecausecompetence importance strata; those with a total of four or five
as a measure of leadershipis available only for are the medium-importance strata; and those with a
Reagan,this measureshould be removedfrom the total greater than five are the high importance
comparativemodel, in this and in other analysesto strata.
follow. Of course, as the incumbent,only Reagan 21. Selective examples show what can transpire
can be evaluatedby the voters on presidentiallead- when political campaigns fail to sustain the candi-
ership performance.For Mondale, voters may ex- date image as strong and competent; for example,
trapolate from Mondale's campaign activity as Gary Hart, Edmund Muskie, and George Romney
reported and seen, current policy positions, and quickly come to mind as candidates who failed in
from his past record as a political leader. Many precisely this fashion. Personal stability and strength
voting theoriesdo assumecomparisonbetween an would appear to be minimal requisites for com-
incumbent'srecord of performanceand the chal- petence.
lenger'spositions.Theoreticalissuesaside,removing 22. The principal focus of Madison's skeptical
Reagan'sjob approval from the analysis does not attention, in Federalist Papers, no. 37, is on the
changethe resultsin any principalway. All coeffi- fallibility and partiality of reason; a common con-
758

Emotional Response
cern in contemporarysocial science (Greenwald Addison-Wesley.

1980;Marcusand Rahnn.d.; Markus1986;Nisbett Fiske, Susan, and M. Pavelchak. 1985. Category-
and Ross 1980). based versus Piecemeal-basedAffective Re-
sponses: Developments in Schema-triggered
Affect. In The Handbook of Motivation and
References Cognition:Foundationsof Social Behavior,ed.
RichardSorrentinoand E. Higgins. New York:
Abelson, Robert,Donald Kinder,MarkPeters,and GuilfordPress.
SusanFiske.1982. Affectiveand SemanticCom- Fonberg, Elzbieta. 1986. Amygdala, Emotions,
ponents in Political Person Perception.Journal Motivation,and DepressiveStates. In Emotion:
of PersonalityandSocialPsychology 42:619-30. Theory, Research,and Experience,ed. Robert
Bay, Christian.1965. The Structureof Freedom. Plutchik and Henry Kellerman.London:Aca-
New York:Athenaeum. demic.
Bollen, Kenneth,and KenneyBarb.1981. Pearson's Fromm, Erich. 1941. Escape from Freedom. San
R and CoarselyCategorizedMeasures.American Francisco:Jossey-Bass.
SociologicalReview 46:232-39. Gray, Jeffrey A. 1970. The Psychophysiological
Bower, Gordon.1981. Mood and Memory.Ameri- Basis of Introversion-Extroversion. Behaviour
can Psychologist36:129-48. Researchand Therapy8:249-66.
Breckler,Steven J. 1984. EmpiricalValidation of Gray, Jeffrey A. 1973. Causal Theories of Per-
Affect, Behavior, and Cognition As Distinct sonalityandHow To TestThem.InMultivariate
Componentsof Attitude.Journalof Personality Analysis and PsychologicalTheory, ed. Joseph
and Social Psychology 47:1191-1205. R. Royce. New York:Academic.
Brody, Richard,and BenjaminPage. 1972. Com- Gray, JeffreyA. 1981. The Psychophysiologyof
ment: The Assessmentof Policy Voting. Ameri- Anxiety. In Dimensionsof Personality:Papersin
can PoliticalScienceReview 66:450-58. Honourof H. J. Eysenck,ed. RichardLynn.New
Brody, Richard,and BenjaminPage. 1973: Indif- York:Pergamon.
ference,Alienation,and RationalDecisions:The Greenwald,Anthony. 1980. The TotalitarianEgo:
Effectsof CandidateEvaluationson Turnoutand Fabricationand Revision of Personal History.
the Vote. Public Choice15:1-17. AmericanPsychologist35:603-18.
Chlopan, Bruce,M. McCain,J. Carbonell,and R. Hastie, Reid. 1986. A Primer of Information-
Hagen. 1985. Empathy:A Review of Available processingTheory for the PoliticalScientist.In
Measures.Journalof Personalityand SocialPsy- Political Cognition,ed. RichardLau and David
chology 48:635-53. Sears. Hillsdale,NJ: LawrenceErlbaum.
Conover, Pamela J., and Stanley Feldman.1986. Izard, Carroll E. 1971. Patterns of Emotions:A
EmotionalReactionsto the Economy:rm Mad New Analysis of Anxiety and Depression.New
as Hell and rmNot GoingTo TakeIt Any More. York:Academic.
AmericanJournalof PoliticalScience30:50-78. Izard, Carroll E. 1977. Human Emotions. New
Craig,StephenC., andMichaelA. Maggiotto.1982. York:Plenum.
MeasuringPoliticalEfficacy.PoliticalMethodol- Jbreskog,Karl G. 1969. A General Approach to
ogy 8:85-110. Confirmatory Maximum Likelihood Factor
Davies, Alan F. 1980. Skills, Outlooks, and Pas- Analysis. Psychometrika34:183-202.
sions: A Psychoanalytic Contributionto the J6reskog, Karl G. 1970. A General Method for
Study of Politics. Cambridge:CambridgeUni- Analysis of CovarianceStructures.Biometrika
versity Press. 57:239-51.
Davis, Mark H. 1983. MeasuringIndividualDif- Jareskog, Karl G. 1973. A General Method for
ferences in Empathy:Evidencefrom a Multi- Estimatinga LinearStructuralEquationSystem.
dimensionalApproach. Journal of Personality In StructuralEquationModels in the Social Sci-
and Social Psychology44:133-26. ences, ed. ArthurS. Goldbergerand 0. D. Dun-
Derryberry,Douglas, and Mary Rothbart. 1984. can. New York:Seminar.
Emotion,Attention,and Temperament.In Emo- Kelley, Stanley, and Thad Mirer.1974. The Simple
tion, Cognition,andBehavior,ed. CarrollIzard, Act of Voting. AmericanPolitical Science Re-
JeromeKagan, and RobertZajonc. New York: view 68:572-91.
CambridgeUniversityPress. Kinder, Donald. 1986. PresidentialCharacterRe-
Diener, Ed, and Robert A. Emmons. 1984. The visited. In Political Cognition, ed. Richard
Independenceof Positive and Negative Affect. Lau and David Sears. Hillsdale, NJ: Lawrence
Journalof Personalityand SocialPsychology47: Erlbaum.
1105-17. Kinder,Donald, and David Sears. 1981. Prejudice
Fishbein, Martin, and Icek Ajzen. 1975. Belief, and Politics: Symbolic Racism versus Racial
Attitude,Intention,and Behavior:An Introduc- Threatto the Good Life.Journalof Personality
tion to Theory and Research. Reading, MA: and Social Psychology40:414-31.
759

Krouse, Richard, and George E. Marcus. 1984. dates. American Political Science Review 80:
Electoral Studies and Democratic Theory Recon- 521-40.
sidered. Political Behavior 6:23-39. Mueller, John. N.d. Trends in Political Tolerance.
Kuklinski, James, E. Ottati, N. Schwarz, and Public Opinion Quarterly. Forthcoming.
Robert Wyler. 1987. Why the Slippage? Explain- Neuman, Russell. 1986. The Paradox of Mass Poli-
ing Why People Support Abstract Principles tics: Knowledge and Opinion in the American
More Than the Application of Them. Presented Electorate. Cambridge: Harvard University
at the annual meeting of the Midwest Political Press.
Science Association, Chicago. Nisbett, Richard, and Lee Ross. 1980. Human
Lane, Robert E. 1978. Autonomy, Felicity, Futility: Inference: Strategies and Shortcomings of Social
The Effects of the Market Economy on Political Judgment. Englewood Cliffs, NJ: Prentice-Hall.
Personality. Journal of Politics 40:2-24. Page, Benjamin. 1976. The Theory of Political
Lau, Richard. 1986. Political Schemata, Candidate Ambiguity. American Political Science Review
Evaluations, and Voting Behavior. In Political 70:742-52.
Cognition, ed. author and David Sears. Hills- Page, Benjamin. 1978. Choices and Echoes in Presi-
dale, NJ: Lawrence Erlbaum. dential Elections. Chicago: Chicago University
Lazarus, Richard. 1982. Thoughts on the Relations Press.
of Emotion and Cognition. American Psycholo- Page, Benjamin, and Richard Brody. 1972. Policy
gist 37:1019-24. Voting and the Electoral Process; The Vietnam
Lazarus, Richard. 1984. On the Primacy of Cogni- War Issue. American Political Science Review
tion. American Psychologist 39:124-29. 66:979-95.
Lewicki, Pawel. 1986. Nonconscious Social Infor- Phares, E. Jerry. 1978. Locus of Control. In Dimen-
mation Processing. Orlando, FL: Academic. sions of Personality, ed. Harvey London and
Lippmann, Walter. 1922. Public Opinion. New John E. Exner, Jr. New York: John Wiley.
York: Macmillan. Pitkin, Hannah. 1967. The Concept of Representa-
Lloyd, Genevieve. 1984. The Man of Reason: tion. Berkeley: University of California Press.
"Male" and "Female" in Western Philosophy. Plutchik, Robert. 1980. Emotion: A Psychoevolu-
Minneapolis: University of Minnesota Press. tionary Synthesis. New York: Harper & Row.
Lushkin, Robert. 1987. Measuring Political Sophis- Pomper, Gerald, Ross Baker, Charles Jacob, Scott
tication. American Journal of Political Science Keeter, Wilson Carey McWilliams, and Henry
31:856-99. Plotkin. 1985. The Election of 1984. Chatham,
Marcus, George E. 1985. A Theory and Methodol- NJ: Chatham.
ogy for Measuring Emotions in Politics. Pre- Rahn, Wendy, John Aldrich, Eugene Borgida, and
sented at the annual meeting of the Political John Sullivan. N.d. A Social-Cognitive Model of
Methodology Society, Berkeley. Candidate Appraisal. In Information and Demo-
Marcus, George E. 1987. The Structure of Emo- cratic Processes, ed. John Ferejohn and James
tional Appraisal: 1984 Presidential Candidates. Kuklinski. Urbana-Champaign: University of
Presented at the annual meeting of the Inter- Illinois Press. Forthcoming.
national Society for Political Psychology, San Roseman, Ira. 1979. Cognitive Aspects of Emotion
Francisco. and Emotional Behavior. Presented at the annual
Marcus, George E., and Wendy Rahn. N.d. Emo- meeting of the American Psychological Associa-
tions and Democratic Politics. In Research in tion, Washington.
Micropolitics, ed. Samuel Long. Greenwich, CT: Roseman, Ira. 1984. Cognitive Determinants of
JAI. Emotions: A Structural Theory. In Emotions,
Markus, Gregory. 1986. Stability and Change in Relationships, and Health, Review of Personality
Political Attitudes: Observed, Recalled, and "Ex- and Social Psychology, no. 5, ed. P. Shaver.
plained." Political Behavior 8:21-45. Beverly Hills: Sage.
Masters, Roger, and Dennis Sullivan. 1986. Non- Roseman, Ira, Robert Abelson, and Michael Ewing.
verbal Displays and Political Leadership in 1986. Emotions and Political Cognition: Emo-
France and the United States. Presented at the tional Appeals in Political Communication. In
1986 annual meeting of the American Political Political Cognition, ed. Richard Lau and David
Science Association, Washington. Sears. Hillsdale, NJ: Lawrence Erlbaum.
Mill, John Stuart. 1956. On Liberty. Indianapolis, Rotter, J. B. 1966. Generalized Expectancies for
IN: Bobbs-Merrill. Internal versus External Control of Reinforce-
Millar, Murray, and Abraham Tesser. 1986. Effects ment. Psychological Monographs 80:1.
of Affective and Cognitive Focus on the Russell, James A. 1980. A Circumplex Model of
Attitude-Behavior Relation. Journal of Personal- Affect. Journal-of Personality and Social Psy-
ity and Social Psychology 51:270-76. chology 39:1161-78.
Miller, Arthur, Martin Wattenburg, and Malanchuk Sears, David O., Richard R. Lau, Tom R. Tyler,
Oskana. 1986. Schematic Assessments of Candi- and Harris M. Allen, Jr. 1980. Self-Interest vs.
760

EmotionalResponse
Symbolic Politicsin Policy Attitudesand Presi- sonality and Their Relevance to Assessing
dential Voting. AmericanPolitical Science Re- Anxiety, with an Emphasis on Self-Report. In
view 74:670-84. Anxiety and the Anxiety Disorders, ed. A.
Scherer,Klaus. 1982. The Nature and Functionof Hussain Tuma and J. D. Maser. Hillsdale, NJ:
Emotion.Social ScienceInformation21:507-9. Erlbaum.
Seligman,Martin. 1975. Helplessness:On Depres- Thompson, Dennis. 1971. The Democratic Citizen:
sion, Development,and Death. San Francisco: Social Science and Democratic Theory in the
Freeman. Twentieth Century. New York: Cambridge Uni-
Shingles, Richard. 1986. Dimensions of Political versity Press.
Efficacy:Tests of New and Old Measures.Pre- Tomkins, Silvan S. 1981. The Quest for Primary
sentedat the AmericanPoliticalScienceAssocia- Motives: Biography and Autobiography of an
tion annualmeeting,Washington. Idea. Journal of Personality and Social Psychol-
Simon, Herbert.1967. Motivationaland Emotional ogy 41:306-29.
Controlsof Cognition.PsychologicalReview74: Tsal, Yehoshua. 1985. On the Relationship between
29-39. Cognitive and Affective Processes: A Critique of
Snyder,Mark,and WilliamIckes.1985. Personality Zajonc and Markus. Journal of Consumer Re-
and SocialBehavior.In TheHandbookof Social search 12:358-62.
Psychology, 3d ed., vol. 2, Special Fields and Warr, Peter, Joanna Barter, and Garry Brown-
Applications, ed. Gardner Lindzey and Elliot bridge. 1983. On the Independence of Positive
Aronson. New York:RandomHouse. and Negative Affect. Journal of Personality and
Storm, Christine,and Tom Storm. 1987. A Taxo- Social Psychology 44:644-51.
nomic Study of the Vocabularyof Emotions. Watson, David, and Auke Tellegen. 1985. Toward a
Journalof Personalityand SocialPsychology53: Consensual Structure of Mood. Psychological
805-16. Bulletin 98:219-35.
Sullivan,Dennis, and RogerMasters.1988. Happy Wiener, B., D. Russell, and D. Lerman. 1978. Affec-
Warriors: Leaders' Facial Displays, Viewers tive Consequences of Causal Ascriptions. In New
Emotions,and PoliticalSupport.AmericanJour- Directions in Attribution Research, 3. vols., ed.
nal of PoliticalScience32:345-68. John H. Harvey and R. F. Kidd. Hillsdale, NJ:
Sullivan, John L., JamesPiereson, and George E. Erlbaum.
Marcus.1982. PoliticalToleranceand American Zajonc, R. B. 1980. Feeling and Thinking: Prefer-
Democracy. Chicago: University of Chicago ences Need No Inferences. American Psycholo-
Press. gist 35:151-75.
Tassinary, Louis, Scott Orr, George Wolford, Zajonc, R. B. 1982. On the Primacy of Affect.
Shirley Napps, and John Lanzetta.1984. The American Psychologist 39:117-23.
Role of Awareness in Affective Information Zevon, M. A., and Auke Tellegen. 1982. The Struc-
Processing: An Exploration of the Zajonc ture of Mood Change: An Ideographic/Nomo-
Hypothesis.Bulletinof the PsychonomicSociety thetic Analysis. Journal of Personality and Social
22:489-92. Psychology 43:111-22.
Tellegen,Auke. 1985. Structuresof Mood and Per-
George E. Marcusis Professorof Political Science,WilliamsCollege, Williamstown,

MA 01267.
761


Sessão3 Marcus1988

Uploaded by

Copyright:

Available Formats

Sessão3 Marcus1988

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Sessão3 Marcus1988

Uploaded by

Copyright:

Available Formats

The Structure of Emotional Response: 1984 Presidential Candidates

Author(s): George E. Marcus

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Over the past two decadespsychologicalmodels of affect

It has been 15 that they add explanatorypower to our

AMERICAN POLITICAL SCIENCE REVIEW

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

studiesas to the numberof dimensionsof pretations and that variations in these

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Figure1. EmotionalResponse to son and Tellegen 1985; see also Masters

Figure2. EmotionalResponse to PreliminaryEmpiricalTest

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Figure 3. Gray-Tellegen Two-Factor

Signalsof Nonpunishment Activation

The BIS System-Appraisal and Evaluationof IntrusiveInformation

afraid of him, proud, disgusted, sympa- sional. These resultscontradictthe expec-

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

(i.e., affect dimensionswould be bipolar scene-reflect the kinds of stimuli that

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

behavior. The locus of control literature become so suddenlyor unexpectedly,the

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Figure 4. Two-Factor Model of Emotional Response: Mondale

angry afri uneasy disgusted feeling hope proud syptetic

.10 .28 -.08

stance to 7 by the number of measures for ULS estimation; however, goodness-

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Figure 5. Two-Factor Model of Emotional Response: Reagan

((Environ ent r (Task-

angry afraid disgusted uneasy feeling hope prou sympathetic

Source: 1984 NES.

remainingvarianceis assumedto be error the two latentfactors,threatand mastery,

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

by using the 16 assessmentsfor Reagan It would seem plausible to argue that

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Table 1. Regressionof Feelings of Threaton CandidateEvaluations

Table 2. Regressionof Feelings of Masteryon CandidateEvaluations

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

characterization of President Reagan evaluationsmay be subjectto persuasion

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Table 3. Regression of Vote Disposition on Comparative Candidate and

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

It is strikingthat issue appraisalshave to define and measureelectoratecompe-

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Table 4. Regression of Vote Disposition on Comparative Candidate and

Low Medium High

Source: 1984 NES.

competenceand moral leadershiphas lit- dent on foreignpolicy issues, or closer to

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

iy significant(Marcus1987). The coeffi- sistent reliance on emotional response in

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

Table A-1. Regression of Vote Disposition on Comparative Candidate and

Appendix: The Measurement be gained by using the feeling thermome-

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM

This content downloaded from 128.235.251.160 on Mon, 26 Jan 2015 19:39:44 PM