Download Enhancement of dominant attitudes in group

Journal oj Personality and Social Psychology
1971, Vol. 20, No. 3, 386-391
ENHANCEMENT OF DOMINANT ATTITUDES
IN GROUP DISCUSSION1
DAVID G. MYERS 2 AND GEORGE D. BISHOP s
Hope College
Two studies were performed in an attempt to generalize the risky-cautiousshift phenomenon to attitudes and to test "information exchange" and "mutual
reinforcement" explanations of discussion-produced shifts. Study I confirmed
an assumption of the information-exchange hypothesis by extending to the
attitude realm the finding that subjects perceive their peers as less in the valued
direction than themselves. In Study II, eight of these items were responded to
before and after treatment by subjects in discussion, information-exchange, and
control conditions. Only discussion-condition groups evidenced significant attitude shifts and polarization effects. Observations of tape recordings of the
discussions indicated that a group's initial mean on an item was a significant
predictor of the direction of the discussion rhetoric, which in turn was a
predictor of group shifts. However, group initial mean did not directly predict
group shifts.
A large amount of research in the past
decade (cf. Kogan & Wallach, 1967) has investigated the effects of group discussion on
responses to questionnaire items on which
subjects recommend the amounts of risk to be
taken by hypothetical persons facing various
life dilemmas (Kogan & Wallach, 1964, Appendix E). It is now well documented that
discussion tends to increase risk taking on
items where the predominant initial response
tendency is in the risky direction, and tends
to enhance caution on items which tend to
elicit cautious initial responses (Myers, Murdoch, & Smith, 1970; Stoner, 1968; Teger &
Pruitt, 1967).
One popular explanation of this phenomenon (Brown, 1965) is that discussion provides
arguments mostly in support of the dominant
value and, also, that mere information about
others' decisions indicates to the average subject that relative to others, he is not as
strongly in the valued direction as he had
1
This research was supported by a grant (MH15999-01) to the senior author from the National
Institute of Mental Health and by a summer College Science Improvement Program grant (GY533S)
to the junior author from the National Science
Foundation. Appreciation is expressed to Neil Blahut
for his assistance in collecting the data.
2
Requests for reprints should be sent to David
Myers, Department of Psychology, Hope College,
Holland, Michigan 49423.
3
Now at Yale University.
supposed. This information is said to be a
stimulus for his shift in the valued direction.
The first suggestion, which could in behavioral
terms be labeled the "mutual reinforcement"
explanation, has not been directly investigated, except for Nordh0y's (1962) observation that the content of arguments in Stoner's
(1962) original groups tended to favor the
direction of group shift. The second explanation, "information exchange," has received
more investigation. Stoner (1968), Wallach
and Wing (1968), and Levinger and Schneider
(1969) all confirmed that individuals do tend
to see themselves as riskier than "fellow students" on items where a shift to risk has been
found. However, other research (Madaras &
Bern, 1968; Murdoch, Myers, & Smith, 1970;
Teger & Pruitt, 1967; Wallach & Kogan,
196S) has yielded conflicting data when mere
information about other's decisions is exchanged without opportunity for discussion.
If Brown's "value enhancement" theory of
group risk-taking effects is correct, it suggests that such discussion effects might generalize to other nonrisk response tendencies.
The present research applies to the attitude
realm the general hypothesis that discussion
enhances the dominant value, if such exists.
Evidence that this hypothesis might prove
fruitful was provided by post hoc examination of group-shift data obtained on discussion of 12 "risk-neutral" items used in a
386
DOMINANT ATTITUDES IN GROUP DISCUSSION
study by Wallach, Kogan, and Burt (1968).*
The mean shift scores on these 12 items were
correlated with the means of individual decisions of a small sample of Hope College students. The resulting correlation of .71 suggested that although few significant shifts occurred, the direction and amount of mean
shift on an item was predictable from the
mean of initial decisions, just as has been
observed with the life-dilemma items. Recent
research by Moscovici and Zavalloni (1969),
Doise (1969), and Myers and Bishop (1970)
also supports the hypothesis.
Study I examined an assumption of the
information-exchange explanation by asking
subjects to respond to 12 attitude items and
estimate the responses of fellow students. It
was hypothesized that subjects would tend to
see others as more neutral than themselves
(less in the valued direction) on these nonrisk materials. On the basis of data provided
by Study I, items were selected for a subsequent study which investigated discussion
effects and the extent to which information
exchange and mutual reinforcement processes
could predict any observed discussion effects.
STUDY I
Method
Subjects, Participants were 118 introductory psychology students at Hope College who completed
the questionnaire at the end of a regular class period.
Materials. From among a number of specially
reacted items, 6 were selected which pretesting suggested would elicit initial responses deviating in a
predictable direction from neutral. Using the same
criterion, 6 items from the Wallach et al. (1968)
nonrisk "Opinion Questionnaire" were also included.
These 12 items had a common format as illustrated
below.
387
loves her parents and also appreciates the fact
that they are financing her education.
Please put a check next to the ONE statement
that best expresses your opinion:
—I strongly favor that Mary N. go home
vacation.
I moderately favor that Mary N. go home
vacation.
I slightly favor that Mary N. go home
vacation.
Undecided
I slightly favor that Mary N. go on the
ology field trip.
I moderately favor that Mary N. go on
biology field trip.
I strongly favor that Mary N. go on the
ology field trip.
for
for
for
bithe
bi-
Procedure. Using verbatim the instructions of
Wallach et al. (1968), subjects completed the 12
items and on the last page of the booklet found the
following instructions.5:
Now that you have finished making your own
recommendations, go back over the Opinion Questionnaire again for a different purpose. This time,
please indicate what you think the majority of
your fellow students here at Hope College would
mark for each item. For each item, write an "M"
in the space that corresponds to the choice that
you believe the majority of your fellow students
would make. You can put the "M" in the same
space you check for yourself, or in a different
space.
Results and Discussion
Responses on the 7-point scale were scored,
ranging from —3 (strongly favor the first
alternative) to +3 (strongly favor the second
alternative). Table 1 summarizes the mean
data for all 12 items. The table indicates that
on 9 of the 12 items, subjects did on the average view their fellow students as significantly
more neutral than themselves. These data extend the results from similar studies cited
above to nonrisk opinion materials, suggesting a general tendency to see oneself as more
in the valued or dominant direction than one's
fellows.
Mary N., a college junior, desires to spend her
spring vacation on a biology field trip to another
region of the country. Her parents, who have not
seen her in several months, would like very much
to have her home over vacation. Mary N. is unsure what to do. Since she is interested in becoming a biologist, she considers the field trip a rare
opportunity for firsthand research experience. It
would also help her test out her interest in becoming a biologist. On the other hand, she knows that
her parents have been looking forward for some
time to having her home for a few days. Mary N.
The present experiment investigated the
effects of group discussion on responses to
selected items from Study I. In addition to
the main discussion condition, two other
* The authors are grateful to Michael Wallach for
providing these items and shift data.
5
These instructions are those used by Ivan D.
Steiner (personal communication, 1967).
STUDY II
DAVID G. MYERS AND GEORGE D. BISHOP
388
TABLE 1
MEAN or OWN RESPONSES, MEAN or RESPONSES
ESTIMATED FOR FELLOW STUDENTS, AND MEAN
DIFFERENCE BETWEEN OWN AND ESTIMATED
FELLOW STUDENTS' RESPONSES
Item no.
3
4
5
6
9
10
1
2
7
8
11
12
Own
Fellow
students
-.70
-2.07
-1.60
-.60
-1.10
-.79
-.08
-.52
-.91
-.19
-.38
-1.14
.63*
1.55* **
.70* *
.41*
.72* **
-.35*
1.22
1.14
-.09
-.20
-.71****
-.37*
-.40***
-1.13****
.73
1.29
1.56
1.74
1.13
.53
.58
1.19
1.34
.00
Difference
Note.—ps were determined by two-tailed t tests of the
deviations of the difference scores from zero. N =118, except
for Items 5 and 12, on which some subjects failed to respond.
*p < .10.
** t < .05.
***p <.01.
»***p < .001.
treatments were used—information exchange
and control. Study I extended the results of
previous research in demonstrating the information-exchange assumption that individuals
mistakenly perceive themselves as more in
the valued direction than others. The present
information-exchange condition provides the
predicted necessary and sufficient condition
for the shift phenomenon by merely allowing
exchange of information about others' decisions without opportunity for discussion. Control subjects completed the instrument twice
with irrelevant intervening activity to ascertain whether shifts occur in the absence of
discussion and information-exchange processes.
Method
Subjects. Volunteer participants were 160 introductory psychology students at Hope College who
signed up for the experiment, 5 for a given time.
Fifteen discussion groups, 10 information-exchange
groups, and 10 control groups resulted. In each condition, three-fifths of the groups were all female, and
two-fifths were all male. Due to the occasional failure of a volunteer to appear, 4 of the discussion and
3 of the information exchange groups had 4 persons.
Of the 10 control groups, 2 had 4 members and 3
had 3 members.
Materials. Since the purpose of the study was to
explore the hypothesis that discussion enhances domi-
nant values as denned by the mean of initial responses, items from Study I which elicited mean
responses more extreme than —1 or +1 were employed. None of the eight items meeting that criterion appeared to elicit initial responses so extreme
as to preclude further shifts toward the extreme. As
in Study I, the format was structured such that on
half of the items the dominant tendency would be
toward the first alternative and on the remaining
half, toward the second listed alternative.
Procedure. Subjects were seated around a round
table and requested to follow along as the experimenter read the instructions for the Opinion Questionnaire. After all were finished the booklets were
collected and the procedures defining the particular
experimental condition were administered.
Discussion groups received new identical booklets
with instructions paralleling those in Wallach and
Kogan's (1965) discussion without consensus condition. Subjects were asked to discuss each situation
for 3 minutes. After 3 minutes had elapsed (or after
a 20-second silent interval if discussion finished
early), the experimenter interrupted and requested a
new decision.
Information-exchange subjects were instructed to
share information about initial decisions but without discussion.
Now we would like to give you an opportunity to
compare notes by telling each other your initial
decisions as best as you can remember them. For
each item I will ask one of you to start by simply
reading the statement which seems to reflect how
you feel. Then you can just continue clockwise
around the table, each person reading his opinion.
Please do not discuss or comment on the items;
just read your choice as it is written. After this
exchange of information on each item I will ask
you to again mark a decision. In making your final
decision you in no way need to feel bound by
your previous decisions. Okay, would you look at
the first item [pause] and think about which
statement best reflects how you feel [20-second
pause]. Why don't we just go around the table,
starting this time with you [name] and simply
read your choice. . . . That's good. When making
your decision now, remember that we're not
interested in your prior opinion, but rather in just
how you feel about the situation now. All right,
go ahead and make your decision for the first
situation.
Control groups were given a brief anagram-solving
task as neutral intervening activity and then asked
to reconsider the situations without feeling bound
by previous decisions.
Results
Responses were again converted to a +3
to —3 scale. This time for simplicity of data
presentation, +3 was scored as strongly favoring the valued alternative and —3 as strongly
389
DOMINANT ATTITUDES IN GROUP DISCUSSION
favoring the less popular alternative. Initial
means on each item averaged across all 160
subjects were calculated. Two items (9 and
12 from Study I) did not elicit the predicted
dominant tendency,6 and hence these items
were not included in the computation of composite shift scores across items.
Enhancement of dominant value. For each
group, the pre- and posttreatment means of
decisions were computed. Table 2 presents
these data and the mean shift scores for each
condition. It indicates that significant enhancement of initial mean tendency occurred
only in the discussion condition. However, an
analysis of variance indicated that shift scores
for the three conditions did not differ significantly from one another (F—2.Q2, df =
2/32).
Tendency toward extremity. Moscovici and
Zavalloni (1969) observed that groups were
more extreme than individuals. In replication
of their observation, Table 2 indicates that
postdiscussion mean distances from zero were
significantly greater than the mean of prediscussion distances from zero. It is possible
that this "polarization effect" of discussion
applies to groups but not to individuals. For
example, a group with — 1, —1, +1, and +3
initial decisions and a +1 final consensus
would be more extreme than its initial mean
distance from zero of +.S. But the group
product of +1 is not more extreme than the
mean initial absolute extremity score of individuals (l.S). (Consistent with this analysis,
Kogan & Wallach, 1966, did not observe
group probability judgments to be more extreme than the mean of initial absolute extremity scores.) Hence, the difference between
the average of pre- and postdiscussion individual extremity scores was calculated for
each group. Table 2 indicates this difference
to be small, but also significant, in the discussion condition.
6
The initial means of .31 and .60 were less than
the arbitrary criterion of 1.00. No significant shift
on these items was observed in any condition. One
other item (which also failed to show significant
shift) had an initial mean of less than 1.00 for discussion subject, but was retained in Table 2, since
across all conditions the initial mean was greater
than 1.00.
TABLE 2
GROUP MEAN SCORES AND GROUP AND INDIVIDUAL
EXTREMITY SCORES BEFORE AND
AFTER TREATMENT
Condition
Discussion
Group M
Group extremity
Individual extremity
Information-exchange
Group M
Group extremity
Individual extremity
Control
Group M
Group extremity
Individual extremity
N> Before After
Difference
.24
.29
.13
15
15
15
1.43
1.47
2.10
1.66
1.76
2.24
10
10
10
1.52
1.56
2.06
1.51
1.55
1.98
10
1.63
1.63
2.08
1.68
1.69
2.13
10
10
-.01
-.01
-.09
.04
.06
.06
t
2.06*
2.73**
2.25"
.21
.23
1.58
.51
.70
1.54
*N is the number of groups.
* p < .06 (two-tailed).
**p <.05 (two-tailed).
Mutual reinforcement effects. The discussion data were further examined to see if
mutual reinforcement processes might help
predict shift scores. To explore mutual reinforcement effects, it was assumed that the
items elicited two conflicting response tendencies and that the expression of an argument
favoring one alternative constituted a reinforcement of that response tendency. It was
predicted that arguments would tend to favor
the dominant (valued) tendency. (See Byrne,
1969, for evidence that hearing an attitude
statement similar to one's own attitude tendency does constitute a reinforcement.)
Audiotapes were made of 14 of the IS
discussion groups (through a ceiling microphone pointed out to subjects before the discussion). For each discussion, two observers
independently counted the number of arguments in support of each alternative, and the
proportion of arguments favoring the dominant alternative was computed from this.
Across all items and groups, the two observers correlated .77 in their observation of these
proportions, so their observations were averaged. The mean proportion of arguments (reinforcements) favoring the dominant alternative was .76, a significant deviation from .50
(t= 10.87, p<.001). The proportion for
each item was then correlated across 84 group
discussions (6 items X 14 groups) with the
group initial means and group shifts. Results
(r = .49 and .37, respectively, N = 84, p <
.001) indicated that the group initial mean
was a significant predictor of the number of
390
DAVID G. MYERS AND GEORGE D. BISHOP
reinforcements in support of the dominant
alternative which was, in turn, a predictor of
the magnitude of shift in the dominant direction. However, a group's initial mean on an
item did not predict the group's shift on that
item (r = —.12). This latter finding has also
been observed on the risk-taking items (Myers & Arenson, 1970), and may be due in
part to regression and ceiling effects—a group
already near the extreme on an item cannot
shift much further toward the extreme. In
summary, although group initial means did
not directly predict group shift scores, the initial means did significantly predict the direction of the discussion rhetoric, and this was a
predictor of shift scores.
DISCUSSION
The present research explored the hypothesis that discussion enhances the dominant
value on attitude materials as well as on lifedilemma risk problems. A second purpose was
to examine the mutual reinforcement and
information-exchange explanations of any
such discussion effect.
Study II revealed marginally significant
discussion enhancement of mean initial tendencies. This is comparable to results obtained by Moscovici and Zavalloni (1969),
who also used a 7-point Likert scale (for preand postdiscussion attitude ratings of French
students concerning policies of General
deGaulle). Moscovici and Zavalloni observed
an average preconsensus-consensus shift of
.28 in comparison with the pre-post discussion
shift of .24 in the present research.
Study II also replicated Moscovici and
Zavalloni's (1969) observation that discussion
"results in a polarization of responses [p.
134]" when one compares the extremity of
group pre- and postdiscussion mean distances
from zero. However, this polarization effect
was of small magnitude when the mean absolute extremity of individuals' pre- and postdiscussion responses were compared.
Why does this small, but apparently reliable, discussion effect occur? Study I confirmed the information-exchange assumption
that people tend to see themselves as more
strongly in the valued direction than others.
Study II therefore provided the necessary
and sufficient conditions for information exchange by having subjects report initial responses without further discussion. Contrary
to the information-exchange explanation, no
shift was observed.
The discussion-based mutual reinforcement hypothesis was examined by content
analysis of tape recordings. As predicted,
arguments did tend to favor the predominant
response tendency. Also, the extent to which
arguments favored the dominant alternative
did significantly predict the group shift
scores. These findings support Brown's (1965)
and Vinokur's (1971) suggestion that shift
effects are due partly to an enhancement of
the prevailing value by the discussion arguments.
REFERENCES
BROWN, R. Social psychology. New York: Free
Press of Glencoe, 1965.
BYRNE, D. Attitudes and attraction. In L. Berkowitz (Ed.), Advances in experimental social psychology. Vol. 4. New York: Academic Press, 1969.
DOISE, W. Intergroup relations and polarization of
individual and collective judgments. Journal of
Personality and Social Psychology, 1969, 12, 136143.
KOGAN, N., & WALLACE, M. A. Risk taking: A study
in cognition and personality. New York: Holt,
Rinehart & Winston, 1964.
KOGAN, N., & WALLACE, M. A. Risk taking as a
function of the situation, the person, and the
group. In G. Handler, P. Musser, N. Kogan, & M.
Wallach (Eds.), New directions in psychology HI.
New York: Holt, Rinehart & Winston, 1967.
LEVINGER, G., & SCHNEIDER, D. J. A test of the "risk
is a value" hypothesis. Journal of Personality and
Social Psychology, 1969, 11, 165-169.
MADARAS, G. R., & BEM, D. J. Risk and conservatism
in group decision-making. Journal of Experimental
Social Psychology, 1968, 4, 350-365.
MOSCOVICI, S., & ZAVALLONI, M. The group as a
polarizer of attitudes. Journal of Personality and
Social Psychology, 1969, 12, 125-135.
MURDOCH, P., MYERS, D. G., & SMITH, G. F. Information effects on cautious and risky shift items.
Psychonomic Science, 1970, 20, 97-98.
MYERS, D. G., & ARENSON, S. J. Enhancement of
initial risk tendencies in group discussion. Unpublished manuscript, Hope College, 1970.
MYERS, D. G., & BISHOP, G. D. Discussion effects on
racial attitudes. Science, 1970, 169, 778-779.
MYERS, D. G., MURDOCH, P. H., & SMITH, G. F. Responsibility diffusion and drive enhancement effects on risky shift. Journal of Personality, 1970,
38, 418-425.
DOMINANT ATTITUDES IN GROUP DISCUSSION
NoRDHjzfsr, F. Group interaction in decision-making
under risk. Unpublished master's thesis, Massachusetts Institute of Technology, School of Industrial Management, 1962.
SIGNER, J. A. F. A comparison of individual and
group decisions involving risk. Unpublished master's thesis, Massachusetts Institute of Technology,
School of Industrial Management, 1961. Cited in
D. G. Marquis, Industrial responsibility and group
decisions involving risk. Industrial Management
Review, 1962, 3, 8-23.
SIGNER, J. A. F. Risky and cautious shifts in group
decisions: The influence of widely held values.
Journal of Experimental Social Psychology, 1968,
4, 442-459.
TEGER, A. I., & PRUITT, D. G. Components of group
risk taking. Journal of Experimental Social Psychology, 1967, 3, 189-205.
391
VINOKUR, A. Review and theoretical analysis of the
effects of group processes upon individual and
group decisions involving risk. Psychological Bulletin, 1971, 76, 231-250.
WALLACE, M. A., & KOGAN, N. The roles of information, discussion, and consensus in group risk
taking. Journal of Experimental Social Psychology,
1965, 1, 1-19.
WALLACE, M. A., KOGAN, N., & BURT, R. B. Are risk
takers more persuasive than conservatives in
group discussion? Journal of Experimental Social
Psychology, 1968, 4, 76-88.
WALLACE, M. A., & WING, C. W., JR. Is risk a value?
Journal of Personality and Social Psychology, 1968,
9, 101-106.
(Received May 20, 1970)