-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
1/20
Psychological Bul le t in
1993,Vol.
113, No.
3,533-552
C o p y r i g h t 1993by the Am e r i can Psychological Association Inc.
0033-2909/93/S3.00
BeyondFiasco:AReappraisal of the Groupthink Phenomenon
and a New
M odel
of
Group
Decision
Processes
Ramon
J. Aldag and
Sally
Riggs
Fuller
This article provides a comprehensive evaluation of the g roup think ph enom enon (I . L. Janis,1971,
1972,1982). Our evalua tion indicates that research does not provide convin cing support for the
validityof the
group think phenomenon
or for the
suggestion that groupthink characteristics lead
to ne gative outcomes. This review, coupled with evidence
from
other l i terature suggested by a
problem-solving perspectiveand adirect exam inationofgrou pthink implici t assumptions, guided
th e
development
of a
new, more general model term ed
t hegeneralgroup problem-solvingmodel .
This model incorporatesavarietyof antecedent conditions, emergent group characteristics, deci-
sion process characteristics, and grou p decision outcome s. Following the re view and m odel devel-
opmen t ,
we
discuss potential con cerns re lating
to our
model, address
t he
al lure
of
group think ,
an d
present implications
of our
analysis
fo r
group problem solving
as
well
as
directions
fo r
fu t ure
research.
In thepastt wodecades, the reha sbeenasubstantial increase
in emphasis on groups in organizations in general(e.g.,Leavitt,
1975;Ouchi,1981;Zander, 1982) and on group problem solving
in particular (e.g., Brandstatter, Davis, & Schuler, 1978;
Del-
becq,
V an de
V e n ,
&
Gustafson, 1975). However, surprisingly
little attention
ha s
been paid
to the
development
o f
comprehen-
sive models
of
grou p problem solving. Models
are
available
t o
suggest when groups, rather than individuals, should be used
for problem solving(Vroom
&
Yetton, 1973)
and the
appro-
priate composition
an d
functioning options
o f
problem-solving
groups (Stumpf, Zand,
&
Freedman,1979),
an d
severalspecific
group problem-solving techniques have been presented (e.g.,
Rohrbaugh, 1979;
Tu roff,
1971;
Van de V en &
Delbecq,
1971).
Furtherm ore, gen eric models of group fun ction ing are also
available bu t do not lend them selves to application
specifically
to grou p problem solving(e.g.,
Romans ,
1950).However, popu-
lar, comprehensive, general models of group problem solving
ar e
lacking.
The g roup think model (Janis,
197 1 ,197 2 ,198 2 )
is the most
prominen t a t tempt
to fill
this void.
In
recent years,
acceptance
of
t he
grou p th in k phe n om e n on
has
become almost un iversal,
and the term groupthink has entered the pop ular vocabulary. It
ha s
been blamed
for
such decision-making
fiascoes as the B ay
of Pigs invasion, the escalation of the Vietnam conflict, the
Watergate
cover-up,
and the Challenger
disaster,
as
well
as for
flawed grou p prob lem solving in bu siness and other organiza-
tions. Despite Longleya n d Pruitt 's (1980) warning that both
Ramon
J.
A ldag, Graduate School
of
Business Adm inistration, Un i-
versity of
WisconsinMadison; Sally Riggs Fuller, School
of Busi-
ness, Un iversity of W ashington .
We
would
like
to
t h a n k
t he
Graduate School
of the
University
of
WisconsinMadison fo r
prov iding
f unding to
support this article.
Ramon J .
Aldag
an d
Sally Riggs Fuller made e qual contributions
to
this article. Names
ar e
presented alphabetically.
Correspondence conce rning this ar ticle should be addressed to Ra-
mon J. Aldag , Gradu ate School of Business, University of Wisconsin,
1155
Observatory Drive, Madison, Wisconsin 53706.
th e
groupthink theory
an d
related research were problematic,
Janis's (1971) original conceptualization has not been altered,
an d groupthink cont inues to beviewedas adefectiveprocess
thatshouldbeguarded against. Articles discussingth edangers
ofgroup think and suggesting remedies have regularly appeared
in p eriodicals aimed at man agers, lawyers, medical profes-
sionals, and the general public (e.g., Cerami, 1972;
Cu lbertson,
1977;
Henderson, 1987; Ludwig, 1973; Rosenblum, 1982;
Sanders, 1980;V onB ergen&Kir k, 1978).Th e
Social Sciences
Citation
Index
showed more than 700 citations of Janis's work
from January1989throu gh J u n e1992.
Janis chose
t he
term
groupthink
because
of its
franklyOrwel-
lian connotation, similar to doublethink an d crimethink.
Janis(1982)wrote that the invidiousness
is
intentional
(p. 9).
This view of groupthink as an undesirable phenomenon con-
t inuesto beevidencedinJanis's most recen t wo rk.Fo rinstance,
Janis (1989) wrote that
wheneverapolicymakingor crisis manage men t group is
funct ion-
in g
as a compatible team
with
a fair or high degree of esprit de
corps, take steps
to
counteract tendencies toward concurrence-
seeking or groupthink. If a leaderdoes this, the decision-mak-
in g
processwill tend to be of higher
quality,
( p.
247)'
He the n presented a l is t ing of remedies for groupthin k.
The widespread acceptance of the groupthink model sug-
gests that it has had considerable he uris tic value. In this sense, it
ha s
served similar roles to models such as those presented by
Maslow
(1943)and Piaget
(1932,1972).
In eachcase,a
relatively
precise model stimulated
further
research that ultimately chal-
lenged its
validity.
For instance, Maslow's hypothesized need
hierarchy,based p rimarily on clinical studies of neu rotic indi-
viduals,met with widespread acceptance and research interest
bu t
failed
to gain substantial empirical support (e.g.,
Alderfer,
1
From Crucial Decisions: Leadership in
Policymaking
and Crisis
Management (p . 247 )b y
Irving
L. Janis, 1989, Ne w
York:
Th e Free
Press. Cop yright
1989by The
Free Press. Reprinted
by
permission
o f
The Free Press, a Division
of
Macm illan, Inc.
53 3
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
2/20
534
R A M O N
1
ALDAG
AND
SALLY R I G G S F U L L E R
1972; Hal l& Nou gaim, 1967; Wahba& Bridw ell, 1976). Simi-
larly,Piaget 's theory, originally based on an in-dep th stu dy of a
small
n um ber of children, served as a seminal s tage theory of
cognitive
development, bu t additional research has suggested
needed revisions
(T. G. R.
Bower,1976; Horn, 1976).
We
argue
that g rou pthin k has p layed a s imilar role to these theor ies . Like
the m, the g rou pth ink model (a) was based on a small and rela-
tively restricted sample
but has
been
widely
general ized;
( b)
posits a precise and de termin istic sequen ce of stages; (c) has had
substantial
intui t ive appeal ;
an d (d) has not
received consistent
empirical suppor t ,
as we
demon st rate .
Con cerns about the val idi ty of the m odels notwithstanding,
each has proved to be extremely valuable. Specifically, th e
grou p th i n k model has stimulated research on group dysfunc-
t i on s and encouraged
viewing
outcomes in problem-solving
t e rms ;
shown how decision processes m ay be relevant to a wide
range
of situations; provided links to other t ype s of literature,
such
as
stress
an d
vigilance;
and
identified potentially impor-
t an t
variables
in
gro up prob lem solving. Also, like
the
Maslow
(1943) an d Piaget (1932, 1972) models , it has provided a rich
base
fo r
fu r t h e r
development
an d
tes t ing.
However, two directions seem necessary to preserve the
model's continued
viability.
First,
th e
g r o up t h ink m o d e l
has a
relatively
nar r o w focus, p rima rily addressing major decision
fiascoes.
As such,
it
would appear
useful to
provide
an ex-
panded perspective applicable
to a
broader range
of
problem
types . Second, the groupthink model has not been revised to
incorporate the emerging literature and empirical findings. In
thisarticlew epresent a new, broader frameworkfo rexamina-
t ion of grou p problem solving, terme d thegeneral group prob-
lem-solving (GGPS)
model.
It captures and recasts some ele-
m ent s
of the gr oup think model . Addit ionally , theGGPSmodel
integrates findings of a
g r o u p t h in k li t e ra t u r e review, material
from
a variety of other lite ratu re suggested by a general
prob-
lem-solving perspect ive, and an e xaminat ion of group think as-
sumptions .
In this article, we first discuss the basic characteristics of the
grou p th i n k model . Next , we summarize and discuss the re-
search that
h as
exam ine d g r o up t h ink , ado p t
a
general problem-
solving perspective and discuss other relevant literature sug-
gested by that perspective, and question groupthink assump-
t ions . We then present the ne w model, note potent ial con cerns
relating to the model, address th e al lure of g r o up t h ink , an d
provide implication s of our literature re view and model devel-
opment for fu ture theory and research on group problem
solving.
Groupthink
Model
At the individual level , Janis and Man n (1977) argued that
th i n k i n g about vital, affect-laden issues generally results in
hot cogn itions, incontrast to the cold cognitionsofrout ine
problem solving,
an d
that such
ho t
cognitions
ca n
result
in
errors
in
scanning
of
alternatives. Such situations inducestress
that results
in
defensive avoidance, characterized
by
lack
of
vigilant search, distortion of the meanings of warning mes-
sages, selective inattention an d forgetting, an d
rationalizing.
Gr o up t h ink
is
seen
as the
group analogue
of
defensive avoid-
ance, or a collective pattern of defensive avoidance (Janis &
M a n n ,1977,
p .129).
According
t o
Janis (1972),
I
u se the te rm group think as aq u ickan d easyway to refer to a
mode of t h i n k i n g that people engage in when they ar e deeply
involved in a
cohesive in-group, when
th e
members ' s tr ivings
fo r
unan imi ty
override their motivationt orealistically appraise al -
ternat ive courses of ac t ion . . . . Group thinkrefersto a deteriora-
t ion of
men ta lefficiency, real i ty testing,
an d
moral judgmen t tha t
results from in -grouppressures ,
(p. 9)
Janis
(1982) explicitly
defined the context in which group-
t h i n k is thou ght to occur and divided these anteceden t condi-
tions into three categories. First,
a
moderate
to
high level
o f
group cohesion
is a
n ecessary
but not
sufficient condition
fo r
g r o up t h ink .
Secondary antecedents thata resaid topredict situ-
at ions in which groupthink is likely to occur are s t ructural
faults and a provocative situational context. In the structural
fault category are insu lation of the grou p, lack of tradition of
impartial leadership, lack of norms requiring methodical pro-
cedures, and hom ogen eity of memb ers' social backgroun ds and
ideology.The provocative situational contextantecedents focus
on
th e
role
o f
stress
as a
situational factor.
The first
stressfactor
ischaracterized
as
extern al threats
of
losses combined with
a
low hope
of finding a
bet ter solution than that
of the
leader.
The
intern al s tress antecedent s tems from temporary
lo wself-
esteem at t r ibutable
to
members' recent failures, perceptions
that the task is too difficult to accomplish, and the perception
that there is no alternative that is morally correct.
There
ar e
several supposed symptoms
of
groupthink (cf .
Janis & Mann, 1977, pp . 130-131), including an i l lusion of
invulnerabil i ty,rat ionalizat ion to discount warn ings and other
negative fee dback , belief in the inh ere nt morality of the grou p,
s tereotyped
views
of mem bers of opposing groups, pressure on
dissenters, self-cen sorship, illusion of un anim ity, and self-ap-
pointed mindguards acting to shield the g roup from adverse
information .
According to Janis
(1 9 71 , 1 9 72 , 1 9 82, 1 9 89 ) ,
g r o up t h ink r e -
sults in a n u m b e ro fconseque nces that inte rfere with
effective
group decision making. For instance, the group limits its dis-
cussion to
only
a few
alternatives.After
a
course
of
action
is
initially
selected, members ignore
n ew
informat ion conce rning
its risks and drawbacks. At the same time, the y avoid infor ma-
t ion con cerning the benefi ts.of rejected alternatives. Members
make l i t t le at tempt
to use
exper ts
to
obtain moreprecise
infor-
mation. In addi t ion, because t hey are so confidentthat things
willt u r n
o u t
well,group members
fail to
consider what
may go
wrong and,
as
such,
do not
develop contingency plans. These
defects
are
seen
as
leading
to
impaired performance
an d
other u ndesirable outcomes. This view
of the
causal sequence
relating to d e t e r m inan t s an d consequences of g r o up t h ink is
sum m ar iz ed in Figure 1 .
Janis
(1 9 71 , 1 9 72 , 1 9 82, 1 9 89 )
suggested methods
to
prevent
or m inimize the supposedly dysfunct ional consequences of
groupthink, including the following: The group leader should
encourage
al l
g r o up m em ber s
to air
their doubts
an d
objec-
tions; leaders should adopt
an
impartial stancerather thanini-
tially stating the ir preference s; mem bers should
be
encouraged
to
discuss
t he
group's delibe rations withtrusted associates
an d
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
3/20
B E Y O N D FIASCO
53 5
ANTECEDENTS
COHESIVENESS
STRUCTURAL FAULTS OF THE
ORGANIZATION:
INSULATION
LACK OF LEADER
IMPARTIALITY
LACK
O F
PROCEDURAL
NORMS
MEMBER HOMOGENEITY
PROVOCATIVE STTUATIONAL
CONTEXT:
HIGHSTRESS FROM
EXTERNALTHREATS WITH
LOW
HOPE
OF
BETTER
SOLUTION THANLEADER S
LOW
TEMPORARY
SELF-ESTEEMINDUCEDB Y:
RECENT FAILURES
DIFFICULTIES
IN
CURRENT
DECISION-MAKING
TASK
MORAL
DILEMMAS
SEEKING
TENDENCY
SYMPTOMS
ILLUSIONOFINVULNERABILITY
BELIEF
IN INHERENTGROUP
MORALITY
COLLECTIVE RATIONALIZATION
STEREOTYPES OF OUTSIDERS
SELF-CENSORSHIP
ILLUSION
OFUNANIMITY
PRESSUREON DISSENTERS
SELF-APPOINTED MINDGUARDS
DEFECTS
INCOMPLETE
SURVEY OF
ALTERNATIVES
INCOMPLETE
SURVEY OF
OBJECTIVES
FAILURE TO RE-EXAMINE
PREFERRED
CHOICE
FAILURE TO RE-EXAMINE
REJECTED ALTERNATIVES
POOR INFORMATION SEARCH
SELECTIVE BIAS
IN
PROCESSING
INFORMATION
FAILURETO DEVELOP
CONTINGENCY PLANS
v
,
LOW PROBAB ILITY OF
SUCCESSFUL OUTCOME
FigureI . Grou pthin k model . (FromDecis ion Making: A
Psychological
Analysis of Conflict, Choice, and
Co mmi t me n t[p 132 ] byIrvingL.JanisandLeon Man n, 1977,NewYork:Th eFree Press. Cop yright 1977
byT he Free Press. Adaptedbypermissionof TheFree Press,aDivisionof Macmillan, Inc.)
report their reactions backto thegrou p; outside expe rts should
be
invitedto meetingsa nd encouraged to challenge members'
views;when a competitor is involved, time should be devoted to
assessment of warnin g signals
from
the competitor and of alter-
native
scenariosof thecompetitor's intentions; when consider-
in g alternatives,t hegroup should split into subgroupst o meet
separately from timetot ime;th egroup should holda second-
chance
meetingafter
a
p re liminary consensus
is
reached
o n a
preferred
alternative;and thegrou p should consider using disso-
nance-inducing
group processes.
Underlying
Assumptions of Groupthink
At base, the prevailingviewof groupthink seems to rest on a
set of generally unstated assumptions: First, th e purpose of
usinga group for problem solving is prim arily to enh ance deci-
sion qu ality. The gro up thin k model is silent con cernin g other
outcomes such
a s
mem ber adherence
to the
solution, member
satisfaction w ith the leader, and
future
motivation of the leader,
to name just a
f e w .
Second, group problem solvingisessen tiallya rational pro-
cess,
with membersunifiedin their p ursuit of a goal and partici-
pating in the group process to facilitate achievement of that
goal. This ration al process involves the selection of thegoal-
maximizingalternative regardless ofwhich particular interests
within theorganizationfavor that alternative (Pfeffer, 1981, p.
22).Such rationalityisbotha description of the decision pro-
cess
and a
value d social ideal (Parsons
&
Smelser, 1956; Pfeffer,
1981). March (1976, p. 69) referred to the theory of rational
decision mak ing usin g term s such as faith and scripture.
Third,
benefits
associated
with
a problem-solving group that
function s wellare assumed to inclu de a wide variety of perspec-
tiveson a problem, more inform ation concerning proposed al-
ternatives, he ightened decision re liability, dam pe nin g of biases,
an d social presenceeffects (cf. Gu zzo, 1986; Maier, 1976).
Fourth, grou pthin k character ist ics are assumed to prevent
th e
actualization
of
these potential group benefits
by
causing
group members to respond passively to external pressures in
undesirable
ways.
Fifth,
it is
assumed that
b y
taking steps
to
prevent
th e
occur-
renceofgroupthink characteristics,amorerationaland effec-
tive group process
will
result, leading to enhanced decision
quality.
Sixth,
an illusionof
well-being
ispresumedto beinherent ly
dysfunctional.
Such
a
position
is
consistent with
the
view that
veridical
perceptions
are
critical
to effective
functioning (e.g.,
Jahoda, 1953; Maslow, 1959). In her review of the organiza-
tional implications of self-efficacy, Gist
(1987)
used the
g r o up t h ink model to argue that self-efficacy can be too high,
thus
producing overconfidence
an d
poor performance .
Al-
though notexplicitly stated, he r view was apparently that fo r
optimal performance ,
self-efficacy
should
reflect
objective
reality.
Finally,
it is
assumed that
th e
concurrence-seeking a t t i tudes
an d
behaviors ( symptoms
an d
defects ) resu ltfrom individ-
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
4/20
536
R A M O N J .
A LD A G
A ND SALLY R I G G S
F U L L E R
uals'
psychological agreeme nt-seeking drives
an d
group pres-
suresf orconsensus.
Review
of the Prior
G r o u p t h in k
Literature
We review
three typ es of literature relating to the group thin k
phe n om e n on :
case analyses, laboratory studies, and concep-
t ua l
articles.
C a se
Analyses
Early su ppor t fo rg r o u p t h i n kw as based almost entirely on
retrospective case studies. For instance, Janis(1971,1972,1982)
focused prima rily on five American foreign-policy crises: the
Marshall Plan,
th e
invasion
of
North Korea,
the Bay of
Pigs
invasion, the Cuban Missile Crisis, and the escalation of the
conflict
in V ietn am. Through case studies, he categorized
these
crises according to whether evidence of groupthink was
revealed. In his 1982
work, Janisalsodiscussed
th e
Watergate
cover-up
as an
example
of
how clever manipu lators
can get
caught
in an
avoidable quagmire
(p .
198).
Tetlock (1979)
examined the same five crises considered by
Janis (1971,
1972).
Content analyses were performed on the
publ icstatementsof keydecision make rs involvedin thecrises.
Consisten t
with
Janis's theory, grou pth ink decision makers
werefound
to be more simplistic in their perceptions of policy
issues
than n ongrou pthink decision makers and to make more
positivereferences to the UnitedStatesand its allies. Inconsis-
t en t
with Janis's theory, gro up thin k decision makers did n ot
make
significantly
m ore negative references to the Com mun ist
states and their allies. Tetlock
(1979)
concluded that the
current findings
converge impressively with
the
conclusions
of
Janis'sintensivecase studie s (p.1322).He attributed the lack of
uniform
support to methodological issues, arguing that content
analyses of pu blic statements may be insensitive to
differences
in evaluationsofopponents in grou p th in kan dn o n g r o u p t h in k
crises.
More recently, several other case studies providing support
for
grou pth ink have been pu blished. For instance , Manz and
Sims (1982) presented three cases to demonstrate groupthink
t endencies in autonomous work-group settings. Smith (1984)
presented case informa tion describing the attempt by the
UnitedStates to rescue its hostages in Tehran as an example of
g r o up t h ink
(this example was noted by Janis in his 1982 edi-
tion).
Hen sley and
Griffin
(1986) applied grou pthin k to a case
analysis of the
1977gym nasiu m controversy
at
Ken t State Uni-
versity, where trustees decided to bu ild an addition to the
school's gym nasiu m on part of the areaof the tragic 1970 con-
frontation
between students and Ohio National Guard
members. Hensley and Griffin (1986) concluded that each
major
condition of the theory was present in the conflict and
that the trustee s were indeed victims of grou pthin k (p. 497).
Esser and Lindoerfer(1989) conducte d a quantitative case
analysis
of the space shuttle Challengeracciden t. They coded
sta tements from th e Report of
th e
Presidential Com mission
on
th e Space Shuttle Challenger Accident (Presidential Commis-
sion
on the
Space Shuttle Challen ger Accident,1986)
as
positive
or ne gative instances of the observable antecedents or conse-
quences of
groupthink. Positive instances
of
group think were
reported to be twice as
frequent
as negative instances. Fu rther-
more, du rin g the 24 hr p rior to lau nch, the ratio of positive to
negative instances increased
an d
then remained high. Esser
and Lindoerfer(1989) noted that it is possible that the many
instance s of decision-makin g defects would be consistent
with
any th eor y of poor decision m aking. W e have no comparison
group withwhich toe valuate this alternative (p .175). Use of
th e
te rminology
positivea nd negative instances
o f
groupthink
wa staken
to
im ply that
an y
sta tement
n ot
direc tly inconsistent
with gro up think was an instance of grou pthin k. The authors
failed
to f ind sta tements re la t ing to some groupthin k symp-
toms or consequences. A lthou gh they speculated that the deci-
sion
makers const i tu ted ahighlycohesive group, no statements
provided information concerning group cohesiveness. Finally,
the finding of a greater relative number of occurrences of in-
stances of grou pt hin k closer to the decision point was inter-
preted to be evidence of poor decision mak ing rather than as
simple
convergence on a solution.
Laboratory Studies
As
recently
as1982,Moorhead
wrote that since
t he
publica-
tion of Janis's book, th ere h as been no systematic research test-
ing any of itspropositions (p .436).Even now, there have been
only a l imited nu mber of studies empir ically examinin g
grou p th in k . W e discussed th e Esser an d Lindoerfer
(1989)
quantitative
case analysis earlier. We now review laboratory
studies of groupthink in chronological order.
Flowers (1977 ) hypothe sized that cohesiveness and leader-
ship style should interact in the decision process, so that
groupthink is
most
likely to
occur when group cohesiveness
is
higha nd
leadership style
is
closed
(i.e.,t he
leader presses
for his
or her own
decision
and a
unanimous decision). Flowers
as -
signed undergraduate students to experimental teams, with
leaders trained to take an open or closed style and cohesiveness
manipu lated by whe ther studen ts kne w each other (high cohe-
sive)
or did not (low cohesive). Teams we re
given
a
crisis
deci-
sion situation involvinga school person nel problem. Open lead-
ership style produced more suggested solutions and u se ofavail-
able
facts
th an did closed leade rship style, but that there was no
support
fo r
Janis's
(1971,1972)
assumption that cohesiveness
is
a
critical variable. Flowers(1977)argued that the results
of
this
experiment suggest that poor decision-making activities can
occur in a wide r variety of crisis situations than originally stud-
ied
by Janis (p. 888).
Cou rtright (1978) placed first-year undergraduates into
teams
an d
gave them
th e
problem , What
is the
best method
for recruiting
new
students
to the
University
of
Iowa? Hal f
were
told that they were chosen to be compatible (high cohe-
sive)and the other half that the y probably would not like each
other (low cohesive). In addition to cohesiveness, Courtright
manipulated decision-making parameters.
In the
freed
condi-
tion, group members were told that arguing is the best way to
solve a problem. In the limited condition, members were told
that cooperation is ne cessary and that time is limited. In the
no-instructions condition , membe rs were told only of the time
limit for discussion. Depen dent variables included the nu mber
of times each mem beroffered solutions, stated agreem ent, and
stated disagreement. The only significant finding was that
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
5/20
B E Y O N D FIASCO
537
groups in the high-cohesive condition an d limited-parameters
condition (the groupthink condition) generated
significantly
less disagreemen t th an
did
groups
in the
low-cohesive
an d
lim-
ited-parameters conditions. However,
C ourtright
(1978)
stated
that
the theory
aspostulatedby Janis is essentiallycorrect
(p .229).
Moorhead an dM ontan ari (1982) developed measuresto tap
g r o up t h ink concepts, including group characteristics,
symptoms, an d defects an dpresented ap re liminary study
to assess the validity of the indexes. B usiness studen ts com-
peted as teams in a management simulation.
After
the competi-
tion, stude nts completed
th e
Moorhead
a n d
Montanari inst ru-
ments. Largely un supp ortive factor analyses led Moorhead and
M o nt anar i(1982)
to conclude that the measurement and vali-
dation
of
grou p th in k
may not be as
simple
as
Janis's clear
an d
allur ing
presentation would suggest (p. 382). They
further
noted that although
th e
group characteristics measures
ap -
peared adequate, the conceptual uniqueness
of the
variables
described under
th e
Symptoms
an d
Defects categories
can be
quest ioned,
based on
these
data
(Moorhead
&
Montanari ,
1982,
p .382). Furthermo re, the y
reported
on the basisof inter-
category correlations that the pre limin ary analysis indicated
that some of the relationships
proposed
by Janis appeared to
hold for the groups used in this study but
that
others do not
appear to be as strong for this data set (Moorhead & Montan-
ari, 1982,p.382).
Callaway
andEsser(1984),
using
undergraduate
psychology
studen ts, manipulated grou p cohesiveness and the adequacy of
decision procedures
in a factorial
design.
The
authors con-
cluded that their findings provided mixed support for the
g r o up t h ink
hypothesis on measures of decision quality and
group processes presumed
to
underlie
th e
groupthink deci-
sions.
In
the irstudy,cohesiveness categories w ere achievedpost
hoc,
and the
degree
o f
support
for
groupthink varied with
th e
group task (the lost-at-sea task was supportive and the horse-
trading task
wa s
not). Given
th e
differential
support across
tasks, the authors reasoned
that
the horse-trading task may be
inappropriate in a
test
of groupthink because it has a single
correct answer that becomes immediately obvious.
Leana(1985)used teams of college studen ts solving a hypo-
thetical business problem. She tested the effectsof group cohe-
sivenessand leader behavior (directive vs. participative) on de-
fective
decision making.
S he
found that teams with directive
leaders proposed and discussed
fewer
alternatives and dis-
cussed
fewer
alternative solutions than did te ams with partici-
pative leaders. However, contrary to Janis's
( 197 1 ,197 2 ,198 2 )
g r o up t h ink
model, members of noncohe sive groups engaged in
more
self-censorship
of
information
th an did membe rs of cohe-
sive groups.
These findings
were
interpreted to be partially
support ive of the group think model.
Gladstein
an d
R eilly (1985),u s in g
a
management simulation
called Tycoon, studied group decision making under threat.
They hypothesized that threat may result in restriction in infor-
mation
processing and a decrease in breadth of distribution of
influence
(termed
constriction
of
control).
Noting th e con-
g r uence
of these hypotheses with the groupthink model, they
stated that no one hasfully tested the groupthink model, al-
though several studies have shown partial support
for it
(Flowers, 197 7; Leana, 1985) (Gladstein &
Reilly,
1985, pp .
615-616).
Altho ugh the results indicated a restriction in
infor-
mation processing
an d
increased
stress
when threat increased,
th e
degree
of
threat m ade
n o
significant
differen ce
in
constric-
tion of control. They argued that group norms and task de-
mands might haveacted
to
moderate
the
relation
between
threat an dconstriction of control.
Callaway,
Marriott, and Esser(1985) reasoned that because
Janis(1972) proposed that gr ou pt hin k is essentially a stress-re-
duction
process,
it might be prevented in cohesive groups if the
stress could be
diffused
by other factors. They investigated the
effects
ofgrou p structure (decision-making procedures)and in-
dividualdo minance on symptoms of groupthink, anxie ty, and
the qu ality of group decision makin g. Students, formed into
groups on the basis of dominance scores, participated in the
lost-at-sea task(Nem iroff& Pasmore, 1975),which permits ob-
jective scoring of the
correctness
of the group decision. Deci-
sion-making procedures were found to
influence
the time re-
quired
to
reach
a
decision
but not
decision quality
or the
pro-
cess variables assumed
to
mediate groupthink.
Groups
composed of highly domin ant members m ade higher quality
decisions, exhibited lower
state
anxiety,
took more time to
reacha decision, tended tomake more statementsofdisagree-
ment and agreement, and reported more group influence on
the members.
Moorhead and Montanari(1986)argued that previous empir-
ical studies of grou pthin k were flawed both because they
failed
to
successfully
produce theantecedent conditions proposedby
Janis (1972,1982) and because they provided incomplete tests
of
the group think theore t ical
framework.
Moorhead and Mon-
tanari (1986) wrote that the most comprehensive empirical
test
to date (Courtright,1978)included measures ofonlytwo of
the seven antecedent conditions, none of the symptoms, three
of the seven decision-making defects, and outcomes measured
on a subjective produ ctivity rating instrum en t (p. 402). To pro-
vide a more complete test, the authors
formed
teams of busi-
ness policy students who completed a highly competitive man -
agement simulation exercise (Moorhead & Mon tanari, 1986,
p.
405).
Refined
versions of the Moorhead and Montanari
(1982)measures were administered, and path analysis was used
to
assess
th e
implied causal sequence.
The
authors concluded
that although several relationships supported the framework
proposed
by
Janis, the em pirically derived mode l suggests that
several
linkages wereopposite to those predicted by the Janis
framework (Moorhead & Montanari, 1986, p. 408). For in-
stance, cohesion had a negative impact on self-censorship, a
positive impact on dissent, and a negative impact on the defect,
few alternatives. Moorhead and Mon tanari
(1986)
noted that
these data do not
show
powerful support for Janis's group-
th in k
hypothesis (p .
408)
bu t
argued
that there
were enough
significant relations amon g variables
to
warrantfurtherinvesti-
gation.
Montanari and Moorhead
(1989)
developed and validated
scales to measure gro upth ink
variables.
2
They included three
2
Although
Moorhead and Montanari 's
(1982)
paper is nowhere
noted
in
Montanar i
a nd
Moorhead's (1989) article,
th e
papers
are es-
sentially
identical (i.e.,
the
latter
is
apparen t ly
a
journ al version
of the
former).
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
6/20
53 8
RA MON J .
A L D A G
A ND
SALLY
RIGGS
F U L L E R
antecedents (cohesion, insu lation, and promotional leadership)
and alleight symptomsan d seven defects. Factor an alyses were
used to reduce eight symptoms to five and seven defects to
three. The analyses were conducted only on variables with in
categories (i.e ., antecedent conditions, symptom s,
an d
defects),
presumably
because th e sample size(N=
61)
did not permita
full factor analysis.
Even with
this restricted approach, there
wa s
relatively little supp ort
for the
s t ru c tu re
proposed
by
Janis.
Montanari and Moorhead
(1989)
concluded that
Janis's
con-
ceptualization of the
grou pthin k variables, when operat ional-
ized via
subscale items, lacks
sufficient
stability
(p .217) .
Conceptual
Articles
In the firstcom prehen sive critical evaluationo fgrou p th in k ,
Longley
and Pr uitt (1980) raised several issues inclu din g the
lack
of
empirical support
an d
deficiencies
in the
the ory itself.
They
called
for a
clearer
definition of
gro upth ink, more r igor-
ou s
empirical research,
an d
increased attention
to
underlying
processes and the relations among variables. Despite these
warnings,
research ongrou p th in k c on t in u e din thesame vein ,
withoutsignificant changest o Janis's
(1972)
original theory.
Whyte(1989)argued that the decision fiascoes attributed to
groupthink may beexplained byprospect polarization, which
uses,inadditionto thepressuresfor
uniformity
ofgrou p th in k ,
th e concepts of
framing
effects
(Tversky
&
K ahneman, 1986)
and g rou p polarization (Myers & Lamm , 1976). Altho ugh
Whyte cited some groupthink research as being more consis-
ten t
with
prospect polarization than with groupthink, he did
not challenge
th e
general constellation
of
characteristics
saidt o
be associated with grou p th in k . Whyte(1989) suggested that
g r o up t h ink although relevant, is an incomplete explanation
for th eoccurrence ofdecision fiascoes (p.54).
Luechauer (1989) provided a brief overviewof grou p th in k
and the
critiques leveled against
it
(citing Longley
&
Pruit t ,
1980,
an d
Whyte ,
1989).
Lu echaue r (1989) then presented
a
revised model
of the
groupt hink phenom enon, which sug-
gests that
th e
self-monitoring propensity (Snyder, 1974)
of
grou p memb ers is a critical personality variable that should be
considered
in the
grou p th in k
formulation an d
that Fantasy
Them es may act as a possible mechan ism wh ichpropelsgroups
toward grou p th in k (p. 5).Luechauer argued that fantasiesen -
able
a grou p to draw a bou ndary aroun d i tse lf that resists in t ru -
sions. L u e c ha u e r (1989) then discussed ho w such additional
variables might act to p u l l th e fantasizing away
from
reality
an d toward gro up thin k (p. 10). He noted that our purpose in
present ing
th is model
is not to
d ebu nk Janis's work. Rather,
it is
to exp and the model by supp lying two e lements that Janis does
not consider in his treatment of the
topic
(Luechauer,
1989, p. 5).
McCauley (1989) used content analysis to review
Janis's
(1982)descriptions of historical
cases.
He sought to de te rmine
whethergrou pthin k concurrence seeking
i s
caused
by
compli-
ance
(public without pr ivate agreement)
or
internalization (pri-
vate acceptance
of
group consensus). McCauley interpreted
Janis
as
arguing
fo r
internalization
bu t
reasoned that compli-
ance migh t have also played a su bstantial role. For
each
of eight
decisions (six groupthink
and two
n ongrou pthink) , McCauley
determined whether compliance and e ight hypothesized
ante-
cedents we re present. The results showed evidence of compli-
ance in two of the six grou pthin k decisions and in neither of the
two decisions without groupthink. Because all eight
cases
re -
flected highdegrees of cohesion and e xternal thre at, these ante-
cedents weren ot useful indiscriminat ing be tween groupthin k
and no ng rou pthin k situations. The W atergate case was the only
example in whichall six of the
other
antecedents were
present .
Across the sixg rou pthin k decisions for the sixdiscriminat ing
antecedents,
21
antecedents were present
an d
15 were absent
(i.e.,
without Watergate, these hypothesized antecedents were
equally
present and absent) . In the two nongroupthink deci-
sions,
five antecede nts were present and seven were absent. In
only
three of the groupthink decisions were more discriminat-
in g
antecedents present than absent. McCauley concluded that
cohesion can lead to both compliance and internalization and
that
future
grou pth ink research should at tempt to distinguish
be tween them.
Park (1990) provided
a
critical evaluation
of
empirical
g r o up t h ink research. He questioned the model's completeness
and the methodologies used to assess it. Criticisms included
researchers'
failure
to test all of the model's variables and the
inappropria te useof
modesof
measuremen t. Park's recommen -
dation was to use structural equation modeling such as
LISREL to test g rou pth ink . Despite these criticisms, Park ap-
peared to accept the phen omen on as be ing valid without sug-
gesting
changes
to the
original
framework.
Hart
(1991)
provided
a
revie w essay
on
grou p th in k ,
focusing
on its relevance to the p olitical
arena.
Following a discussion of
the groupthink phenomenon, i ts theore t ical context , and a
brief review
of related research, he presen ted a v ariety of con-
clusions and recommendat ions. For example , he caut ioned
against
relianceo n a narrowset of
criteria
fo rdecision quality
and en couraged a more inte grative perspective, combin ing p sy-
chological with political and organizational paradigms. Fur-
thermore ,
he
called
fo r
greater precision
in
establishing
th e
anteceden ts and dy namics of grou pthin k and for arrival at bet-
te r grounded diagnost ics of when, how,and whygrou p th in k
occurs.
Discussion of Prior Group think Literature
In
this section,
we first
present some general comments
re -
gardingpast grou pth ink literature and the n discuss specific
lessons th at
may be
drawnfrom thatliterature.
General
Co mmen ts
As suggested by the earlier review, most support fo rgroup-
th in k ha s come from retrospective case studies that havefo -
cusedondecisionfiascoesrather than comparing th edecision-
making
processes
associated with good versus
b ad
decisions.
Experimental studies of groupthink have
considered
only a
small
portion
of the m odel,often without a cohesive group and
in situations inconsistent with Janis's ( 1 9 71 ,1 9 72,1 9 82,1 9 89 )
antecedents.Fu rthermore , they have relied exclusively on stu-
dent samples de aling with hypothetical or simulateddecisions,
with
potential
resultant problems fo rexte rnal validity. Military
strategists,
m anagers, politicians,
orother
real-world decision
makers
have never been used. In the laboratory, many real-
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
7/20
BEY OND
FIASCO
539
world
g rou p characteristics, including ongoing power relation-
ships
an d
political maneuverings, have been necessarily
ig-
nored. Although student samples in laboratory settings may be
valuable
to
address ma ny issues relating
to
grou p problem solv-
ing,
their
use to
examine
groupthinkis
problematic.
Empirical studies of groupthink have tended to use short-
term decision quality as the sole outcome measure and have
contained serious threats
to
internal validity.
On the
basis
o f a
cursory review
ofonly three laboratory studies of groupthink
(Cou rtright, 1978; Flowers,
1977;
Tetlock, 1979), Posner-Weber
(1987)concluded
that
in
the s tudy of group think, each experimenter has tended to
start
fromscratch, either arb itrarily or intuitively deciding what aspects
of
the group think phenomenon
will
be
studied..
. .Janis'stheory
is
largely based on intuition; he is well-informed, but his
state-
ments are not clearly
confirmed,
(p .124)
Interestingly, Janis apparently discounted these experiments.
For instance, Jan is (1989) cited n one of the exp erimen tal stud-
ie sof grou pthin k; his discussion of groupthin k referred only to
case stu dies and ane cdotal p ieces. However, Janis
(1985)
explic-
itly
recognized
the
dangers
of
reliance
on
anecdote
an d
sug-
gested the modus
operandi
approach (Scriven, 1976) as a
more
rigorou s approach to the an alysis of case studies. Stud ies
of
grou pthink that
use
this methodology have
notappeared.
3
Studies
of
groupthink
have focusedon
various combinations
of
Janis's (1972, 1982) original antece dents
an d
were rarely
faithful to the
original conceptualization,
in
which cohesion
wa s
a n ecessary condition. A lso, themajorityof studies did not
operationalize the m odel as Janis(1971,1972,1982)denned i t.
An othe rimpo rtant aspect of prior research conce rns the situa-
tions chosen for study. Th e group think conceptual ization is
restrictive, primarily because of the model's antecedents and
th ein ternation al policy contextforwhich group thinkwa sdevel-
oped. The narrow
focus
of the model's antecedents is not sur-
prising
given
th e
fact that Janis extracted them directly from
th e majorpolicy decisionfiascoes hee xamin ed. Early case evi-
dence
provided
in
support
of
groupthink
wa s
drawn
largely
from
these and similar hot decision situations. The ability to
generalize from
such situations
i s
questionable.
Fo r
ins tance,
th e
e xtreme ly high group cohesiveness
an d
clearly
un ified
goals
in
m a n y
major
military policy decisions
or in
groups facing
a
natural
disaster are unlikely to
often
be encountered in other
settings. Nevertheless, laboratory and survey research on
grou p thin k
ha s typically used settings considerably different
from Janis's conceptualization. Such issues call into question
the appropriateness of most past groupthink research.
Lessons
o f
Prior Groupthink
Research
Our discussion of prior research suggests that conclusions
should bedrawnfrom that re search only with caution. Never-
theless, the following lessons
appear
to emerge from th e
groupthink research.
Genera l Suppor t for the M o d e l
To
assess
th e levelof
overall supp ort
for
groupth i nk ,
it
would
be
tempting
to
conduct
a
meta-an alysis
of
prior research. How-
ever,the
relatively small nu mbe r
of
empirical studies,
t he
vari-
et yof
methodologies used,
an d
methodological
difficulties as-
sociated with many studies make this approach infeasible
(Camp bell, 1986; Park, 1990).
On the basis of our
review,
it seems clear that there is little
supportfor the full groupthink model. In fact, to our knowl-
edge, no study of grou pthink hasfullytestedthe model, and in
no stud y were all results consistent with the m odel. McCauley's
(1989)
review raised doubts abou t
Janis's
(1982)conclusions re-
garding
the prevalence of the antecedent conditions in his ini-
tial
scenarios.
Furthermore,
th e
cen tral variable
of
cohesiveness
has not been
found
to play a
consistent role.
Flowers (1977)
went
so far as to state tha t a revision of Janis's theory m ay be
justified,
on e w hich would eliminate cohesiveness as a critical
variable (p. 895). This suggestion is diame tricallyopposed to
Janis's
(1982)
view that high cohesiveness and an accompanying
concurrence-seeking tendency that interferes with critical
think ing are the central features of grou pthin k (p. 9).
Th e
variable that
ha s
received
th e
most consistent
supportis
the
directive leadership antecedent
(i.e.,
lack
of
tradition
of im-
partial leadership). However, almost by definition, leader be-
haviors that promote the
leader's
own views and do not allow
open exploration of alternatives
will
be associated with
grou p thin k
defects such
as
incomplete survey
of
alternatives
an d
failure
to reexamine preferred and rejected alternatives.
Such domain overlap calls into question research that shows
links
between impartial leadership and decision-making de-
fects.
Evidence
of
Grouping
of
Characteristics
Grou pthink's proposed causal chain is based on the premise
that sets of variables such as symptom s and de fects group to-
gether. However, support for the grou pin g of those characteris-
tics
derives from anecdote, casual observation, and intuitive
appeal rather than
from
rigorous research. Because there has
beenn ofull factor analysisofgrou pthin k variables,it isimpos-
sible
a t
this point
t o
determine whether
th e
groupth i nk
factor
s tructure
is as
hypothesized.
As
previously noted,
Moorhead
and Montanari 's
(1986)
restricted factor analyses
appeared
to
support a simpler model than that presented by Janis (1982).
Thus, sup por t for this aspect of the m odel is lacking.
Presumed
Negative
Outcomes
Grou pthin k has been overwhelmingly viewed as an unal-
layed
evil, leading to
uniformly
negative outcomes. Indeed,
such aviewis
universally implicit
in the
language
of
groupth i nk
(e.g., th e
common references
to
symptoms
of
groupthink,
3
This approach attempts tode te rminet he probable causeo funde-
sirable events w ith high certainty even th oug h the available
data
do not
come
from
control led experiments or quasi-experiments. It involves
formulating
al l of the know n causal sequences that might account for
an
observed outcome
an d
then attempting
to find
which
of
them
ap-
pears
to be
implicated. However, althou gh this
approachis
intriguing,
it
continues to
focus
only on sources of error. Also, because many
causal sequences
may
lead
to
similar consequences, expected causal
sequences may be confirmed when other unexamine d sequences are
actually p laying roles.
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
8/20
540
RA MON J . A LD A G A ND SA LLY
RIGGS
F U L L E R
victims of group think, and defects of grou pthink ) . When
used in groupthink research, such negative terminology can
invite
distortions in responses caused by scale-use tendencies
and related psychometric
difficulties
and can also result in
framing
effects.
Individuals
(whether subjects or researchers) presented with
negatively framed terminology may adopt the readily available
negative
frame
and respond accordingly (Bazerman, 1990;
Tversky
&
Ka hn e m a n ,
1986).
Therefore, even simple attempts
by
the subjects to give responses that are consistent with the
tone of the questions wou ld result in n egatively oriented re-
sponses. In man y cases, faileddecisions are examined and char-
acteristics of group thin k are the n sought. There is evidence that
when individuals are provided with knowledge of a negative
outcome, they
infer
a negative process (Guzzo, Wagner, Ma-
guire,Herr,
&
Hawley,
1986).
Furthermore ,
afocus
only
on the
conjunction of groupthink characteristics and negative out-
comes invites illusory correlation (cf. Ein horn ,
1980;
Hogarth,
1980; Kle inmu ntz ,1990).
On a more fundamental level, thisframing has resulted in a
focus
on error rather than on decision quality per se. Janis
(1982)
noted that he began studyin g fiascoes for the purposes
of
studyingsources of error inforeign policy decision-making
(p .9). However, there is more to the perform ance of a football
team than the absence offumbles and interceptions, and there
ismore
to
group decision quality than
th e
absence
of
error.
A
focus on negative outcomes of group processes may divert at-
ten t ion
from
group synergies.On eexampleis the assemblyef-
fect
bonus, which , as noted by Collins and Guetzkow(1964), is
productivity wh ich exceeds the p otential of the most capable
member and also exceeds the sum of the efforts of the group
m em ber s
workin g separately (p. 58). There is considerable evi-
dence
for
th is assembly
effect
bonus,
at
least
in
some contexts
(e.g.,
Bu rleson, Levine ,
&
Samter,
1984).
Thu s, researchers
ma y
learn lit tle about supe rior group pe rformance by a
focus
solely
on
fiascoes.Instead,a focus on decisions withabroad rangeof
outcomes, includ ing superior performance , isnecessary.
The consequences of the groupthink model's focus on fias-
coes are doubly ironic. First, the consideration only of fiascoes
precludes generalization to other decision situations used in
virtually
all attempts to assess the validity of group think . Sec-
on d, the focus on fiascoes makes it impossible to say any thin g
even
about
th e
de t e rm in a n t s
of fiascoes.
As
our discussion suggests, recent scale-development efforts
such as those of Mon tana ri and Moorhead (1989) are badly
needed. T hat is, Mon tanari and Moorhead have attempted to
develop
psychometrically adequate indexes while removing the
negative
labels associated wit h Janis's
(1972,1982)
variable cate-
goriesand the associated problems.We seethisas anencourag-
in g
a n d
critical development.
Group th ink Interpretation
Strong and weak interpretations of groupthink have
emerged.
The
strong interpretation, commonly
cited,
suggests
that groupthink
is an
integrated
set of
characteristicswith
d e-
termin istic linkages. That is, grou pthin k
characteristics
cluster
because
of
their commo n ties
to
specific anteceden t
conditions
and arelinkedin acausal chain
from
those antecedent condi-
t ionstosymptomstodefectstooutcomes.Th eweak interpreta-
tion is sometimes presented in response to the failureof empiri-
cal examination s to provide results wholly consistent with
groupthink. It suggests that groupthink may be evidenced by
th epresence
of some
subset of
these
characteristics and
that
th e
causal ordering posited by Janis(1972,1982)may be suggestive
rather than necessary.
When the weak in te rpre ta t ion is
presented,
it is crucial to ask
what
characteristics, if
any,
are unique to groupthink.
That
is,
some g rou pth ink characteristics, such as belief in the
Tightness
ofadecisiona n d pressurest oachieve consensus,ar e common
to many group processes and in fact may be beneficial. For
instance, Nemeth (1986)h as argued that convergence to the
majority viewpoint is desirable if the majority viewpoint is
correct. Oth er characteristics, such as mindgu ards and self-cen-
sorship, better discriminate between groupthink an d other
group processes. It isdifficult to accept the argument that the
presence
of
characteristics common
to
most problem-solving
group s is supportive of the existence of gro up thin k.
Janis
(1989)
apparently did not accept this weak
form
of
group think. To the contrary, he wrote that
it
does not suffice merelyto see if a few of the eight te ll ta le symp-
toms
o f
g r o u p t h i n k
can bedetected.
Rather ,
it is
ne cessary
to see
if
practically all the symptoms were man ifes ted and alsoto see if
th e antecedent condit ions and the expected immediate conse-
quencesthe symptoms of defective decis ionmakingare also
present. (Janis, 1989,p. 60)
4
Comprehensiveness
Several researchers and theorists have suggested that the
groupthink model isincomplete as amodel of group problem
solving. Some have proposed specific additions, as we discuss
later, whereas othe rs have noted that other un explore d variables
may
be playing roles (Courtwright, 1978; Moorhead & Monta-
nari, 1986). Amo ng the proposed additions are norms, leader
power, task characteristics,
an d
stage
of
grou p
development.
5
The importance of these factors is evidenced both in the
groupthink research and in the more general literature on
group problem solving.
Norms. Moorhead (1982)noted after a brief review of the
research relating cohesivenesstogroup pe rformance ,an d con-
sistent with the prior arguments of Seashore (1954)and others,
that group normswill moderate the influence of cohesiveness
on performance . Wh en the normsfavorhigh pe rformance, co-
hesiveness should enhance performance. When the norms fa-
vo rlow pe rforman ce, cohesiveness should lower performan ce.
This suggests that examination only of cohesiveness per se,
4
From
Crucial Decisions: Leadership
in
Policymaking
an d
Crisis
Management (p. 60) byI r v ingL.Janis,1989, New\brk:Th eFree Press.
Copyright 1989 by The Free Press. Reprinted bype rmission of The
Free Press,adivisiono fMacMillan Inc .
' A l t h o u g h
others (e.g., Luechauer, 1989; Whyte, 1989) have sug-
gested
additional variables, those variables were
presenteda s
alterna-
tive
explanations fo r g r o upth ink r a th e r th a n as supplements to the
model.
Fo r
instance, Luechauer argued that self-monitoringpropen-
sity
may be
important ,whereas Whyte emphasized
th e potent ial
role
ofprospect
polarization.
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
9/20
BEYOND FIASCO
541
rather than asking
Cohesiveness
to what? , may be misleading.
Furthermore, Gladstein and Reilly
(1985)
reasoned that the
natu re of group norm s might have prevented the constrictionof
control they had hypothesized. They noted that working as a
team was astrong n orm in theschool inwhicht hestudywa s
performed
an dthatthegroupsin thestudy appeared toadhere
to that norm, making decisions as a group and striving for
consensus.
The
need
t o
consider norm s
ha s
also bee n widely recognized
an d
evidenced
in the
more gen eral group problem-solving liter-
ature. For instance, Bazerman, Magliozzi, and
Neale
(1985)
noted how norms
influence
group members' interpretation of
problems
an d
the ir subsequent attention
to
distributive
o r
inte-
grative
solutions. Similarly, Bazerman (1990) discussed
how
norms influence
the choice of third-party interven tionproce-
dure s in grou p decision con flict as well as memb er risk
prefer-
ences. Furthermore, Howell and Frost (1989) demonstrated
ho w
norms and leadership style may interact to
influenceper-
formance
and othe r outcomes in group decision making. Thus,
norms mayplaypervasive roles in group problem solving and
should be explicitly considered.
Leaderpower . Flowers
(1977)
wrote that
on e
pote ntial ly important factor
n ot
dealt with
b y
Janis
was the
degree of power held by the leader over group members. Janis's
groups had
powerful
leaders who could exercise reward andpun-
ishmenta swella s
legitimate, expert,
an dperhaps
referent power
. . . over their m embers, whereasthe
leaders
in this experiment
held
on ly certain legi timate power. Add ing
the
variable
o f
power
to the operational defini tion of group think m ight create in highly
cohesive groups an increased thru st toward grou pthink (p.895).
Similarly,
McCauley (1989),
in his
review
of the Bay of
Pigs
decision, concluded that compliance with
a
group norm
pro-
mulgated by a
powerful
and attractive groupleader (p .
254)
contributed to
faulty
decision making.
The group problem-solving literature has regularly noted the
role of leader power. For example, Huber (1980) argued that a
powerful
leader
i s
needed
to
provide
th e
opportuni ty
forcon-
trol
if
problem-solving groups be gin
to
engage
indysfunctional
behaviors. Clearly, th e
degree
an d
locus
of
power
in a
group
setting
may have
important consequences
for
group decision
processes
a nd
outcomes.
Task
characteristics.
Callaway
a nd
Esser (1984) reasoned
that
the
nature
of the
task
may be
important .
In
particular, they
argued tha t a e ure ka-type task, in which a single correct answer
becomes immediately obvious once insight is achieved by any
group member, may be inappropriate in atestof groupthink.
This is because the grou p interaction necessary for groupthin k
characteristics to occur is circumven ted. Thus, whether the task
requires a pooling of inpu ts, or interaction atall,may be impor-
tan t. Fu rthermo re, Gladstein and Reilly (1985) noted that the
task the y used allowed for a high degree of specialization and
that
it was
t h u s difficult
for one
person
to
make
all of the
decisions. This suggests that characteristics such
as
distribu-
tion of
information
and task comple xity, aswell as specializa-
tion per se, may be relevant task characteristics.
Task
characteristics have
often
been examined
in
group
problem solving. For instance , researchers have considered how
group
influencesaffect
decision performance
in
additive, com -
pensatory,conjun ctive,a nd
disjun ctive tasks (e.g., Stein er,
1976;
Weldon
& Gargano, 1985). Similarly, the literature on such al-
ternatives to traditional interacting groups as the nominal
group technique and the Delphi process suggest that group
decision processes should be fitted to the task (e.g., Stumpf,
Freedman,
&Zand,1979).
Stage
of group development.
Lean a (1985) argued that
th e
stage of group development may be relevant because i t may
moderate the role of cohesiveness. She
noted,
bui lding on
Longley and Pru itt (1980), that ad hoc group s with no p revious
experience together may be highly susceptible to groupthink
symptoms because ofinsecurity concerning me mber rolesa nd
group norms. However, groups with a tradition of working to-
gether
ma y
exhibit
far
fewer symptoms
of
defective decision
makingbecause membersa resecure enou ghintheir rolesand
status
to challenge one another but have also developed ways of
reaching agreement. Thus,
th e
potential benefits
an dcostsof
alternative techniques
fo r
grou p problem solving, including dis-
sonance-induction procedures, may depend on the stage of
group development .
AGeneral Problem-Solving Perspective
In
the p revious section, we highlighted several importa nt fac-
tors emergingfromresearch on group think that merit incorpo-
ration in a general model of group problem solving. In each
case,
we also noted exp licit lin ks to the problem-solving litera-
ture. In this section, we suggest how a problem-solving perspec-
tive
may be
fur ther
used to guide the development of such a
general model. In so doing, we first indicate how the frame-
work
of the p roblem-solving process may be used to recast an d
expand on the grou pthin k model. We then discuss additional
sets of variables suggested by the literature related to problem
solving
and question certain assumptions of the groupthink
model.
ProcessOrientation
The
large literature relating
to the
problem-solvingprocess
suggests that m any group think variables
are
relevant
bu t
that
a
recasting and expan sion is appropriate. Problem solving may
be
viewed
as a
m ultistage process that includes problem identi-
fication, alternative generation, alternative evaluation and
choice, decision implementation, an d decision control (Bass,
1983;
Elbing , 1978). Explicit recognition of these stages is im-
portant for several reasons.
First, group think defects
can be
vie wed simply
as difficulties
occurring
through the stages of the problem-solving
process.
That is, the incomplete su rvey of objectives represen ts a
fail-
ure
at the
problem identification stage,
th e
incomplete survey
of alternatives is a failureat the alternative generation stage,
the failure to examin e risks of preferred choice is afailureat
the evaluation and choice stage, and so on.
Second, other variables not explicitly noted by Janis are
knownto be
important
in the
problem-solving
process.
These
include explicit problem
definition,
qu al i ty
of
alternatives gen-
erated, source of the solution, group decision rule, gathering of
control-related information,
an d
t i mi ng
of
solution conver-
gence. Conce rning
th e
last variable,
t he
decision stages
differ in
th e
degree
t o
which they rely
on
divergent
or
convergent think-
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
10/20
542
RAMON J. ALDAG AND SALLY
RIGGS
FULLER
ing.
For instance , the a l terna t ive genera t ion s tage i s p r im ari ly
divergent, whereas the evaluation and choice stage is
largely
convergent.
A s
such, premature convergence, ra ther than con-
vergence per se, isun desi rable .
Th i r d ,
t he
va r i o us
stagesof the
proble m-solving process
m ay
requi re
different
social arrangements,
an d
those appropriate
at
on e
stage
may be
troublesome
at
another .
Fo r
instance,
al -
though interac t ion may be
facilitative
a t the a l tern a t ive evalua-
tion an d choice stage, i t is often dysfu nct iona l at the al ternative
generation
stage
(D. WTaylor, B erry,&B lock,
1958;
V a n de Ve n
&
Delbecq,
1971;
Vroom, Grant , & Cot ton, 1969).
Four th , i t becomes apparent tha t groupthink defec ts focus
primari ly on the first
three stages
of the
problem-solving pro-
cess, en ding with evaluation
an d
choice. Janis
(1972,1982)
gave
less em phasist o implementa t ion an d control . Onlythe failure-
to-work-out-contingency-plans symptom relates
to the
imple-
menta t ion s tage,
an d
control
is not
exp lici tly considered. This
focus
is restrictive. For instance, in the event of a
faulty
deci-
sion,
if
i mp l emen ta t io n
has not
been considered,
it
w o ul d
be
impossible
to sort ou t the relative roles of
faulty
decision mak-
in g
an d
faulty
implementation.
As an
example, Janis (1985)
cited the
ill-fated
a t tempt by President C ar ter to use mi l i ta r y
force
to rescue the Ame rican hostages in I r an as an e xamp le of
th e
u n s q u e a k y
wheeltrap (p .
81).
That is, Janis reasoned that
th e
decision makers neglected
to
consider
th e
possibility
of
early abortion of the rescue mission because the squeaky
wheel on which a t tent ion was focused was the obvious dange r
once
t he
m ission reached Tehran.
Infact, it is not
c lear whethe r
t h i sreflectsfaultye valua t ion, lack of a t tent ion to impleme nta-
t i o n ,
or
bo th .
Finally,a n
explici t
focus on the
full
problem-solving process
emphasizes
the
importance
of multiple-decisionoutcomes.For
instan ce, alon g with decision qu ali ty, the problem-solving l i tera-
t u r e
stresses
t he
need
t o
consider decision
acceptability,
satisfac-
t ion with
th e
decision process,
and so on
(e.g., Huber, 1980).
T he
l i tera ture
fur the r
indicates that
the
degree
of
g r o up member s '
i n p u t s to
problem solving,
a n d
w h e th er
t he
member s
feel
tha t
the ul t imate dec ision reflects th e i r i np u t s ,
influences
these ad-
dition al outcome s (e.g., V room & Y etton , 1973).
Lessons
From
Related
Literature
The
problem-solving literature
suggests
other
elements that
should be recognized in developing an d interpret ing a more
general model . Those elements include organizational power
an d
poli tics, group
cohesiveness,
social control,
an d
directive
leadership.
Organizational
Power
and Politics
The l i tera ture concerning organiza tional poweran d
politics
isan
integral part
of the
study
of
organizational decision mak-
ing. Organizations have been described
as
political
coalitions
(March, 1962),
an d
Pfeffer
(1981)
stated
that
power,
influence,
an d political activityare all inevitablea n d important e lements
of admin istrative activi ty. Ofte n, th e development and use of
powera reevidenced not byovert confronta t ionsbut by exhibi-
tions
of acquiescence. One
type
of political
alignment
is the
politics
of
patronage,
in
which l i t tle
or no
conflict
is
manifest
becau se of the ab il i ty of powerful forces to smooth over
differ-
ences and ga in consen sus (Lawler & Bacharach,
1983).
Viewed
in
this l ight , cer ta in grou pthin k charac ter is t ics
may
reflect
n ot
passive
a n d
maladapt ive responses
to
stressful antecedent con-
ditions but active, conscious
efforts
to a t ta in personal out-
comes. Thus,a broader problem-solving perspective would in -
clude considera t ion
o f
such proactive poli tical
action.
G roup
Cohesiveness
Group cohesiveness isseen as a primary antecedent condi-
t ion for grou pthin k and i s know n to resul t in dysfunc t ion al
forms
of
conformity
behavior. However, there
is
also consider-
able evidence that h igher levels
of
cohesiveness
may
have
a
vari-
e ty
of desi rable consequences. These potent ia l bene f its inc lude
enh a n c ed c o mmun i c a ti o n a mo ng member s, h i g h er
m e m b e r
sat-
isfaction,
decreased member tension
an d
anxiety,
an d
higher
levels of
group task accompl ishment (Shaw,
1981).
Fur ther-
more, i t wou ld appe ar th at si tuation al, task, and goal character-
istics may
modera te
th e
desirabil i ty
of
cohesiveness
in
g r o up
problem
solving
and
that positive effects
should not be
ignored.
Social
Control
Social con trol hasbeen d ef i neda s social a r r a ng ementse m-
ployed
to
keep
the
behavior
of
some people
in
l ine wi th
th e
exp ectation s of others (He witt & Hewitt, 1986, p.
155). Infor-
mal
social controls
a re
unofficial , subtle pressures
to
c o nf o r m
to
no r ms . Nemeth
an d
Staw (1989) identified
five
ways
in
which
social control is exercised to achieve
uniformity.
Several
of
these mechanisms
are
p ar t icular ly relevant
to thediscussion
of
groupthink
and
group problem
solving.
First,
conformity
in
ag r o upm ay
reflect
a g r eementon the solution that most closely
a l igns
with
establ i shed norms. Second,
powerful
individuals
often have the abi l i ty to achieve conform i ty to the i r posit ions.
Finally,
u n i f o r m i t y
can be the
result
of a
more
tacit form of
in f luence , tha t of agreement wi th majori ty viewpoints .These
th r ee
forms
of social control can serve as al ternative explana-
t ions for the occurren ce of groupthink's symptoms and defec ts.
Nemeth and Staw
(1989)
also noted that con formit y has
both necessary and desi rable elements , par t icular ly wi th regard
to
a t t a inme nt
of
goals
a n d
h a r mo ny
(p .
175).
W h e n
th e
no r m
of
the
organization
or the
wishes
ofthosein
power
isadopted,
th e r e
is
less strain
on individuals as
well
as
greater efficiency
becauseth eg r o u pca nmove toward it sgoals. Howe ver, Nem eth
and Staw also noted that al though conformity may serve
useful
funct ions ,
it may
also
stifledissent.
They wrote
that there
is
evidence tha t d issent , even when erroneous, con tr ibutes
to the
detection
of
t r u t h
and to the
i mp r o vemen t
of
both perform ance
and decision m akin g (Ne me th & Staw, 1989, p.
196).
Thus, th is
l i terature suggests that th e d e t e r mi na n t s an d roles of confor-
mi ty
in
group problem solving
a re
more complex than
is
sug-
gested by the g r o up th i nk mo d el .
Directive
Leadership
Directive leadership
is
viewed
as an
antecedent condition
of
groupthink
and
thus
of
negative
decision outcomes.
The
possi-
ble effectsof directive leadership have been widely examine d in
-
8/10/2019 (1993) Ramon J. Aldag and Sally Riggs Fuller , Beyond Fiasco a Reappraisal of the Groupthink Phenomenon and a
11/20
BEY OND FIASCO
543
the problem -solving literatu re, perhap s most notably by Vroom
an d
Yetton (1973).
Th e
model presented
by
those
authors, as
wella srelated resear ch (e.g., Field, 1982; Vro om&Jago, 1978),
indicates that autocratic styles sometimes result
in
effective
de -
cision making. Furthermore, situational theories of leadership,
especiallypath-goal theory (House &Mitchell, 1974), argue
that directive leader behaviors may in some instances result in
enhanced group performance and member satisfaction.There
is
considerable empirical support
for
such views (e.g., Schrie-
sheim & Von Glinow,1977).Thus,itshouldnot beassumed that
directive leadership isin heren tly dysfunctional.
The ben efits of leader impartiality are also challenge d by the
literatureontransformational leadership(e.g.,Bass,1990).That
literatureargues
that
successful leaders commu nicate theirvi-
sion
to followers and inspire and stimulate them toward the
accomplishment of
that vision.
Finally, itappe ars that Janis's lack oftradition of impartial
leadership antecedent ha sbeen used invarious,
distinctways
in gr oup think research, oftenconfoundi ng
th e
leader's pro mo-
tion of a correct decision process with promotion of a preferred
outcome. Althoughth eformerisoften espoused asdesirablefor
the p roper function ing of a decision group (e.g., Huber, 1980),
th e
latter
m ay
inhibit member inputs.
Examining Past Assumptions
Earlier,
we
presented several implicit assumptions
of
groupthink. Our review of the groupthink research and the
problem-solving literature poses several theoretical challenges
to those assumptions.
First, although grou pt hin k assumes that groups are used p ri-
marilytoenhanceth equalityofdecisions,group decision m ak-
ing isoften
used specifically
to
increase
th e
acceptance
of a
decision P f e f f e r , 1981). Also, organizational members ma y
choose
to use
groups
for
other
reasons,
including
socioemo-
tional benefitsan d accomplishment ofmembers' ow nsecond-
ary goals
(e.g.,
hidden agendas).
Second, the assumption that the group process is a rational
pursuit
of a
unified goal
is
inconsistent with
the
known com-
plexities of grou p processes. Such processes are characte rized
by interplays
of
covert
an d
overt motivations, concern
not
jus t
about the current problem but also with residues of past prob-
lems and the
anticipation
of
future problem-related interac-
tions,an d man y other subtleties outsideof the typical
focus
of
th e
rational mode l. Processes
a nd
outcomes that app ear irratio-
nal to objective observersmay in
fact
be fun ctional from th e
viewpoint
o findividual membersa n devenof theentire group.
For instance, such processes an d ou tcomes may serve to m ain-
tain the motivation of the group leader, help ensu re the contin-
ued use of group processes, prevent th e defection of certain
group members, serveas a means formembers to show alle-
giance to coalitions, an d min imize the likelihood of the use of
still other p olitical acts. Organizations are inher ent ly political,
with
shifting
coalitionsan dinte rest groups,adiversityofgoals,
an d
active
use of
power (cf. Hardy,
1985;
Lawler
&
Bacharach,
1983;Pfeffer,
1981).According
to
thisview,political activity
is
not