Evolution of the interpersonal conflict paradigm

Text-only Preview

Judgment and Decision Making, Vol. 3, No. 7, October 2008, pp. 547–569
Evolution of the interpersonal con?ict paradigm
Mandeep K. Dhami?
Henrik Olsson
University of Cambridge
Max Planck Institute for Human Development
Using Brunswik’s (1952) lens model framework, Hammond (1965) proposed interpersonal con?ict theory to explain
the nature, source, and resolution of disagreement or “cognitive con?ict” between parties performing judgment tasks. An
early review by Brehmer (1976) highlighted the potential of this approach in, for example, understanding the structure
of cognitive con?icts, and the effect of task and person variables on judgment policy change and con?ict resolution.
However, our bibliographic and content reviews from 1976 to the present day demonstrate that research on cognitive
con?ict using the lens model has declined sharply, while research on “task con?ict” has grown dramatically. There
has also been a shift to less theoretical precision and methodological rigor. We discuss possible reasons for these
developments, and suggest ways in which lens model research on cognitive con?ict can be revitalized by borrowing
from recent theoretical and methodological advances in the ?eld of judgment and decision making.
Keywords: interpersonal con?ict theory, lens model, cognitive con?ict, disagreement, task con?ict, cognitive continuum
theory, simple heuristics.
ent from con?ict caused by motivational and value dif-
ferences among parties, cognitive differences can evolve
It was during the cognitive revolution in psychology and
into motivational and value-laden con?icts. Under these
the cold war period in political history when Hammond
circumstances, the underlying cognitive differences can
(1965) proposed that con?icts between parties perform-
be very dif?cult to detect and resolve.
ing judgment tasks could be viewed as purely cogni-
An early review of research using IPC theory published
tive, thus making it unnecessary to examine the moti-
in Psychological Bulletin by Brehmer (1976) highlighted
vations and values of con?icting parties as social psy-
the potential of this approach in advancing our under-
chologists might do. In interpersonal con?ict (IPC) the-
standing of cognitive con?ict in both laboratory and real
ory, Hammond (1965) outlined how this cognitive con-
world settings. Despite this, since that time, IPC the-
?ict could be construed within Brunswik’s (1952) lens
ory appears to have featured little in the growing ?eld
model framework, as well as the experimental methods
of judgment and decision making (JDM). For example,
that researchers could use to study the nature, source,
in a historical review of theories in the ?eld, Goldstein
and resolution of disagreement between parties perform-
and Hogarth (1997) provide only a passing mention of
ing judgment tasks. Brie?y, cognitive con?ict represents
IPC theory when considering developments in judgment
differences in how parties conceptualize the solution to
research. The need for con?ict theories, including cog-
a problem. For instance, different parties may have dif-
nitive con?ict, seems apparent in an era characterized
ferent policies for solving a judgment problem in terms of
by international terrorism where, amongst other things,
the information they rely on. Inconsistency in how parties
parties disagree about the level of threat, and how to
apply their judgment policies can also lead to disagree-
manage and minimize it (Mandel, 2005), and in an era
ment. Thus, according to IPC theory, parties striving to
characterized by a movement towards greater use of al-
make a joint judgment on the same task could con?ict be-
ternative dispute resolution. In fact, today, a theory of
cause they disagree in principle (in that they have differ-
cognitive con?ict could bene?t from recent theoretical
ent policies for how to solve the problem) and/or in prac-
and methodological advances in the ?eld of JDM. For
tice (in that they are inconsistent in the application of their
instance, JDM researchers have shown that individuals
policies). Importantly, while cognitive con?ict is differ-
are likely to use simple process models when performing
judgment tasks (e.g., Dhami & Harries, 2001; Garcia-
?We would like to thank Robin Hogarth, Jeryl Mumpower, Jon
Retamero & Dhami, in press; Rieskamp & Hoffrage,
Baron, and an anonymous reviewer for their helpful comments on an
1999); and that non-cognitive factors such as emotions
earlier draft of this paper. Send correspondence to Mandeep K. Dhami,
University of Cambridge, Institute of Criminology, Sidgwick Avenue,
may affect how individuals make judgments (e.g., see
Cambridge, England, UK, CB3 9DT. E-mail: [email protected]
Loewenstein & Lerner, 2003). In addition, researchers

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
Achievement for Party 1
have employed new tools such as virtual environments
and computer simulations when studying judgment be-
havior (e.g., see Brehmer, 1992; Mosler, Schwarz, Am-
Cue utilization
validities, Party 1
mann, & Gutscher, 2001).
Our goal is to evaluate the evolution of IPC theory from
Party 1’s
its inception to the present day. Speci?cally, we con-
sider how research on cognitive con?ict has developed in
terms of its theoretical underpinnings and methodologi-
cal stance, and we review the ?ndings of empirical re-
search on cognitive con?ict. Our goal is modest in that
Parties 1 and 2
we focus our efforts on cognitive con?ict as it directly
emerged from the IPC paradigm and related lens model
Party 2’s
framework. A review of con?ict theories and research
more generally are not within the scope of the present pa-
per. The article is organized into three main parts. First,
Cue utilization
validities, Party 2
we consider the emergence of IPC theory from 1965 to
1976 by outlining its roots in Brunswikian psychology,
Achievement for Party 2
the experimental methodology employed, and early re-
search ?ndings. Second, from 1976 to the present day, we
Figure 1: Lens model for study of interpersonal con-
trace the evolution of IPC theory and cognitive con?ict
?ict and interpersonal learning (adapted from Hammond
research by conducting bibliographic and content reviews
[1965] and Hammond et al., [1966b]).
of publications that cite central articles by Hammond
(1965) and Brehmer (1976). Finally, we discuss the fu-
ture of IPC theory and cognitive con?ict research by con-
tation of the lens model to the study of cognitive con?ict
sidering opportunities for theoretical advancement and
(simpli?ed for our purposes; see also Cooksey, 1996).
methodological innovation offered currently in the ?eld
For readers unfamiliar with this framework it is worth
of JDM. We hope these will inspire future researchers.
pointing out that the model shows a collection of cues
diverging from a criterion in the environment, and these
2 Interpersonal con?ict theory from
cues can be used by the different parties to predict the cri-
terion. To the extent that a party’s cue utilization validi-
Hammond, 1965 to Brehmer, 1976
ties match the ecological validities of the cues, the party
will be able to achieve the criterion (i.e., make accurate
In this section, we review the development of IPC the-
decisions). Con?ict can also occur in the absence of an
ory from 1965 to 1976. We consider the roots of Ham-
outcome criterion, and to the extent that the cue utiliza-
mond’s (1965) IPC theory in Brunswik’s (1952) lens
tion validities differ across the different parties they will
model framework, the experimental methods proposed
be in con?ict (i.e., disagree in their decisions). In the real
to study cognitive con?ict, and the main ?ndings of the
world, the environment is often complex in that there are
early body of research on cognitive con?ict as reviewed
multiple, inter-correlated cues that are only probabilisti-
by Brehmer (1976).
cally related to the criterion.
Analysis of cognitive con?ict involves comparing the
Cognitive con?ict and the lens model
cognitive systems of the con?icting parties i.e., the right
side of the lens model shown in Figure 1. In situations
Using Brunswik’s (1952) lens model framework as a ba-
where there is no outcome criterion analysis would be
sis for theory and method, Hammond (1965) introduced
restricted to the right side. There could of course be
IPC theory for understanding the nature, source, and res-
more than two parties in which case the model would in-
olution of cognitive con?ict.1 Figure 1 presents an adap-
clude N-systems on the right side (Cooksey, 1996), and
a party could also refer to a dyad or group of individuals
Hammond (1965) applied the earlier developed multiple cue proba-
bility learning paradigm based on the lens model framework (Hammond
(Rohrbaugh, 1988).
& Summers, 1965) and the technique of cognitive feedback (Todd &
The lens model equation shown below (Tucker, 1964;
Hammond, 1965) to judgment in social situations, namely con?ict sit-
uations (see also Hammond, Wilkins, and Todd [1966b] for the related
see also Cooksey, 1996) details how a comparison of two
study of interpersonal learning). These historical antecedents of IPC
theory differ from those noted by others. For instance, Brehmer (1976)
claimed that IPC theory was guided by the conceptual framework of
1975), and Mumpower and Stewart (1996) stated that cognitive con?ict
social judgment theory (Hammond, Stewart, Brehmer, & Steinmann,
research was rooted in cognitive continuum theory (Hammond, 1996a).

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
cognitive systems can be formally done:
According to Hammond (1965, and later Brehmer,
1976), in cognitive con?ict research, the researcher’s task
is to measure the nature and extent of con?ict between
A = GR1R2 + C
(1 ? R2)
(1 ? R2)
parties; document their efforts to agree; measure the na-
This equation points out that agreement between par-
ture and extent of compromise/resolution; measure the
ties, r
nature and extent of changes in the cognitive systems
is a function of two components, namely
of con?icting parties; and document the effect of task-
1R2, which is the linearly predictable component
(when using multiple linear regression analysis) of each
and person-related factors on con?ict, compromise, and
party’s judgments contributing to overall agreement, and
change. Such analyses are not only of theoretical import,
(1 ? R2)
(1 ? R2), which is the unmodeled com-
but can also contribute to strategies for dispute resolution.
ponent of each party’s judgments contributing to overall
agreement. Equation 1 can be, and often is, reduced to
Methodology for cognitive con?ict re-
the ?rst component if one assumes that the unmodeled
search in the lens model framework
component of agreement is zero. Policy similarity is mea-
sured by G, while R1 and R2 are measures of each party’s
From the perspective of IPC theory, the method used
cognitive control over their judgment policies. The inter-
to study cognitive con?ict involves experimentation
pretation of C is more dif?cult as it could refer to several
(Brehmer, 1976; Cooksey, 1996; Hammond, 1965; 1973;
things such as the extent to which both party’s policies are
see also Rohrbaugh, 1988, for group-based research
similar but unmodeled, the extent to which both party’s
The standard experiment is divided into a
policies are different and unmodeled, or a lack of unmod-
training stage where parties are trained to think differ-
eled response variance in one or both parties.
ently about a judgment task (i.e., develop a different set
Con?ict may be due to systematic and non-systematic
of cue-dependencies), and a con?ict stage where the par-
cognitive differences in the way parties solve the problem
ties are brought together to attempt to arrive at a mutually
(Brehmer, 1976). Systematic differences refer to stable
agreeable solution to the problem.4 More speci?cally, af-
or predictable features of policies such as differences in
ter each party has learned to solve the task alone they are
relative cue weights, form of function relating cue val-
brought together, unaware that they have different poli-
ues to judgments, organizing principles (i.e., how cues
cies. The parties are then asked to co-operate on solv-
are combined), and policy consistency/cognitive control.
ing another set of problems which are actually different
Here, the lack of policy similarity means that parties may
from the ones they each learned.5 On every trial or judg-
disagree both in principle and practice. Non-systematic
ment problem, they study the available information and
differences introduce randomness or unreliability into the
make judgments of the criterion variable alone and then
application of policies. Here, the lack of cognitive control
communicate these to one another (overt individual judg-
means that parties may disagree in practice even though
ment). If they disagree, they must discuss the problem
they agree in principle (false disagreement) or they may
until they reach an acceptable joint response (joint judg-
agree in practice even though they disagree in principle
ment). They are then asked to reconsider their original
(false agreement; Hammond & Grassia, 1985).2
decisions, and these revisions remain private (covert in-
The nature and extent of the con?ict may change as
dividual judgment). Finally, if there is an environmental
parties interact with each other and the task, thus high-
criterion, they are presented with the correct solution. So,
lighting the importance of studying interpersonal learn-
the parties must adapt to one another as well as adapt to
ing and task characteristics when understanding cogni-
the task in order to agree and achieve.
tive con?ict. Indeed, an individual’s ability to learn about
The researcher can precisely de?ne and manipulate the
another person’s behavior is central to con?ict resolution
quantity and quality of cognitive differences, and ob-
(Hammond et al., 1966b), as is his/her ability to learn
jectively measure cognitive con?ict, compromise, and
about the characteristics of the task.3
sion, as would be done in a study on cognitive con?ict, participants are
2Mumpower and Stewart (1996) point out that systematic differ-
asked to predict the other person’s response. Comparison of the predic-
ences in policies may be due to missing or poor feedback, missing or
tion with the other person’s actual response provides a measure of in-
poor information, bias in subjective evaluations of one’s own policy, and
terpersonal knowledge. Research on multiple-cue probability learning
redundancy of information. Non-systematic differences may increase
demonstrates how people learn about the task (Hammond & Summers,
when the task is unpredictable or requires use of a large amount of in-
1965). Characteristics of the task that individuals must learn include the
formation, especially in a nonlinear way, and when the party is learning
cues, cue values, cue distributions, cue inter-correlations, and ecologi-
to solve a novel task or a familiar task in a novel way. (See also Ham-
cal validities (Dhami, Hertwig, & Hoffrage, 2004).
mond’s [1996a] cognitive continuum theory for how task characteristics
4Participants may also be selected because they have con?icting
may in?uence judgment).
policies, and so the training stage is eliminated.
3Research on interpersonal learning is also conducted within the
5Alternatively, the parties may be presented with a set of problems
lens model framework shown in Figure 1 (Earle, 1973; Hammond et
that only one party has learned, thereby requiring the other party to
al., 1966b; Hammond, 1972). Here, instead of making a joint deci-

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
change. Furthermore, the researcher can add complex-
First, con?ict may persist due to non-systematic cogni-
ity to the experiment by, for example (as Brehmer, 1976,
tive differences even when parties are motivated to agree,
noted), introducing payoffs, manipulating feedback, and
and actually do agree in principle. Indeed, while parties
involving groups. Thus, although this paradigm may not
reduce the systematic differences in their policies (i.e.,
fully represent all relevant features of what are typically
there is policy similarity), over time the inconsistency of
complex problems, it can provide a reasonable analysis
their policies increases thus leading to little reduction in
of some de?nable aspects. As such, Brehmer (1976) and
the amount of con?ict although the structure of the con-
Hammond (1965) both claimed IPC theory may be used
?ict has altered (e.g., Brehmer, 1969). This is because
to guide research into real world con?icts.
parties tend to decrease their dependency on their old
The basic data collected from a typical experiment in-
policies at a faster rate than they increase their applica-
cludes the joint judgment, and the overt and covert indi-
tion of a new policy that is compatible with each others’
vidual judgments made by the parties before and after this
(e.g., Brehmer, 1972).
(Hammond, 1965). As Hammond (1965) noted, these
Second, policy change itself does not signify will-
measures can be used to study the extent and nature of
ingness to compromise but rather a desire to achieve,
cognitive con?ict, compromise, and change with respect
although compromise is sought when accuracy is not
to the task, and with respect to the other party. (There
clearly observable/obtainable. When one party is initially
are overt and covert measures of compromise, con?ict,
trained in the optimal policy and the other is not, the latter
and change). For instance, con?ict can be measured by
will learn from the former if the task is highly predictable
comparison of each party’s overt individual judgments.
(e.g., Brehmer, 1973a). However, if task predictability
A comparison of each party’s overt and covert individ-
is low, the parties start off by decreasing dependency on
ual judgments and the joint judgments provides a mea-
their initial policies. Here, based on feedback, the party
sure of compromise at the overt and covert levels, re-
with the optimal policy soon appropriately switches back
spectively. Furthermore, a comparison of each party’s
to his/her original policy, and the other party also learns
(overt and covert) individual judgments and the criterion
from feedback (e.g., Brehmer, 1974). When there is no
(where available) and the other party’s judgments pro-
feedback, parties may compromise: this reduces con?ict
vides a measure of cognitive change with respect to the
without leading to observable inaccuracy (e.g., Brehmer,
task and other party, respectively. Hammond (1965) also
pointed out that the measures could be derived on both an
Third, formal (surface and system) task characteristics
inter- and intra-trial basis (i.e., comparison of each party’s
can in?uence each party’s policy development and the
responses averaged across trials or comparison of each
ease with which they can achieve, and such characteris-
party’s response on each trial, respectively), and that anal-
tics alone can explain cognitive con?ict. Hammond and
yses could examine both external and internal dynamics
Brehmer (1973) did not ?nd much evidence for substan-
such as the effect of interpersonal learning (Hammond et
tive or content task characteristics in?uencing cognitive
al., 1966b) and feedback (Todd, Hammond, & Wilkins,
con?ict. Surface characteristics refer to the number of
1966). Indeed, the early research conducted by Ham-
cues, the metric level of cues, and the inter-cue correla-
mond and colleagues focused on such topics (Hammond
tions, while system characteristics refer to the distribu-
et al., 1966a).
tion of cue validities, forms of functions relating cues to
the criterion, organizing principles, and task predictabil-
Early ?ndings of cognitive con?ict re-
For example, there is greater agreement despite
less reduction of policy differences when the cues are
search in the lens model tradition
inter-correlated than when they are orthogonal (Brehmer,
After Hammond’s initial research on cognitive con?ict,
1975). This may be because cue inter-correlations enable
Brehmer and colleagues conducted an intensive series of
the parties to achieve with little change of their original
studies. In 1976 Brehmer reviewed the research that had
policies (Mumpower & Hammond, 1974). In addition,
been conducted on cognitive con?ict using IPC theory.
there is less agreement between parties when task pre-
By then, research had examined issues concerning: (a)
dictability is low because each party’s policies are less
the structure of cognitive con?icts; (b) the relative im-
consistent rather because of any systematic differences
portance of the task and the other party in affecting pol-
in their policies (e.g., Brehmer, 1975). Similar ?ndings
icy change and con?ict resolution; (c) the effect of task
have been observed for tasks that require policies with
characteristics on cognitive con?ict; and (d) the effect of
nonlinear function forms which tend to be more dif?cult
person characteristics on cognitive con?ict. Research had
to develop (e.g., Brehmer, 1973b).
also begun to study (e) how cognitive con?ict could be re-
Fourth, traditional individual difference variables such
solved via supports/aids. We describe the main ?ndings
as gender do not affect measures of cognitive con?ict
(Hammond & Brehmer, 1973).

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
Finally, cognitive aids may be useful for reducing con-
of the lens model that are guiding cognitive con?ict re-
?ict. Hammond and Brehmer (1973) applied the tech-
search today? To answer these questions, we used a com-
nique of cognitive feedback (Todd & Hammond, 1965)
bination of bibliographic and content reviews of publica-
and developed a cognitive aid to con?ict resolution called
tions since 1976 that cite the central articles by Hammond
POLICY.6 This interactive computer program enables
(1965) and Brehmer (1976). Thus, we focus on cogni-
parties to express their policies, compare them, change
tive con?ict research as it directly emerged from the IPC
them, and discover the effects of such changes on con-
paradigm and related lens model framework. While the
?ict (see Rohrbaugh, 1988, for group decision support
content review can shed light on the theoretical, empiri-
systems). Cognitive feedback involves providing infor-
cal, and methodological contributions made since 1976,
mation about the task (i.e., ecological validities, inter-
the bibliographic review indicates the “in?uence” or “im-
cue correlations, predictability, and cue-criterion func-
portance” of the contributions. The bibliographic review
tion forms), the party’s judgment policy (i.e., utiliza-
also helps us to identify new research fronts in cognitive
tion validities, cognitive control/consistency, and cue-
con?ict research emerging from the work of Hammond
judgment function forms), and the match between them
(1965) and Brehmer (1976). The main limitation of this
(i.e., achievement, and its linear and nonlinear compo-
approach, however, is that it can exclude relevant publi-
nents) (Balzer, Doherty, & O’Connor, 1989; Doherty &
cations by virtue of them not citing the central articles of
Balzer, 1988). Such feedback can help to speed con?ict
interest. Later, we discuss how this limitation excluded
reduction (Balke, Hammond, & Meyer, 1973).
potentially relevant work on negotiation.
In 1969, Leon Rappoport warned that “if the cognitive
We conducted a “cited reference” search on the ISI
con?ict model is to serve as anything more than a labora-
Web of Knowledge, Web of Sciences Databases (Science
tory analogue, it must be determined whether socially-
Citation Index Expanded, Social Sciences Citation Index,
induced (i.e., “natural”) cognitive differences generate
and Arts & Humanities Citation Index) to identify rele-
the same con?ict phenomena as laboratory induced (i.e.,
vant journal publications in the period after 1976, to 2007
“arti?cial”) cognitive differences” (p. 143). In fact, as
that cited Hammond (1965) or Brehmer (1976).7 Publi-
Brehmer (1976) noted, many of the ?ndings that were
cations before 1976 were also added in order to provide a
observed in the laboratory on simulated tasks were also
full picture of the evolution of cognitive con?ict research
obtained in naturalistic environments or real tasks, par-
in the lens model tradition. Overall, our searches resulted
ticularly for use in policy development (e.g., Adelman,
in 192 hits, with 141 publications dating after 1976. After
Stewart, & Hammond, 1975; Balke et al., 1973; Brown
1976, 39 publications cited Hammond (1965), 102 cited
& Hammond, 1968; Steinmann, Smith, Jurdem, & Ham-
Brehmer (1976), and 15 cited both authors (i.e., were
mond, 1975). Brehmer (1976) concluded his review with
repeats). Thus, excluding the repeats there were a total
avenues for future research including examining the an-
of 177 publications (192 minus the 15 repeats) with 126
tecedents and consequences of policy inconsistency, and
publications dated after 1976. (A list of the 177 publica-
further analysis of real world con?icts.
tions is available from the second author.)
First, we conducted a bibliographic review of the
177 publications using CiteSpace II (Chen, 2004, 2006)
Interpersonal con?ict theory and
which is a bibliometric tool that visualizes trends and
cognitive con?ict research post
turning points in scienti?c literatures based on citations.
The input was bibliographic records from the publica-
tions and the outputs include illustrations of co-citation
networks either in a cluster view or in a time zone view.8
Here, we trace the evolution of IPC theory after 1976
In CiteSpace II, the entire time interval is sliced into equal
to the present day to determine what further contribu-
length segments in which citations and co-citations are
tions cognitive con?ict research in the lens model tra-
calculated. In our analysis we used two year segments.
dition has made since Brehmer’s 1976 review. For ex-
In each time slice the co-citation network is determined
ample, have researchers followed up on the suggestions
by three thresholds, citation (c), co-citation (cc) and co-
initially made by Hammond (1965) that IPC theory can
citation coef?cient thresholds (ccc; this threshold deter-
tell us something about real world political con?icts?
mines the cosine coef?cients in the normalization of the
Have researchers conducted research on the antecedents
co-citation counts). The thresholds can be set for three
and consequences of policy inconsistency as suggested
by Brehmer (1976)? Are there other ideas beyond those
7Searches were as of October 27, 2007.
8Here, we provide only a brief overview of the steps involved in
6This was originally called COGNOGRAPH. The emphasis is on
analyses using CiteSpace II, since only some of it’s basic features were
teaching consistent new policies. However, the effectiveness of this aid
required for present purposes. The reader is referred to Chen (2006) for
has not been empirically tested (Brehmer & Brehmer, 1988).
a detailed overview of CiteSpace II.

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
points in time with linear interpolation between them.
The upper left and the upper right of Figure 2 shows
The resulting networks in each time slice can then be
publications largely concerned with JDM, only some of
pruned by using the Path?nder algorithm or the minimum
which are related to cognitive con?ict (but not directly on
spanning tree algorithm. The networks in each time slice
the topic itself). Here, for instance, researchers have ex-
are then merged into a synthesized network. As our main
amined how cognitive con?ict may affect a third person’s
objective was to illustrate the network of the most cen-
judgments. For example, Cosier (1978) studied the effect
tral publications, we present ?gures with pruned (using
of different ways in which expert advice could con?ict
the Path?nder algorithm) co-citation networks based on
and the effect of their degree of accurate knowledge of
high thresholds. That is, the resulting merged network
the environment on subjects’ predictions of the criterion
shows only the most important publications in terms of
(see also Schwenk & Cosier, 1980). Cosier, Ruble, and
citations and co-citations during the time period. In the
Aplin (1978, Study 1) examined perceived helpfulness of
merged network, individual publications are represented
expert advice under high and low con?ict. Researchers
as tree rings where the thickness of a ring is proportional
also investigated factors that may impact judgment poli-
to the number of citations in a given time slice. The size
cies which have implications for future research on cog-
of the outermost ring and the size of the font of the publi-
nitive con?ict (Hagafors & Brehmer, 1983), and shown
cation label are proportional to the betweenness centrality
how judgment analysis can be used to study expert judg-
of the publication. The betweenness centrality measure is
ment (Adelman & Mumpower, 1979). However, most
a graph theoretical property that speci?es the importance
of the publications on the upper left and right of Figure
of a node’s position in a network (Chen, 2006). The color
2 are unrelated to cognitive con?ict. For instance, Dink-
of the connecting lines between the citation trees repre-
age and Ziller (1989) explored US and German children’s
sents the year of the ?rst co-citation of the publication.
conceptualizations of war and peace via photographs.
Second, we conducted a content review of those pub-
Most interestingly, the bottom right of Figure 2 shows
lications since 1976 on cognitive con?ict in the lens
that a new research front on group con?ict appears to
model tradition, and which had cited Hammond (1965)
have emerged which also apparently examines cognitive
or Brehmer (1976). After examining the 126 publica-
con?ict. It is in the mid-1990s, after the publication of
tions, only 17 were deemed relevant to this review. As
Jehn’s (1995) article on the bene?ts and detriments of in-
described below, the remainder (109) were either pub-
tragroup con?ict, and the earlier book by McGrath (1984)
lications on cognitive con?ict but not in the lens model
on the interaction and performance of groups, that we can
tradition or on topics related to (but not directly on) cog-
observe this new research front. These new central arti-
nitive con?ict such as interpersonal learning, group deci-
cles and their offshoots are at the bottom right of Figure
sion making, and decision aids. Our content review sum-
2. As we will discuss later, this new research front is
marizes the methods and main ?ndings of the 17 relevant
not grounded in the lens model tradition and, although
they still occasionally cite Brehmer (1976; and, rarely,
Hammond, 1965), these researchers use different theo-
retical frameworks and research tools than those used by
Bibliographic review
researchers studying cognitive con?ict in the lens model
The main results of the bibliometric analysis are pre-
tradition. In fact, these researchers are not as interested in
sented in Figure 2, which shows the cluster view of a
cognitive con?ict as de?ned in the lens model tradition.
co-citation analysis from 1965 to 2007. Brehmer (1976)
and Hammond (1965) are the two most central articles as
Content review
they were the basis of the selection procedure. The pub-
lications on IPC theory or cognitive con?ict research in
As mentioned above, we also conducted a content review
the lens model tradition can be found on the left. These
of the 17 (out of 126) publications classi?ed as being on
are mostly from the 1960s and 70s. Indeed, there appears
cognitive con?ict in the lens model tradition, which cited
in recent years to be a decline in cognitive con?ict re-
Hammond (1965) or Brehmer (1976).
The Appendix
search using the lens model tradition, and few central ar-
presents a summary of the main aims, methods, and ?nd-
ticles were published in the years after Brehmer (1976).
ings of these studies. (The main publications before 1976
Although 126 publications have cited Brehmer (1976) or
were reviewed in Section 1).
Hammond (1965) over the past 30 years, few of these
All 17 publications reported studies that appeared to
actually examine cognitive con?ict in the lens model tra-
have moved beyond the theoretical issues reviewed by
dition. Of the 17 publications that we classi?ed as rel-
Brehmer in 1976 to investigate a new set of problems
evant to the content review we report below, there were
(except perhaps Rose et al., 1982).
First, nine stud-
two highs of 3 publications in 1977 and 1979 and then a
ies examined the effect of some form of intervention on
sharp decline to one or zero each year following that.
cognitive con?ict or judgment performance. Cosier and

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
Figure 2: Co-citation network of publications 1965–2007 (2 years slice, parameters c, cc, ccv: 3, 2, 25; 3, 3, 25; 4, 4,

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
Rose (1977) examined the effect of cognitive con?ict
couples resolved con?ict differently than couples with
and goal con?ict on judgment performance, and found
dissimilar perceptions.
less prediction error under high (than low) cognitive con-
Finally, some studies also included measures of in-
?ict in earlier trials, and under no-goal con?ict. Holz-
terpersonal learning as well as interpersonal con?ict
worth’s (1983) study measured the impact of task pre-
(Alexander, 1979; Gillis, 1979b; Gillis & Moss, 1978;
dictability and mediation on con?ict reduction, and re-
McCarthy, 1977).
ported that, while there was no signi?cant effect of me-
Methodologically, most researchers diverged from the
diation, agreement was greater under more (than less)
experimental method proposed by Hammond (1965) in
predictable tasks. Alexander (1979) measured the ef-
several ways. First, in seven studies there was no training
fect of communication technique on con?ict reduction,
stage where participants learned to perform the judgment
and found that dyads trained in the “region of validity”
task (Dhir & Markman, 1984; Harmon, 1998; Harmon &
technique showed greater con?ict reduction than those
Rohrbaugh, 1990; McCarthy, 1977; Reagan-Cirincione,
not trained as such. Harmon (1998) studied the effect of
1994; Summers et al., 1977; Qualls & Jaffe, 1992). Sec-
decision making method and communication medium on
ond, and relatedly, in over half of the studies parties were
group satisfaction and agreement, and found that audio-
not trained to hold different judgment policies. Rather,
communication (as opposed to face-to-face communica-
in some studies parties were brought together based on
tion) increased satisfaction while policy modeling deci-
their existing policy differences (Bose & Paradice, 1999;
sion methods improved agreement over conventional de-
Harmon, 1998; Harmon & Rohrbaugh, 1990; McCarthy,
cision making methods. Harmon and Rohrbaugh (1990)
1977; Reagan-Cirincione, 1994). Gillis and colleagues
and Sengupta and Te’eni (1993) studied the effect of cog-
paired participants according to the medication they were
nitive feedback on group JDM. Whereas cognitive feed-
prescribed (Gillis, 1979a; 1979b; Gillis & Moss, 1978).
back increased group cognitive control, it did not increase
Dhir and Markman (1984) and Qualls and Jaffe (1992)
agreement, and shared feedback did not improve group
studied married couples. These methodological depar-
judgment accuracy over individual feedback/no feed-
tures represent more than super?cial deviations. Rather,
back, but it did increase agreement. Reagan-Cirincione
they can reduce the researcher’s control over the study of
(1994) and Bose and Paradice (1999) measured the ef-
cognitive con?ict by, for instance, introducing unwanted
fectiveness of group decision aids or support systems on
(and potentially unknown) variability in how different
group performance, which revealed that such aids were
parties perform the task and in the degree of existing con-
effective. Andersson and Brehmer (1979) compared the
?ict between parties.
effect of individual and interpersonal learning on policy
Finally, in seven studies parties did not interact at the
change, and reported no signi?cant differential effects of
con?ict stage. Rather, participants were either given a
these types of learning.
simulated person’s judgments in con?ict to their own
Second, ?ve studies investigated group con?ict (Bose
(Cosier & Rose, 1977; Rose et al., 1982) or participants’
& Paradice, 1999; Reagan-Cirincione, 1994; Harmon,
responses were paired (Summers et al., 1977). In Mc-
1998; Harmon & Rohrbaugh, 1990; Sengupta & Te’eni,
Carthy’s (1977) study, joint judgments were optional, and
1993). These reported on the effectiveness of cognitive
Andersson and Brehmer (1979) examined how individ-
feedback (i.e., availability of feedback and whether it is
ual learning compared to interpersonal learning. Dhir
shared), and group decision aids (where group discus-
and Markman (1984) and Qualls and Jaffe (1992) simply
sion was aided by a facilitator and computer analyst) or
paired individuals’ judgments. This elimination of the
support systems (where there is computerized collection
interpersonal communication between con?icting parties
and communication of individual judgments, amongst
means that relevant issues such as interpersonal learning
other things), as well as decision making method (i.e.,
cannot be addressed in the study of cognitive con?ict,
structured policy modeling or not) and communication
compromise, and change.
medium (i.e., audio or face-to-face).
Before summarizing the ?ndings of this content re-
Third, two studies examined potential perceptual in-
view, it is worth pointing out that, since the bibliographic
?uences on cognitive con?ict. Dhir and Markman (1984)
review technique used for initial selection of publications
studied marital con?ict in task de?nition rather than judg-
was limited to those that cited the articles by Hammond
ment performance. They found that feedback of their
(1965) and Brehmer (1976), some potentially relevant
spouses’ perception of the task had a differential im-
work on negotiation was excluded (Darling, Mumpower,
pact on husbands’ and wives’ ability to correctly predict
Rohrbaugh, & Vari, 1999; Milter, Darling, & Mumpower,
their spouses’ judgment policies. Qualls and Jaffe (1992)
1996; Mumpower & Rohrbaugh, 1996). This work reit-
examined how husbands’ and wives’ pre-existing per-
erates the importance of the task environment when un-
ceptions in?uenced con?ict in joint purchase decisions.
derstanding negotiation or con?ict behavior (Mumpower,
Here, similar perceptions led to less con?ict and these
1988; 1991). Negotiation tasks do not always have an

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
outcome criterion, or it may be irrelevant. Characteris-
the mid-1990s there was an emerging research front on
tics of negotiation tasks are often subjectively interpreted
group con?ict that apparently examines cognitive con-
by the con?icting parties, and these characteristics (inter-
?ict. The central publications were by McGrath (1984)
pretations) may change as the parties interact. The task
and Jehn (1995). However, this new research front is not
structure in turn affects the most appropriate negotiation
grounded in the lens model tradition and, although they
strategy. Thus, in negotiation tasks parties must agree on
still occasionally cite Brehmer (1976; and rarely Ham-
what the task is and how to solve it. This work expands
mond, 1965), these researchers use different theoretical
or rede?nes the terminology for discussing con?ict res-
frameworks and research tools than those used by re-
olution: for example, settlements may be ef?cient, have
searchers studying cognitive con?ict in the lens model
joint utility or equality, and strategies may involve com-
tradition. It is worth brie?y reviewing the new central
promise or logrolling/horsetrading (where parties make
publications in order to assess the degree to which this
trade-offs so they each obtain a desirable outcome). Con-
research front, which has attracted more researchers than
trolling for formal task characteristics, substantive task
the lens model tradition, marks a theoretical and method-
characteristics (i.e., cover story) can affect negotiators’
ological advance in cognitive con?ict research.
ability to reach ef?cient settlements (Milter et al., 1996).
McGrath’s (1984) book reviews the theoretically
This work has also described procedures to support con-
grounded empirical literature on small groups, and sum-
?ict resolution in multi-party negotiations in real-world
marizes the methods used to study small groups. He notes
public policy settings (Darling et al., 1999).
that when a group’s task is to resolve con?icts, as is often
In sum, although our bibliographic review indicates
the case, IPC theory (which he confusingly refers to as
that after 1976 relatively few studies were published on
“social judgment theory” throughout) is relevant to un-
cognitive con?ict in the lens model tradition our content
derstanding the negotiation process. IPC theory is thus
review suggests that several new contributions were made
reviewed in a chapter entitled “Cognitive con?ict tasks:
by this small body of literature. In fact, the literature
Resolving con?icts of viewpoint within the group.” Here,
went beyond the issues studied in the earlier work re-
a passing reference is made to Brunswik’s (1955; whose
viewed by Brehmer (1976) in several interesting ways.
name is misspelled throughout) lens model, and articles
However, for unknown reasons, no-one followed up on
by Brehmer (1976) and Hammond et al. (1966a, 1975)
the suggestions initially made by Hammond (1965) that
are summarized. The experimental method associated
IPC theory can tell us something about political con?icts,
with IPC theory is also summarized. In addition, with ref-
which nowadays may focus on identifying and manag-
erence to a study by Rohrbaugh (1979), McGrath (1984)
ing threats to national and global security, although the
concludes that the cognitive feedback approach used by
work on group con?ict and negotiation sometimes deals
IPC theorists to improve group judgment is not very ef-
with public policy issues (e.g., Darling et al., 1999; see
fective. Overall, McGrath (1984, p. 66, p. 89, p. 93) calls
also Hammond and Grassia, 1985 for public policy ex-
the work on IPC “limited,” noting that much of the re-
amples). Similarly, few researchers directly examined
search has been conducted only on “two-person groups,”
the antecedents and consequences of policy inconsistency
and he calls the method used “very elaborate.”
as suggested by Brehmer (1976). Karelaia and Hogarth
Thus, McGrath’s (1984) book introduced IPC theory
(2008) recently examined the impact of several factors
and its associated method to researchers interested in
such as outcome feedback and cue redundancy on pol-
studying group JDM. However, this was just one of sev-
icy inconsistency, which may be worth exploring in the
eral approaches reviewed by McGrath, and he was some-
context of cognitive con?ict research. Researchers also
what critical of it. It is no surprise therefore, that few re-
often departed from the experimental method described
searchers interested in group JDM have studied con?ict
by Hammond (1965). Rather than representing useful
in the lens model tradition. In fact, later, Jehn’s (1995)
innovations these departures appear to dilute the control
reference to Brehmer (1976) is merely to point out that he
that the researcher has over the experimental situation in,
(and others) suggest that the relationship between con?ict
for instance, knowing the precise sources of con?ict, and
and performance is in?uenced by the type of task a group
limit the study of important issues in cognitive con?ict
performs. Similarly, others refer to Brehmer (1976) sim-
such as interpersonal learning.
ply as a means of suggesting that cognitive con?ict may
result in affective con?ict (Amason, 1996). Generally,
3.3 Research on group con?ict: A paradigm
the research questions, theoretical insights, and exper-
shift in cognitive con?ict research
imental method of IPC theory were overlooked in the
central articles by McGrath (1984) and Jehn (1995), and
Beyond the small body of published literature on cog-
with the exception of work by Rohrbaugh and colleagues
nitive con?ict in the lens model tradition conducted af-
(Rohrbaugh, 1988, and Harmon & Rohrbaugh, 1990),
ter 1976, the bottom right of Figure 2 revealed that in
abandoned in recent research on group con?ict (e.g.,

Judgment and Decision Making, Vol. 3, No. 7, October 2008
Interpersonal con?ict theory
Pelled, Eisenhardt, & Xin, 1999).
have focused on con?ict in real world settings such as or-
As the central article by Jehn (1995) demonstrates,
ganizations which lends external validity to their ?ndings,
cognitive con?ict is often de?ned in terms of “task con-
the main focus in this new research front on group con-
?ict.” According to Jehn (1995, p. 258)
?ict is not necessarily cognitive con?ict, but task con?ict.
“Task con?ict exists when there are disagreements
It could be argued that the methods employed do not rep-
among group members about the content of the tasks be-
resent an advance, and that the theories, as they currently
ing performed, including differences in viewpoints, ideas,
stand, are limited. As such we believe the research front
and opinions.”
on group con?ict revealed at the bottom right of Figure 2
This concept was measured by Jehn (1995, p. 268) us-
does not really represent a constructive paradigm shift in
ing a short scale that includes items such as “How of-
cognitive con?ict research.
ten do people in your work unit disagree about opinions
regarding the work being done?” “How frequently are
there con?icts about ideas in your work unit?”
The future of interpersonal con-
“How much con?ict about the work you do is there in
?ict theory and cognitive con?ict
your work unit?”
Responses are provided on 5-point
scales anchored by 1 = “none” and 5 = “a lot.” Oth-
ers have used similar measures (e.g., Pelled et al., 1999).
Therefore, the new concept of task con?ict is somewhat
In this ?nal part, we offer possible reasons for the sharp
vague and ill-de?ned, and its measurement is not very
decline of cognitive con?ict research in the lens model
precise. For instance, it is unclear how participants inter-
tradition, and we discuss the future of this research and
pret concepts such as “con?ict,” “opinions,” and “ideas,”
IPC theory in the context of the growing ?eld of JDM.
and there is no clear differentiation of different aspects
Speci?cally, we consider some of the opportunities for
of the phenomenon of task con?ict. Finally, its measure-
theoretical advancement and methodological innovation
ment is on a short, subjectively interpreted scale. This
in cognitive con?ict research.
approach clearly departs from the precise de?nition of
At a meeting in 2006 of the Brunswik Society, a small
cognitive con?ict provided by IPC theory as the relation
international group of scholars dedicated to Brunswikian
between cognitive systems which is measured quantita-
psychology, researchers offered possible explanations for
tively in terms of agreement (rA), policy similarity (G),
the historical decline of cognitive con?ict research in
and cognitive control (R1 and R2), and which clearly dif-
the lens model tradition and the neglect of IPC theory.
ferentiates between different aspects of the phenomenon
In particular, researchers recalled that they felt most of
(e.g., policy similarity versus cognitive control).
the important and interesting questions concerning cog-
Furthermore, this group of researchers are largely in-
nitive con?ict had already been suf?ciently addressed
terested in questions pertaining to the impact of task (cog-
in Hammond’s and Brehmer’s early work, thus leaving
nitive) con?ict on outcomes such as work satisfaction,
little scope for new insights.
Researchers also remi-
liking of other group members, intentions to remain in the
nisced that at the time there were several other areas
group, performance (Jehn, 1995), and emotional con?ict
of Brunswikian-related research available for exploration
(Pelled et al., 1999). They have typically used quantita-
which were more appealing such as the study of clini-
tive questionnaire and qualitative observational and inter-
cal judgment (Hammond, 1955), multiple cue probabil-
view methods as well as archival analysis to examine the
ity learning (Hammond & Summers, 1965), cognitive
nature and effects of existing con?ict within work groups.
feedback (Todd & Hammond, 1965), interpersonal learn-
For instance, Jehn (1995) measured individuals’ perfor-
ing (Hammond, 1972; Hammond et al., 1966b), and so-
mance via appraisal ratings, departmental records, and
cial judgment theory (Hammond et al., 1975). From a
supervisors’ ratings. The experimental method is rarely
more practical perspective, researchers noted that cogni-
employed. This makes it dif?cult to advance causal theo-
tive con?ict research in the lens model tradition was chal-
ries, and their ?ndings remain limited to description and
lenging. For instance, echoing McGrath’s (1984) criti-
prediction. Indeed, these researchers typically use multi-
cism, researchers complained that the proposed experi-
ple regression and other correlational techniques for data
mental method was inef?cient since it required many par-
analysis, which is why the statistical textbook by Cohen
ticipants and much time. Furthermore, researchers noted
and Cohen (1983) also appears in the bottom right of Fig-
that later generations of students were not always suf?-
ure 2.9
ciently trained to conduct the relatively complex statisti-
Therefore, while researchers working on group con?ict
cal analysis required by the lens model equation.
It is however, premature to conclude that cognitive
9Although research in the lens model tradition has also traditionally
employed correlational tools, these are used in conjunction with exper-
con?ict research in the lens model tradition conducted to
imental techniques.
date has provided a complete picture of cognitive con-