Professional Documents
Culture Documents
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=aapor.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
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 support@jstor.org.
American Association for Public Opinion Research and Oxford University Press are collaborating with JSTOR
to digitize, preserve and extend access to The Public Opinion Quarterly.
http://www.jstor.org
Public Opinion du Jour:
An Examination of the Spiral of Silence
CARROLL J. GLYNN AND JACK M. McLEOD
Abstract This study investigates the spiral of silence theory using perceived vote outcomes
as well as actual vote behaviors. The first major hypothesis-that those who see their
position as gaining support will be more likely to discuss while those who see their position
as losing support will be less likely to discuss-was supported by the data. Some support was
also found for the hypothesis that individuals perceiving support for a certain candidate
would be more likely to express a preference for the candidate. Results also seemed to
justify the Noelle-Neumann distinction between the resistant hardcore and other types of
citizens. The relative independence of perception of opinion judgments from own prefer-
ences also was investigated and support was found for this difference.
Carroll J. Glynn is an Assistant Professor in the Department of Communication Arts,
Cornell University. Jack M. McLeod is Baier-Bascom Professor in the School of Journalism
and Mass Communication, University of Wisconsin.
Study Design
Telephone interviews were conducted with a panel of 98 eligible voters
in Dane County, Wisconsin at three time points-two prior to and one
after the 1980 presidential election. The first interview was taken before
the initial September debate, the second in late October, and the third in
the week after the November election. The second and third waves also
included a replenishment sample of 367 additional respondents not pre-
viously interviewed. Random digit dialing procedures were used with
proportionate probability sampling stratified according to the number of
private lines at each exchange. Respondents were screened at each
selected number to ensure their eligibility to vote.
Two types of perceptions of public opinion were used to form the
critical independent variables for the study. In both the September and
October interview waves, respondents were asked to estimate what per-
centage of the vote each of the three major candidates would receive in
November (Estimated Percent). They were also asked whether each
candidate was gaining support, staying the same, or losing support (Sup-
port Change) in both the September and October waves. The data thus
included separate estimated percentage and support change measures for
each of the three candidates at two time points.
Separate measures of support change were combined into a single
index of Support Discrepancy. Each respondent's presidential choice
(Carter, Reagan, or Anderson) was compared to that person's judgment
of October support change for the preferred candidate. High discrepancy
(3) was assigned cases where the preferred candidate was seen as losing
support, moderate discrepancy (2) was given where that candidate was
judged as staying the same, and low discrepancy (1) was attributed where
the chosen candidate was perceived as gaining support.
Noelle-Neumann's conceptual distinction between the hardcore opin-
ion holders and the rest of society was also investigated despite a lack
of operationalization in her previous research. The variable was indexed
here by using the degree of vote consistency over time. For the larger
panel (N = 465), those who said in October that they were certain to
vote for a given candidate and who did in fact vote for him in November
were called Hardcore, while those less certain or who changed were
combined into a Combined Leaner and Nonhardcore category. A slightly
734 CARROLL J. GLYNN AND JACK M. MCLEOD
more sensitive index was used with the small panel (N = 98), for whom
we had September vote preference measures. Those who unequivocally
preferred a given candidate in both September and October and who
reported voting for that same candidate in November were classified as
Hardcore. Those who were leaning toward a candidate in September
and/or October and who voted for that candidate in November were
called Leaners, and those showing greater change, switching candidates
between rounds, for example, were labeled Nonhardcore.
At first glance there would appear to be a problem with conceptualiz-
ing Hardcore in this fashion; aspects of voting behavior enter into both
the contingent Hardcore variable and the dependent Vote Preference
variable. However, the confounding' is not real. The measurement of
Nonhardcore and Leaners reflects the degree of change or lability of vote
preference over time regardless of direction, whereas vote preference
and actual vote reflect direction and intensity of the intention at a given
point in time.
In order to test Noelle-Neumann's first major hypothesis of the spiral
of silence, a special set of questions was devised based on hypothetical
social situations comprised of other people, potentially congruent or
noncongruent with the respondent's vote preference, discussing these
preferences. We asked:
If you were at a social gatheringwith people you know, and they were
discussingthe presidentialcampaign,how likelywouldyou be to enterinto
the conversationif they were Reagansupporters?Wouldit be very likely,
somewhatlikely, not very likely, or not at all likelythatyou wouldjoin the
conversation?
The question was repeated for the situation where the social gather-
ing consisted of Carter supporters. For respondents supporting either
Reagan or Carter in October, the Reagan and Carter social gathering
questions were identified as either congruent or noncongruent according
to the vote preference (or nonpreference) of the respondent. For sup-
porters of the two leading candidates, then, the dependent variable could
be assessed by comparing the likelihood of entering the discussion in the
two types of social gatherings. Time limitations prevented the use of this
question for a gathering of Anderson supporters. Anderson backers were
thus excluded from this analysis because a congruent situation could not
be analyzed for them. Sample size limitations led us to use the larger
panel only for this analysis.
The major dependent variable, Vote Preference, was measured in
September and October by creating an index for each of the three major
candidates using a five-point scale: sure of voting for the given candidate
(5), leaning toward the candidate (4), undecided (3), leaning toward
PUBLIC OPINION AND THE SPIRAL OF SILENCE 735
another candidate (2), and sure of voting for another candidate (1). In the
postelection survey, voting was measured by reports of having either
voted for or not voted for a given candidate.
Hierarchical least-squares regression was the basic method used in the
analysis of vote preferences. The control variable, party affiliation, was
introduced as the first predictor. The main effects of the two perception
of opinion variables (Estimated Percent and Support Change for the
given candidate) were then entered as a second predictor block. The third
block was Hardcoreness as a main effect. Finally, a fourth block com-
prised of the two interactions of Hardcoreness with the perception of
opinion variables was introduced. Significance tests (Fvalues) were used
for individual Betas and for blocks of predictor variables.
Two relatively stable political variables, Party Affiliation and Political
Interest, served primarily as control variables. Party affiliation was in-
dexed as a seven-point scale as follows: strong affiliation with the Re-
publican party, 7; Moderate (not strong) Republican, 6; Independent
Republican, 5; Independents without a leaning, 4; Independent Demo-
crats 3, and so forth. Political interest was measured by asking: "Gener-
ally, would you say you are very interested in politics, interested in poli-
tics, somewhat interested in politics, or not at all interested in politics?"
In order to examine various communication sources for their potential
influence on perceptions of public opinion, respondents were asked:
"We're interested in which source you use most for political information.
Would you say you use newspapers, friends or family, television, maga-
zines or some other source most for current events and political informa-
tion?" After choosing which they used most, the respondents then ranked
the other sources. The rankings provided data for three indices: Tele-
vision Reliance, Newspaper Reliance, and Interpersonal Reliance.
Results
PUBLIC EXPRESSION OF OPINION
The first hypothesis was tested by examining the effects of support
discrepancy on the likelihood of entering into a political discussion with
people with similar (congruent) or dissimilar (noncongruent) viewpoints
in hypothetical social situations (Table 1). As noted previously, support
discrepancy was measured on a three-point scale.1
1
Noelle-Neumann traditionally has used a hypothetical situation where respondents
assume they are sitting next to a stranger in a train. However, because trains are not a
primary mode of transportation or a common method for meeting individuals in the United
States, the authors felt it was preferable to use a hypothetical social gathering for the present
study. In this fashion, respondents could be placed in a situation where they are faced with
individuals with similar opinions and in situations with individuals with opposite opinions.
736 CARROLL J. GLYNN AND JACK M. MCLEOD
Social Gathering
Table 2. September and October Vote Preferences (Perceived Outcome) and November
Vote Regressed on Party Affi'liation, Estimates of Candidates' Vote Percentages, Support
Change, Hardcoreness, and Hardcoreness-Percentage/Support Interactions (N = 78)
Discussion
It is obvious from the results of this study that the nature and develop-
ment of public opinion is a complex cognitive and societal process. The
first major hypothesis-that those who see their position as gaining
support will be more likely to discuss while those who see their position as
losing support will be less likely to discuss-was supported by the data.
Some support was also found for the second major hypotheses-that
individuals perceiving support for a certain candidate would be more
likely to express a preference for that candidate. In the present study
these perceptual variables were found to influence not only vote prefer-
ences for Carter in September and October, but also reports of actual
vote for him in November. They also predicted to vote preferences for
Reagan in September and for Anderson in October. The interaction of
Hardcoreness with the two perception of public opinion variables signifi-
cantly affected vote preferences for Carter but not for Reagan or Ander-
son. While this is somewhat surprising (we would have expected percep-
tual variables to influence Anderson votes), perhaps for the relatively
well-informed liberal voters in this sample it was a matter of estimating
whether Carter had a chance because they knew Anderson did not.
Although such substantial connections between perceptions of public
opinion and personal vote preferences have been uncovered, it is impor-
tant to be wary in asserting the direction of causation. Vote preference
may lead to holding distorted estimates of public opinion, although panel
analyses in previous research make a case for treating vote preference as a
dependent variable (Glynn and McLeod, 1982). The lack of relationship
between party affiliation and perceptions of opinions also argues for this
casual direction.
The results seem to justify the Noelle-Neumann distinction between
the resistant Hardcore and other types of citizens. Interactions of Hard-
coreness with other variables were frequent enough to argue for this
distinction. In addition, other analyses not shown here have found the
Hardcore to be different from other respondents in their higher level of
campaign discussion, in their reports of lesser influence of this discussion
on their voting decisions, and in their attribution of more importance to
PUBLIC OPINION AND THE SPIRAL OF SILENCE 739
candidates' issue stands and personal qualities. They also were less likely
to see the candidates' chances of winning as a factor in their choice,
providing some validation to our findings reported here.
Most of the variance in voting preferences was not accounted for by the
perception of public opinion variables. That is, the perceptions of public
opinion seem to operate with relative independence from the individual's
own position. To illustrate this further we have compared the aggregation
of individual opinion (Own Preference) to five other aggregated expres-
sions of opinion: individual perceptions of current opinion (Perceived
Opinion); perceptions of vote outcome (Perceived Outcome); current
national Gallup poll data (Poll Data); panel respondents' November vote
(Own Vote); and national vote in November (Actual Vote) (Table 4).
Table 4 does provide support for the separation of perceptions of opinion
from aggregated individual opinion (usually known as "public opinion").
In September, for example, while the aggregate of individual responses
gave Carter a clear lead over Reagan (41 percent vs. 32 percent) with
Anderson a close third (27 percent), respondents perceived the current
opinion (Perceived Opinion) as being close between Carter (42 percent)
and Reagan (40 percent), and placed Anderson a more distant third (18
percent). All three estimates were within 3 percent of the most recent
September Gallup poll prior to the first interview wave, considerably
closer than to the aggregate of the respondents' own preferences.
The relative independence of perception of opinion judgments from
own preferences also is seen in the results for the respondents' percep-
References
Allport, Floyd
1937 "Toward a science of public opinion." Public Opinion Quarterly 1:7-23.
Almond, Gabriel
1950 The American People and Foreign Policy. New York: Harcourt, Brace.
Glynn, Carroll, and Jack McLeod
1982 "Public opinion, communication processes and voting decisions." Pp. 759-74 in
Michael Burgoon (ed.), Communication Yearbook 6. International Communica-
tion Association. Beverly Hills: Sage.
In press "Implications of the spiral of silence theory for communication and public
opinion research." In Dan Nimmo and Keith Sanders (eds.), Political Communi-
cation Yearbook Vol. 1. Carbondale: Southern Illinois University Press.
Noelle-Neuman, Elisabeth
1974 "The spiral of silence: a theory of public opinion." Journal of Communication 24
(2):43-51.
1977 "Turbulences in the climate of opinion: methodological applications of the spiral
of silence theory." Public Opinion Quarterly 41:143-58.
Page, Benjamin, and Robert Shapiro
1982 "Changes in Americans' policy preferences, 1935-1979." Public Opinion Quar-
terly 46:24-42.
Taylor, D. Garth
1982 "Pluralistic ignorance and the spiral of silence." Public Opinion Quarterly 46:
311-35.