Professional Documents
Culture Documents
Ebook Carnevale, P. Ed., de Dreu, C.K.W. Ed. Methods of Negotiation Research International Negotiation Series
Ebook Carnevale, P. Ed., de Dreu, C.K.W. Ed. Methods of Negotiation Research International Negotiation Series
edited by
ISBN 10 90 04 14858 2
ISBN 13 978 90 04 14858 1
2006 by Koninklijke Brill NV, Leiden, The Netherlands.
Koninklijke Brill NV incorporates the imprints Brill, Hotei Publishers,
IDC Publishers, Martinus Nijhoff Publishers and VSP.
http://www.brill.nl
All rights reserved. No part of this publication may be reproduced, translated, stored in a
retrieval system, or transmitted in any form or by any means,
electronic, mechanical, photocopying, recording or otherwise, without
prior written permission from the Publisher.
Contributors ................................................................................................ ix
Chapter 8
Quantitative Coding of Negotiation Behavior ...................................... 105
Laurie R. Weingart, Mara Olekalns, and Philip L. Smith
Chapter 9
The Use of Questionnaires in Conict Research .................................. 121
Aukje Nauta and Esther Kluwer
Chapter 10
A Multilevel Approach to Investigating Cross-National Differences in
Negotiation Processes .............................................................................. 135
Xu Huang and Evert Van de Vliert
Chapter 11
Methodological Challenges in the Study of Negotiator Affect ............ 149
Bruce Barry and Ingrid Smithey Fulmer
Chapter 12
Comparative Case Studies ...................................................................... 165
I. William Zartman
Chapter 13
Discourse Analysis: Mucking Around with Negotiation Data ............ 177
Linda Putnam
Chapter 14
Field Experiments on Social Conict .................................................... 193
Dean G. Pruitt
Chapter 15
Laboratory Experiments on Social Conict .......................................... 211
Peter J. Carnevale and Carsten K.W. De Dreu
Chapter 16
Managing Conict in the Literature: Meta-analysis as a Research
Method .................................................................................................... 227
Alice F. Stuhlmacher and Treena L. Gillespie
Chapter 17
When, Where and How: The Use of Multidimensional Scaling
Methods in the Study of Negotiation and Social Conict .................. 239
Robin S. Pinkley, Michele J. Gelfand, and Lili Duan
CONTENTS vii
Chapter 18
Markov Chain Models of Communication Processes in
Negotiation .............................................................................................. 257
Philip L. Smith, Mara Olekalns, and Laurie R. Weingart
Chapter 19
All that Glitters is Not Gold: Examining the Perils of Collecting
Data on the Internet ................................................................................ 273
Yeow Siah Cha
Chapter 20
The Method of Experimental Economics .............................................. 289
Rachel Croson
Chapter 21
Empirical Research in Law and Negotiation ........................................ 307
Rebecca Hollander-Blumoff
Chapter 22
Methodologies for Studying Personality Processes in
Interpersonal Conict .............................................................................. 323
Lauri A. Jensen-Campbell and William G. Graziano
Chapter 23
The Heart of Darkness: Advice on Navigating Cross-Cultural
Research .................................................................................................. 341
Catherine H. Tinsley
Chapter 24
Disparate Methods and Common Findings in the Study of
Negotiation .............................................................................................. 351
Carsten K.W. De Dreu and Peter J. Carnevale
Bruce Barry:
Owen Graduate School of Management, Vanderbilt University, Nashville, TN
37203 USA
E-mail: bruce.barry@vanderbilt.edu
Jacob Bercovitch:
Department of Political Science, University of Canterbury, Private Bag 4800,
Christchurch, New Zealand
E-mail: jacob.bercovitch@canterbury.ac.nz
Peter J. Carnevale:
Department of Psychology, New York University, 6 Washington Place, Room
577, New York, NY 10003 USA
E-mail: peter.carnevale@nyu.edu
Rachel Croson:
The Wharton School, University of Pennsylvania, 567 Huntsman Hall, 3730
Walnut Street, Philadelphia, PA 191046340 USA
E-mail: crosonr@wharton.upenn.edu
Daniel Druckman:
Institute for Conict Analysis and Resolution (ICAR), George Mason Univer-
sity, Fairfax, VA 22030 USA
E-mail: ddruckma@gmu.edu
Lili Duan:
Department of Psychology, University of Maryland, College Park, MD 20742
USA
E-mail: lduan@psyc.umd.edu)
x CONTRIBUTORS
Ronald J. Fisher:
International Peace and Conict Resolution Program, School of International
Service, American University, 4400 Massachusetts Avenue, NW, Washington,
DC 20016 USA
E-mail: rsher@american.edu
Ray Friedman:
Owen Graduate School of Management, Vanderbilt University, Nashville, TN
37203 USA
E-mail: ray.friedman@owen.vanderbilt.edu
Michele J. Gelfand:
Department of Psychology, University of Maryland, College Park, MD 20742
USA
E-mail: mgelfand@psyc.umd.edu
Treena L. Gillespie:
Department of Management, California State University-Fullerton, P.O. Box
6848, Fullerton, CA 928346848 USA
E-mail: tgillespie@fullerton.edu
Rebecca Hollander-Blumoff:
New York University School of Law, Lawyering Program, 245 Sullivan Street,
New York, NY 100121301 USA
E-mail: hollande@juris.law.nyu.edu
Xu Huang:
Department of Management & Marketing, The Hong Kong Polytechnic Uni-
versity, Hung Hom, Kowloon, Hong Kong
E-mail: mshuangx@polyu.edu.hk
Esther Kluwer:
Department of Social and Organisational Psychology, Utrecht University, P.O.
Box 80, 140 NL 3508 TC Utrecht, The Netherlands
E-mail: E.S.Kluwer@fss.uu.nl
David Matz:
Graduate Programs in Dispute Resolution, University of Massachusetts Boston,
100 Morrissey Boulevard, Boston, MA 02125 USA
E-mail: David.Matz@Umb.edu
CONTRIBUTORS xi
Aukje Nauta:
TNO Work and Employment, P.O. Box 718, 2130 AS Hoofddorp, the Nether-
lands E-mail: a.nauta@arbeid.tno.nl
Mara Olekalns:
Melbourne Business School, University of Melbourne, 200 Leicester Street
Carlton VIC 3053 Australia
E-mail: m.olekalns@mbs.edu
Robin L. Pinkley:
Director of the American Airlines Center for Labor Relations and Conict
Resolution, Edwin L. Cox School of Business, Southern Methodist University,
Fincher Building, P.O. Box 750333, Dallas, TX 75275 USA
E-mail: rpinkley@mail.cox.smu.edu
Dean G. Pruitt:
Institute for Conict Analysis and Resolution, George Mason University, 9006
Friars Road Bethesda, Maryland 20817 USA
E-mail: dean@pruittfamily.com
Linda L. Putnam:
Department of Communication, Texas A&M University, 4234 TAMU, College
Station, TX 778434234 USA
E-mail: lputnam@tamu.edu
Chayeow Siah:
Department of Social Work and Psychology, National University of Singa-
pore, 11 Law Link, Singapore 117570
E-mail: swkchays@nus.edu.sg
Philip L. Smith:
Department of Psychology, University of Melbourne, Victoria, 3010 Australia
E-mail: philipls@unimelb.edu.au
Alice F. Stuhlmacher:
Department of Psychology, DePaul University, 2219 N. Kenmore Avenue,
Chicago, IL 60614
USA (E-mail: astuhlma@depaul.edu)
Catherine H. Tinsley:
The McDonough School of Business, Georgetown University, Washington,
DC 20057 USA
E-mail: Tinsleyc@georgetown.edu
Laurie R. Weingart:
David A. Tepper School of Business Carnegie Mellon University, 5000 Forbes
Avenue, Pittsburgh, PA 15213 USA
E-mail: weingart@cmu.edu
Jonathan Wilkenfeld:
Department of Government and Politics, University of Maryland, College
Park, Maryland 20742 USA
E-mail: Jwilkenf@gvpt.umd.edu
I. William Zartman:
School of Advanced International Studies, the Johns Hopkins University, 1740
Massachusetts Ave, Washington, DC 20036 USA
E-mail: zartman@jhu.edu
Methods of Negotiation Research: Introduction
innovative methods, common and less than common, all that move the eld
forward. With these articles, we make the point that there is a wealth of
methodological tools that negotiation and conict researchers have at hand,
and each has strengths and weaknesses. Our specic objectives include the
following: provide an introduction to a variety of methods and their utility;
identify issues and controversies with various methods; increase the accessi-
bility of works in one empirical domain to readers in another, and thus
broaden the scope of research and theory; improve communication between
domains so the collective enterprise is improved; provide stimulus for yet
unknown approaches and procedures that further contribute to the validity and
vitality of research in this domain; and stimulate the application of a method
used in one domain into another domain.
Another goal that we have for these special issues is to extract valuable
insights about conict phenomena from data collected with diverse methods in
different settings. Since there are synergies among the methods (multi-method
approaches are greater in their impact than the sum of the parts), a synthesis
is needed. To that effect, at the conclusion, we present a review article that high-
lights common features of effects in negotiation that have been obtained with
diverse methodological tools. For example, many studies show that a forgiv-
ing strategy in negotiation can be exploited, and a tough strategy can backre
by producing a competitive response, and this has been obtained in both lab-
oratory studies of university students and in studies of international disputes.
There are of course many examples of such cross-domain consistency in
effects, for example in work on aggression (Anderson & Bushman 1997).
Our concluding article reects Don Campbells notion that validity is
achieved through triangulation, that is, by using a variety of methodological
approaches and procedures. In navigation, triangulation is the technique where
two visible points are used to determine the location of a third point. Applied
to validity and reliability of measurement, triangulation is the use of multiple,
different indicators in such a way that errors can be excluded and underlying
constructs can be identied (Campbell & Fiske 1959). The concept applies
quite well not only to measurement, but to all aspects of method. When two
or more methods or data sources converge on a construct, we have greater
assurance that our conclusions are not driven by an error or artifact of any
one procedure. Each method has strengths and weaknesses, and to the extent
they do not overlap, we can stand on more solid ground with conclusions
about theory.
METHODS OF NEGOTIATION RESEARCH: INTRODUCTION 3
References
Anderson, C.A., & Bushman, B.J. (1997). External validity of trivial experiments: The case
of laboratory aggression, Review of General Psychology 1:1941.
Campbell, D.T., & Fiske, D.W. (1959). Convergent and discriminant validation by the multi-
trait-multimethod matrix, Psychological Bulletin 56:81105.
Methods of Negotiation Research II
San Sebastian, and Bert Spector who guided the entire effort. We are grateful
to the many people who helped with reviews, including Bianca Beersma, Eric
van Dijk, Ray Friedman, Andrea Hollingshead, Peter Molenaar. Bernard
Nijstad, Kathleen OConnor, Dean Pruitt, Andy Schotter, Donna Shestowsky,
Kees van Veen, Jim Wall, Ching Wan, and several others who indicated a
preference for remaining anonymous.
The Joys of Field Research
A previous mission had been a disaster. My fathers B-17 squadron had put
up 13 planes. Over Germany, the ghters came out of the sun and in 10 sec-
onds eight planes 72 airmen went down, like burning leaves.
The current mission had gone better; fewer comrades had been lost; and
Dads bomber was somewhere back over England, but in a cloud bank, iso-
lated from the other planes. Lost. And only one person knew they were lost,
the navigator. As for the others, the crew was celebrating a completed mis-
sion, and the pilot was focused on the low fuel gauge.
The compass wasnt working; neither was the radio (B-17s seemed to be
allocated to crews in alphabetical order and with a pilot named Wallace and
a navigator named Wall, this crew usually drew a decrepit bomber). The pilot
called back, Navigator, report our current position, we have to land now!
Give me a minute, Dad said. As he stalled, the clouds parted for a
moment to reveal the plane was directly over the base. Bank right and take
her down; were home. Then, in relief, he threw up.
Because of several experiences like this, my father does not like being lost,
but his rstborn does. I enjoy being lost in the woods, mountains, caves and
on rivers, lakes, and oceans (okay, maybe not on the ocean). I also like being
lost when I conduct eld research. It, along with mistakes, scientism, the par-
ticipants, stuckness, and discovery are what I refer to as the joys of eld
research.
Lost
When I refer to being lost in eld research, I mean physically lost as well as
conceptually lost. My rst recognition that I liked lostness came on a blaz-
ing August afternoon in Nanjing during the mid-1980s. My translator and
I were pedaling up and down countless hills through a laoconian tangle
of streets in the city, lost. We were seeking an inner-city community a
street whose mediator had agreed to an interview, but we were hopelessly
lost. I do not believe we ever found that mediator. Rather, my translator
probably bribed a local mediator to talk with me. The entire search was, in
retrospect, a great deal of fun.
On a cooler day, a dozen or so years later, I was lost in Canada. I had dri-
ven across the U.S.-Canadian border on an expressway, traveling to Kingston,
to nd some Canadian peacekeepers. On the return trip, I could not nd my
way back into the U.S. This is stupid, I thought. If I came across the bor-
der on a four-lane highway, surely I can nd my way back. But I could not,
and as I spent hours exploring the back roads of Canada, I found it was
enjoyable and concluded this was a ne physical representation a metaphor
for being lost in eld research.
Usually, a eld researcher is to some extent lost. Youre not always certain
what the process is that youre studying. An independent variable seems to be
missing. There are a confusing number of dependent variables. Cause and
effect are difcult to tease out. Results and conclusions are frequently con-
tradictory. As for personal examples of being lost, my past is replete with
them, but a current one is that I am lost as I consider the effects of conict
in business-to-business e-commerce. But I am having fun.
Mistakes
Mistakes are somewhat akin to being lost. To me, conducting eld research
can be likened to rowing a single, a skiff. You are rowing backwards; there-
fore, you cannot see where you are going. If you attempt to overcome this
inconvenience by looking over your shoulder, you will ip, because the boat
is only ten inches wide, and very unstable. So in order to row well, you relax,
look straight ahead (backwards), focusing on your stroke, your legs, and your
speed. Conse-quently you run into obstacles such as rocks, boats, stumps, and
shores only to nd yourself in the cool water. That is, you make mistakes. It
is a lot of fun.
In eld research, mistakes are not only fun; they also are instructive. A
heart transplant surgeon emphasized this to me when I was interviewing lead-
ers for my book Bosses (Wall 1986). In my eld, he noted, we obviously
try to avoid as many mistakes as possible. But any mistake I make with one
patient makes me a better surgeon with the next.
I agree. Each specic mistake teaches an individual lesson; moreover, mis-
takes in general are quite helpful because they eradicate or at least lower
researcher hubris. In eld research, arrogance is the researchers worst enemy.
If you are overly condent, if you know that you are investigating the correct
THE JOYS OF FIELD RESEARCH 9
problem, from the correct perspective, with the correct procedure, with the
correct sample, at the correct time you will probably nd yourself with a
years worth of useless data. A few mistakes, early in the research, will bring
you back to reality and reduce the long run costs.
Mistakes also teach the eld researcher that good enough is good
enough. When the Russian general Kalashnikov was recovering from battle
wounds and designing the AK-47, he remembered a local saying, Perfection
is the enemy of good enough. Therefore, he designed his automatic rie to
be only good enough. It is not extremely accurate. Its bullets do not go as far
as those of some other machines. The muzzle velocity is relatively slow.
When it is red, it sounds like a child rumbling through the kitchen pots and
pans. But it always res, under almost any condition; seldom does it jam; rust
and abuse does not stop it or even slow it down, all because Kalashnikov designed
its parts to t together good enough, not precisely.
Why did Kalashnikov design the AK-47 this way? I honestly do not know,
but my guess is that he did not sit in a hospital room thinking, I want to
build a weapon that will re under almost any condition and can easily be
operated by a technological dummy. My guess is that mistakes had taught
him a lesson. Probably he, or his troops, had experienced weapons that did
not re when they became dirty or rusty or frigid weapons that may have
been too complicated to quickly disassemble and reassemble.
For eld researchers, mistakes usually have the same effect as the soldiers
experience in that they teach the researcher not to be overly precise. I learned
this lesson in a humorous experience. When in Nanjing, I found there was
one mediator per 1000 citizens; therefore, in a city of two million there was
a large potential sample. Awed by the number, my aspirations soared. As they
did, I decided to precisely measure the techniques employed by Chinese
mediators. I knew that self-reports by the mediators would be good enough,
since this was the rst empirical study of Chinese mediation. But I wanted
precision.
Accompanied by two translators and a pack of Marlboros (as a gift to the
mediator), I begin my rst observation. We mediator, two disputants, two
translators and I sat at a very short, rough table. Initially, it went well; yet,
a couple dozen faces soon peered in the open windows. As the mediator noted
these, his voice grew louder. Then people started climbing through the win-
dows and sitting on the oor; we were up to 25 spectators. Rising to the
occasion, the mediator began mediating with gusto. As the oor space became
inadequate, observers sat in the windows and three decided to join us at the
table. Now the mediator stood, pointed his hand, addressed the audience and
10 JAMES A. WALL, JR.
raised his shrillness about three octaves. After two hours of this, I expressed
my appreciation, delivered the Marlboros, mounted my bike and muttered,
Self-reports should be good enough for this project.
Finally, and most importantly, mistakes emphasize the value of laboratory
procedures. That is, mistakes make it quite obvious that a lot of variance is
occurring, and you had better take it seriously. Therefore, you learn/relearn to
keep the study as simple as possible. Control as many factors as possible and
randomize the rest. Match within conditions and obtain as large a set of
observations as possible.
Scientism
Perhaps the major contribution and invigorating joy from eld research is that
it allows the researcher to engage in scientism which is a combination of
empiricism and reason (Shermer 2002). When describing this term or process,
Michael Shermer notes that it embraces (these) twin pillars of a philosophy
of life appropriate for an Age of Science. He holds that scientism is/was
practiced by Jacob Bronowski, Stephen Hawking, Carl Sagan, Stephen Jay Gould,
and Jared Diamond, some of my intellectual heroes.
While I doubt if I adequately understand this process or practice it as well
as the above intellectuals, I do nd it quite useful to focus, rst, on my method
of study and its results (empiricism), then shift to reasoning and return to the
results.
My best example of this centers upon the number 1.30. In a study on com-
munity mediation in India, I was examining the mediation techniques used by
a formal mediation group, the panchayat (group of ve). One of the tech-
niques was a dictated agreement point, and the average use for this technique
was 1.30. That is, this mediation group, on average, dictated about one and a
third agreement points (or concessions) per mediation. Initially, I recorded this
gure in a table, compared it to the average usage of individual mediators
an elder in India and moved on with my writing.
Then I started to use some reasoning. The average is 1.30 per mediation.
If there are two or more disputants in each mediation, should not the average
number of directives be about 2.00, with one side being told what to do and
likewise, the other? In some cases, the panchayat could have issued one rul-
ing that covered both parties (e.g., farmer A is to quit throwing dead rats into
farmer Bs irrigation ditch), but still the 1.3 seemed too low.
As I thought about this nding, it occurred to me that ideas about the set-
tlement might be coming from the disputants as well as from the mediator
that self-direction might be supplementing mediator direction. So I examined
THE JOYS OF FIELD RESEARCH 11
Participants
For me, the greatest joy in eld research comes from the interaction with the
participants, that is, with the individuals being studied. Here there is a brio
because the participants are excited that you are studying them and what they
12 JAMES A. WALL, JR.
do. You are interested! And you perhaps can help them improve. I vividly
recall a conversation with an American appeals judge, who for years had been
mediating in settlement conferences prior to trial. (A settlement conference is
a meeting in which the judge, the plaintiffs attorney and the defendants attor-
ney prepare the case for trial. Here points of agreement are noted, disagree-
ments are acknowledged, points of law are discussed. If the judge wishes, he
or she can attempt to mediate the case.) He noted he had been taking notes
on how he mediated his cases for the last two years. His goal was to deter-
mine how his efforts affected the plaintiffs and defendants decisions on
whether or not to settle out of court. When I told him that was exactly what
I was studying, he beamed and devoted the rest of the day to delineating his
mediations from the last few months. As we discussed these, he gave his
opinion as to what he felt worked and what did not. Then he asked me for
my opinion.
Being asked for your opinion, to evaluate the performance of the people
you are interviewing judges, peacekeepers, CEOs, e-commerce directors,
imams, panchayat leaders and who usually are performing quite well is
extremely invigorating. Even though this interaction occurred more than a
decade ago, I still recall the emotional boost it gave me. A few months ago,
a similar question from an experienced mediator produced the same effect.
This mediator, whom I was shadowing, had mediated over 2000 cases and
was currently ghting an uphill battle in a case where a 16 year-old girl had
stepped into the side of a moving speeding? truck. About four hours into
the ordeal, the mediator detoured me into a storage room and asked, How
am I doing?
Such interactions are not only invigorating, they also provide valuable
information about the processes under study. For example, in a recent study
of peacekeeping, Dan Druckman and I had predicted that a time constraint
in the form of impending darkness would have an effect on peacekeepers
mediations in a role-playing case presented to them. (Concisely stated, the
prediction was that the time constraint would reduce the amount of the inter-
vention.) After a dozen or so interviews, I concluded that the time constraint/
darkness manipulation was having no effect. So at the end of each interview
I began to ask the peacekeepers (in the time constraint condition) if the loom-
ing darkness had any effect on their mediations. Some replied afrmatively,
which was consistent with the theory and our previous eld observations.
Others said there was no effect because they and their subordinates were
trained to operate 24/7 and were outtted with night-vision equipment to do
so. Most informative was the third group, who held that darkness would elim-
THE JOYS OF FIELD RESEARCH 13
inate the need to mediate. One peacekeeper from this group noted that dark-
ness, rain, cold, wind and snow reduce the pressure. When these come, dis-
putants go home and there is no conict to mediate.
Stuckness
Somewhere among the pages of Zen and the Art of Motorcycle Maintenance
(Pirsig 1974), Robert Pirsig discusses the value of being stuck. Stuck-
ness, he notes, is the beginning of true insight because when you are stuck,
and realize that this is the case, you will begin to analyze and understand the
situation so as to get unstuck. This stuck-then-understand process is superior
to erroneously concluding that you are not stuck and advancing in the wrong
direction.
Guided by Pirsigs logic, I have attempted to develop a taste for stuckness.
While it has improved my eld research and while I have frequently been
able to tolerate the condition, it does not bring me joy.
This might seem odd if not a bit inconsistent for a fellow who likes to
be lost. Consider though, when you are lost, there is a dynamic ow. You are
moving around, looking, thinking, testing, and experimenting. Admittedly,
these movements are accompanied by a touch of fear as well as slight chill
or coolness and an odd feeling in the stomach. But overall, it is an enjoyable
process.
By contrast, stuckness brings me a sense of immobilization, of going
nowhere as when I get my Forerunner stuck in deep snow or mud. And the
emotion? It is one of impatience, irritation, and internal roasting.
The difference here is worth explaining; therefore, let me give you a mun-
dane comparison and then one from eld research. First, the mundane. A few
weeks ago my wife and I locked ourselves out of our house (the responsibil-
ity for this blunder will forever be disputed) which is rather burglar-proof.
When I realized we were on the outside looking in, I felt lost not stuck
because I knew that I would eventually nd a way into the house. Feeling
rather cool/calm but somewhat stupid, I enjoyed thinking about the problem
from different perspectives.
Initially, I thought like a thief; then I took the perspective of a mountain
climber. Subsequently, I played John Nash and thought, take the option that
keeps the joint costs for you and your wife as low as possible. As I was
thinking this, I also approached the problem as a carpenter, who has a lot of
wood, tools, and some skill. A strategy came to me: I cut a very large piece
14 JAMES A. WALL, JR.
of wood from the doorframe (large pieces are easier to replace than are small
ones) near the lock. This gave me direct access to the bolt which I slid over
with a long putty knife. Later I repaired the doorframe. Total cost = 0. The
time spent, including repair, was one hour.
Compare this to an example of being stuck. One spring morning when I
was a graduate student in Durham, I decided to mow the yard prior to taking
a hike with my wife and our Irish Setter. First, I drove to get some fresh
gasoline; secondly, I took out the grinder to sharpen the blade and picked out
a 1/2-inch wrench to take off the blade. Then I started to pull the lawnmower
out from under the house. Or I should say, that I started looking for the lawn
mower, because it was not there.
I was stuck; my face was hot and my feet felt stuck to the ground. Despite
all of Pirsigs advice stare at the problem, push thoughts out of your mind,
start over, go back to basics, write down your ideas, take a break, drink a cup
of coffee, spend additional time thinking about the problem I could not get
unstuck nor develop a positive approach to the problem. So I quit looking and
was irritable for the rest of the day. (A month or so later I discovered the
mower in the garage of my wifes mentor.)
Now to an analogous comparison in eld research. For me, the sensations
of being lost and stuck both center upon the same set of variables: the tech-
niques used by U.S. mediators as they handle disputes.
As for being lost and content I feel this way when I broach the general
question of what determines these techniques; that is, what are the indepen-
dent variables that affect or determine which techniques (dependent variables)
are used by U.S. mediators.
Drifting, with this question in mind, I can contently review the literature,
shadow mediators, interview disputants, ask mediators to recall mediations
and discuss the question openly with professional and informal mediators
around the country. For about ve years, I have been lost but it has been great
fun and currently I have some tentative answers. The rst one is that we have
in the United States high inter-mediator variation in the use of the techniques
because there is no model of mediation in the U.S. that guides or reduces the
variation among mediators. This contrasts with mediation in China where
there is PLA-dictatorial model; in Korea, where there is a Confucian model;
in Malaysia, which has an imam (Muslim) model; and in India where there
is a panchayat (formal group of ve) model.
The second conclusion is that mediator power is important, because it
affects how the disputants treat the mediator, rather than how the mediator
treats the disputants. Finally, there is the role of reinforcement and supersti-
tion. Because about 80 percent of the mediations in the U.S. result in agree-
THE JOYS OF FIELD RESEARCH 15
ment, mediators here are reinforced for whatever approach they take. There-
fore, they continue to employ the same techniques from mediation to media-
tion (i.e., there is low intra-mediator variation.) In addition, the mediators
superstitiously associate the techniques they use with the successful settlement
of the dispute. This combination of reinforcement and superstition gives rise
to the various styles that mediators report and champion.
Turning from lost to stuck, I nd myself in the latter state when I treat
the mediators techniques as the independent variable and attempt to investi-
gate its effects on other variables (e.g., on rate of agreement). Specically, I
am most stuck and irritated that I have been unable to relate any single tech-
nique or group of techniques (e.g., a strategy) to disputant agreement. It is
generally acknowledged that mediation (versus no mediation) does result in a
higher level of settlements; yet, no one including me has been able to
determine which techniques are responsible for the success. Here, I am stuck,
and irritated.
Having expressed my ire with stuckness, I will come full circle to admit
that it does have its benets in eld research. It reduces ones hubris. It makes
one think; snuffs out mediocre ideas; eliminates false starts; and along with
mistakes and being lost, it often underpins discovery, a major joy of eld
research.
Discovery
mosque leaders) and those of the ketua kampungs (village mayors). With the
Malaysian model as a starting point, we attempted to compare the mediations
of the temple monks in Thailand to that of the Malaysian village mayor or a
respected elder. Unfortunately, the work with the monks was not proceeding
smoothly. The data were odd and we were confused as to why. (I think Ronda
was lost and I was stuck.) Then a pattern emerged for Ronda as she pieced
together many of our observations: The monks seemed uncooperative or con-
fused. Some insisted they did not mediate. Our interviewer seemed to be
making no progress in asking people about disputes they took to the monks
or in obtaining reports from the monks. The pattern/insight was that villagers
were not taking important disputes to the monks because they did not respect
them. In Thailand, most, if not all, young men are expected to be monks for
a certain period of time and some of them continue as monks for a lifetime.
As a result, monks are not generally perceived as wise, or as ones who should
be asked to mediate a dispute.
While the above pattern is simple and unwanted, the one currently forming
in a study of legal mediation has more positive aspects. It is more complex
because a chain of causation is involved, and it is benecial in that it charts
out the direction for future research. My beginning hypothesis in the study of
attorneys and judges as mediators was that judges would mediate differently
than attorneys. Specically, they would reduce the clients perceived BATNAs
(i.e., the outcomes from a trial) by citing their own experience on the bench
and they would use their status to control attorney behavior.
The emerging pattern, however, seems to have the causation coming from
the opposite direction, from the clients and attorneys behaviors toward the
judges. I noted that prior to any of the judges actions, the clients and attor-
neys paid substantial deference to him, in that they addressed him as judge
(not by his rst name), waited for him to shake hands, observed where he sat,
and were quiet when he spoke. Occasionally the clients even asked the judge
for his opinion as to what they should do. Over coffee breaks the attorneys
told the judge how helpful he was in a previous case.
These types of behavior by the clients and attorneys tended to shape the
judges behavior, rather than vice versa. The specic pattern seems to be that
the clients seek guidance from the judge, who supplies it in a professional
way. In turn, the clients instruct the attorneys, who after their own input
make their offers to the judge. Observing this causal chain, the judge realizes
the client is in the drivers seat; therefore, when an impasse is reached, or the
concession-making slows, the judge addresses the clients.
A related causation exists in attorney mediations, but the pattern and tone
are different. Here the client views the mediating attorney as a lawyer who at
18 JAMES A. WALL, JR.
best is neutral, but who teams up with the opposing lawyers against him or
her. Therefore, the client does not want the mediators advice and their attor-
neys must proffer the offers and concessions. Noting this, the mediator con-
cludes the attorney is in the drivers seat and addresses him or her.
Conclusion
A balanced article would follow the discussion of the joys of eld research
with a delineation of the costs, subsequently compare the two, and announce
the net benets or costs. As you might expect, this is not to be a balanced
article. I do not make cost-benet analyses when considering eld research,
and I am in good company: Michelangelo didnt consider the costs and
benets of sculpturing David; rather he took a chisel to the marble and
chipped away everything that did not look like David. Bruce Springsteen does
not think like an accountant when he chooses to belt out, This Guns for
Hire. Al Pachino probably did not conduct a cost-benet analysis as he
starred in A Scent of a Woman. Bill Cosby and Robin Williams certainly
do not choose to continue in stand-up comedy routines because of the net benets.
And John LeCarre, just as Stephen King, writes because of the passion for
the art, not because the benets outweigh the costs.
Likewise, I have a passion for eld research. When I rst begin a project,
I think about it continually day and night with a resultant number of
sleepless nights. Then I begin to investigate to subject, losing more sleep
because of my excitement. Excitement grows as I ddle with a real-life phe-
nomenon that I can see, hear, smell, and feel, interacting with participants
who are similarly excited that I am studying them. They want a better under-
standing of their activities; they want to improve them and usually they want
me to be successful. This initial excitement, cultivated by the interaction with
the participants, tends to ow toward a passion as my waywardness, mistakes,
and stuckness give way to discovery and progress.
Yet, experience has shown me that this zeal for eld research like most
passions must be guided with several cautions so that it serves research
rather than making the researcher its victim. With this goal in mind I proffer
the following admonitions:
1) Do not get excited for the sake of excitement; because if you do, you will
waste a great deal of time and energy. When I was a graduate student, I asked
my mentor, Stacy Adams, how a social scientist maintains excitement about
social phenomena for a lengthy period of time. The leathery wrinkles of his right
cheek indicated that I had asked the wrong question and his response veried
THE JOYS OF FIELD RESEARCH 19
my conclusion. Im very surprised you asked that question, were his exact
words.
I had posed it because we, at the time, were studying personal space, which
was a dull topic to me. Concluding that I had asked the wrong question which
perhaps indicated some laziness on my part, I knew I had to offset these with
a born again excitement in the project. So I became excited and then
wasted about a year on a dull topic instead of moving to boundary role and
negotiation research, where I had a truly exciting dissertation (Wall & Adams
1974).
2) The second admonition is to avoid believing that a topic is important because
you have a passion for it. Cognitive dissonance explains why each of us believes
there is high value in what we do. Yet it can lead us to self-delusion; that is, we
can be studying something that is unimportant, but we think it is important.
Let me be more precise. Studying unimportant phenomena is ne, because
there is a possibility that the unimportant will become important. Are we not
glad, for example, that Flemmings mentor did not tell him to quit fooling
around with mold and to move on the something bigger?
Also, it is acceptable to study unimportant phenomena when it is fun to do
so. To me, this perquisite along with not having to maintain a xed sched-
ule in the ofce is a major luxury of our trade. Personally, I enjoy study-
ing unimportant phenomena, such as the electron-microscope images of the
protein crystals that my wife has isolated from some of her bacteria. The
crystals are bright orange and more interestingly have the same angles as a
large calcite crystal on my desk. I enjoy studying the interaction between cats
and snakes. A female cat, I have noted, will cautiously avoid a snake or a
rope that moves like a snake; however, she will very contently observe her
kittens playing with a snake. I enjoy studying BATNA levels. In tinkering
with these, I have found that negotiators do not abandon a negotiation when
the BATNA exceeds the probable net outcome of the negotiation. Rather, the
BATNA level must be about 11/3 times the negotiation outcomes.
Returning to my admonition, from these odd examples, I contend it is
appropriate to be excited about unimportant phenomenon because it is fun and
may eventually be fruitful. But it is inappropriate to believe that a phenome-
non is important simply because you are excited about it. To engage in such
delusion sacrices ones professional objectivity.
3) As a third suggestion: Abandon some projects. Admittedly, I am better at
advising this than doing it, but I know that for some eld research projects there
comes a time for abandonment, because, for example, the phenomenon cannot
be understood or a process cannot be divided into manageable parts. It is too
20 JAMES A. WALL, JR.
complex or so simple that its trivial. The process unfolds so quickly that the
independent variables and dependent variables seem meshed. The participants are
uncooperative. There are an inadequate number of research sites. You run out of
money, or the project quits being fun.
How do you do it? My approach is to put all my notes, articles, books, and
data on the project into some Xerox boxes, walk to the nearest dumpster, toss
them in and go for a long hike.
Two activities allow you to make such a clean break. One is a back-up
eld project. Always have two or more projects going at once so that you can
abandon one without remorse or panic when it is necessary. The second
activity is teaching. This is not only a service to students and probably your
major contribution to society. It also provides you with emotional backing
when you must abandon a eld project.
4) As a nal area for caution, do not get so excited about the subject matter that
you are sloppy in writing it up for publication. Because most readers, reviewers
and editors are not as excited about your project as you are, they need to be con-
vinced. Their conversion requires that you get their attention, explain why the
phenomena under study is important, present lucid details on what you did nd
and tie up the package neatly at the end. Moreover, remember that writing is
as Stephen King puts it work. Set aside adequate time for writing because
it takes ample time. Then focus, revise your work, have young colleagues review
it, and then revise, revise, revise. Lay it aside for a month, then revise, revise,
revise.
In closing, I shift from the four cautions back to my major theme: Field
research not only allows scholars to contribute to our knowledge pool; it can
also be a lot of fun. So, relax or get excited and enjoy it.
References
Arunachalem, Vairam, Lytle, Ann and Wall, James (2001). An evaluation of two mediation
techniques, negotiator power, and culture in negotiation. Journal of Applied Social Psycho-
logy, 31: 951980.
Pirsig, Robert (1974). Zen and the Art of Motorcycle Maintenance. London: Head.
Sherman, Michael (2002). The shamans of scientism. Scientic American, June, p. 35.
Wall, James (1981). Mediation: An analysis, review, and proposed research. Journal of
Conict Resolution, 25: 157180.
Wall, James (1986). Bosses. Lexington, Mass.; D.C. Heath.
Wall, James and Adams, Stacy (1974). Some variables affecting a constituents evaluations of
and behavior toward a boundary role occupant. Organizational Behavior and Human
Performance, 11: 390408.
Wall, James and Callister, Ronda (1999). Malaysian community mediation. Journal of
Conict Resolution, 43: 343365.
THE JOYS OF FIELD RESEARCH 21
Wall, James and Lynn, Ann (1993). Mediation: A current review. Journal of Conict
Resolution, 36: 160194.
Wall, James and Druckman, Daniel (2003). Mediation in peacekeeping missions. Journal of
Conict Resolution, 47: 693705.
Wall, James, Stark, John and Standifer, Rhetta (2001). Mediation: A current review and the-
ory development. Journal of Conict Resolution, 45: 370391.
How Much Do We Know About Real Negotiations?
Problems in Constructing Case Studies
DAVID MATZ
the published reports. The parties in fact were much closer to agreement than
reported, and many of the negotiators believed that with four more days of
negotiating a framework agreement on most though not all of the major
issues could have been reached. The newspaper account of modest progress
had thus muted what, in my research, became the central question: Who ended
the negotiation and why? Elsewhere, I have written about the actual results
of the negotiation and my conclusions about what ended it (Matz, 2003a,
2003b). Here I want to focus on some of the problems I encountered in learn-
ing about those results.
My research began eight months after Taba ended. There were by then two
memoirs and several newspaper interviews and analyses by and with mostly
Israeli participants (Beilin 2001; Sher 2001); there was also one scholarly account,
done by an Israeli with strong contacts among the Palestinians (Klein 2001).
There were twenty eight participants at Taba; I interviewed seventeen, of
whom 11 were Israeli. The Palestinians were harder to reach because some
were traveling out of the region, and because the politics in the fall of 2001
made travel difcult. (One telephone interview proved to be an unsatisfactory
solution.) Interviewees from both sides were equally forthcoming about issues
in the negotiation process, but the Palestinians were much more circumspect
when asked about the larger context. This imbalance of sources certainly
made me wary of imparting bias to my results, but my practice of cross-
checking within teams and between teams gives me some assurance that this
weighting did not itself lead to errors in my narration or interpretation.
Almost all negotiations experience the tension between publicity and secrecy.
Occasionally, as at Oslo, the entire process including its very existence was
secret until it is completed. This is rare. Nonetheless, negotiators most often
work to keep most of their moves out of the public eye during the negotiation
period itself, and sometimes permanently. Privacy is an essential component
of good negotiating (Pruitt 1999). The tension exists because of the pressure
applied by a negotiators constituencies or audiences, groups that may hold a
range of views about the negotiation, from full support to full hostility. Pri-
vacy blunts the capacity of these constituents to shoot down a component of
an agreement before it can be presented in the attering light of a full agree-
ment. Privacy, moreover, allows negotiators to explore in a tentative, testing
manner, what the other side will offer and accept. It allows either side to hint
at a willingness to change position, without committing itself to such a change.
HOW MUCH DO WE KNOW ABOUT REAL NEGOTIATIONS? 25
Privacy thus enables the parties to use, even create, their own subtle commu-
nication process. Taba negotiators told me that in four eyes (i.e. in private
conversation) they knew what the other side was willing to do, though there
was no specic sentence uttered by the party in question to which such an
inference could be attached. In an interview in Haaretz eight months after Taba,
Shlomo Ben Ami (Israeli Foreign Minister under Barak) described a meeting
with Abu Ala (chief Palestinian negotiator at Taba), using telling phrases to
explain how he had inferred receptivity among the Palestinians: they showed
readiness, the feeling was, it could be assumed (Ben Ami in Haaretz
2001)). Though this process of inference is vulnerable to misunderstanding and
duplicity, exploration without commitment is nonetheless essential for good
negotiating. And only privacy enables this to occur.
One result of this secrecy is that scholars and journalists are dependent in
large measure on what the negotiators are willing to tell. And negotiators have
a number of incentives not to tell the public the truth. In addition to the usual
problems of faulty memory, self serving perspectives, and scores to settle, rec-
ollections of negotiations raise special problems. Whatever the collaborative
energy in a negotiation, no such process can be free of competition. It may be
that people drawn to negotiation have already a well developed competitive
sense, and it may be that the negotiation process (with its inevitable supply
of omniscient Monday-morning-quarterbacks) further exaggerates that sense.
(Perhaps, the act of memoir writing enhances the competitive spirit yet fur-
ther.) Whatever the cause, a perusal of negotiator memoirs makes clear that
competition does not end with the negotiation (see Kissinger 1982). Negotia-
tors memoirs often exhibit an apparent need to show superiority to the other
side and frequently to ones own teammates; to show that one was not duped,
and to show that one did the best that anyone could. These incentives exist
with all negotiations, but they are especially strong when the negotiation, as
at Taba, is one chapter in an on-going process.
Participants efforts to communicate what happened in a negotiation are not
made easier by the subtlety of the forms of communication inherent in the artof ne
gotiation, as suggested above. Body language, tone of voice, pacing, energy level,
the impact of humor all have much to do with negotiating success, but they
are extremely difcult for authors to recapture after the event. The bold judg-
ment (He never wanted a deal.) or the quoted dialogue (When he said that to
me, I told him that. . . .) have an impact on the reader that can overshadow
what may have been a more signicant ow of communication at the table.
On some occasions the competition between negotiating teams is subordi-
nated to the competition between the negotiators on the one hand and the
audiences each negotiator needs to face on the other. Whatever the degree of
26 DAVID MATZ
hostility and distrust between two sides, negotiators need each other and this
often creates strong bonds of allegiance between them. One upshot can be a
collaborative falsication to the public. Oslo is of course one such instance
in which both sides conspired to hide from, and lie to, the press. Another
occurred at the press conference closing Taba where the leadership of both
sides agreed explicitly to tell the press, falsely, that the sides would not budge
on the Palestinian commitment to a right of return and the Israeli resistance
to it. In reality, they had reached most of an agreement.
Attempting to learn what actually occurred in a negotiation is inuenced by the
availability of writings produced during the negotiation. Sometimes there are
position papers and proposals, drafts and redrafts. But often these do not exist
or are not available. At Taba there was a marked aversion to putting anything
in writing. As negotiating progress was made, the idea of drafting proposals
was proposed and was usually shot down, by both sides at different times.
But some writings were produced. Since the location of particular bound-
aries was a central part of the negotiation, the use of maps would seem to
have been inevitable. But which maps were put on the table has been a mat-
ter of assertion and denial. Though there has been much talk of various maps,
only half of one Palestinian map has so far as of 2001 been allowed into
the public light (Klein 2001); the Israelis put one map on the table, though
this has been disputed by the Palestinians (Enderlin 2003).
One group of negotiators at Taba those dealing with the refugee ques-
tions, did produce a great deal of writing, mainly in the form of proposals
and counter proposals, some drafted by one side, some drafted jointly. There
would appear to have been at least four such documents, and there are sug-
gestions that there were yet more. Of these, only one has appeared in print,
showing up in Le Monde, eight months after the negotiation ended (2001).
In short, scholars and journalists have no direct access to what occurs in a
negotiation, important documents may be misleading or missing, and the eye
witness participant/reporters have numerous incentives to distort their reports.
Of course, cross checking of sources (interviews, documents, memoirs) when
there is a high level of consistency does allow one to develop a reasonably
reliable picture of at least some portions of the negotiation. But inconsisten-
cies in the sources are frequent, and thus unsurprisingly one is rarely free of
the need to use a standard of plausibility as well as ones imagination.
Let me now give an account of one episode in the Taba negotiation in which
inconsistency of reports raised questions, and then describe how I have tried
to determine what really happened.
HOW MUCH DO WE KNOW ABOUT REAL NEGOTIATIONS? 27
The Taba negotiation began on a Sunday evening, and for two days things went
very well. On Tuesday afternoon however, two Israeli restaurant owners vis-
ited a Palestinian town and were killed by Palestinians. Prime Minister Barak
then made the decision to recall the cabinet ministers on the negotiating team.
Gilad Sher (Baraks chief of staff, and one of the Israeli heads of the nego-
tiation delegation) describes a dinner on Wednesday evening at which he, two
other members of the Israeli delegation (both former members of Israeli secu-
rity services), Abu Ala (head of the Palestinian delegation) and Mohammed
Dahlan (head of Gaza security and member of the Palestinian delegation)
attended (2001). The dinner occurred at a sh restaurant 10 miles south of Taba.
Sher makes clear that no one else, other than some Egyptian coordinators,
knew of the dinner, and no one (other than Barak and Arafat) was told of the
dinner after it occurred. Sher reports that the purpose in arranging the dinner
was to learn if Arafat was prepared to reach an agreement. The Israelis offered
Abu Ala an airplane to y immediately to Arafat for an answer. According to
Sher, Abu Ala answered: The master of the house does not want an agreement.
Reading of this dinner I was puzzled. Sher presented himself as having
seen Taba, from the time it was rst proposed, as a somewhat empty negoti-
ating effort (too little too late), at best an opportunity to freeze dry the posi-
tions of the parties for use in later, presumably much later, negotiations. With
such a pessimistic view, why call a melodramatic meeting? In our interview
Sher sloughed off any inconsistency and said it had made good sense to have
the dinner.
A week or so later I interviewed Israel Hasson, one of the Israeli former
security ofcials present at the dinner. Even before my rst question, Hasson
volunteered to tell me of the dinner, but never mentioned Shers presence,
though I asked him directly about it. Hasson described the dinner in almost
the identical words used by Sher. When this difference in attendance surfaced
some weeks later Hasson called to say that he, Hasson, had initiated the
dinner, and after Abu Ala had answered as quoted above, Abu Ala asked
to have Sher join them. Hasson then contacted Sher, who drove down and
had the same dialog again. The issue of who was present at the dinner is not
in itself important for the issues raised by the Taba negotiation, but this
conict in the testimony does support skepticism about some of the versions
of what actually happened at this dinner.
In an interview, Abu Ala recalled the dinner but could not recall who was
there. When asked about the master of the house comment he said that
Arafat had told him the night before Taba began that if Abu Ala could reach
28 DAVID MATZ
an agreement and would recommend it, he, Arafat, would sign it. Since one of
the core questions at Taba and throughout all Israeli-Palestinian negotiation
does Arafat want a deal? is raised by this event, the question of what Abu
Ala said at the dinner justies some scrutiny.
There are, I think, three possible interpretations of these accounts. One is,
that the Israelis invented Abu Alas admission altogether in order to blame
Arafat should the negotiation fail. A second is that, the Israelis may have
asked a question about Arafats intent, the Palestinian leader, a master of the
shrug, may have responded with a physical gesture, and the Israelis may have
interpreted this to mean Arafat would sign nothing. And third is that, the story
occurred just as dramatically as the Israelis told it.
What to do with the various contradictions? Though in retrospect, Sher pre-
sents a blas account of his own expectation of Taba and of his own perfor-
mance there, on Monday evening Sher saw that some agreements at Taba
were possible. He called Barak and told him so. (On Tuesday some Palestin-
ians were also seeing agreement possibilities and called their Ramallah head-
quarters asking for additional staff.) On Wednesday, following the killings,
Barak left open whether he would be willing to re-start the talks. But for
Barak to end the talks would have left the blame for termination solely on
him (though he of course could in turn blame the terrorists and thus Arafat).
As such, an admission from Abu Ala that Arafat would accept no agreement
would be handy to balance the blame. An alternative but not inconsistent
Israeli goal for the dinner would have been to see if the professional inten-
sity of the Palestinian negotiators at Taba (commented on by several Israeli
negotiators at the time) reected a directive from Arafat to reach a deal.
However, asking Abu Ala the question directly would have risked falling
into a trap of their own devising: if the answer had been that Arafat was open
to a deal that week, any failure in the negotiation would have fallen more
heavily on Barak. It is possible that the Israelis did raise the question of
Arafats intent, though perhaps indirectly. What is not plausible is that Abu
Ala, with Dahlan sitting there, would answer it in the blunt, explicit way he
is quoted. As a constant in Israeli-Palestinian negotiations has been the need
to blame the other side for any failure, how likely is it that the head of the
Palestinian delegation in the middle of the negotiation would tell major play-
ers on the other side that his chief wants no agreement? What could he gain
by such an admission? Though tensions within the Palestinian team were so
intense that anyone at some moment might have blurted out anything about
anyone even Arafat, I think it unlikely that Abu Ala, a very experienced pro-
fessional, would do so. Also, why would Sher and Hasson keep such a strik-
ing admission secret from the Israeli negotiating team for the rest of the
HOW MUCH DO WE KNOW ABOUT REAL NEGOTIATIONS? 29
week, and from the press even after Taba was over? Considering the impor-
tance of blame in the negotiations, this silence is striking. The story showed
up in public only in Shers memoirs published 8 or 9 months after Taba.
My view is that at the dinner in the desert the Israelis opened the topic of
Arafats intentions and Abu Ala responded with a shrug. Since Taba ended
without a mutual blaming orgy as there had been after Camp David this
story stayed in Shers notes until he came to write his memoir. In the writing,
as is not uncommon, the narrative became less ambiguous than the original
event, i.e. the shrug became a speech. With what clarity Sher conveyed the
story to Barak at the time of Taba is uncertain, though Baraks failure to men-
tion it to me when it would have been useful (see below) suggests that Sher
did not relate it to Barak in as stark a way as he did in his memoirs.
I would like now to turn to a more central concern of my research. There
are two generally accepted and inconsistent understandings about where the
parties stood when Taba ended. According to one, they made progress and ran
out of time; according to the other, no progress occurred because the only goal
of Taba negotiators (shared by both sides) was to make Barak look good for
some voters in the coming election. My understanding, based on interviews and
on piecing together the various accounts, is far more optimistic than either
story. I asked the negotiators: If you had four more days in which to nego-
tiate, could you have reached a framework agreement on your topic? On all
issues, except two very important ones, the answer was afrmative. (The two
exceptions were control of the holy places in Jerusalem, and the process for
ending the conict.) This assessment was created by matching each sides
estimation, issue by issue, of what a settlement could look like.1 It was also
supported, if obliquely, by the joint zeal of the parties in keeping secret the
documents and maps produced at the negotiation. Such systematic secrecy can
be explained only by assuming that these papers contained serious concessions.
Even I, however, optimist-in-chief for the Taba negotiation, have to ac-
knowledge several qualications. A negotiators inference from the behavior
of his opposite number that an agreement was possible leaves open the pos-
sibility as had happened before Taba that the negotiator was reading the
signals incorrectly, that his opposite number would change his mind, or that
the other side was intentionally deceptive. Even if the signals were read cor-
rectly, no one on either side had, when the negotiation ended on Saturday, yet
pulled together all the potential agreements and reviewed the entire package.
This process would have been full of politics, drama, and guessing about
what the other side would do. In short, it would have been unpredictable.
Perhaps most importantly, no one believed that an agreement on the holy
places was possible, and this failure was so signicant that it could have
30 DAVID MATZ
derailed everything else. Nonetheless, I believe the evidence shows that the
parties at Taba were much closer to an agreement than is generally believed,
and that a framework agreement was possible with a few more days of nego-
tiating. Since the record is consistent that the negotiation ended on Saturday
at the initiative of the Israelis with the uncomplaining concurrence of the Palestin-
ians, the question is: Why?
wanted no deal. That he did not mention the Wednesday night meeting in the
desert is a hint, as suggested above, that he had not been aware of it at the
end of Taba, and perhaps, if he had not read Shers memoir, he was not aware
of it at all.
Believing that there was more to learn about Baraks thinking than I could
get directly from him, I decided to ask some experts. I began a tour of people
who claimed to be Barak insiders, people who knew his mind. From these
sources I gathered thirteen explanations for why he ended the Taba negotiation.
His wife did it. She never liked the idea of Taba and nally persuaded
Barak that it was a political mistake.
His political advisor, James Carville, did it. He convinced Barak that an
agreement at Taba would cost him any chance at the election.
Even before Taba began, he had lost the conviction that a historic break-
through was possible. He instead focused on negotiation only as an elec-
toral ploy.
He had a history of getting cold feet when an agreement started to look
possible. Thats what he did with Syria and he did it several times with the
Palestinians. Put more positively, when he got very close to an agreement,
the concessions made him worry deeply for the future of Israel.
His characteristic pattern as prime minister had been the zigzag: move
toward a goal and then away from it. Who knows why?
He really wanted an agreement with the Palestinians but was so maladroit
as both politician and negotiator that he missed opportunity after opportu-
nity. (This item has many examples and subcategories.)
He had wanted an agreement at Taba, but having learned (on Wednesday
night) that Arafat didnt want a deal, he changed his focus to one of mak-
ing the world understand Arafats real intent.
He decided that no deal was possible at Taba and that negotiations would
have to continue later, before and after the election.
Ben Ami (Israeli Foreign Minister) and Abu Ala (Head of the Palestinian
delegation) agreed at Taba that any further progress in negotiating would
require breakthroughs at a leader-to-leader summit, and this advice con-
vinced Barak.
His negotiators at Taba were too desperate to achieve an agreement and
were giving away the store. They were making deals neither he nor the
Israeli public could or should accept.
Individual Israeli negotiators were making what amounted to private deals
and he was not sure what they were. He felt himself to be losing control
of the process, worried that he would be trapped with whatever they agreed
HOW MUCH DO WE KNOW ABOUT REAL NEGOTIATIONS? 33
to, and thus ended the negotiation to avoid being responsible for a situation
he had not anticipated and could not control.
The Israeli public would never accept any deal from Taba and a negative
popular vote would clearly present Israel as the reason for no peace.
If the Israeli public approved an agreement by a small margin, it would cre-
ate a storm of opposition from the nationalist right wing that could lead
Israel into civil war.
What does one do with this motivational smorgasbord? Each has at least some
inherent plausibility, each can claim someone up close who says its true, and
each has some smattering of facts that can be adduced to support it.
Selecting a few elements of the context in which Barak made his decision
may help. We know that Barak expressed no optimism about sending a team
to Taba in the rst place. He seems never to have given his negotiating team
one clear mandate, and his various utterances were thus interpreted by differ-
ent negotiators according to their own predispositions. Pulling back his main
negotiators on Tuesday suggests a softness in his commitment to reach an
agreement at Taba; still, though he could easily have let the negotiation end
right there, he let the team return to Taba on Thursday. Many of the negotia-
tors at Taba were important to his election campaign and, being at Taba, were
not able to carry out their campaign functions. Some of those negotiators
recalled the pressure to get back to work.
Pulling this together into an interpretation of Baraks thinking on that
Saturday, I would conclude that he just wore out. His optimism and drive to
reach an agreement were not enough to overcome the pessimism and skepti-
cism all around him, and in him. Put differently, nearly everything on the
above list weighed on him. Perhaps we should not choose from the list, but
rather see it as an array of lures beckoning Barak not to negotiate further at
Taba. Taken together (even where they are inconsistent with each other) they
describe a climate resisting further negotiation, impacting his viewpoint and
thus his decision. In this climate it would have taken a leader of extraordi-
nary strength, willing to take larger risks, with something larger than Baraks
tiny and diminishing base of popular support, to continue the negotiation and
push for agreement. It is thus not surprising that Barak would see it as pru-
dent to adopt the conventional campaign stance: assert that he had done
everything possible for peace.
About Arafats intentions I have less evidence. Abu Ala says that Arafat
was willing to sign an agreement. Arafat sent a good negotiating team, a team
that clearly thought agreements were possible and that worked diligently through-
out the week, a team that tried to keep the Israelis at the table on Tuesday
34 DAVID MATZ
afternoon when Barak was calling his cabinet ministers back from the nego-
tiation. On the other hand, we have the puzzling account of the Wednesday
night dinner that might suggest Arafats lack of commitment. And we have,
nally, the ready acquiescence by the Palestinian leadership on Saturday to
the Israeli decision to end the negotiation, suggesting that the Palestinians
too whether reecting Arafat or not did not choose to negotiate further.
Arafats vicious speech in Switzerland the day after Taba ended attacked
Barak and the Israelis, and made no mention of Taba. When I asked Pales-
tinians close to Arafat what they made of this confusing picture and why he
seemed willing to end the negotiation, they professed no insight into his inten-
tions. As Arafat is a master of the masked motive, I take their confession
seriously and can only echo it.
Conclusions
One goal for this essay has been to identify special problems in understand-
ing what happens in a negotiation. The imputation of motive is one concern.
Another is the necessity of comparing many sources (interviews, memoirs,
documents) to learn what happened, of cautiously accepting a consensus of
sources when there is one, and of assessing inconsistent accounts by explicitly
stated tests of plausibility. Any account that seems to rely on one source, or
that does not identify differences among sources is, I would suggest, suspect
on those grounds alone. These are hardly novel insights, but they have not
necessarily been the rule in studying negotiations. Let me close with two
examples, from other negotiations, of (mis)understandings that have become
part of our scholarly lore.
Social Conict (Rubin, Pruitt, and Kim 1994), in its analysis of interests,
cites the example of Henry Kissinger mediating the provision of food and
medicine to Egyptian troops surrounded by Israeli troops after the Yom
Kippur War. Kissinger in this account after careful analysis. . . . concluded
that Israel wanted actual control of the road, whereas Egypt wanted only the
appearance that Israel did not control it. . . . A bridging solution was found
that involved stationing Israeli soldiers unobtrusively on the sides of the road
(so that they actually controlled it) and having UN checkpoints on the road
itself (to give the impression of international control). The authors cite Golan
(1976) but Golan actually provides a rather different picture, especially when
supplemented by the report of that negotiation done by the the Insight Team
of the London Sunday Times (1974). First, on a minor matter, Kissingers
careful analysis came after he was told by an Israeli aid that the solution
HOW MUCH DO WE KNOW ABOUT REAL NEGOTIATIONS? 35
lay in separating the issue of control of the road from allowing aid through
to the Egyptian army (Golan 1976). More importantly, this part of the nego-
tiation was, as described by Golan and The Times, a four way power ght.
The parties were Israel, Egypt, the US, and the UN, and the topic was the
control of the road to the Egyptian soldiers. Egypt gave in early in the
process, recognizing as Golan saw it, the limits of its power. Kissinger rst
proposed that the UN control the road, and Israel objected. Then he proposed
one UN inspection station and Israel agreed. Kissinger later converted that to
three UN stations and pressed Israel very hard to take it, though there is no
indication that Sadat sought this (Brecher and Geist 1980).2 Furthermore, as
the checkpoints were being constructed there were st ghts between Israeli
and UN soldiers over their construction (Sunday Times 1974). The result of
these various tussles was that Israelis were in the UN checkpoints inspecting
whether the material being sent to the Egyptian soldiers was only humanitar-
ian in nature. Moreover, according to Kissinger what made these concessions
effective was that, as part of the package the Israelis achieved the return of
their prisoners (Kissinger 1982). In summary, the bridging solution gave
Israel everything it had originally sought, gave Kissinger a deal that allowed
the process to go forward and gave him much of the credit for achieving it,
gave the UN a hollow role that might charitably be called symbolic, gave
Sadat not so much a g leaf as the barely visible appearance of a g leaf,
and can, with no undue cynicism, be seen as a conspiracy among the Egypt-
ian, Israeli, and US governments to fool the Egyptian people. The term
bridging solution suggests that each side used problem solving technique to
satisfy its interests. I think it accurate to say that this bridging solution
was a thin cover that reected perfectly the power relations of the parties.
To focus now on a later chapter in the aftermath of the Yom Kippur War,
the 1978 Camp David negotiations dealt with the Israeli withdrawal from the
Sinai. In their account in Getting To Yes, Fisher, Ury, and Patton (1981), con-
trast the parties opening positions (Egypt wanted all the Sinai returned and
Israel wanted to keep a part of it) with the negotiated result (Egypt got full
sovereignty, but with limits placed on where Egypt could put any weaponry).
This development is used to illustrate the value for seeking a settlement of
interests, Israels being security, Egypts being sovereignty. The difculty
here is that the basic principles that satised these interests (limits on the use
of weapons, demilitarized zones, U.N. control of yet other areas) were already
in place on the ground when Camp David began. It was the formula the
Egyptians and Israelis had agreed to with Henry Kissinger as mediator in
1974. Signicant modications of the plan were indeed negotiated at Camp
David, including moving the lines separating the parties, substitution of an
36 DAVID MATZ
international force for the UN, and most contentiously the time table for
phasing out this plan. But the concepts were already in the note books of
President Carters mediating team before the summit began, and it is difcult
to believe that the two sides had not themselves anticipated the Carter team
proposal in their own planning for the negotiation.3 Again, this was not prob-
lem solving and an exploration of interests. It was taking an obvious solution
staring the parties in the face and modifying it to t the situation.
These examples, and the Taba account more generally, raise the question of
what we know about what goes on in real negotiations. They suggest that
we may well know less than we have assumed. As a result we may need to
review some of our theory to see how much of it may be based on such
assumptions.
Notes
1. The negotiation concerning Maale Adumim provides an illustration of how this matching
process can be done. Maale Adumim is a large Israeli settlement (25,000 people) situated
east of Jerusalem which blocks much north-south trafc for a new state of Palestine. The
Palestinians had been intensely fearful that a Palestinian state would be divided by Israeli
lands and roads, leaving it without a coherent land mass, and Maale Adumim raised this
specter. Nonetheless, at one point, the Palestinian negotiators conceded that Maale Adumim
would be annexed to Israel as part of the overall agreement. Israel argued for more area
around the settlement, and connection to other settlements, and the Palestinians then
rescinded their concession. Unofcially however, the Palestinians acknowledged that
Maale Adumim could be part of the annexation if the parties could work out the scope of
the surrounding area. Thus, when the negotiation ended, the ofcial Palestinian position
refused Israeli annexation of Maale Adumim, but unofcially negotiators on both sides saw
the trade that would have reversed that position (access roads for the Palestinians which
the Israelis would grant and limited growth space for Maale Adumim which the
Palestinians would grant).
2. Brecher conrms Kissingers own interests in getting the results he thought best for the US.
3. Compare Quandt, W. (1986) Camp David: Peacemaking and Politics. Washington D.C.:
Brookings Institution (p. 382) with Kissingers Years of Upheaval (p. 839.) Supplemented
with an interview with Herman Eilts, then U.S. Ambassador to Egypt, who was present at
Camp David.
References
Asfour, Hanan (2001). Ben Amis Occupation Syndrome. Haaretz Magazine, October 19.
Beilin, Yossi (2001). Manual For a Wounded Dove. (Hebrew) Tel Aviv. Miskal-Yidioth
Ahronoth Books and Chemed Books.
Ben Ami, Shlomo (2001). End of a Journey. Haaretz Magazine, September 14.
HOW MUCH DO WE KNOW ABOUT REAL NEGOTIATIONS? 37
Bernard, H. Russell (1994). Methods Belong To All Of Us, in Robert Borofsky, editor,
Assessing Cultural Anthropology. New York: McGraw-Hill.
Brecher, Michael, with Geist, Benjamin (1980). Decisions in Crisis: Israel 1967 and 1973.
Berkeley, California: University of California.
Eldar, Akiva (2001a). How to Solve the Palestinian Refugee Problem. Haaretz, May 29.
Eldar, Akiva (2001b). Sir, If You Please, Tear Down Your House. Haaretz, May 31.
Enderlin, Charles (2003). Shattered Dreams. Paris. The Other Press.
Fisher, Roger and Ury, William and Patton, Bruce (1991). Getting To Yes. New York. Penguin.
Golan, M (1976). The Secret Conversations of Henry Kissinger. New York. Quadrangle.
Holbrooke, Richard (1999). To End a War. New York. The Modern Library.
Insight Team of The Sunday Times (1974). The Yom Kippur War. Garden City, New York.
Doubleday.
Kissinger, Henry (1982). Years of Upheaval. Boston, MA. Little Brown and Co.
Klein, Menachem (2001). Shattering a Taboo: The Contacts Toward A Permanent Status Agreement
in Jerusalem 19942001. (Hebrew.) Jerusalem: The Jerusalem Institute for Israel Studies.
Malley, Robert and Agha, Hussein (2001). The New York Review of Books. (August 9)
Matz, David (2003a). Trying To Understand The Taba Talks. Palestine-Israel Journal of
Politics, Economics and Culture 3.
Matz, David (2003b). Why Did Taba End? Palestine-Israel Journal of Politics and Culture 4.
Pruitt, Dean (1981). Negotiating Behavior. New York. Academic Press.
Pruitt, Dean (1999). Secrecy, Ripeness, and Working Trust at Oslo, in Deborah Kolb, editor,
Negotiation Eclectics. Cambridge, MA. PON Books.
Quandt, William (1986). Camp David: Peacemaking and Politics. Washington, D.C. Brookings
Institution.
Rubin, Jeffrey, Pruitt Dean, and Kim, Sung Hee (1994). Social Conict: Escalation, Stalemate,
and Settlement, second edition. New York. McGraw-Hill.
Schiff, Zeev (2001). What Was Obtained at Taba Regarding Palestinian Refugees. Haaretz,
September 12.
Sher, Gilad (2001). Just Beyond Reach. (Hebrew) Tel Aviv. Miskal-Yidioth Ahronoth Books
and Chemed Books.
Studying Negotiations in Context:
An Ethnographic Approach
RAY FRIEDMAN
Ethnographic research methods have provided powerful insights into the way
negotiations and dispute resolution occurs. It was through the use of ethno-
graphic methods that Walton and McKersie (1965) developed their insight that
there are four dimensions of labor-management negotiations: integrative bar-
gaining, distributive bargaining, attitudinal restructuring, and intra-organizational
bargaining. In my work, ethnographic observations led me to see the impor-
tance of rituals in labor negotiations, and the way those rituals derive from
the role structure within which negotiators have to work (Friedman 1994;
Friedman & Gal 1991). In a study of mediation, participant observation en-
abled Kolb (1983) to see how mediators brought particular frames to their
work, rather than choosing frames based on the particular situation. Close
observations of mediators led Silbey and Merry (1986) to distinguish between
bargaining and therapeutic styles of mediation. Merrys (1989) use of anthro-
pological reports allowed her to distinguish between the legalistic model of
American mediators and the more socially-embedded structure of mediation
common to non-industrial societies.
This approach to negotiation research stands in sharp contrast to the more
frequently-used experimental method for studying negotiation. While experi-
ments are designed to isolate specic phenomenon of interest and con-
trol all other factors, ethnographic research looks at behavior in the total
negotiation context. While experiments can be done with nave negotiators,
ethnographic research targets experienced negotiators who carry with them
communal norms, ongoing relationships, and detailed knowledge about the
institutional context. While experiments are done in brief periods of time,
ethnographic research follows real negotiations over natural periods of time
often weeks or months. Experimental research can teach us a great deal about,
for example, psychological biases (Neale & Bazerman 1985), the effects of
personality (Barry & Friedman 1998), or the inuence of social motivations
(De Dreu, Weingart, and Kwon 2000). But what we can not learn from these
experiments is how experienced negotiators and mediators act in real, com-
plex situations. Nor do experiments, as Hopmann (2002) points out, capture
some of the most important, albeit subtle aspects of negotiation [such as]. . . . iden-
tify formulation (p. 73). For ve years, I studied labor negotiations by sit-
ting in negotiation sessions, attending caucuses on both sides, debrieng lead
bargainers after negotiation sessions, interviewing participants, and follow-
ing the negotiations over weeks and months (Friedman 1994). This pro-
cess has several signicant disadvantages it takes a great deal of time (in
classic studies, such as Whyte [1934] and Suttles [1968] researchers spend
years living in the communities they studied), and does not produce data of
a type preferred by many journal editors. Nonetheless, ethnographic research
can generate unique insights into negotiation that can not be gained in any
other way.
Consider, for example, the issue of constituent pressures. We know from
experimental research that when negotiators are observed by and accountable
to constituents they are more aggressive in their negotiation style (Carnevale
1985), at least within the U.S. (Gelfand & Realo 1999). But the relationship
between constituents and negotiators is far more complex and subtle. Lead
bargainers can control external pressures through appropriate choice of team
members. One union negotiator reported that he puts his most difcult con-
stituents (what he calls the bomb tossers) on the committee:
They are the guys that stand in the back of the meeting and they throw the bomb
out there that gets everybody riled up and then they sneak out of the meeting
when it comes to the strike vote. You grab them and you put them on your team
and you put them on the committee, put them on the hot spot. Get the guys with
the biggest mouth. Put them on the committee; put them on your side.
Another union negotiator talked about the need to continually test her ability
to control her side through saber rattling:
It is not rattling sabers to scare an employer. It is doing two things: one is test-
ing my organization to make sure I have the power. I want to know if I have
the power or not. My bottom line is going to be affected by how much mem-
bership support there is for those issues. The other reason is to remind the
employer that this is not my agenda, it is our members agenda.
The relationship between a negotiator and his or her constituents requires
careful management and relationship building. As one other negotiator put it:
You have to feel it. Each committee is different. Have they worked with me
before? A lot of it is trust. If they trust you, then they trust you to lead them.
You nd that most rank and le people dont want to say anything at the table,
so you give them all the freedom they want and you tell them that they are
encouraged to speak and most wont. Yet if I am talking about a custodian issue,
it is far more effective hearing it from the custodian than it is for me.
STUDYING NEGOTIATIONS IN CONTEXT 41
Yes, constituent pressures do affect negotiators, but that is just the starting
point. By studying real negotiators in context, we see how they manage those
pressures.
Similarly, we know that relationships affect negotiators, with different tac-
tics being used for friends and strangers (Sondak, Neale and Pinkley 1999).
We must bear in mind again that relationships between negotiators are very
complex. It is not just that you care more about the other side when you
know them well, but also that you have more tactical options if you know the
other side well. One negotiator used his relationship with the other side to
reign in an overly aggressive lawyer: John a manager called, and he said
this is b.s. I said I know, but no one told the lawyer that. He said, go back
to work and Ill settle the rest. I did. Then we got one of our best contracts.
If there is a strong relationship, there is more trust in both the validity of
the proposals made and less need for formal contracting. Two negotiators
explained:
Summers he knows me, I know him. Whatever he can extract from the com-
pany he puts on the table as a nal proposal. This is a guy I would trust implic-
itly. We will not bull shit each other.
Friedrich and Lamberts CEO is Tom OBrian. If he says this is what we agreed
on, Id sign it tomorrow. If I have a question about a clause I can call him and
hell x it. He has integrity.
Relationships can be a source of leverage, and provide alternative avenues for
communication. One union lawyer reported, discussing a particular manage-
ment labor lawyer, said I have only dealt with him myself one or twice I
think. He tends to be very legalistic actually. But the nice thing is he has a
client who is open to a getting-to-yes style. So you can kind of get around
Fred sometimes by convincing his principal. Fred is much more likely to say
dont do that, it will get you in trouble. Yes, relationships matter, but what
is also important is how negotiators assess the quality of relationships, the fact
that relationships with the other side may include both good and bad rela-
tionships, and how negotiators depend on and use good relationships in nego-
tiations. As these examples show, we gain insights from ethnographic research
that can not be obtained through experimental methods.
Before proceeding, I should be clear about what is distinctive about ethnog-
raphy. Ethnography has four features (Friedman & McDaniel 1998). First, ethnog-
raphy requires direct observation of the activity being studied in the case
of negotiations, sitting in on some or all of the preparation meetings, negoti-
ations, and caucuses. Second, it requires that the observation be in a natural
setting so that the researcher can uncover special languages, unique and
42 RAY FRIEDMAN
peculiar problems, and, more generally, distinct patterns of thought and action
(Van Maanen 1983: 13). Natural observation also allows the researcher to
see how people manage in situations where there are multiple, concurrent
pressures. Third, ethnography gives prominence to the words, interpretations,
and experiences of the people studied. Ethnographers want this information
because it allows them to examine the cultural knowledge, behavior, and
artifacts that participants share and use to interpret their experiences (Schwartzman
1993). Fourth, observations of how people behave, and how they interpret
their experiences, are shared with the reader. The goal is to enable the reader
to experience the situation much as the negotiator does.
That being said, ethnography is not at all the most common approach to
the study of negotiations. Ethnographic methods are very time consuming,
and require a different set of research and data analysis skills. The rest of this
article will be devoted to explaining some of the challenges in studying nego-
tiations ethnographically. What I will report is my own personal experiences
and research strategies.
Gaining Access
The main problem with gaining access is that fact that you, the researcher,
pose risks for the negotiators. Typically, information is tightly controlled in
labor negotiations (and, indeed, most negotiations; see Hopmann [2002] who
explains that there is often a legal responsibility to maintain condentiality in
negotiations). If you were to reveal to the other side what you hear in cau-
cuses, you could cause enormous damage to the negotiations. While reveal-
ing information will certainly not be a researchers intention, there are usually
social opportunities (such as when everyone gathers to get coffee before nego-
tiations begin). On such occasions, an inadvertent comment could reveal
something, or ones body movement and facial expression during negotiations
might reveal information. Also, the very act of having an observer present
may change behavior, making people self-conscious or hesitant to say what
they want. For these reasons it takes a tremendous amount trust to gain access
to a negotiation. Negotiators will want reassurances about condentiality and
ones credibility as a researcher.
Further complicating the situation is the fact that, in labor negotiations
at least, many people are involved. On several occasions, a lead bargainer
approved my study, but I had to explain myself all over again when I rst met
STUDYING NEGOTIATIONS IN CONTEXT 43
Being Unobtrusive
Once negotiations begin, great care must be taken not to attract too much
attention. On the one hand, it is impossible not to develop relationships with
people. You are with them for weeks or months, and they will want to get to
know you and ask your opinion and advice. Indeed, one of the really fun
parts of ethnographic research is that you encounter very interesting people.
Also, you do want to keep track of how they see the situation, what their
goals are, and how they interpret negotiating tactics on both sides. Still, the
more you interact, the more you risk becoming an actor rather than an
observer. If you take stands on issues or strategies, participants may not be
as open and accepting of your presence.
Another way that you may get unwarranted attention is through note-tak-
ing. It is already an intrusion for you to be present, but when they see you
suddenly whip out your pen and pad and start scribbling, that reminds them
that they are being watched. Some ethnographers prefer to wait until after
a meeting before taking notes, or go to the backroom frequently to write
down quick reminders. I prefer to take notes at the bargaining table, but I try
to take notes continuously even if nothing is really note-worthy at the
moment. If I am seen as just writing all the time, them my note taking gets
less attention.
What to Observe
what proposals were made. In some cases this is factual, while in other cases
it is interpretive. Fred may tell the other lead negotiator that the company can
not provide a certain level of benets, but it is also important to note the
intensity of this comment (is it matter-of-fact, or with great passion?), who
Fred is looking at when he is making that comment (is it just the other lead
bargainer, or each member of the union team?), and the reaction of the other
team (do they appear surprised or angry?). At the second level, I am record-
ing my questions and analysis. Why is Fred making this comment now? It
seems that he is trying to change the topic because he wants to avoid another
difcult issue for the moment. He is suddenly taking a new approach that is
different than his plan expressed in caucus maybe the unions prior com-
ment really was a surprise? So, notes can be very extensive, and one eye is
always on data gathering while the other is on interpretation.
Reporting Results
Unlike more traditional research methods, there are few established conven-
tions for reporting ethnographic research results. Van Maanen (1988) says that
much of it comes down to an ability to tell a compelling story. A compelling
story both makes the situation clear to the reader and makes the reader more
likely to trust the ndings a good story is hard to tell if there is not real,
in-depth knowledge of and understanding of a situation. As Golden-Biddle
and Locke (1993) put it, one must convince the reader that the research was
authentic, and that the interpretations are plausible. Validity depends not on
Alphas or Chi-squares, but in-depth explanation and clarity of writing.
The challenge is nding a way to use observations, quotes, and stories in
such a way that it leads towards a conceptual argument. Good ethnographic
work should lead to new theories. In their book, Walton and McKersie (1965)
used their eld observations to build a conceptual model of negotiation processes,
and connected their ideas to social science theory. In my book, I focused on
the ways role constraints affected attempts to change how people negotiate.
Kolbs (1983) study builds towards a model explaining factors that inuence
mediators choice of mediation strategy. Note, however, that ethnographic
work tends to be seen most often in books, not articles. It often takes more
room to provide the data for an ethnographic study than for a statistical study,
and many journals and their reviewers seem to be less accepting of ethno-
graphic data than survey data. There has been some publishing of ethno-
graphic studies of organizational issues in top journals (see, e.g., Barley
1996), but I do not know of a single ethnographic study of negotiation in
either OBHDP or JAP, two of the leading journals for negotiation research.
Scholars typically write books, or look for alternative outlets for their work.
Conclusion
Traditional research methods for the study of negotiation can be very pow-
erful, but that research has been criticized for decontextualizing conict
(Barley 1991). Ethnographic research offers an alternative approach which can
provide insights into the types of complex situations that negotiators really
STUDYING NEGOTIATIONS IN CONTEXT 47
face. This approach is not easy it can be more time consuming, costly, and
difcult than other research methods but the payoff comes from the way
these in-depth studies challenge scholars to develop new ideas and theories,
based on what really happens in negotiations rather than on the logical next
step in a series of experiments. Hopefully, this article will serve to encourage
others to venture down this path, and provide them with a few more ideas for
managing their ethnographic study of negotiations.
References
(* included general references for ethnographic research that are not cited)
Neal, M.A. and Bazerman, M.H. (1985) The effects of framing and negotiator overcondence
on bargaining behaviors and outcomes. Academy of Management Journal, 28(1): 3449.
* Schutz, A. (1967) The Phenomenology of the Social World. Evanston: Northwestern
University Press.
Silbey, S.S. and Merry, S. 1986. Mediator settlement strategies, Law and Policy, 8: 732.
Suttles, G.D. (1968) The Social Order of the Slum, Chicago: University of Chicago Press.
Sutton, R. (1994) The Virtues of Closet Qualitative Research. Stanford University: Draft.
Schwartzman, H. (1993) Ethnography in Organizations. Newbury Park, CA: Sage.
Sondak, H., Neale, M.A., and Pinkley, R.L. (1999) Relationship, contribution, and resource
constraints: Determinants of distributive justice in individual preferences and negotiated
agreements. Group Decision and Negotiation, 8(6): 489510.
* Van Maanen, J. (1988) Tales of the Field. Chicago: The University of Chicago Press.
Van Maanen, J. (1983) The Fact of Fiction in Organizational Ethnography. Pp. 3756 in
J. Van Maanen, ed., Qualitative Methodology. Beverly Hills, CA: Sage.
Walton, R.E. and McKersie, R. (1965) A Behavioral Theory of Labor Negotiations: An Analysis
of a Social Interaction System. New York: McGraw-Hill.
Whyte, W.F. (1943) Street Corner Society, Chicago: University of Chicago Press.
The Problem-Solving Workshop as a Method of Research
RONALD J. FISHER
Introduction
Authors Note: The author wishes to thank Herbert Kelman and an anonymous reviewer for
helpful comments on an earlier draft of this article.
research has been severely limited, and the potential of the approach in this
domain has not been realized. This article will identify the ways that the
problem-solving workshop can serve as a research method, and identify the
issues that must be addressed to increase its usage and utility in this direction.
rules, the third party provides a rough agenda for the sessions, starting with
an initial exchange of perceptions. The third party then moves on to an analy-
sis of the attributions, interests, and needs underlying incompatible positions
and escalatory interactions. Next, the participants engage in application and
development of insights and models of understanding and the creation of
ideas for peacebuilding and resolution. Finally, the third party urges partici-
pants to consider the constraints and resistances to these options. In other
words, the workshop follows the sequence of problem solving (with recycling
as required), which is common in many forms of human endeavor. Casting
the workshop as part of a research enterprise with an academic base was part
of the initial rationale that invited the participants into the role of conict ana-
lyst and collaborator in problem solving (Burton 1969; Kelman 1972). Subsequent
to the initial explications, more detailed and systematic guides have been
developed for organizing and facilitating problem-solving workshops (Burton
1987; Mitchell & Banks 1996). What has been largely ignored in the press
to develop the theory of practice for conducting workshops is their investiga-
tive potential for validating and creating theoretical notions about protracted
and violent conicts, and for operationalizing a form of action research with
implications for social change.
the general agreement of the participants that it was helping to fuel the arms
buildup on the island.
Thus, a generic third party intervention identied by Kelman and Cohen
(1976) is the provision of theoretical inputs to enrich the analysis, a function
identied as diagnosing the conict in my model of third party consultation
(Fisher 1972). Moreover, one of the objectives in this initial model was the
study of conict, by which the members of the third party team could learn
about the nature of escalated and destructive conict, partly as it is portrayed
in the analysis and interaction of the participants. This learning can include
the development of new understandings or new applications of concepts from
other domains to the study of conict. For example, in one of the Cyprus
workshops I organized, a member of the third party team developed the notion
of an identity dilemma, which the Greek and Turkish Cypriot participants appeared
to be exhibiting in their discussions. While there was a joint appreciation of
the importance of supporting and exhibiting their shared Cypriot identity,
movement in this direction put them into potential conict with their Greek
or Turkish identity. Thus, if becoming more Cypriot meant becoming less
Greek or Turkish, this left them vulnerable to charges by nationalists in both
communities that they were compromising their primary cultural identica-
tion. On a grander scale, it appears that the motivation to apply and develop
human needs theory, which arose mainly in international development, to the
study of conict occurred because participants in workshops repeatedly
revealed such needs when asked to engage in conict analysis. In other
words, when asked to go beneath the surface positions and interests in the
conict, participants would typically move to a discussion of needs for secu-
rity, recognition of identity, distributive justice, and so on that paralleled the
typologies of basic human needs developed in sociology and elsewhere. Thus,
the frustration of these fundamental requirements for human development and
their related fears (e.g., over a lack of security), came to center stage in the
theoretical analyses of protracted or deep-rooted, i.e., intractable, conicts offered
by theorists such as Burton (1990) and Azar (1990). Human needs theory is
now regarded as an essential element of conict theory and a common tool
for conict analysis in problem-solving workshops.
Problem-solving workshops have been seen from the beginning as a useful
complement to negotiations between conicting parties, assuming that trans-
fer effects occur between the unofcial and ofcial tracks. Kelman and Cohen
(1976) initially outlined the ways in which workshops could make contribu-
tions at all stages of negotiation from encouraging parties to enter negotia-
tions, to working on issues that are hampering negotiations, to discussing long
THE PROBLEM-SOLVING WORKSHOP AS A METHOD OF RESEARCH 53
nature of the dispute, the focus was on the issues that seemed to be sus-
taining the conict, both in terms of making it very difcult to resolve and
in terms of being resistances in the negotiation process. The agenda also
included the underlying needs and fears of the two communities, acknowl-
edgements and assurances that would assist in moving into a shared future,
the principles and qualities of a renewed relationship, and the development of
peacebuilding activities that would support a shift from peacekeeping to
peacemaking. By all accounts, the workshop intervention was a success in
terms of both process and outcomes. As a result of the workshop, a series of
cross-line meetings and joint committees of business leaders were initiated, a
bicommunal art exhibit was launched for a period of several years, and the
seeds were sown for the organization of a bicommunal committee to organize
peacebuilding activities, including training in conict resolution (Diamond &
Fisher 1995). The third party was established as an acceptable and competent
facilitator of intercommunal cooperation, and two further workshops were
held before funding dried up in the aftermath of the closure of CIIPS.
To evaluate the effects of the workshop as a social intervention, a variety
of indictors were used. Closing comments and post-workshop questionnaires
evidenced high levels of satisfaction and perceived utility. Participants indi-
cated that they had increased their understanding of the perspectives, ratio-
nales, and sensitivities of the other community, and that the third party team
had facilitated the interaction and injected new ideas that helped them better
understand the difcult issues in the conict. Pre and post-workshop ques-
tionnaires indicated that the increased understanding and respect led to
increased involvement in intercommunal contacts. Follow-up interviews one
and a half years later demonstrated that participants had maintained positive
evaluations of the workshop, and that most had continued their engagement
in bicommunal peacebuilding activities. In addition, six of the eight reported
that they had conveyed a positive evaluation of the workshop project to their
respective leaderships. This outcome was likely critical in garnering continu-
ing support for the workshop project.
Conceiving of the workshops as a form of action research was useful in
conceptualizing the project and in understanding its place in the ongoing con-
text of the conict. It also assisted in supporting the importance of evalua-
tion, both formal and informal, at each step of the way, so that future actions
could be planned in a collaborative and effective manner. For example, dis-
cussions at the main workshop and site visits following it were instrumental
in deciding to hold further workshops focusing on the role of education in
helping to maintain the conict and its potential for helping to de-escalate it
and rebuild the relationship between the two communities (Fisher 1997).
56 RONALD J. FISHER
The stance taken here is strongly supported by the work of Kelman, who
describes problem-solving workshops as the central tool in his program of
action research on the Arab-Israeli conict (Kelman 1979). In this context, the
goal of the workshops is to understand and overcome the psychological bar-
riers to change, to create new possibilities for negotiation, and to contribute
to reaching a settlement at the political level. According to Kelman:
Our work can best be described as an action research program, which combines
and integrates efforts at conict resolution in the Middle East with unique oppor-
tunities to observe and learn about the dynamics of the Middle East conict
and of international conict in general. Our action is designed to contribute to
conict resolution by creating opportunities for communication between inuen-
tial Arabs and Israelis under the auspices of our group of social scientists (1979,
p. 104).
In addition to workshops, the program of action research involves site visits
to the region, interviews with a variety of individuals, organizing and attend-
ing conferences, and other activities to learn about the conict and to estab-
lish the action researcher role.
Although the program involves both action and research, Kelman (1986)
notes that the action requirements have taken precedence over the research
requirements. In other words, the procedural and ethical necessities of orga-
nizing, implementing, and evaluating workshops are given priority, because
these ensure the viability of workshops and their contribution to conict res-
olution. For example, Kelman precludes manipulating workshop procedures in
order to test the effects on participants. Nonetheless, the research aspect is
critical to the enterprise. As Kelman observes:
. . . it is our role as researchers that provides the rationale and legitimacy of our
action involvement and that allows representatives of conicting parties to inter-
act with each other under auspices in ways that deviate from the norms gener-
ally governing their relationship . . . Thus our action requires involvement in a
research program just as our research requires involvement in an action program.
The unique advantage of this type of research is that it provides us the oppor-
tunity to make rich, detailed observations of ongoing processes of conict and
conict resolution, which would not be accessible to us unless we were engaged
in an action program (1986, p. 297).
The outcome of this form of action research for Kelman and his colleagues
is that it produces learning that support negotiations in a variety of ways and
at all stages of the negotiation process. For example, in 1986, Kelman noted
that Israeli and Palestinian participants had gained insight into each others
perspectives (concerns, areas of exibility, constraints, priorities), had learned
that there was someone to negotiate with, and had learned of positive changes
in the adversary and how they could promote further change through their
THE PROBLEM-SOLVING WORKSHOP AS A METHOD OF RESEARCH 57
own actions. These kinds of lessons are clear contributors to the prenegotia-
tion process that helped pave the way to the Oslo accords (Kelman 1995).
Concluding Issues
References
Azar, Edward E. (1990). The Management of Protracted Social Conict. Hampshire, UK: Dartmouth
Publishing.
Burton, John W. (1969). Conict and Communication: The Use of Controlled Communication
in International Relations. London: MacMillan.
Burton, John W. (1987). Resolving Deep-Rooted Conict: A Handbook. Lanham, MD: Univer-
sity Press of America.
Burton, John W. (ed.) (1990). Conict: Human Needs Theory. New York: St. Martins Press.
Chufrin, Gennady I. and Saunders, Harold H. (1993). A Public Peace Process. Negotiation
Journal 9, 2:155177.
de Reuck, Anthony V.S. (1974). Controlled Communication: Rationale and Dynamics. The
Human Context 6, 1:6480.
Diamond, Louise and Fisher, Ronald J. (1995). Integrating Conict Resolution Training and
Consultation: A Cyprus Example. Negotiation Journal 11, 3: 287301.
Doob, Leonard W. (ed.) (1970). Resolving Conict in Africa: The Fermeda Workshop. New
Haven, CT: Yale University Press.
Fisher, Ronald J. (1972). Third Party Consultation: A Method for the Study and Resolution of
Conict. Journal of Conict Resolution 16, 1:6794.
Fisher, Ronald J. (1982). Social Psychology: An Applied Approach. New York: St. Martins
Press.
Fisher, Ronald J. (1989). Prenegotiation Problem-Solving Discussions: Enhancing the Potential
for Successful Negotiation. In Janice G. Stein, editor, Getting to the Table: The Process of
International Prenegotiation. Baltimore, MD: Johns Hopkins University Press.
Fisher, Ronald J. (1997). Interactive Conict Resolution. Syracuse, NY: Syracuse University
Press.
Fisher, Ronald J. (In Press). Interactive Conict Resolution. In I. William Zartman, editor,
Peacemaking in International Conict (rev. ed.). Washington, DC: United States Institute of
Peace.
Kelman, Herbert C. (1972). The Problem-Solving Workshop in Conict Resolution. In Robert
L. Merritt, editor, Communication in International Politics. Urbana: University of Illinois
Press.
Kelman, Herbert C. (1978). Israelis and Palestinians: Psychological Prerequisites for Mutual
Acceptance. International Security 3, 1:162186.
Kelman, Herbert C. (1979). An Interactional Approach to Conict Resolution and its
Application to Israeli-Palestinian Relations. International Interactions 6, 2:99122.
Kelman, Herbert C. (1982). Creating the Conditions for Israeli-Palestinian Negotiations.
Journal of Conict Resolution 26, 1:3975.
Kelman, Herbert C. (1986). Interactive Problem Solving: A Social-Psychological Approach to
Conict Resolution. In William Klassen, editor, Dialogue Toward Inter-Faith Understand-
ing. Jerusalem: Ecu-menical Institute for Theological Research.
Kelman, Herbert C. (1987). The Political Psychology of the Israeli-Palestinian Conict: How
Can We Overcome the Barriers to a Negotiated Solution? Political Psychology 8, 3:347
363.
Kelman, Herbert C. (1995). Contributions of an Unofcial Conict Resolution Effort to the
Israeli-Palestinian Breakthrough. Negotiation Journal 11, 1:1927.
Kelman, Herbert C. (2000). The Role of the Scholar-Practitioner in International Conict
Resolution. International Studies Perspectives 1, 3: 273288.
THE PROBLEM-SOLVING WORKSHOP AS A METHOD OF RESEARCH 59
Kelman, Herbert C. (2002). Interactive Problem Solving as a Tool for Second Track
Diplomacy. In John Davies and Edy Kaufman, editors, Second Track/Citizens Diplomacy:
Concepts and Techniques for Conict Transformation. Lanham, MD: Rowman & Littleeld.
Kelman, Herbert C. and Cohen, Stephen P. (1976). The Problem-Solving Workshop: A Social-
Psychological Contribution to the Resolution of International Conict. Journal of Peace
Research 13, 2:7990.
Kelman, Herbert C. and Cohen, Stephen P. (1986). Resolution of International Conict: An
Interactional Approach. In Stephen Worchel and William G. Austin, editors, Psychology of
Intergroup Relations (2nd ed.). Chicago, IL: Nelson-Hall.
Lewin, Kurt (1948). Resolving Social Conicts. New York: Harper.
Mitchell, Christopher (2001). From Controlled Communication to Problem Solving: The
Origins of Facilitated Conict Resolution. International Journal of Peace Studies 6,
1:5967.
Mitchell, Christopher and Banks, Michael (1996). Handbook of Conict Resolution: The
Analytical Problem-Solving Approach. New York/London: Pinter.
Ross, Lee and Stillinger, Constance (1991). Barriers to Conict Resolution. Negotiation
Journal 7, 4:389404.
Rouhana, Nadim N. (2000). Interactive Conict Resolution: Issues in Theory, Methodology,
and Evaluation. In Paul C. Stern and Daniel Druckman, editors, International Conict Resolution
after the Cold War. Washington, DC: National Academy Press.
Saunders, Harold H. (1999). A Public Peace Process: Sustained Dialogue to Transform Racial
and Ethnic Conicts. New York: St. Martins Press.
Saunders, Harold H. and Slim, Randa M. (2002). Moderating Sustained Dialogue in
Tajikistan. Paper presented at Conducting Dialogues for Peace: A Symposium on Best Practices,
United States Institute of Peace, Washington, DC, November.
Walton, Richard E. (1970). A Problem-Solving Workshop on Border Conicts in Eastern
Africa. Journal of Applied Behavioral Science 6, 5:453489.
Time-Series Designs and Analyses
DANIEL DRUCKMAN
design. For technical details on calculations for many of these techniques, see
the chapters on forecasting in Frei and Ruloff (1989). The rst section shows
how time-series data can be used for post-diction with correlation and regres-
sion techniques. The next section presents the idea of interrupted time-series
analysis. The following section on Bayesian approaches to analysis includes
examples of the way that the calculations are done. The nal section provides
examples of several qualitative time-series analyses of conict-resolving pro-
cesses. Each of these applications addresses a research problem in the analysis
of negotiation and related interactions. Together, the examples call attention to
the range of applications afforded by time-series techniques.
were asked: Was there a pattern in the events that preceded the crises? How
were the impasses resolved? Was there a pattern in the events that occurred
following impasse resolution? Time-series analysis was used to address the
rst and third questions.
Statements made by both delegations during the sessions were coded into
categories reecting hard and soft negotiating rhetoric. The categories were
from a coding system known as bargaining process analysis (BPA). Designed
originally by Walcott and Hopmann (1978), the BPA is intended to facilitate
analyses of verbal behavior in international negotiation. Ratios indicating the
percentage of the total statements made by a delegation during a session (total
of 22) within a round (total of 10) were calculated and summed for an aggre-
gate score for each delegation by session. The aggregate scores were then
plotted in two dimensions, percent hard statements (plotted on the vertical
dimension as the sum of retractions, commitments, threats, and accusations)
and sessions (the horizontal dimension). The resulting trends provided a time
series for each delegation. By placing each of the six crisis sessions on the
graph, it was possible to analyze patterns of verbal behavior before and after
the event.
Correlational analyses revealed distinct patterns of behavior before and
after the crises. Succinctly stated, the pattern took the form of a gap between
the delegations in percent hard during the session before the crisis. The gap
was closed by the softer delegation. It increased the number of hard state-
ments, matching the harder delegation, resulting in mutual hardness and an
impasse. This pattern was discovered through inductive analysis rather than as
a test of an hypothesis. It became known as comparative responsiveness and
formed the basis for further research with multiple cases to be discussed
below. It also corroborated a pattern of reciprocity found in earlier bargaining
research with children (Druckman & Bonoma 1976). (The base-rights study
also investigated the way these crises were resolved and the patterns of
behavior that followed resolution; for details see Druckman 1986.)
A companion study reported by Stoll and McAndrew (1986) investigated
patterns of responsiveness through ten years (19691979) of talks between the
Soviet Union and the U.S. on strategic arms limitations (SALT). Performed
with a deductive approach, these investigators evaluated the t of three alter-
native models of reciprocity, referred to as directional, trend, and comparative
reciprocity. Each of these models predicts the reactions to concessions made
by the other negotiating team. The directional model species that negotiators
will either match (a concession for a concession) or mismatch (a retraction in
response to a concession) the others move. The trend model indicates that
negotiators will either increase or decrease their concession in response to a
64 DANIEL DRUCKMAN
None of the six mediations was effective in changing the parties behavior.
The mediators did not create opportunities for settlements by reducing the
violence and bringing about the conditions for a ceasere. On the other hand,
the combat did provide the condition for reduced violence and a ceasere
agreement. Escalation proceeded to the large-scale use of military force, re-
sulting in a mutually-hurting stalemate followed by a period of de-escalation
(Zartman 2000). Reasons for this outcome are discussed in Mooradian and
Druckman (1999).
Although this study provided insights into a question that has both theo-
retical and practical implications, it is a weak experimental design. The dif-
ferences found between the mediated interventions and the combat suggest
that the latter had a stronger impact on violent behavior than the former. It
does not suggest that escalation (combat) caused de-escalation (reduced violence).
A causal interpretation of these ndings would be more plausible if a com-
parable case without the interruptions (mediations, combat) was included in
the design; for example, de-escalation occurs only following combat in time-
series that compare cases with and without combat that includes casualties.
By limiting the analysis to the single case of Nagorno-Karabakh, these sorts
of inferences are unwarranted. By extending the analysis to another case, they
become more plausible. This extension is made in the context of focused or
matched case comparisons. (See, for example, Faure 1994 for discussion of
the approach.)
The idea that interruptions are treatments suggests that they have certain
characteristics. These include being relatively abrupt, although not necessarily
sudden, relevant to an ongoing process, attended to by the parties involved in
the process, and hypothesized to have an impact on that process. Other kinds
of interruptions may consist of worker strikes, government or company col-
lapse, a natural disaster such as an earthquake, the insertion of a peacekeeping
force, a summit conference among national leaders, an inux of foreign eco-
nomic aid, or an entirely new insight into the source of a conict.
The key point made in this section is that planned or unplanned events can
interrupt an ongoing process, turning it towards or away from agreements.
The impacts of the interruptions can be evaluated with before and after quasi-
experimental designs, allowing causal inferences to be made. This discussion
concludes our treatment of quantitative retrospective analyses. We turn now
to an approach that examines the way changing situations can alter the esti-
mates of probabilities for future (unknown) events.
68 DANIEL DRUCKMAN
Bayesian Analysis
Another type of analysis that captures change over time, invented by Rev-
erend Thomas Bayes in the eighteenth century, is based on assumptions about
how the past and the present can be used to estimate probabilities about the
occurrence of future events. Bayes theory of probability is described in an
essay published posthumously in a 1764 issue of the Philosophical Trans-
actions of the Royal Society of London. It was sent to the Royal Society by
Richard Price who wrote:
I now send you an essay which I have found among the papers of our deceased
friend Mr. Bayes, and which, in my opinion, has great merit. In an introduction
which he has writ to this Essay, he says, that his design at rst in thinking on
the subject of it was, to nd out a method by which we might judge concern-
ing the probability that an event has to happen, in given circumstances, upon
supposition that we know nothing concerning it but that, under the same cir-
cumstances, it has happened a certain number of times, and failed a certain
number of times (for more on Bayes see the web page, Bayes Ways to Gaze
into the Future . . . and how it pays (at www.vma.bme.hu/mathhis/Mathematicians/
Bayes.html).
The analysis consists of evaluating alternative hypotheses about the occur-
rence (Hi) or non-occurrence (Hj) of an event, for example, whether or not a
coup will occur within a dened period such as one year. The rst estimates
consist of initial probabilities that a coup will (p[H1]) or will not occur (p[H2]).
These probabilities can be estimated from statistics on coups that have
occurred in the country over the last 30 years. The next estimate is referred
to as a conditional probability and consists of the chances that an event (coup
or no coup) will occur if a particular symptom is present (p[EHi]). Examples
of symptoms related to coups, which are listed in Frei and Ruloff (1989), are
cabinet reshufing (.4 that coup will occur, .3 that it will not occur if there
is reshufing), call for free elections (.6 that a coup will occur, .5 that it will
not occur), and purges (.1 and .8 respectively). The estimates for symptoms
often come from experts or from expert panels. When the conditional proba-
bilities are taken into account, a revised initial probability results (p[HiE]) as
calculated by the following formula:
P(HiE) = P(Hi).P(EHi)/ S (P[Hi],P(EHi])
By inserting the initial (.25, .75) and conditional probabilities for reshufing
(.4, .3) into this formula and performing the calculations, the result is .31 that
a coup will occur in the next year and .69 that it will not occur. Note that
the initial probability for a coup, based on past evidence, increases somewhat
when the symptom of reshufing is taken into account (from .25 to .31).
TIME-SERIES DESIGNS AND ANALYSES 69
Similar calculations can be made for each of the other symptoms, free elec-
tions and purges.
An example more closely related to negotiation is the chance that a peace
agreement will hold through time. Suppose that past experience suggests that
peace agreements have held for at least ve years in one-third of the cases
examined; thus, p(Hi) is .33 and p(Hj) is .67. Suppose also that an expert
panel decided that there are ve factors that affect the longevity of agree-
ments. They assigned conditional probabilities to the factors as follows: fac-
tions of spoilers (.1 supporting Hi; .6 supporting Hj), schisms within the
parties (.2 and .4), regime continuity (.5 and .3), international pressure (.7 and
.2), and availability of arms (.2 and .5). The effect of the symptom referred
to as spoilers is to reduce the chances of sustaining the agreement from .33
(initial probability) to .08 (revised probability). Schisms reduce the chances
for longevity from .33 to .21. Regime continuity increases the chances from
.33 to .46. International pressure has the strongest positive impact on lon-
gevity, from .33 to .63. And, arms availability reduces the chances from .33
to .17. When the conditional probabilities of all ve symptoms are combined
in the formula the revised probability is .30, roughly the same as the initial
probability of .33; apparently the symptoms are off-setting with two favoring
longevity regime continuity and international pressure and three jeopar-
dizing the agreement spoilers, schisms, and availability of arms.
Bayesian analysis is an example of time-series analysis that focuses on
changes in the probabilities of events (revised probabilities) due to circum-
stances. It can also be understood as an alternative to regression for fore-
casting events. The beta weights that derive from a regression analysis are
estimates of the relative importance of variables which may be symptoms
in a Bayesian analysis in predicting a dependent variable (an event in
Bayesian terms). When used for forecasting, the regression equation is a lin-
ear extrapolation of past trends, considered as the initial probabilities or pri-
ors in a Bayesian analysis. Unlike regression, Bayesian forecasts are updated
probabilities for events that take both past (initial probabilities or priors)
and present (conditional probabilities of symptoms) into account. The updat-
ing can be continuous as a result of monitoring situations on a regular basis.
Expert judgments can also be rened as new information becomes available
to analysts. Regression is backward-looking in the sense of relying on a past
which cannot be changed, although memories of the past are often adjusted.
Bayesian analysis is an attempt to capture present events or processes, focus-
ing attention on changes which have implications for forecasts. Despite its appeal,
Bayesian techniques have rarely been used in addressing problems in negoti-
ation or conict resolution.
70 DANIEL DRUCKMAN
each rivalry in terms of the dominant inuence strategy used by the two
nations. The strategies were bullying, reciprocating, trial and error, and
stonewalling. For example, the four crises in the U.S.-Soviet relationship with
corresponding strategies and outcomes are as follows: Berlin Blockade,
194849 (Soviet bullying, U.S. reciprocating leading to a U.S. victory); Berlin
Wall, 1961 (Soviet bullying, U.S. reciprocating leading to a stalemate); Cuban
Missile, 1962 (Soviet trial and error, U.S. bullying leading to a U.S. victory),
and Middle East Alert, 1973 (Soviet trail and error, U.S. reciprocating lead-
ing to a U.S. victory). The other two rivalries relied primarily on bullying
strategies leading usually to wars or, in one crisis over Kashmir, stalemate.
Based on these patterns, Leng concluded that reciprocating strategies are
the most effective means of promoting cooperation in militarized crises. These
ndings are consistent with quantitative time-series analyses, calculated as weekly
average hostility scores during the period of the crisis, also performed by the
author on these three rivalries, as well as with the authors earlier ndings
based on a larger sample of crises occurring between 1816 and 1980 (Leng
1993). These are particularly interesting ndings in light of the selection of
cases. They challenge a prevailing assumption that the most effective means
of prevailing in interstate militarized disputes is through the use of escalating
coercion. They also reinforce the value of combined analyses for bolstering
condence in ndings obtained by only one approach.
Process Tracing
The process tracings illustrated here have several features in common. Each
depicts a linear process through chronological time. They all proceed toward
an outcome that serves to conclude a process. And, each involves a relatively
small number of actors or parties. Of interest is the question whether the tech-
nique can be used to depict non-linear interactions among many actors in
situations where the process does not conclude with a single outcome. An
example is the problem of coordination among different intervenors (or NGOs)
working to reduce conict in the same region. Data collected by Allen Nan
(1999) on the activities of many organizations operating in three former
Soviet republics make evident the difculty of constructing paths to outcomes.
For this problem, multiple paths involving various types of intervenors and
interventions converged and diverged toward or away from reduced levels of
conict: There were too many actors, activities, and varieties of coordination
and complementarity to discern a path proceeding from initial to end states.
This sort of complexity is captured better by frameworks that show the inter-
play among the actors, activities, and processes, including recursive relation-
ships between them. (See, for example, Allen Nan 1999: Figure 8.8.1.) Both
process tracing and framework construction are examples of ways to envision
information to enhance understanding of social processes.
A variety of qualitative time-series analyses were illustrated in this section.
They included the use of typologies for depicting relationships and inuence
strategies used at different points in time, tracing of processes that occur in
simulated and actual negotiations, and extended interactions between multiple
parties or organizations through the course of staged peace processes or non-
linear activities intended to address ongoing conicts. All of these examples
illuminate the value of using frameworks or category systems for longitudinal
analysis. They capture changes through time that would be missed by cross-
sectional (same point in time) comparisons and complement the quantitative
analyses discussed earlier. This concludes our treatment of the family of time-
series approaches. We turn now to a summary of the applications discussed
and an idea for combining several of the techniques.
Conclusions
conclusions about the likelihood that the event will occur, its impact on a
process, and the way it emerged from a sequence of prior events. These are
the facets of analysis that contribute to the goal of explaining the way that
negotiation processes unfold toward outcomes. (See Irner 2003 for an exam-
ple of a multi-method analysis of negotiation process-outcome relationships.)
References
Allen Nan, S. (1999). Complementarity and coordination of conict resolution efforts in the
conicts over Abkhasia, South Ossetia, and Transdniestria. Unpublished doctoral disserta-
tion, George Mason University.
Blalock, H.M. (1960). Social Statistics. New York: McGraw-Hill.
Carstarphen, N.M. (2003). Shift happens: Transformations during small group interactions in
protracted social conicts. Unpublished doctoral dissertation, George Mason University.
Cook, T., and Campbell, D.T. (1979). Quasi-Experimentation: Designs and Analysis Issues for
Social Research in Field Settings. Boston: Houghton Mifin.
Druckman, D. (2001). Turning points in international negotiation: A comparative analysis,
Journal of Conict Resolution 45:519544.
Druckman, D. (1993). The situational levers of negotiating exibility, Journal of Conict
Resolution 37:236276.
Druckman, D. (1986). Stages, turning points, and crises: Negotiating military base rights,
Spain and the United States, Journal of Conict Resolution 30:327360.
Druckman, D., and Bonoma, T.V. (1976). Determinants of bargaining behavior in a bilateral
monopoly situation II: Opponents concession rate and similarity, Behavioral Science
21:252262.
Druckman D., and Druckman, J.N. (1996). Visibility and negotiating exibility, Journal of
Social Psychology 136:117120.
Druckman D., and Green, J. (1995). Playing two games: Internal negotiations in the
Philippines. In I.W. Zartman (Ed.) Elusive Peace. Washington, DC: Brookings.
Druckman, D., and Harris, R. (1990). Alternative models of responsiveness in international
negotiation, Journal of Conict Resolution 34:234251.
Faure, A.M. (1994). Some methodological problems in comparative politics, Journal of Theoretical
Politics 6:307322.
Frei, D., and Ruloff, D. (1989). Handbook of Foreign Policy Analysis. Boston, MA: Martinus
Nijhoff.
George, A.L., and Bennett, A. (2004). Case Study and Theory Development. Cambridge, MA:
MIT Press.
Guilford, J.P. (1954). Psychometric Methods. New York: McGraw-Hill.
Irner, C.G. (2003). The promise of process: Evidence on ending violent international conict.
Unpublished doctoral dissertation, GeorgeMason University.
Leng, R.J. (1998). Reciprocity in recurring crises, International Negotiation 3:197226.
Leng, R.J. (1993). Interstate Crisis Behavior, 18161980: Realism versus Reciprocity. Cam-
bridge: Cambridge University Press.
Lepgold, J., and Shambaugh, G. (1998). Rethinking the notion of reciprocal exchange in inter-
national negotiation: Sino-American relations, 19691997, International Negotiation 3:227252.
78 DANIEL DRUCKMAN
Mooradian, M., and Druckman, D. (1999). Hurting stalemate or mediation?: The conict over
Nagorno-Karabakh, 199095. Journal of Peace Research 36:709727.
Pew Case Studies in International Affairs (1999). The ISD compendium of case study
abstracts and indexes, the Institute for the Study of Diplomacy. Edmund A. Walsh School
of Foreign Service, Georgetown University.
Pruitt, D.G. (2004). Escalation, readiness for negotiation, and third party functions. In I.W.
Zartman and G.O. Faure (Eds.) Escalation and Negotiation. Cambridge, UK: Cambridge
University Press.
Stern, P.C., and Druckman, D. (2000) Evaluating interventions in history: The case of inter-
national conict resolution, International Studies Review 2:3363.
Stoll, R.J., and McAndrew, W. (1986). Negotiating strategic arms control, 19691979.
Journal of Conict Resolution 30:315326.
Walcott, C., and Hopmann P.T. (1978). Interaction analysis and negotiating behavior. In R.T.
Golembiewski (Ed.) The Small Group in Political Science. Athens, GA: The University of
Georgia Press.
Zartman, I.W. (2000). Ripeness: The Hurting Stalemate and Beyond. In P.C. Stern and
D. Druckman (Eds.) International Conict Resolution after the Cold War. Washington, DC:
National Academy Press.
Social Research and the Study of Mediation:
Designing and Implementing Systematic Archival Research
JACOB BERCOVITCH
Introduction
reports of real events, by real mediators and negotiators, in real conict situ-
ations, and use that data in a systematic fashion to develop useful theories,
and answer real questions on the process and outcomes of mediation. To do
otherwise would be to mistake our wishful thinking for a complex reality.
Mediation is practiced in a variety of elds, ranging from domestic disputes
through legal conicts and community disputes to international conicts. Thus,
mediation may be studied by sociologists, psychologists, political scientists,
anthropologists, economists, and others. These scholars borrow concepts and
methods from each other. Whichever approach or discipline one adopts, the
study of mediation is something of a research puzzle for which we must nd
less than perfect answers through the analysis of data. Some of the questions
that make up the mediation puzzle include the following:
When should a mediator enter a dispute? Should one wait till a certain
threshold of violence has been crossed, or is it better to intervene before the
parties engage in openly hostile behavior?
Should a mediator be, neutral (could a mediator be neutral)? Would a biased
mediator be acceptable to the parties under some conditions? Would a
biased mediator be successful?
What exactly do mediators do when they intervene in a dispute? Are there
some recognizable patterns of behavior, and are such patterns associated
with any measure of effectiveness? How do mediators even choose one
strategy over another? Do they have an identiable decision calculus?
Which disputes are more amenable to mediation?
What factors or conditions affect the process and outcome of mediation?
Can any of these be changed so as to help mediation?
Are some mediators more likely to achieve their objectives than others? Are
factors such as personal attributes, rank, position and prestige related at all
to mediation outcomes, and if so how?
Do mediated outcomes endure? Do they satisfy the parties basic needs, or
do they merely provide an interlude for the parties to regroup, marshal resources
and engage in further aggression?
How best can we organize our research efforts, answer these questions, and
resolve the puzzle? Several approaches may be identied. A signicant part
of the literature dealing with mediation and purporting to answer these ques-
tions has been largely descriptive, relying mostly on single cases. The best
exemplars of this approach (e.g., Princen 1992; Touval 1982; Touval and
SOCIAL RESEARCH AND THE STUDY OF MEDIATION 81
Zartman 1985) used analytical frameworks and examined their chosen cases
with great care. However, it was, and is, difcult to draw general inferences
from these studies, no matter how carefully crafted they may have been. Des-
criptive approaches make good sense of idiosyncratic situations, they encour-
age policy makers and practitioners to publish their memoirs, but they do
have too wide a diversity of focal points, and it is impossible to generalize
from them or assess their validity. All that we can know is what happened in
the specic case and why mediation failed or succeeded. To answer the ques-
tions above, we need to treat mediation as a class of public events, not as a
unique set of circumstances.
An approach that has attempted to answer broad questions and deal with
issues of causes, effects, and outcomes in a large number of cases is the
experimental approach. Experimental approaches have universal rather than
particular goals. Experimental approaches, frequently conducted in laboratory
settings, test for causal relations. They permit one to manipulate values of
independent variables and determine how each affects the dependent vari-
able. Above all, they offer the possibility of replication and reinterpretation.
Some of the best studies in this tradition (e.g., Brookmire & Sistrunk 1980;
Carnevale 1986; Carnevale et al. 1989; Kochan & Jick 1978; Pruitt and
Johnson 1970. For a fuller review see Rubin 1980) exemplify how a pheno-
menon can be studied scientically, and done so in ingenious ways using
exciting research designs and challenge some of our basic notions about the
process.
Experimental approaches represent a systematic way of collecting informa-
tion about the world. We usually know how the data is recorded and gener-
ated. We know the hypotheses tested and the causal inferences made. We
know that when more data is needed, further experiments may be undertaken
to improve data quality. What we do not know is how we can overcome the
sense of articiality that inevitably characterizes these studies. In other words,
the data is real, but the real world is so very different form the data. To con-
duct an experiment with undergraduate psychology students in any university
on the validity of threats under conditions of multilateral deterrence may tell
us very little indeed about the behavior of policy-makers in hostile countries
with nuclear weapons at their disposal. There is a considerable leap of faith
required of us if we are to presume that the micro processes observed by
social psychologists, under quite well controlled situations, resemble the
stressful, often chaotic, and usually unpredictable political environment of
international relations. Experimental approaches offer the prospect of excellent
data collection methods, but their reliability and external (not internal) valid-
ity is at best doubtful.
82 JACOB BERCOVITCH
Mediation 2194
(58.7%)
Negotiation 1249
(33.4%)
Arbitration 29
(0.8%)
Referral to an International or Regional Organization 166
(4.4%)
Multilateral Conference 38
(1.0%)
Total Number of Conict Management Strategies 3737
(100%)
Some Findings
Making use of available data, we now have close to one million observations
on mediation and other forms of conict management over a fty year period.
Acquiring knowledge about any form of social relations may not be easy.
Acquiring knowledge about a process that is usually conducted behind closed
doors, where stakes and loss of face may be high, is particularly difcult. In
trying to understand the reality of mediation, we do, I believe, encounter
fewer difculties in constructing and validating hypotheses with this form of
research than any other form. We are less likely to err with our inferences and
interpretations when our knowledge is grounded in incidents and records of
real human behavior. A good research design and systematic use of archival
data (which after all are explicit and public) maximize internal and external
validity and our ability to demonstrate real patterns, replicate them, or refute
them. Better data are almost always a precursor to better theories. As social
scientists we usually have to make the best of what data we have. In the case
of mediation and negotiation, I doubt we have better data than that which
may be obtained through archival research.
Once colleted and codied, the data are arranged in the form of a large
matrix, with rows containing information about our units of analysis, media-
tion events, and the columns with different bits of information about each of
the independent variables. At this stage we can engage in a variety of quan-
titative or qualitative procedures to examine distribution of the data, relation-
ships between two or more variables, and develop causal inferences and
models, and answer the questions or puzzles posed above. Some of the most
important ndings of the project are highlighted below:
SOCIAL RESEARCH AND THE STUDY OF MEDIATION 89
Nature of Conict
Many studies posit a relationship between a given conict and mediation out-
comes. We found that conict intensity impacts adversely on mediation effec-
tiveness (Bercovitch 1986). High fatalities and high intensity conicts are not
really amenable to mediation. Issues in conict showed a marked effect on
mediation, with tangible issues such as resources and territory showing a
much higher likelihood of successful mediation than intangible issues, such as
identity or sovereignty (Bercovitch & Houston 1996). Conict complexity, exam-
ined as numbers of issues per conict, also showed a strong relationship with
ineffective mediation.
In a series of studies we were able to show that who the parties are has a
great impact on how a conict is likely to be mediated. Democratic parties
are more likely to use mediation and achieve a positive outcome than non-
democratic parties (Bercovitch & Houston 1996). Similar patterns were found
between economically developed countries and mediation. Domestic obser-
vance of human rights and respect for liberty correlate very highly with suc-
cessful mediation, as does the level of internal homogeneity (Bercovitch &
Houston 2000). Parties of equal power resources or small states make the
most of mediation services offered to them (Bercovitch 1986). Mediation is
not a method suited for conicts involving superpowers.
In a series of studies we were able to show that mediator neutrality, that old
chestnut in the literature, is far less important than mediator ability, experi-
ence and resources (Bercovitch & Houston 2000). We were also able to show
that the best time to initiate mediation is some months after a conict had
manifested itself (Bercovitch 1986). The most prominent mediators in inter-
national conicts are not regional or international organizations, but states,
and larger states at that. The most successful mediators are not private indi-
viduals but high ranking mediators (Bercovitch & Schneider 2000). A neutral
environment is highly conducive to successful mediation, as is the use of an
active intervention strategy (Bercovitch & Houston 2000).
This kind of archival analysis, based on a sound theoretical design, in
which the data has been carefully and consistently checked, where we deal
with a large number of case studies, and where we try and offer as complete
a description of each case by examining as many explanatory variables as we
90 JACOB BERCOVITCH
Conclusion
Acknowledgements
I would like to thank Judith Fretter and Robert Trappl for all their help. This
article has beneted considerably from the comments of Peter Carnevale and
an anonymous reviewer.
References
Bercovitch, J. (1986). International Mediation: A Study of Incidence, Strategies and Condi-
tions of Successful Outcomes, Cooperation and Conict 21(3):155168.
Bercovitch J., and Houston A. (1996). The Study of International Mediation: Theoretical
Issues and Empirical Evidence. In J. Bercovitch (Ed.) Resolving International Conicts. Boulder.
Westview Press.
Bercovitch, J., and Jackson, R. (1997). International Conict: Encyclopedia of Conicts and
their Management, 19451995. Washington, D.C.: Congressional Quarterly.
Bercovitch, J., and Houston, A. (2000). Why Do They Do It Like This? An Analysis of the
Factors Inuencing Mediation Behavior in International Conicts, Journal of Conict Resolution
44:170202.
Bercovitch J., and Schneider, G. (2000). Who Mediates: The Political Economy of Inter-
national Conict Management, Journal of Peace Research 37:145165.
Brookmire, D., and Sistrunk, F. (1980). The effects of perceived ability and impartiality of
mediator and time pressure on negotiation, Journal of Conict Resolution 24:31127.
Butterworth, R.L. (1976). Managing Interstate Conict, 19451974. University Center for
International Studies, University of Pittsburgh, PA, USA.
Carnevale, P. (1986). Strategic choice in mediation, Negotiation Journal 2:4156.
Carnevale, P.J., Conlon, D., Hanisch, K., and Harris, K. (1989). Experimental research on the
strategic choice model of mediation. In K. Kressel and D.G. Pruitt (Eds.) Mediation
Research: The Process and Effectiveness of Third Party Intervention. San Francisco: Jossey-
Bass, Inc.
Kochan, T.A., and Jick, T. (1978). A theory of public sector mediation process, Journal of
Conict Resolution 22:209240.
Nortehdge, F., and Donelan, M. (1971). International Disputes: Political Aspects. London:
Europa Publications.
Pruitt, D.G., and Johnson, D.F. (1970). Mediation as an aid to facesaving in negotiation,
Journal of Personality and Social Psychology 14:239246.
Princen, T. (1992). Intermediaries in International Conict. New Jersey: Princeton University
Press.
Rubin, J.Z. (1980). Experimental research on third party intervention in conict, Psycho-
logical Bulletin 87:379391.
Small, M., and David Singer, J.D. (1982). Resort to Arms Beverly Hills: Sage.
92 JACOB BERCOVITCH
Tillema, H.K. (1991). International Armed Conict Since 1945. Boulder: Westview Press.
Touval, S. (1982). The peace brokers Mediators in the Arab-Israeli conict 19481979. New
Jersey: Princeton University Press.
Touval, S., and I.W. Zartman, (1985). (Eds.) International Mediation in Theory and Practice.
Boulder: Westview Press.
Wallensteen, P. (2002). Understanding Conict Resolution. London: Sage.
Zacher, M. (1979). International Conict and Collective Security. New York: Praeger.
Reections on Simulation and Experimentation in the Study
of Negotiation1
JONATHAN WILKENFELD
About 15 years ago, I began a lengthy collaboration with Sarit Kraus, a com-
puter scientist specializing in Distributed Articial Intelligence (DAI). (I had
REFLECTIONS ON SIMULATION AND EXPERIMENTATION OF NEGOTIATION 95
barely heard of Articial Intelligence at that point). Kraus had been working
on developing an automated agent capable of participating with other auto-
mated agents or with human players in the Diplomacy board game. With a
strong interest in modeling the negotiation process from a bargaining and
game theoretic perspective, she was directed to me by her associates at the
University of Maryland Institute for Advanced Computer Studies, some of
whom had heard of my work on ICONS. It was apparent from the beginning
that although we had very different training, we had a joint interest in better
understanding how humans (in my case) and machines (in Krauss case)
negotiated in difcult (crisis) situations in order to reach mutually benecial
agreements. Whereas my objective was to understand the conditions under
which negotiations ended without resort to violence, Krauss objective was to
understand how intelligent agents (machines) could be endowed with negoti-
ation skills such that they could share scarce resources (tools like printers, Internet
access, etc.) or cooperate to perform tasks. In both cases, negotiations are
employed in situations where time is critical to some or all parties, resources
are scarce, and tasks may require cooperative behavior. We wanted to mini-
mize the amount of time spent on negotiation, and maximize the likelihood
of reaching mutually benecial outcomes (Kraus et al. 1995; Kraus 2001;
Wilkenfeld et al. 1995).
Our experimental work has been characterized by several features that tend
to set it apart from other experimental work in foreign policy analysis (see
Geva and Mayhar 1997; Mintz, Geva, Redd, and Carnes 1997; Tetlock &
Belkin 1996). First, we have focused exclusively on behavior in crisis (Brecher
& Wilkenfeld 2000), with a particular emphasis on the effects shortness of
time and high level of threat have on crisis dynamics and outcomes. Second,
our subjects had access to an elaborate Decision Support Systems (DSS)
which, when properly used, could provide information on the utilities associ-
ated with various outcomes and hence enhance the subjects ability to maxi-
mize expected utility, even under conditions of incomplete information.3
Third, all communications among the participants have taken place in con-
trolled network environments, allowing for detailed analysis of the types of
tools they consulted as well as the content and form of their interactions dur-
ing the negotiation. Together, these features provide a rich environment in
which experiments can be run and their results assessed.4
At the core of our experimental work to date is the development of a strate-
gic model of negotiation with an accompanying decision support system
(DSS) based on three different scenarios: a hostage crisis involving an Indian
plane hijacked by Sikh separatists and forced to land in Pakistan (see Kraus,
Wilkenfeld and Zlotkin 1995, Wilkenfeld, Kraus and Holley 1999), a shing
96 JONATHAN WILKENFELD
solution) (Touval & Zartman 1985, Zartman & Touval 1996). Hopmann indi-
cates that only a powerful mediator can play this role, and he cites two modes
of mediator involvement as manipulator: Inuencing the direction of the
negotiation process by raising the costs of disagreement and the rewards for
agreement; Manipulating the international environment to push the parties toward
agreement.
From the experimental perspective, the question to be explored is how
different types of mediation, as opposed to no mediation, impact the utility
maximizing behavior of negotiators, the pace of crisis abatement, and the sat-
isfaction with mediation.6
associated with crisis termination and on the duration of a crisis. Data from
the Ecuador/Peru simulations indicate that only manipulative mediation meets
the negotiators expectations of leading them to a more benecial outcome
than they could have otherwise secured, while facilitation may actually lower
average benets a situation that could lead to discontent among disputants
and, possibly, recurrent crises. In addition, the more rapid conclusion of a cri-
sis brought about only by manipulation is an essential component of cri-
sis management, given the relationship between prolonged crisis negotiations
and the likelihood of escalating violence and war. We conclude that manipu-
lative mediation, as compared with facilitation or no mediation, is an effec-
tive means of crisis management.
Our endorsement of manipulation must be tempered, though, by at least
two factors that need to be explored more rigorously in the future. First, our
analysis has only looked at the two extreme styles of mediation: facilitation
and manipulation. It may be the case that formulative mediation, which falls
between these two extreme styles, is an even more effective management tool
under certain circumstances.
A second concern with manipulative mediation is about both the short
and long-term implications of solutions developed and forced upon parties by
an outsider, fearing that these situations can lead to feelings of alienation and
resentment of the mediator, of the process of negotiation, and even of the
other parties involved in the negotiation (Princen 1992; Kelman 1992; Keashly
& Fisher 1996). While our ndings on negotiator satisfaction do not reveal
resentment of manipulation vis--vis facilitation, greater consideration of the
unintended consequences that could accompany the adoption of manipulative
mediation as a crisis management tool is necessary before we can conclude
that manipulation is an appropriate perhaps the most appropriate means
of managing international crises. Our current ndings, though, do indicate that
manipulation shows potential as a key approach to mitigating the violence and
instability associated with international crises.10
expense and time involved in the design and execution of a series of experi-
ments, it is not surprising that relatively few political scientists venture into
this domain.
Given this rather bleak picture, where do I see simulation and experimen-
tal design tting into the array of approaches now dominating the study of
foreign policy analysis and negotiations? At the level of political psychology,
there are simply some characteristics of decision-makers and decision-making
that we are never going to be able to study systematically with anything approach-
ing a large enough n to allow for careful statistical work. It is here, under
controlled laboratory environments, that we can begin to make progress in
understanding how decision-makers cope with and process information in cri-
sis versus non-crisis environments, how third parties can impact the course of
and outcomes to contentious situations, how actor attributes such as power (of
various sorts) and willingness to compromise affect their negotiation strategies
and behavior, and the circumstances under which violent strategies are chosen.
While nothing can substitute for the data we can glean from in-depth com-
parative case studies, the experimental environment when properly speci-
ed can help us ll in the gaps in our knowledge and ultimately allow us
to generalize.
But in order to do this right, we must begin to properly train the next gen-
eration of political scientists in simulation and experimental techniques. This
is not everyones cup of tea, nor should it be. It does offer an important alter-
native to the cross national and case study approaches that dominate the eld
of foreign policy analysis today. But we lack the experimental infrastructure,
the culture of the laboratory, and perhaps most importantly, the channels by
which experimental results can become important sources of information as
we seek to better understand the behavior of political decision-makers. And
perhaps most importantly, we need to be able to demonstrate more convinc-
ingly that results obtained in experimental simulation environments can help
us better understand the behavior of real decision-makers in critical situations.
Much work remains to be done on this front.
Notes
1. This work has been supported by the National Science Foundation under grant IIS0208608,
and the US Institute of Peace (under grants SG-3595 and SG-5200). Parts of this paper
are drawn from Wilkenfeld (2002) and Wilkenfeld, Young, Asal, and Quinn (forthcoming).
2. For additional information on the ICONS simulations, see http://www.icons.umd.edu.
3. For a discussion of decision support systems in general, and their application to negotiation
in particular, see Wilkenfeld et al. 1995.
REFLECTIONS ON SIMULATION AND EXPERIMENTATION OF NEGOTIATION 101
4. For details on our experimental procedures, see Santmire et al. (1998).
5. Hopmann (1996) qualies the notion of the manipulative mediator being a party to the dis-
pute by stating that this mediator Aalmost@ assumes the role of a disputant in the nego-
tiation process.
6. In the simulations conducted thus far, only facilitation and manipulation are examined.
7. The experimental environment is described more fully in Santmire et al. (1998), Wilkenfeld
et al. (1995, 2001).
8. For a discussion of these crises in detail, see Brecher and Wilkenfeld (2000). The scenario
and decision support systems, developed by Kathleen Young, David Quinn, and Chris
Frain, are available from the author upon request. A more extended discussion of this
research project appears in Wilkenfeld, Young, and Quinn (2001).
9. For details on the experimental procedures, see Wilkenfeld, Young, Asal, and Quinn (forth-
coming).
10. For a more extended discussion of these experiments on mediator style in international
crises, see Wilkenfeld, Young, Asal, and Quinn (forthcoming).
11. I am reminded of one of the very rst works I became familiar with in simulation, the
study by Hermann and Hermann (1967), which used the Inter-nation Simulation (INS) to
attempt to replication the conditions leading to the outbreak of World War I. They checked
for seemingly everything, yet curiously they couldnt get war to break out. As it turned
out, the student playing one of the key roles was an avowed pacist, and simply wouldnt
go to war under any circumstances.
References
Bercovitch, J.J., Anagnoson, T., and Wille, D.L. (1991). Some conceptual issues and empirical
trends in the study of successful mediation in international relations, Journal of Peace
Research 28 (1): 717.
Bercovitch, J., and Langley J. (1993). The nature of the dispute and the effectiveness of inter-
national mediation, Journal of Conict Resolution 37(4): 67099.
Bositis, D.A. (1990). Research Designs for Political Science: Contrivance and Demonstration
in Theory and Practice. Carbondale: Southern Illinois University Press.
Brecher, M., and Wilkenfeld, J. (2000). A Study of Crisis. Ann Arbor, MI: University of
Michigan Press, paperback with CD-Rom edition.
Brody, R. (1969). The Study of International Politics qua Science: The Emphasis on Methods
and Techniques. In K. Knorr and J.N. Rosenau (Eds.) Contending Approaches to Inter-
national Politics. Princeton, NJ: Princeton University Press. 110128.
Burton, J.W. (1984). Global Conict: The Domestic Sources of International Crisis. Brighton,
Sussex, Great Britain: Wheatsheaf Books Ltd.
Campbell, D.T., and Stanley, J.C. (1963). Experimental and Quasi-Experimental Designs for
Research. Chicago: Rand-McNally.
Fisher, R.J. (1972). Third Party Consultation: A Method for the Study and Resolution of
Conict, Journal of Conict Resolution 16:1:6794.
Franklin, A. (1996). Experiment, Right or Wrong. New York: Cambridge University Press.
Geva, N., and Mayhar, J. (1997). The Cognitive Calculus of Decisions on the Use of Force.
Paper presented at the annual meeting of the American Political Science Association, Washington,
DC.
102 JONATHAN WILKENFELD
Guetzkow, H., et al. (1963). Simulations in International Relations: Developments for Research
and Teaching. Englewood Cliffs, NJ: Prentice-Hall.
Hermann, C., and Hermann, M. (1967). An Attempt to Simulate the Outbreak of World
War I, American Political Science Review 61:406416.
Hopmann, P.T. (1996). The Negotiation Process and the Resolution of International Conicts.
Columbia, SC: University of South Carolina Press.
Keashly, L., and Fisher, R.J. (1996). A Contingency Perspective on Conict Interventions:
Theoretical and Practical Considerations. In J. Bercovitch (Ed.) Resolving International Conicts:
The Theory and Practice of Mediation. Boulder, CO: Lynne Rienner Publishers. 235261.
Kelman, H.C. (1992). Informal Mediation by the Scholar/Practitioner. In J. Bercovitch and
J.Z. Rubin (Eds.) Mediation in International Relations. New York: St. Martins Press. 6496.
Kinder, D., and Palfrey, T. (1993). On Behalf of an Experimental Political Science. In
D. Kinder and T. Palfrey (Eds.) Experimental Foundations of Political Science. Ann Arbor:
University of Michigan Press. 139.
Kraus, S. (2001). Strategic Negotiation in Multiagent Environments. Cambridge: MIT Press.
Kraus, S., Wilkenfeld, J., and Zlotkin, G. (1995). Multiagent Negotiation under Time Con-
straints Journal of Articial Intelligence 75:297345.
Laponce, J.A. (1972). Experimenting: A Two-Person Game between Man and Nature. In
J.A. Laponce and P. Smoker (Eds.) Experimentation and Simulation in Political Science.
Toronto, ON: University of Toronto Press. 315.
McGraw, K. (1993). Political Methodology: Research Design and Experimental Methods. In
R.E. Goodin and H.D. Klingemann (Eds.) A New Handbook of Political Science. New York:
Oxford University Press. 769786.
Mintz, A., Geva, N., Redd, S., and Carnes, A. (1997). The Effect of Dynamic and Static
Choice Sets on Political Decision Making, American Political Science Review 91:553566.
Princen, T. (1992). Intermediaries in International Conict. Princeton, NJ: Princeton University
Press.
Santmire, Tara, Wilkenfeld, J., Kraus, S., Holley, K., Santmire, Toni, and Gleditsch, K. (1998).
The Impact of Cognitive Diversity on Crisis Negotiations, Political Psychology 19(4):721748.
Snyder, R. (1963) Some Perspectives on the Use of Experimental Techniques in the Study of
International Relations. In H. Guetzkow et al. (Eds.) Simulations in International Relations.
Englewood Cliffs, NJ: Prentice-Hall. 123.
Starkey, B. and Blake, E. (2001). Simulation in International Relations Education, Simulation
and Gaming 32(4): 537551.
Starkey, B. and Wilkenfeld, J. (1996). Project ICONS: Computer-assisted Negotiations for the
IR Classroom, International Studies Notes 21(1):2529.
Tetlock, P., and Belkin, A. (Eds.) (1996). Counterfactual Thought Experiments in World
Politics: Logical, Methodological, and Psychological Perspectives. Princeton, NJ: Princeton
University Press.
Torney-Purta, J. (1992). Cognitive Representations of the Political System in Adolescents: The
Continuum from Pre-Novice to Expert, New Directions for Child Development 56:1124.
Torney-Purta, J. (1996). Conceptual Change Among Adolescents using Computer Networks
and Peer Collaboration in Studying International Political Issues. In S. Vosniadou et al.
(Eds.) International Perspectives on the Design of Technology Supported Environments. Hills-
dale, NJ: Lawrence Erlbaum. 203219.
Torney-Purta, J. (1998). Evaluating Programs Designed to Teach International Content and
Negotiation Skills, International Negotiation: A Journal of Theory and Practice 3(1):
7797.
REFLECTIONS ON SIMULATION AND EXPERIMENTATION OF NEGOTIATION 103
Touval, S., and Zartman, I.W. (1984). Introduction: Mediation in Theory. In S. Touval and
I.W. Zartman (Eds.) International Mediation in Theory and Practice. Boulder, CO: Westview
Press. 717.
Wilkenfeld, J., and Kaufman, J. (1993). Political Science: Network Simulation in International
Politics, Social Science Computer Review 11(4):464476.
Wilkenfeld, J., Kraus, S., and Holley, K. (1999). The Use of Decision Support Systems in
Crisis Negotiations. In Kent, A. (Ed.) Encyclopedia of Library and Information Science.
New York: Marcel Dekker.
Wilkenfeld, J., Kraus, S., Holley, K., and Harris, M. (1995). GENIE: A Decision Support
System for Crisis Negotiations, Decision Support Systems 14:369391.
Wilkenfeld, J., Kraus, S., Santmire, Tara, and Frain, C. (1999). The Role of Mediation in
Conict Management: Conditions for Successful Resolution. Paper presented at the annual
meeting, International Studies Association.
Wilkenfeld, J., Young, K., Asal, V. and Quinn, D. (Forthcoming). Mediating International
Crises: Cross-National and Experimental Perspectives, Journal of Conict Resolution.
Wilkenfeld, J., Young, K., and Quinn, D. (2001). Mediating International Crises in the 20th
Century. Paper presented at the International Studies Association Meetings Hong Kong.
Zartman, I.W., and Touval, S. (1996). International Mediation in the Post-Cold War Era. In
C.A. Crocker and F.O. Hampson with P.A. all, Managing Global Chaos: Sources of and
Responses to International Conict. Washington, D.C.: United States Institute of Peace Press.
445461.
Quantitative Coding of Negotiation Behavior
Level of coding
a
Numbers correspond to those in Table 1.
Coding data
Unitizing
Coders must rst identify the unit of analysis for coding. This can vary from
a single utterance to an interaction between two individuals. Speaking turns,
that is, units that include all actions and/or statements made by an individual
while he or she holds the oor, are often used when studying negotiation
because they allow us to examine how one negotiator responds to another.
Speaking turns can be subdivided into acts, that is, identiable ideas or
thoughts, which can be used when interested in the content of the dialogue
said rather than the interactive nature of the negotiation. In the Towers II data,
we unitized at the level of acts, that is, we coded each identiable idea, but
also retained information about speaking turns to maximize our exibility.
Note that one speaking turn can be made up of several thought units.
Decisions regarding choice of unit type must be linked to the research
question being asked, with special emphasis on the appropriate level of analy-
sis and to the question of where in action and speech relevant meaning
resides. Meaning can be lost if we select too small a unit, because the indi-
vidual statements convey a different meaning to that conveyed by a speaking
turn. We can also add redundancy as a result of separating immediate restate-
ments, that which simply repeat previous tactics rather than adding new infor-
mation. In contrast, by selecting too large a unit, information can be lost. If
multiple categories of statements are made during a speaking turn, the
researcher must decide which code best represents the behavior within a given
unit. Dominance schemes, which identify the kinds of behavior that are
expected to have the greatest impact on the interaction, can be developed to
QUANTITATIVE CODING OF NEGOTIATION BEHAVIOR 111
provide rules of thumb for these decisions (e.g., Weingart et al. 1993).
Alternatively, the rst or the last code within the unit might be retained. Regardless
of the approach that is used, the risk of losing valuable information remains.
Coders must meet two criteria of reliability: unitizing reliability and inter-
pretive reliability. In both cases, the level of reliability places an upper bound
on a coding schemes predictive ability. Unitizing reliability, the degree of
agreement regarding identication of the units to be categorized, is needed
to ensure that multiple coders of the same interaction will be coding from
the same set of units. Unitizing reliability is most typically assessed using
Guetzkows U, which calculates the difference between the number of units
identied by an independent coder and the true number of units (the aver-
age of two coders estimates) using a simple equation
U = (O1 O2)/(O1 + O2 )
where U * 100 is the percent of disagreement and O is the observer or coder
(Folger et al. 1984; Guetzkow 1950). For example, if coder 1 identied 75
units and coder 2 identied 85 units in the same transcript, Guetzkows U =
(85 75)/(85 + 75) = .062, meaning that there was a 6% discrepancy between
the number of units identied by either coder and the true number of units.
Interpretive reliability refers to consistency in applying labels to the units and
is typically measured across coders. That is, multiple coders independently
code the same text, and their assignment of categories is compared for con-
sistency. Thus, consistency of interpretation is assessed in terms of the reli-
able application of interpretive rules. It is also useful to obtain consistency
measures within coders across time to check for drift in application of the
coding scheme. The most common and complete global measure is Cohens
Kappa (1960), which determines the level of agreement corrected for agree-
ments due to chance. Cohen presents the following formula to calculate
Kappa:
Kappa = (P' PC)/(1 PC)
where P' is the observed percentage agreement among coders and PC is the
proportion of chance agreement. Assuming that each unit to be coded has the
same probability of accurate classication, then the probability of chance
agreement (PC) is 1/k, where k is the number of categories in the coding
scheme (Folger et al. 1984). For example, if there are 5 categories and the
coders agreed on 75% of the codes assigned (P' = .75), then Kappa =
(.75 .2)/(1 .2) = .6875. Additional details for how to calculate Kappa can
be found in Cohens original article (1960). More detailed discussions of
Kappa can be found in Fleiss (1971) and Folger et al. (1984).
112 LAURIE R. WEINGART, MARA OLEKALNS AND PHILIP L. SMITH
Interpretive validity
Interpretive validity refers to the degree to which a coding scheme taps into
the information it was designed to obtain. Different approaches have been
used to validate coding schemes, depending on the goals of the research.
Poole and McPhee (1994) identify three types of claims that might be made
in an attempt to demonstrate validity of a coding scheme. These claims are
linked to the source of interpretation of the actions being studied: 1) observer-
privileged: where the researcher seeks to explain interaction from the outside
without reference to subjects perspectives, 2) generalized subject-privileged:
focusing on shared meanings for members of a culture, and 3) restricted
subject-privileged: with the goal of identifying idiosyncratic meanings for
people in a particular group or relationship. As one moves from observer to
restricted subject-privileged approaches, more input from participants and more
intimate knowledge about the work group (as might be experienced by a group
member) is required.
So far, we have provided a how to for coding data. Having gotten this far,
the obvious question is what next? Before one can start analyzing the data,
there are still a number of issues that need to be resolved. How this is done
will be determined, in part, by the kinds of research questions that one is ask-
ing. In this section, we consider three related issues: aggregating data, level
of analysis and time segmentation.
Aggregating data
important to know how many times each negotiator in a dyad or group uses
a specic tactic (e.g. making a demand) or can the hypotheses be tested by
knowing, overall, the number of times that demands were made in a negotia-
tion? In our research, we have largely focused on behaviors aggregated to the
level of the dyad or group. This has enabled us to provide a global descrip-
tion of regularities in negotiators behaviors and the relationships between the
negotiating context, negotiators behavior, and their outcomes. It allows us to
ask whether, in general, integrative outcomes are associated with different
kinds of strategies rather than distributive outcomes or impasses. It also
enables us to question whether, in general, cooperatively-motivated negotia-
tors behave differently from individualistically-motivated negotiators.
Our experiences suggest that while this level of analysis captures many
important aspects of negotiators behavior there are circumstances under
which an individual level analysis is important, for example when negotiators
differ in terms of power (Giebels, DeDreu, and Van de Vliert 2000) or social
motives (Weingart, Brett, and Olekalns 2002). In our research, we have mani-
pulated the composition of negotiating dyads and groups. At the level of the
dyad, this means that two negotiators can have similar attributes, for exam-
ple, both are cooperatively-motivated, or they can have mixed attributes, that
is, one negotiator is cooperatively motivated while the other is individualisti-
cally motivated. An interesting question is how dyad or group composition
affects negotiation processes: Do cooperatively-oriented negotiators behave
the same way regardless of the characteristics of the other negotiator or do
they adapt their behaviors? To answer questions such as this, it is necessary
to shift from analyzing dyadic or group behaviors to individual behaviors
within those groups.
Independent of the content of the coding scheme, researchers need to
decide how detailed their coding scheme should be. The answer to this ques-
tion again depends on the level of your theory whether global or specic
behavior is of interest. If your hypotheses are about integrative or distributive
behavior generally, you might want to classify comments using those broad
categories. We nd it is more effective to use more detailed coding categories
and then aggregate these categories to the more general level during the
analysis phase. This approach makes it less likely that two functionally dif-
ferent behaviors will be subsumed under one code. However, the more
detailed the coding scheme the more likely we are to see coding errors and
low reliability. Increasing the number of categories also means that the fre-
quency of each category decreases, resulting in sparse data sets that create
problems for most statistical methods.
There are several approaches for aggregating tactics into more general
groupings. You may choose to adopt a theory-driven approach, aggregating
114 LAURIE R. WEINGART, MARA OLEKALNS AND PHILIP L. SMITH
Levels of analysis
The second issue concerns the level of analysis. In analyzing your data, you
can frame your hypotheses in terms of the frequency with which strategies
are used, how they are sequenced or how strategies evolve over time. Each
of these levels of analysis answers slightly different questions. In this section,
we provide a brief overview of the issues associated with each level of analy-
sis (see Brett, Weingart and Olekalns, 2004; Weingart & Olekalns, 2004). Our
other article, to be published in the next issue of International Negotiation
QUANTITATIVE CODING OF NEGOTIATION BEHAVIOR 115
Time segmentation
Researchers who want to examine negotiation phases are faced with a choice
between a stage model and an episodic model. The principal advantage of a
stage approach is that it ensures comparability across research; its principal
disadvantage is that, by arbitrarily determining when stages begin and end,
it is likely to miss naturally occurring shifts in the negotiation process. If
one chooses to work within a stage approach, one still needs to make several
decisions. The rst is how many stages in which to divide the negotiation.
Negotiation theory identies anywhere between three and 12 stages. In our
research, we have used six stages as the number of segments that provides a
reasonably ne-grained analysis of negotiation while maintaining a reasonable
number of observations within each segment (e.g., Olekalns, Smith and Walsh
1996). The second is how to determine these stages. You can do this either
on the basis of time (divide each negotiation into, for example, ve minute
intervals) or on the basis of speaking turns (for example, if the negotiation is
made up of 300 speaking turns, then each 50 speaking turns become a stage).
The principal disadvantage of a time-based segmentation is that negotiations
do not all take the same amount of time. You may thus nd yourself com-
paring strategies that occur in the 1015 minute interval and which represent
the midpoint for one negotiating dyad but the endgame for another. We there-
QUANTITATIVE CODING OF NEGOTIATION BEHAVIOR 117
Conclusion
References
Bales, R.F. (1950). Interaction process analysis: A method for the study of small groups. Cambridge,
MA: Addison-Wesley.
Baxter, L.A. (1992). Conict management: An episodic approach, Small Group Behavior
13:2342.
Brett, J.M., Weingart, L.R., and Olekalns, M. (2004). Baubles, Bangles and Beads: Modeling
the Evolution of Negotiating Groups over Time. In M.A. Neale, E.A. Mannix, and
S. Blount-Lyons (Eds.) Research in Managing Groups and Teams: Time in Groups (Volume 6,
pp. 39-64). New York: Elsevier Science.
Carnevale, P.J.D., Pruitt, D.G., and Seilheimer, S.D. (1981). Looking and competing: Accountability
and visual access in integrative bargaining, Journal of Personality and Social Psychology
40:111120.
Cohen, Jacob (1960). A coefcient of agreement for nominal scales. Educational and
Psychological Measurement, 20, 3746.
Donohue, W.A., Diez, M.E., and Hamilton, M. (1984). Coding naturalistic negotiation inter-
action, Human Communication Research 10:403425.
Fleiss, J.L. (1971). Measuring nominal scale agreement among many raters, Psychological
Bulletin 76:378382.
Folger, J.P., Hewes, D.E., and Poole, M.S. (1984). Coding social interaction. In B. Dervin
and M. Voight (Eds.) Progress in communication sciences (Vol. 4, pp. 115161). Norwood:
Ablex.
118 LAURIE R. WEINGART, MARA OLEKALNS AND PHILIP L. SMITH
Giebels, E., De Dreu, C.K.W., and Van de Vliert, E. (2000). Interdependence in negotiation:
Effects of exit options and social motive on distributive and integrative negotiation,
European Journal of Social Psychology 30:255272.
Greenacre, M.J. (1993). Correspondence analysis in practice. London: Academic Press.
Guetzkow, H. (1950). Unitizing and categorizing problems in coding qualitative data, Journal
of Clinical Psychology 6:4758.
Hewes, D.E. (1979). The sequential analysis of social interaction, Quarterly Journal of
Speech 65:5673.
Hyder, E.B., Prietula, M.J., and Weingart, L.R. (2000). Getting to best: Efciency versus opti-
mality in negotiation, Cognitive Science 24(2):169204.
Lewicki, R.J., Saunders, D.M., and Minton, J.W. (1985). Negotiation. Irwin: Boston.
Olekalns, M., and Smith, P.L. (1999). Social value orientations and strategy choices in com-
petitive negotiations, Personality and Social Psychology Bulletin 25:657668.
Olekalns, M., and Smith, P.L. (2000). Negotiating optimal outcomes: The role of strategic
sequences in competitive negotiations, Human Communication Research 24:528560.
Olekalns, M., and Smith, P.L. (2003). Testing the relationships among negotiators motiva-
tional orientations, strategy choices, and outcomes. Journal of Experimental Social
Psychology.
Olekalns, M., Smith, P.L., and Walsh, T. (1996). The process of negotiating: Strategies, tim-
ing and outcomes, Organizational Behavior and Human Decision Processes 67:6177.
Poole, M.S., Folger, J.P., and Hewes, D.E. (1987). Analyzing interpersonal interaction. In
M.E. Roloff and G.R. Miller (Eds.) Interpersonal processes: New directions in communica-
tion research (pp. 220256). Newbury Park, CA: Sage.
Poole, M.S., and McPhee, R.D. (1994). Methodology in interpersonal communication. In
M. Knapp and G.R. Miller (Eds.) Handbook of interpersonal communication. Newbury Park,
CA: Sage.
Pruitt, D.G. (1981). Negotiation behavior. Academic Press: New York.
Pruitt, D.G., and Carnevale, P.J. (1993). Negotiation in social conict. Pacic Grove: Brooks/
Cole.
Pruitt, D.G., and Lewis, S.A. (1975). Development of integrative solutions in bilateral nego-
tiation, Journal of Personality and Social Psychology 31:621633.
Putnam, L.L., and Jones, T.S. (1982). The role of communication in bargaining, Communi-
cation Monographs 3:262282.
Smith, P.L., Olekalns, M., Weingart, L.R. (in press). Markov claim Models of communication
processes in negotiation. International Negotiation.
Trujillo, N. (1986). Toward a taxonomy of small group interaction coding systems, Small
Group Behavior 17:371394.
Walton. R.E., and McKersie, R.B. (1965). A behavioral theory of labor negotiations,
McGraw-Hill: Boston.
Weingart, L.R. (1997). How did they do that? The ways and means of studying group
processes. In L.L. Cummings and B.M. Staw (Eds.) Research in organizational behavior
(Vol. 19 pp. 184239). Greenwich: JAI.
Weingart, L.R., Bennett, R.J., and Brett, J.M. (1993). The impact of consideration of issues and
motivational orientation on group negotiation process and outcome. Journal of Applied Psy-
chology 78(3):504517.
Weingart, L.R., Brett, J.M., and Olekalns, M. (2002). Conicting social motives in negotiating
groups. Paper presented at the 62nd Annual meeting of the Academy of Management,
Denver, Colorado.
QUANTITATIVE CODING OF NEGOTIATION BEHAVIOR 119
Weingart, L.R., Hyder, E.B., and Prietula, M.J. (1996). Knowledge matters: The effect of tac-
tical descriptions on negotiation behavior and outcome, Journal of Personality and Social
Psychology 70:12051217.
Weingart, L.R., and Olekalns, M. (2004). Communication Processes in Conict and Negotia-
tion. In M. Gelfand and J. Brett (Eds.) The Handbook of Negotiation and Culture: Research
Perspectives. Palo Alto, CA: Stanford University Press.
Weingart, L.R., Thompson, L.L., Bazerman, M.H., and Carroll, J.S. (1990). Tactical behavior
and negotiation outcomes, The International Journal of Conict Management 1:731.
The Use of Questionnaires in Conict Research
Many books and articles have been written about conducting questionnaire
research (Belson 1981; De Vaus 1991; Dillman 1978; Kidder & Judd 1986;
Schuman & Scott 1987; Shaughnessy, Zechmeister & Zechmeister 2000;
Tanur 1992). For example, Schaughnessy and colleagues (2000) write that the
wording of questions can be a threat to validity, because how a question is
phrased has implications for how that question is answered. Good question-
naire items are short (20 or fewer words), simple, direct, clear, specic, and
familiar to all respondents. Furthermore, they do not involve leading, loaded
or double-barreled questions. Other issues that Schaughnessy and colleagues
(2000) deal with is questionnaire design (e.g., attractiveness), ordering of
questions (e.g., interesting questions rst), and enhancing response rate in a
mail survey (e.g., mail one or more follow-up questionnaires).
Despite the attention for questionnaire research, books and articles on
general research methods pay little attention to the specic challenges encoun-
tered when studying conict. In this article, we will therefore discuss the
strengths and weaknesses of using questionnaires in conict research. We do
not aim at completeness in summing up all the challenges. Instead, we will
restrict ourselves to some strengths and weaknesses that we have encountered
when performing our own research on conict in organizations and in close
relationships.
The main focus is on measuring conict behavior, although there are many
other relevant aspects to conict, such as conict issues, causes, and outcomes
(De Dreu, Harinck & Van Vianen 1999; Van de Vliert 1998), as well as the
motives, emotions, and cognitions that accompany conict (De Dreu & Carne-
vale, 2003). We believe conict behavior is a key factor of the conict pro-
cess because it largely determines the (positive or negative) consequences of
conict. At the same time, conict behavior may be one of the most difcult
conict features to measure through questionnaires, because questionnaires by
denition measure perceptions of current or past behavior instead of the
actual behavior itself.
We will address ve issues concerning the use of questionnaires in conict
research. The rst is that conict is a highly sensitive topic, which has con-
sequences for the acquisition of participants and the response to surveys. The
second issue concerns the validity of conict behavior questionnaires. Third,
we will discuss self-serving bias (i.e., the tendency to perceive oneself as
more positive than the opponent), which poses a specic threat to validity that
deserves special attention. Fourth, we will address the fact that conict
surveys usually entail correlational, and often cross-sectional, designs that
prohibit conclusions about causality. The fth and last issue concerns the fact
that data on conict behavior are by denition dyadic in nature; we discuss
THE USE OF QUESTIONNAIRES IN CONFLICT RESEARCH 123
the practical and methodological problems that go along with this. We will
illustrate these issues with our own research on conict in organizations and
in close relationships and include recommendations for doing questionnaire
research.
Sensitivity
Conict is a sensitive subject. People generally do not like to admit that they
have conicts with others because it is associated with negative consequences,
as Nauta experienced when interviewing employees in organizations about con-
ict and negotiation (Nauta & Sanders 2000, 2001; Nauta, De Vries, and
Wijngaard 2001). Because of this sensitivity, people are often reluctant to par-
ticipate in research on conict. For example, Nauta (2003) tried to recruit four
different Dutch organizations for a study on organizational determinants and
consequences of conict. Although several organizations showed an interest in
the project, many organizations ultimately refused to participate due to their
fear of launching discussions and dissatisfaction among their employees.
Instead of speaking about conict in the introduction of the study, Nauta
therefore chose to frame the research more positively by stressing how orga-
nizations could improve interpersonal and inter-team collaboration. This pre-
vented participants from experiencing some of the unpleasant connotations
that may accompany conict such as ghting and violence and aggres-
sion, facilitating the decision to participate in the study.
The fact that conict is a sensitive topic may cause another research pro-
blem as well: Response may be low, due to the fear of participants that their
answers become public. For example, in the same research project described
above, Nauta (2003) performed a survey in a medium-sized organization that
measured several biographic variables like gender, age, education, and orga-
nizational unit. Some participants remarked that their anonymity was not
guaranteed by asking these specic background questions. For other partici-
pants, this may have been a reason not to return the questionnaire, because
of their fear of being recognized. Perhaps people who experienced relatively
frequent and severe conicts were less likely to return the questionnaire,
resulting in a selective sample of participants.
When using a questionnaire to study organizational conict, it is therefore
important to ask as few demographic questions as possible. Also, researchers
must promise (and hold their promise) that individual data will never be made
available to the management of the organization. Moreover, careful attention
must be paid to the introduction of the research and further communications
124 AUKJE NAUTA AND ESTHER KLUWER
about the research. Researchers must be clear about the goals and expected
results of the research, in order to motivate people to participate.
Nevertheless, questionnaires can also have an important advantage when study-
ing a delicate subject. For example, in research on close relationships, ques-
tionnaires are often chosen deliberately to study topics that are sensitive, such
as relationship problems, sex, and conict (e.g., Kluwer, Heesink, and Van de
Vliert 1996). People generally feel more comfortable lling out an anonymous
questionnaire rather than talking face-to-face to an interviewer or a researcher
about these topics.
Validity
it out for free and directly receive the results on their preferred mode of
conict management, compared to a reference base.
A questionnaire that is widely used in the Netherlands is the Dutch Test
for Conict Handling (DUTCH; Euwema & Van de Vliert 1990; Janssen &
Van de Vliert 1996; Van de Vliert 1997; see Table 1). De Dreu et al. (2001)
showed that self-ratings on the DUTCH in a role-playing experiment corre-
lated highly with both ratings of the opponent and observer ratings. However,
the correlations for avoiding behavior were not signicant. Avoiding behavior
is difcult to observe, especially in the laboratory, because participants are not
likely to avoid conict when they are asked to act out a conict. In addition,
avoiding is ambiguous behavior open to multiple attributions. For example,
someone who consistently downplays the importance of the conict issue may
do this in order to avoid the issue and to reduce interaction to a minimum.
The opponent, however, may perceive such behavior as a cunning way to get
ones way, to buy time, and to impose ones will on others (i.e., forcing).
Perhaps avoidance more than any other conict behavior involves acts
that are difcult to judge, which in turn makes accurate understanding of
underlying intentions more important. Because individuals have better knowledge
about their own intentions than opponents and neutral observers do, conver-
gence between self-reports and other-reports of avoiding is likely to be low.
Nevertheless, it is still a question whether conict questionnaires such as
the DUTCH really measure behavior in all circumstances. In the study by the
Dreu et al. (2001), the DUTCH was administered immediately after the role-
play was over. Participants thus reected upon the behavior they had just
exhibited. However, in eld studies, the DUTCH is generally used to measure
the behavior one usually or generally shows when engaged in conict. In
this setting, the DUTCH appears to measure a general intention to certain
conict behavior or even a general style, instead of actual behavior. A possi-
ble solution to this problem is the critical incident method (Flanagan 1954;
see also Kluwer 2000), in which participants are asked to recall a recent
conict and answer questions about this particular incident, including behav-
ioral measures. This yields a more valid measure of actual conict behavior
and participants nd it easier to answer questions about their behavior in a
specic real-life situation than questions about their general behavioral style
(see for example, Kluwer 2000). Nevertheless, one must realize that the
answers may be inuenced by the specic context in which the critical inci-
dent took place.
Another option is to use vignettes or scenarios. For example, Kluwer, De
Dreu and Buunk (1998) asked participants to read a scenario, in which they
126 AUKJE NAUTA AND ESTHER KLUWER
Kluwer, Heesink, and Van de Vliert (2000) used vignettes to manipulate the
topic of the conict (housework, paid work, or child care) and each partners
role in the conict (complainant versus defender of the status quo; see Table
2 for an example).
Table 2. Example of a Conict Scenario (Kluwer, Heesink, and Van de Vliert 2000)
You are dissatised with the time your spouse spends on household tasks. For exam-
ple, you are dissatised with how much attention your spouse pays to housework,
how often your spouse tidies the house, cleans, does the dishes, does groceries, or the
way your spouse carries out chores, etcetera. In other words, you want a change in
the time your spouse spends on housework. However, your spouse is satised with the
situation as it is.
Whenever researchers measure conict behavior, chances are high that they
will nd some sort of self-serving bias, that is, a tendency to see ones own
conict behavior as more constructive and less destructive than the conict
behavior of ones opponent (De Dreu, Nauta & Van de Vliert 1995; Kluwer
et al. 1998; Nauta and Van Sloten 2003). Although it is an interesting theo-
retical phenomenon, self-serving bias is a threat to validity, because partici-
pants report biased perceptions of their own and their opponents behaviors.
Self-serving bias may also imply that the measures are inuenced by a gen-
eral tendency towards social desirability, which is more likely to occur when
one rates ones own behavior than when one rates others behavior. This may
also occur when measuring conict frequency, for example in close relation-
ships. People are generally reluctant to admit that they have relationship
conicts, because it might imply that their relationship is in trouble. In addi-
tion, happy couples tend to distort their appraisal of their relationship in pos-
itive ways (e.g., Murray & Holmes 1997). Hence, people may not admit that
they have conicts with their partner due to social desirability biases or they
may not report conict due to positive illusions they hold about their rela-
tionship. For both reasons, measures of absolute conict frequency are likely
to be unreliable.
A possible solution to biased measurements of conict behavior and con-
ict frequency is to measure these concepts from different perspectives: Self
reports, opponent reports and observer reports of conict behavior and conict
frequency. When interrater reliability is high, it can be assumed that the
average of these three perspectives is an accurate (intersubjective) measure.
128 AUKJE NAUTA AND ESTHER KLUWER
Cross-sectional Data
Conict data that are gathered through questionnaires are often correlational
in nature, which prohibits conclusions about causality. For example, Nauta,
De Dreu and Van der Vaart (2002) examined whether social value orientations
predicted the degree to which employees paid attention to the goals of their
colleagues working in other departments, which in turn would lead to
problem-solving behavior during interdepartmental conicts. Although they
used multiple methods (i.e., questionnaires and oral interviews), it was still
difcult to conclude whether a pro-social orientation indeed led to a higher
concern for others goals which in turn resulted in problem-solving behavior.
They could not rule out that a fourth variable (e.g., altruism or differences
in the tendency to give socially desirable answers) explained why respon-
dents scored either high or low on pro-social orientation, goal concerns, and
problem-solving behavior.
A solution to the problem of causality is to use experimental designs. When
this is not possible in the eld, quasi-experimental designs as described by
Cook and Campbell (1979) (e.g., the use of vignettes or scenarios) are a good
alternative. A general recommendation is to use multiple data collection
methods (e.g., both observation and self-report) or longitudinal designs. How-
ever, many practical arguments that have to do with time, energy, and money
hinder researchers from using these designs. Still, we recommend the use
of at least one extra measurement method. For example, in their study on
conicts about working relationships between employees and their superiors
in ten different organizations, Nauta and Van Sloten (2004) used question-
naires, together with data from personnel les about absenteeism. Moreover,
they distributed surveys among both employees and their superiors, which enabled
them to aggregate data of superiors about determinants of conict behavior to
the organizational level, and then examine whether these organizational vari-
ables (like communication culture, perceived common ground) predicted the
conict behaviors of employees.
Dyadic Data
The last issue that we address concerns the dyadic nature of conict data. By
denition, conict is an interpersonal concept: Conict is experienced when
one feels frustrated or hindered by some other party (Van de Vliert 1998). It
is therefore interesting to study both sides of the conict. However, in orga-
nizations, it is often difcult to couple data of two conict parties while at
the same time guaranteeing anonymity. Therefore, studies of dyadic conict
THE USE OF QUESTIONNAIRES IN CONFLICT RESEARCH 129
than observing actual interaction. It simply measures both parties joint per-
ceptions regarding interaction patterns that occur during conict. A weakness
of this approach is that information about spouses individual perceptions is
lost by aggregating to the dyad level.
One way to use information at both the individual level and the dyadic
level is to use multi-level analysis (Raudenbush & Bryk 1986; Snijders &
Bosker 1999). With multi-level analysis, it is possible to examine the impact
of variables at both the dyadic and the individual level upon a dependent vari-
able at the individual level. Moreover, multi-level analysis takes into account
that individuals within dyads are interdependent.
Multi-level analysis is comparable to regression analysis: Its goal is to
build a model that expresses how the dependent variable can be explained by
multiple independent variables. An important difference, however, is that the
variance in the dependent variable is split up into variance at the individual
level and variance at the dyadic (or group) level. This variance is explained
by both variables at the group level (e.g., dyadic means on conict behavior)
and variables at the individual level (e.g., individual measures of conict
behavior).
In sum, we recommend that, when it is theoretically necessary to couple
data of both sides of a conict, researchers should use analytic techniques that
take into account the nested structure of the data.
Conclusions
In this article, we have described some of the issues that researchers may
come across when they examine conict through the use of questionnaires.
We discussed the sensitivity of conict as a research topic which may hinder
the recruitment of participants. Nevertheless, questionnaires may sometimes
be more suitable to measure touchy subjects than more confrontational
research methods such as face-to-face interviews. Furthermore, we discussed
the validity of conict questionnaires and the fact that conict behavior is eas-
ily confounded with behavioral intentions. Specically, we noted that ques-
tionnaire measures of conict behavior are likely to be biased, because participants
tend to give self-serving and/or socially desirable answers. In addition, we
discussed the correlational nature of most questionnaire research designs.
Finally, we described that conict data are often dyadic in nature, which leads
to specic methodological problems due to the dependency of dyad members.
We would like to conclude with a few recommendations to conict re-
searchers who want to use questionnaires. First, conict research generally needs
THE USE OF QUESTIONNAIRES IN CONFLICT RESEARCH 131
References
Acitelli, L.K., and Antonucci, T.C. (1994). Gender differences in the link between marital sup-
port and satisfaction in older couples, Journal of Personality and Social Psychology 67:688698.
Belson, W.A. (1981). The design and understanding of survey questions. Aldershof: Gower.
Blake, R., and Mouton, J.S. (1964). The Managerial Grid. Houston, TX: Gulf.
Christensen, A., and Heavey, C.L. (1993). Gender differences in marital conict: The demand/with-
draw interaction pattern. In S. Oskamp and M. Constanzo (Eds.), Gender issues in con-
temporary society. London: Sage. 113141.
Cook, T.D., and Campbell, D.T. (1979). Quasi-experimentation. Design and analysis issues for
eld settings. Boston: Houghton Mifin.
De Dreu, C.K.W., Evers, A., Beersma, B., Kluwer, E.S., and Nauta, A. (2001). A theory based
measure of conict management strategies in the work place, Journal of Organizational
Behavior 22:645668.
De Dreu, C.K.W., and Carnevale, P.J. (2003). Motivational bases of information processing
and strategy in conict and negotiation. In M.P. Zanna (Ed.), Advances in Experimental
Social Psychology. Vol. 35. New York: Academic Press 235291.
De Dreu, C.K.W., Harinck, F., and Van Vianen, A.E.M. (1999). Conict and performance in
groups and organizations. In C.L. Cooper, and I.T. Robertson (Eds.), International Review
of Industrial and Organizational Psychology. Vol. 14. Chichester, UK: Wiley. 376405.
De Dreu, C.K.W., Nauta, A., and Van de Vliert, E. (1995). Self-serving evaluation of conict
behavior and escalation of the dispute, Journal of Applied Social Psychology, 25, 20492066.
De Vaus, D.A. (1991). Surveys in social research. London: UCL Press/Allen and Unwin.
Dillman, D.A. (1978). Mail and telephone surveys: The total design method. New York: Wiley.
Euwema, M.C., and Van de Vliert, E. (1990). Gedrag en escalatie bij hirarchische conicten
[Behavior and escalation in hierarchical conicts], Toegepaste Sociale Psychologie, 4:2841.
Flanagan, J.C. (1954). The critical incident technique, Psychological Bulletin 51:327359.
132 AUKJE NAUTA AND ESTHER KLUWER
Pruitt, D.G., Peirce, R.S., McGillicuddy, N.B., Welton, G.L., and Castrianno, L.M. (1993).
Long-term success in mediation. Law and Human Behavior 17:313330.
Putnam, L.L., and Wilson, C.E. (1982). Communicative strategies in organizational conicts:
Reliability and validity of a measurement scale. In M. Burgoon (Ed.), Communication Yearbook.
Vol. 6. Beverly Hills, CA: Sage. 629652.
Rahim, A., and Magner, N.R. (1995). Conrmatory factor analysis of the styles of handling
interpersonal conict: First-order factor model and its invariance across groups, Journal of
Applied Psychology 80:122132.
Raudenbush, S.W., and Bryk, A.S. (1986). A hierarchical model for studying school effects,
Sociology of Education 59:117.
Rubin, J.Z., Pruitt, D.G., and Kim, S. (1994). Social conict: Escalation, stalemate, and
settlement. New York: McGraw-Hill.
Schuman, H. and Scott, J. (1987). Problems in the use of survey questions to measure public
opinion. Science, 236, 957959.
Shaughnessy, J.J., Zechmeister, E.B., and Zechmeister, J.S. (2000). Research methods in psy-
chology. Boston: McGraw Hill.
Shockley-Zalabak, P. (1988). Assessing the Hall Conict Management Survey, Management
Communication Quarterly 1:302320.
Snijders, T.A.B., and Bosker, R.J. (1999). Multi-level analysis. London: Sage.
Tanur, J.M. (1992). Questions about questions: Inquiries into the cognitive bases of surveys.
New York: Russell Sage Foundation.
Thomas, K.W. (1992). Conict and negotiation processes in organizations. In M.D. Dunnette
and L.M. Hough (Eds.), Handbook of industrial and organizational psychology 2nd ed.,
Vol. 3. Palo Alto, CA: Consulting Psychologists Press. 651717.
Thomas, K.W., and Killman, R.H. (1974). The Thomas-Kilmann conict mode instrument.
Tuxedo, N.Y.: Xicom.
Thomas, K.W., and Killman, R.H. (1978). Comparison of ve instruments measuring conict
behavior, Psychological Reports 42:11391145.
Van de Vliert, E. (1997). Complex interpersonal conict behaviour. Theoretical frontiers. Hove:
Psychology Press.
Van de Vliert, E. Conict and conict management. In P.J.D. Drenth and H. Thierry (Eds.)
Handbook of work and organizational psychology 2nd ed., Vol. 3. Personnel psychology.
Hove, UK: Psychology Press. 351376.
A Multilevel Approach to Investigating Cross-National
Differences in Negotiation Processes
Multilevel Modeling
These residual variances will decrease if the independent variables from the
two levels have a signicant impact on the dependent variable. Mlwin can
further split the country-level residual variance of the dependent variable into
the residual variance of intercepts and the residual variance of slopes. The
residual variance of intercepts (cf. means) refers to the variations in the level
of the dependent variable across nations. The residual variance of slopes (cf.
correlations) refers to the cross-national variations in the relationship between
an independent variable and the dependent variable. The residual variance of
slopes will also decrease if a cross-level interaction term has a signicant
impact on the dependent variable.
Little Opportunity to Use Skills and Much Opportunity to Use Skills and
Abilities Abilities
4 4
Individualism
Job Satisfaction
Job Satisfaction
3.5 3.5
Collectivism
Collectivism
3 3
Individualism
2.5 2.5
1.5 0.5 0.5 1.5 1.5 1 0.5 0 0.5 1 1.5
Blue-Collar White-Collar Blue-Collar White-Collar
a sampling universe consisting of all adult citizens, aged 18 and older (after
a listwise deletion of missing values N = 13,584; M = 412 per nation; Mage =
40, SDage = 15.2; 47 percent female; 65 percent with paid jobs).
There are two kinds of reasons for doing voluntary work. People may take up
volunteer jobs for either self-serving or egoistic reasons, or for other-serving
or altruistic reasons, or for both. We argued that the two types of motives
might cooperate or compete with each other depending on national contextual
variables. In particular, in rich nations, self-serving and other-serving motiva-
tions might well be positively linked because self-sacricing is a way of
establishing ones self-identity, conrming ones notion of the sort of person one
sees oneself to be, and expressing the values appropriate to this self-concept
(Katz & Kahn 1978: 361). By contrast, in poor nations, volunteer workers
might have to make a relatively serious choice between serving their own
interests and acting out of altruistic or humanitarian concerns. In addition, as
people have to consume more resources to cope with extremely cold or hot
climates, we hypothesized that, in countries with more demanding cold or hot
climates, national wealth produces competition rather than cooperation
between self-serving and other-serving motivations. In countries with more
temperate climates, however, the degree of national wealth was supposed to
have less inuence on the cooperative or competitive nature of the complex
motivation to engage in volunteer work. Depicted in Figure 3 is the path dia-
gram of the variant of the cross-level nation-as-moderator model that repre-
sents those expectations.
A stepwise multilevel analysis was conducted to predict other-serving moti-
vations for doing voluntary work. In step 1, country-level individualism was
controlled for because this cultural dimension is positively associated with
both colder climates and national wealth. In steps 2 and 3, the individual-level
motivation to serve ones own interests was entered (the degree of self-serving
motivation in step 2, the cross-national variation in the relationship between
self-serving motivation and other-serving motivation in step 3). Step 4 con-
tained average climatic temperature, temperature-squared and gross national
income per capita. The two-way interactions followed in step 5, and the three-
way interactions in step 6.
The analysis revealed that cultural individualism at the country level does
not play a part (step 1), that self-serving motivation at the individual level is
positively related to other-serving motivation (step 2), and that the link
between self-serving motivation and other-serving motivation varies consider-
ably from one nation to another (step 3). The most important nding, in step
6, represented in Figure 4, showed that a voluntary workers self-serving and
other-serving motivations tend to be positively linked in high-income regions
CROSS-NATIONAL DIFFERENCES IN NEGOTIATION PROCESSES 143
Other-Serving Motivation
Other-Serving Motivation
3.5 3.5 3.5
Poor Countries
2.5 2.5 2.5 Poor Countries
1.5 1.5 Rich Countries 1.5
0.5 0.5 0.5
0.5 0.5 0.5
1.5 1.5 Poor Countries 1.5
2.5 Rich Countries 2.5 2.5
3.5 3.5 Rich Countries
3.5
4.5 4.5 4.5
1.5 0.5 0.5 1.5 1.5 0.5 0.5 1.5 1.5 0.5 0.5 1.5
Self-Serving Motivation Self-Serving Motivation Self-Serving Motivation
Figure 4. National Wealth and Thermoclimate Jointly Inuence the Link between
Self-Serving and Other-Serving Motivations
particular psychological relationship, the link between self-esteem and life sat-
isfaction, is systematically related to cultural individualism. However, this
two-step approach neglects the problem of differential reliability.
Differential reliability occurs when the number of respondents varies con-
siderably across the nations under investigation. Oishi, Diener, Lucas, and Suh
(1999) pointed out that the number of respondents in Diener and Dieners
(1995) study ranged from 29 to 985. Consequently, the correlation coefcient
computed in Cameroon with 29 respondents is less reliable than the correla-
tion coefcient obtained in Canada with 985 respondents. Oishi et al. (1999)
further argued that multilevel modeling resolves the problem of differential
reliability caused by directly relating correlation coefcients drawn from sep-
arate countries to the national characteristics of these countries. In multilevel
analysis, the differential reliability becomes less problematic because regres-
sion lines instead of correlation coefcients are used as dependent variables.
In short, the studies on job satisfaction and motives for volunteer work dis-
cussed in the preceding paragraphs demonstrate that each of the variants of
the cross-level model shown in Figures 1 and 3 can be used to tackle specic
issues in cross-national research. More importantly, by linking national char-
acteristics with the cross-national variations in individual-level relationships
in a cross-level model, one can predict and explain the variations in these
individual-level relationships across nations. Last, but not least, using multi-
level modeling, researchers can control for a number of rival explanatory vari-
ables at the country level, and can even examine the joint effects of different
country-level variables on individual-level relationships.
CROSS-NATIONAL DIFFERENCES IN NEGOTIATION PROCESSES 145
& Bosker 1999). However, in practice, it is costly to collect data from such
a large number of countries. Another limitation is that only rather simple indi-
vidual-level relationships can be tested. As shown in the studies reported
above (Huang & Van de Vliert, forthcoming; Van de Vliert et al., 2004), mul-
tilevel modeling only allows for tests of the moderating effects of national
contextual variables on a single individual-level relationship or on an indi-
vidual-level two-way interaction effect. As a rule, negotiation models pro-
posed by various researchers are characterized by a rather complex structure
(Gelfand & Dyer 2000; Morris & Fu 2001). It is difcult to test these nego-
tiation models using multilevel analysis. Lastly, the research model proposed
here can be used to study and explain cross-national variations in negotiating
behavior and outcomes, but not the behavior and outcomes of people from
different countries during negotiation.
These limitations notwithstanding, the cross-level nation-as-moderator
model may bring about new perspectives on theory building. The basic
assumption of the cross-level model is that the impact of the negotiators psy-
chological states, social conditions of negotiation, and behavior of the nego-
tiators on negotiation outcomes may systematically vary across nations. In
some nations, some strategies may have a positive effect on outcomes; in oth-
ers, there may be no effect at all. There may even be a negative effect in cer-
tain nations. These cross-national variations may exist in a systematic manner.
And this systematic variation may well be explained by some national con-
textual variables. Put in a different way, using multilevel modeling, we can
predict the applicability of certain theoretical relationships across nations. The
cross-level nation-as-moderator model permits researchers to generalize about
the cross-national applicability of individual-level negotiation processes. We
CROSS-NATIONAL DIFFERENCES IN NEGOTIATION PROCESSES 147
References
Brett, J.M. (2000). Culture and Negotiation, International Journal of Psychology 35(2):
97104.
Brett, J.M., and Okumura, T. (1998). Inter and intracultural negotiations: US and Japanese
negotiators, Academy of Management Journal 41: 495510.
Bryk, A.S. and Raudenbush, S.W. (1992). Hierarchical Linear Models: Applications and Data
Analysis Methods. London/New Delhi: Sage.
Bryk, A.S., Raudenbush, S.W., and Congdon, R.T., Jr. (1994). Hierarchical Linear Modeling
with the HLM/2L and HLM/3L Programs. Chicago: Scientic Software International.
Carnevale, P.J., Pruitt, D.G., and Britton, S.D. (1979). Looking tough: The negotiator under
constituent surveillance, Personality and Social Psychology Bulletin 5: 118121.
Chiu, L.H. (1993). Self-esteem in American and Chinese (Taiwanese) children, Current
Psychology: Research and Reviews 11: 309313.
Diener, E., and Diener, M. (1995). Cross-cultural correlates of life satisfaction and self-
esteem, Journal of Personality and Social Psychology 68(4): 653663.
Diener, E., Diener, M., and Diener, C. (1995). Factors predicting the subjective well-being of
nations, Journal of Personality and Social Psychology 69(5): 851864.
Fu, H.Y. and Morris, M.W. (2000). Need for closure fosters adherence to cultural norms:
Evidence from cross-cultural studies of conict resolution choices, Stanford GSB Research
Paper, #1649.
Gelfand, M.J., and Realo, A. (1999). Individualism-collectivism and accountability in inter-
group negotiations, Journal of Applied Psychology 84(5): 721736.
Goldstein, H., Rasbash, J., Plewis, I., Draper, D., Browne, W., Yang, M., Woodhouse, G., and
Healy, M. (1998) A users guide to Mlwin. University of London.
Graham, J.L., Kim, D.K., Lin, C.Y., and Robinson, M. (1988). Buyer-seller negotiations
around the Pacic Rim: Differences in fundamental exchange processes, Journal of
Consumer Research 15: 4854.
Hofstede, G.H. (1991). Cultures and Organizations: Software of the Mind. London: McGraw-Hill.
Huang, X., and Van de Vliert, E. (2004). Job level and national culture as joint roots of job
satisfaction, Applied Psychology: An International Review, 53: 329348.
Inglehart, R. (1997). Modernization and Postmodernization: Cultural, Economic and Political
Change in 43 societies, Princeton: Princeton University Press.
Kreft, I., and De Leeuw, J. (1998). Introducing Multilevel Modeling. London, Thousand Oaks;
New Delhi: Sage Publications.
Locke, E.A. (1976). The nature and causes of job satisfaction. In M.D. Dunnette, (Ed.)
Handbook of Industrial and Organizational Psychology. Chicago: Rand McNally College
Publication.
148 XU HUANG AND EVERT VAN DE VLIERT
Lytle, A.L., Brett, J.M., Barsness, Z.I., Tinsley, C.H., and Janssens, M. (1995). A paradigm
for conrmatory cross-cultural research in organizational behavior, Research in Organi-
zational Behavior 17: 167214.
Maslow, A.H. (1954). Motivation and Personality. New York: Harper & Row.
McCusker, C.R. (1994). Individualism-collectivism and relationships in distributive negotia-
tion: An experimental analysis, Unpublished Dissertation. University of Illinois at Urbana-
Champaign.
Oishi, S., Diener, E.F., Lucas, R.E., and Suh, E.M. (1999). Cross-cultural variations in pre-
dictors of life satisfaction: Perspectives from needs and values, Personality and Social
Psychology Bulletin 25(8): 980990.
Snijders, T.A.B., and Bosker, R.J. (1999) Multilevel Analysis: An Introduction to Basic and
Advanced Multilevel Modeling. London: Sage.
Stigler, J.W., Smith, S., and Mao, L. (1985). The self-perception of competence by Chinese
children, Child Development, 56: 12591270.
Van de Vijver, F.J.R. and Poortinga, Y.H. (2002). Structural equivalence in multilevel research,
Journal of Cross-Cultural Psychology 33: 141156.
Van de Vliert, E., Huang, X., and Levine, R.V. (2004). National wealth and thermal climate
as predictors of motives for volunteer work, Journal of Cross-Cultural Psychology, 35:
6273.
World Values Study Group (1994). World values survey, 19811984 and 19901993. Ann
Arbor: Inter-university Consortium for Political and Social Research (ICSPR).
Methodological Challenges in the Study of Negotiator Affect
In the pages that follow, we rst briey review existing research on affect
in negotiation with particular attention to methodological approaches taken in
empirical studies. We then point to methodological issues raised in these stud-
ies and consider possible remedies. The latter discussion is organized around
the distinction between affective experience and affect strategy in negotiation.
Our objective is to augment the likelihood that researchers can capture the
real emotions that individuals experience, express, mask, and strategically
deploy within the negotiation encounter.
Summary
The most prevalent empirical technique for the study of affect in negotiation
is the manipulation or measurement of affect as a precursor to interaction.
Typically the form of affect that is induced or measured is a diffuse mood
state (rather than a more ephemeral and intense emotional state). A few stud-
ies have assessed affect that arises within the negotiation encounter or as a
consequence of a negotiated outcome. There has been almost no empirical
attention to the use of emotion strategically as a negotiating gambit.
With respect to the strategic side of affect, one issue (among others) is
whether the tactical use of emotion can be reliably detected and measured.
Content coding of audiotapes, videotapes, or transcripts has been used to
study negotiation tactics and process (e.g., Brett, Shapiro and Lytle 1998).
While there is evidence that coders can reliably code facial emotional expres-
sion (Kring & Gordon 1998), it is less certain that coders can accurately
detect whether emotion is being systematically suppressed, exaggerated or
simulated as a tactical ploy. Another problem arises when we consider that
the duration of genuine emotions and mood states varies considerably.
Measuring the life span of individual emotions is difcult but probably nec-
essary in order to study their effects on negotiation processes. As difcult as
it is to measure the persistence of genuine emotion, establishing the temporal
boundaries of strategically generated expressions of affect through observation
is even more challenging.
Remedies
So what are researchers to do? It seems that a shift toward the study of con-
text-rich negotiation is not only inevitable, but also crucial to the advance-
ment of knowledge about the role of emotion in negotiation. The discussion
that follows offers examples of several approaches that have been utilized suc-
cessfully in other research domains, and which may prove adaptable to the
study of affect in negotiation.
The remedy that is probably most straightforward to imagine but rather
more difcult to implement is to place a greater emphasis on the observational
study of live (not simulated) negotiations. As with all eld studies, gaining
access to observe such negotiations is half the battle. Some targets for obser-
vational studies might include: organizations with departments or employees
whose primary function involves negotiation (e.g., sales, purchasing, customer
service, employee recruitment/hiring, labor relations), community mediation
METHODOLOGICAL CHALLENGES IN THE STUDY OF NEGOTIATOR AFFECT 157
centers, judicial settings (e.g., plea negotiation conferences, small claims court
mediation), and other nonprot or governmental agencies where disputes are
frequently negotiated or mediated.
For the observational approach to bear fruit, one must accept that a re-
searcher can accurately detect and reliably classify the experience of emo-
tion by interactants. As we mentioned above, there is evidence that observers
can reliably assess visible emotional expression but expression and experi-
ence are not the same thing. As studies of affect in retrospectively constructed
episodes (Scanzoni & Godwin 1990) attest, recall techniques can elicit from
interacting parties a sense of affective experience that occurred in the past.
Clearly, however, the veracity of recalled emotion states is likely to be en-
hanced to the extent that the temporal distance between emotional experience
and recall is diminished.
A tool that has been employed to address this problem is experience sam-
pling methodology (ESM) (also referred to as ecological momentary assess-
ment or augmented diary methods). Study participants are asked to report on
their current mental states (cognitive and/or emotional) at frequent intervals
(randomly, episodically, or at particular time intervals prompted by signals
from the researcher) over some time period (Bolger, Davis, and Rafaeli 2003;
Kubey, Larson, and Csikszentmihalyi 1996). For example, Weiss, Nicholas,
and Daus (1999) used an ESM approach to study the dynamic relationships
among affect, expectancy theory-based job beliefs, and job satisfaction. They
utilized written questionnaires (Current Mood Report, Larsen & Kasimatis
1990) and beepers to prompt responses, nding that average mood contributed
to job satisfaction over and above job beliefs and dispositional tendencies
toward happiness. Ilies and Judge (2002) explored the longitudinal interrela-
tionships among mood (positive and negative affect), personality and job
satisfaction. Four times a day over a 19-day period, participants completed a
web-based mood adjective questionnaire; other measures of personality
and job satisfaction were also collected at the same time (responses were prompted
by email). They found that level and variability of mood predicted level and
variability of job satisfaction, and that within-person, these varied together
over time.
Given issues with compliance when paper diaries are used (e.g., Feldman
Barrett and Barrett 2001), electronic devices such as cellular telephones, per-
sonal digital assistants or palmtop computers are being used to collect ques-
tionnaire responses in ESM studies, particularly in health studies (e.g.,
Collins, Kashdan, and Golnish 2003; Kamarck, Shiffman, Smithline, Goodie,
Paty, Gnys and Jong 1998). Bolger, Davis and Rafaeli (2003) and Feldman
Barrett and Barrett (2001) discuss practical and technological considerations
related to use of such electronic devices in ESM.
158 BRUCE BARRY AND INGRID SMITHEY FULMER
which the traders were responding, and found, among other things, prelimi-
nary evidence suggestive of experience effects (i.e., experienced traders
seemed to be less emotionally sensitive to certain market changes than less
experienced traders). A recent health psychology study combined ambulatory
physiological monitoring and electronic experience sampling methods to
examine the effect of everyday stress and emotion on blood pressure (ABP)
(Kamarck et al., 1998). One hundred twenty participants wore ABP monitors
that inated and took blood pressure readings every 45 minutes during the
day for six days; when the cuff inated, respondents also responded to ques-
tionnaires (Diary of Ambulatory Behavioral States) on palm-top computers.
Although the effects of negative affect and emotional arousal on blood pres-
sure were generally modest, researchers noted substantial individual differ-
ences in blood pressure responsiveness to emotion.1
Realistically, the collection of physiological data is probably not feasible in
all eld studies of emotion, but as the data collection process becomes tech-
nically easier to do, collecting at least some basic physiological data from
negotiators could provide additional insights into the biological processes that
accompany negotiators emotional responses. At the present time it is not
clear that physiological data (including brain imaging) are precise enough to
differentiate among different discrete emotions; they are better at distinguish-
ing between broad positive and negative emotions (Cacioppo et al. 2000).
However, these approaches may be useful in conjunction with other approaches
to further corroborate whether emotional arousal was indeed experienced physio-
logically, which could prove especially valuable in detecting the use of sup-
pressive emotion-based tactics.
In situations where negotiation transcript data are available, techniques of
natural language analysis may be useful. On the one hand, some researchers
urge caution in using counts of emotion words to study emotion:
From an evolutionary perspective, language did not emerge as a vehicle to
express emotion. In natural speech we generally use intonation, facial expres-
sion, or other nonverbal cues to convey how we feel. Emotional tone is also
expressed through metaphor and other means not directly related to emotion
words. Taken together, it is our sense that emotion researchers should hesitate
before embarking on studies that rely exclusively on the natural production of
emotion words (Pennebaker, Mehl and Niederhoffer 2003: 571).
On the other hand, there is preliminary evidence (Rogan & Hammer 1995),
that patterns of language intensity (use of metaphors, obscure words, profan-
ity statements, qualiers, etc.) and valence may reect negotiator affective
patterns and be related to crisis negotiation outcomes.
160 BRUCE BARRY AND INGRID SMITHEY FULMER
Shifting from the quantitative end of the research design spectrum to the
qualitative end, we discover still more tools for exploring the role of emotion
for the negotiator-in-context. A hallmark of ethnography and related forms of
qualitative research is the ability to collect data unconstrained by prior expec-
tations about the information that can be acquired and the variables that can
be examined (Becker 1996/2001). Emotion is arguably more socially and cul-
turally embedded, and more contextually inuenced than cognitive process-
ing and so may lend itself more to qualitative methods in order to obtain
sufciently rich or thick description (Geertz 1973) about how emotion man-
ifests itself in vivo. Ethnographic methods such as interviews and participant-
observation are particularly well suited to this task, with researchers acting
both as observers of interaction and as interviewers regarding the emotional
tenor and experience of interaction that has just occurred. Friedmans (1994)
study of labor negotiations is an example of the extensive application of
ethnography to negotiation. Other examples of this approach used to study
emotion-related phenomena (although not in negotiation contexts per se)
include: the work of Thoits (1996), who utilized participant-observation to
study how people manage the emotions of others in support groups; a study
of emotional contrast strategies of bill collectors that employed interviews and
participant-observation (Rafaeli & Sutton 1991); and an interview-based study
of reciprocal emotion management by paralegals (Lively 2000).
Open-ended interview methods could conceivably be applied to laboratory
negotiations as well, with experimenters poised to debrief subjects about the
affective components of the interaction just experienced, perhaps in the form
of verbal protocol analysis that accompanies a replay of a tape of the inter-
action although again, one is mindful that the range of emotional experi-
ence in the lab may be constrained by the articiality of the task or its
attendant incentives, risks, and consequences. Still, there are a strong basis to
infer from the literature on the psychology of emotions that individuals can
do quite well at reporting their own emotions during and following events that
elicit emotional experience and expression (e.g., Kring & Gordon 1998).
Given the strengths and limitations inherent in each of these approaches, it
may make sense to consider the application of several methods in the same
study. This could be accomplished by simultaneously combining several quan-
titatively oriented approaches such as self-report measures and/or physiologi-
cal monitoring (e.g., Kring & Gordon 1998). Taking advantage of the full
spectrum of research approaches, negotiation researchers might also consider
utilizing mixed-methods approaches that combine, either sequentially or con-
currently, both quantitative methods and qualitative approaches (see Creswell
2003, for a review of mixed methods research strategies).
METHODOLOGICAL CHALLENGES IN THE STUDY OF NEGOTIATOR AFFECT 161
Conclusion
The position taken in this article can be distilled to six observations: (1) emo-
tion is fundamental to the negotiation encounter; (2) research on negotiation
has focused on cognitive processes and underemphasized the role of emotion;
(3) emotion is relevant to the negotiation encounter as anterior, experiential,
and strategic phenomena; (4) conceptualizations of how emotion ts into
negotiation are far out in front of empirical investigation in terms of sophis-
tication and complexity; (5) prevailing methods in the empirical literature on
negotiation are not well suited to the investigation of emotional processes;
and (6) creative new methods (in the sense of being novel to research on the
psychology of negotiation) are needed to do empirical justice to the subject
of affect in negotiation. In this article, we survey a variety of research
approaches already used to study emotion in other elds that might also be
adaptable to the study of affect in context-rich negotiation. We encourage
negotiation scholars to consider the use of multiple methods and varied tech-
niques such as those described here as we progress toward the testing of
existing theories, and as a stepping stone toward future theory building
regarding the role of emotion in negotiation.
Note
1. See Bolger, Davis and Rafaeli (2003), for more on physiological data collection in con-
junction with diary studies.
References
Alliger, G.M., and Williams, K.J. (1993). Using signal-contingent experience sampling
methodology to study work in the eld: A discussion and illustration examining task per-
ceptions and mood, Personnel Psychology 46:525549.
Allred, K.G. (1999). Anger and retaliation: Toward an understanding of impassioned conict
in organizations. In R.J. Bies, R.J. Lewicki, and B.H. Sheppard, (Eds.) Research on nego-
tiation in organizations, Vol. 7. Stamford, CT: JAI Press.
Allred, K.G., Mallozzi, J.S., Matsui, F., and Raia, C.P. (1997). The inuence of anger and
compassion on negotiation performance, Organizational Behavior and Human Decision
Processes 70:175187.
Baron, R.A. (1990). Environmentally induced positive affect: Its impact on self-efcacy, task
performance, negotiation, and conict, Journal of Applied Social Psychology 20:368384.
Barry, B. (1999). The tactical use of emotion in negotiation. In R.J. Bies, R.J. Lewicki, and
B.H. Sheppard (Eds.) Research on negotiation in organizations, Vol. 7. Stamford, CT: JAI
Press.
162 BRUCE BARRY AND INGRID SMITHEY FULMER
Barry, B., Fulmer, I.S., and Van Kleef, G.A. (2004). I laughed, I cried, I settled: The role of
emotion in negotiation. In M. Gelfand and J. Brett (Eds.) The handbook of negotiation and
culture: Theoretical advances and cross-cultural perspectives. Palo Alto, CA: Stanford University
Press.
Barry, B., and Oliver, R.L. (1996). Affect in dyadic negotiation: A model and propositions,
Organizational Behavior and Human Decision Processes 67:127143.
Bazerman, M.H., Curhan, J.R., Moore, D.A., and Valley, K. (2000). Negotiation, Annual
Review of Psychology 51:279314.
Becker, H.S. (2001). The epistemology of qualitative research. In R. Emerson, (Ed.) Contemporary
eld research: Perspectives and formulations. Prospect Heights, IL: Waveland Press.
(Reprinted from Jessor et al., [Eds.] [1996]. Ethnography and human development. Chicago:
University of Chicago Press.)
Bolger, N., Davis, A., and Rafaeli, E. (2003). Diary methods: Capturing life as it is lived.
Annual Review of Psychology 54:579616.
Brett, J.M., Shapiro, D.L., and Lytle, A.L. (1998). Breaking the bonds of reciprocity in nego-
tiations, Academy of Management Journal 41(4):410424.
Brief, A.P., and Weiss, H.M. (2002). Organizational behavior: Affect in the workplace.
Annual Review of Psychology 53:279307.
Cacioppo, J.T., Berntson, G.G., Larsen, J.T., Poehlmann, K.M., and Ito, T.A. (2000). The psy-
chophysiology of emotion. In M. Lewis, and J.J. Haviland-Jones, (Eds.) 2nd edition.
Handbook of Emotions. New York: Guilford Press.
Cacioppo, J.T., and Gardner, W.L. (1999). Emotion. Annual Review of Psychology 50:
191214.
Carnevale, P.J.D., and Isen, A.M. (1986). The inuence of positive affect and visual access
on the discovery of integrative solutions in bilateral negotiation, Organizational Behavior
and Human Decision Processes 37:113.
Collins, R.L., Kashdan, T.B., and Gollnish, G. (2003). The feasibility of using cellular phones
to collect ecological momentary assessment data: application to alcohol consumption.
Experimental and Clinical Psychopharmacology 11, 1:7378.
Creswell, J.W. (2003). Research design: Qualitative, quantitative and mixed methods appro-
aches. Thousand Oaks, CA: Sage.
Davidson, R.J. (1994). On emotion, mood, and related affective constructs. In P. Ekman and
R.J. Davidson (Eds.) The nature of emotion: Fundamental questions. New York: Oxford
University Press.
Feldman Barrett, L. (1998). Discrete emotions or dimensions? The role of valence focus and
arousal focus. Cognition and Emotion 12, 4:579599.
Feldman Barrett, L., and Barrett, D.J. (2001). An introduction to computerized experience
sampling in psychology, Social Science Computer Review 19(2):175185.
Forgas, J.P. (1992). Affect in social judgments and decisions: A multiprocess model,
Advances in Experimental Social Psychology 25:227275.
Forgas, J.P. (1998). On feeling good and getting your way: Mood effects on negotiator
cognition and behavior, Journal of Personality and Social Psychology 74:565577.
Friedman, R.A. (1994). Front stage, backstage: The dramatic structure of labor negotiations.
Cambridge: MIT Press.
Geertz, C. (1973). The interpretation of cultures: Selected essays. New York: Basic Books.
Goates, N., Barry, B., and Friedman, R.A. (2003). Good karma: How individuals construct
schemas of reputation in negotiation contexts. Paper for the annual meeting of the Inter-
national Association for Conict Management, Melbourne, Australia.
METHODOLOGICAL CHALLENGES IN THE STUDY OF NEGOTIATOR AFFECT 163
Hagemann, D., Waldstein, S.R., and Thayer, J.F. (2003). Central and autonomic nervous sys-
tem integration in emotion. Brain and Cognition 52:7987.
Hegtved, K.A., and Killian, C. (1999). Fairness and emotions: Reactions to the process and
outcomes of negotiations, Social Forces 78:269303.
Hochschild, A.R. (1983). The managed heart: Commercialization of human feeling. Berkeley:
University of California Press.
Ilies, R., and Judge, T.A. (2002). Understanding the dynamic relationships among personality,
mood, and job satisfaction: A eld experience sampling study, Organizational Behavior and
Human Decision Processes, 89:11191139.
Isen, A. (1987). Positive affect, cognitive processes and social behaviour. In L. Berkowitz
(Ed.) Advances in experimental social psychology, Vol. 20. New York: Academic Press.
Kamarck, T.W., Shiffman, S.M., Smithline, L., Goodie, J.L., Paty, J.A., Gnys, M., and Jong,
J.Y. (1998). Effects of task strain, social conict, and emotional activation on ambulatory
cardiovascular activity: Daily life consequences of recurring stress in a multiethnic adult
sample, Health Psychology 17:1729.
Kimmel, M.J., Pruitt, D.G., Magenau, J.M., Konar-Goldband, E., and Carnevale, P.J.D. (1980).
Effects of trust, aspiration, and gender on negotiation tactics, Journal of Personality and
Social Psychology 38: 922.
Kramer, R.M., Newton, E., and Pommerenke, P.L. (1993). Self-enhancement biases and nego-
tiator judgment: Effects of self-esteem and mood, Organizational Behavior and Human Decision
Processes 56:110133.
Kring, A.M. and Gordon, A.H. (1998). Sex differences in emotion: Expression, experience,
and physiology, Journal of Personality and Social Psychology 74:686703.
Kubey, R., Larson, R., and Csikszentmihalyi, M. (1996). Experience sampling method appli-
cations to communication research questions, Journal of Communication 46:99120.
Lane, R.D., and Nadel, L. (2000). Cognitive neuroscience of emotion. New York: Oxford University
Press.
Larsen, R.J., and Kasimatis, M. (1990). Individual differences in entrainment of mood to the
weekly calendar, Journal of Personality and Social Psychology 58:164171.
Lawler, E.J., and Yoon, J. (1995). Structural power and emotional processes in negotiation.
In R.M. Kramer and D.M. Messick (Eds.) Negotiation as a social process. Thousand Oaks,
CA: Sage.
Lewicki, R.J., Barry, B., Saunders, D.M., and Minton, J.W. (2003). Negotiation. 4th Edition.
Boston: McGraw Hill/Irwin.
Lively, K.J. (2000). Reciprocal emotion management: Working together to maintain strati-
cation in private law rms, Work and Occupations 27(1):3263.
Lo, A.W., and Repin, D.V. (2002). The psychophysiology of real-time nancial risk process-
ing, Journal of Cognitive Neuroscience 14(3):323339.
Moore, D.A., Kurtzberg, T.R., Thompson, L.L., and Morris, M.W. (1999). Long and short
routes to success in electronically mediated negotiations: Group afliations and good vibra-
tions, Organizational Behavior and Human Decision Processes 77:2243.
Morris, M.W., and Keltner, D. (2000). How emotions work: The social functions of emotional
expression in negotiations. In B.M. Staw and R.I. Sutton (Eds.) Research in organizational
behavior, Vol. 11. Amsterdam: JAI.
Oatley, K., and Duncan, E. (1992). Incidents of emotion in daily life. In K.T. Strongman
(Ed.) International Review of Studies on Emotion, Vol. 2. Chichester: John Wiley and
Sons.
164 BRUCE BARRY AND INGRID SMITHEY FULMER
OConnor, K.M., and Arnold, J.A. (2001). Distributive spirals: Negotiation impasses and the
moderating role of disputant self-efcacy, Organizational Behavior and Human Decision
Processes 84:148176.
Park, O.S., Sims, H.P., and Motowidlo, S.J. (1986). Affect in organizations: How feelings and
emotions inuence managerial judgment. In H.P. Sims and D.A. Gioia (Eds.) The thinking
organization. San Francisco: Jossey-Bass.
Pennebaker, J.W., Mehl, M.R., and Niederhoffer, K.G. (2003). Psychological aspects of nat-
ural language use: Our words, our selves, Annual Review of Psychology 54:547577.
Pillutla, M.M., and Murnighan, J.K. (1996). Unfairness, anger, and spite: Emotional rejec-
tions of ultimatum offers. Organizational Behavior and Human Decision Processes 68:208224.
Rafaeli, A., and Sutton, R.J. (1991). Emotional inuence strategies as means of social
inuence: Lessons from criminal interrogators and bill collectors, Academy of Management
Journal 34: 749775.
Rhoades, J.A., Arnold, J., and Jay, C. (2001). The role of affective traits and affective states
in disputants motivation and behavior during episodes of organizational conict, Journal of
Organizational Behavior 22:329345.
Rogan, R.G., and Hammer, M.R. (1995). Assessing message affect in crisis negotiations: An
exploratory study, Human Communication Research 21(4):553574.
Russell, J.A., Bachorowski, J., and Fernandez-Dols, J. (2003). Facial and vocal expressions of
emotion, Annual Review of Psychology 54:32949.
Scanzoni, J., and Godwin, D.D. (1990). Negotiation effectiveness and acceptable outcomes,
Social Psychology Quarterly 53:239251.
Thoits, P.A. (1996). Managing the emotions of others, Symbolic Interaction 19(2):85109.
Thompson, L.L., Nadler, J., and Kim, P.H. (1999). Some like it hot: The case for the emo-
tional negotiator. In L.L. Thompson, J.M. Levine, and D.M. Messick, (Eds.) Shared cogni-
tion in organizations: The management of knowledge. Mahwah: Erlbaum.
Watson, D., Clark, L.A., and Tellegen, A. (1988). Development and validation of brief mea-
sures of positive and negative affect: The PANAS scales. Journal of Personality and Social
Psychology 54: 10631070.
Weingart, L.R., Bennett, R.J., and Brett, J.M. (1993). The impact of consideration of issues
and motivational orientation on group negotiation processes and outcome, Journal of
Applied Psychology 78:504517.
Weingart, L.R., Hyder, E.B., and Prietula, M.J. (1996). Knowledge matters: The effect of tac-
tical descriptions on negotiation behavior and outcome, Journal of Personality and Social
Psychology 70:12051217.
Weingart, L.R., Thompson, L.L., Bazerman, M.H., and Carroll, J.S. (1990). Tactical behavior
and negotiation outcomes, International Journal of Conict Management 1:731.
Weiss, H.M., Nicholas, J.P. and Daus, C.S. (1999). An examination of the joint effects of
affective experiences and job beliefs on job satisfaction and variations an affective experi-
ences over time, Organizational Behavior and Human Decision Processes 78(1):124.
Comparative Case Studies
I. WILLIAM ZARTMAN
Introduction
Case studies are one of, if not the, most frequently used methods for con-
ducting research on negotiation. They can vary from purely historical studies
that seek to establish all the relevant facts of the encounter to analytical stud-
ies chosen to illustrate specic theoretical propositions. The most successful
cases for the purposes of generating useful knowledge are those located some-
where in the middle of that spectrum, leaning toward the latter side of cen-
ter. Case is used here to refer to the story of negotiations on a single
conict or problem, either as a single set of encounters or as a number of
successive instances. Cases are the best way of combining empirical data (a
redundant expression) with theory and concepts, but their use raises further,
more interesting methodological questions. These questions concern the num-
ber of cases used by a research project and the type of data to be drawn from
them, questions that are at the forefront of importance in the current wave of
scholarship on the subject.
Any research activity designed to create knowledge involves a question
requiring answers or a category of events needing an explanation, a theory
embodying those answers and explanations, and a method for gathering and
using data. Each step in the research process poses its challenges. Given the
need to bridge idiosyncrasies and to combine depth of Weberian understand-
ing with the breadth of multiple instances, there is much to be said for using
a comparative case method to answer questions and provide explanations
about negotiations, focusing on the basic question of how outcomes are
obtained. Case studies can be exploratory or conrmatory, providing inductive
ideas for generalized explanations or deductive testing of logical constructs.
Case studies can show causal links; they shed light on process and allow an
exploration of the dynamic path from components to results, thus satisfying
the needs of both analysts and practitioners. Comparative case studies lie at
the crossroads of reality and theory; they present their evidence through the
eyes of a knowledgeable specialist and they test it against the hypothetical
constructs of a creative conceptualist. So the Janus-faced challenge to case
study authors is dual and the standards of quality are high.
International Negotiation Series 2:
P. Carnevale and C.K.W de Dreu (eds.) Methods of Negotiation Research, 165176
2006 Koninklijke Brill NV. Printed in the Netherlands.
166 I. WILLIAM ZARTMAN
have been named). I have been associated with such concepts as formulas and
ripeness, but there are plenty of others exibility, prenegotiation, BATNA
and security points, integration and distribution, toughness dilemma, value
making and value taking, among many others that have made the under-
standing of the negotiation process and the answering of the crucial question
move ahead. But for all their logic, concepts and effects such as these (to
dodge the big word theory for the moment) need to be tested against data
to make sure that they work and are not just gments of creative imagination.
Testing logic against real data has three purposes: to see if the logic is real,
to provide an explanation of real events if it is, and to incite a search for
alternative explanations alternative logics for exceptions. Where do these
data come from? The best source of data is historical reality, from negotia-
tions that have actually happened, been recorded and analyzed in case studies.
Case data are authoritative; they record what happened, not what could or
might have happened (although there is value in counterfactual analysis as
well).1 There are many good historical case studies of negotiation that seek
no particular conceptual guidance or verication but work merely to establish
a good record and understanding of the event itself, nding an explanation of
outcomes in the events (and often personalities) of the case itself. Fine exam-
ples of such historical studies are numerous, ranging from Nicolson (1946) to
Preeg (1970).2 Often such accounts provide generalized bits of wisdom that
emerge from the events and can be helpful in understanding and producing
successes or failures in other (or the same) cases, beginning the inductive
process of theory- and concept-building, as in the very studies just cited. The
contrast is evident in the two oldest accounts of negotiations: the negotiations
recorded in Genesis between God and Abraham over Sodom, which is rich in
concepts but only implicitly, and the negotiations recorded in Thucydides
(1960) between Sparta and Athens, where the purported lessons are made
most explicitly. Other essentially historical studies refer to concepts already
formulated, either induced from other cases or deduced from logical premises.
But whether one is the writer of Genesis, Thucydides, Preeg, Nicolson, or
any other author of quality about cases of negotiation, s/he has been steeped
in the ambiance and context of the case, as a participant observer (as in the
case of the rst three names cited) or as a diligent researcher, and has devel-
oped a feel for the subject that makes a deep understanding analysis of the
case possible. Whatever the level of conceptual sophistication, a case study
writer never stops being a diligent historian too if s/he does a good job, and
that feel for the case allows the writer to get behind the data to give it con-
text and meaning and achieve a Weberian understanding of its dynamics
(Skocpol and Somers 1980/1994).
168 I. WILLIAM ZARTMAN
Choice of Cases
The simplest answer to the choice of data sources is to pick the case in which
the analyst is interested, and to seek guidance in an explanation of its out-
come from the available concepts and theories. In single case studies, the
choice is generally directed by considerations external to the concept; the pur-
pose is to nd explanations for the case, not tests of the concept. The case is
the dependent variable or explanans, the thing being explained, whereas the
concept is the independent variable or explanandum, the thing explaining.
Such use of cases assumes either that the concepts are already well estab-
lished or that the case can be used inductively to derive them. Two excellent
examples are studies of multilateral negotiations, one by Bunn (1992) on arms
control negotiations and the other by Winham (1986) on the GATT Tokyo
Round of negotiations. While focusing on their subject per se, both found the
concept of formulas useful in their analysis. Another insightful pair is Rubins
(1981) collection of interpretations of the Kissinger shuttle negotiations in the
Arab-Israeli conict and Pruitts (1997) collection of analyses of the Oslo
negotiations on the Palestine-Israeli conict; in these cases, the concepts
tested and applied varied among contributing authors.
However, single case studies are of inherently limited utility in producing
knowledge about negotiation as opposed to data on the unique case. Things
that happened once, however engrossing as a story, have no way of telling us
whether they represent regularities or exceptions; truth is stranger than ction
not because it is exceptional but because the story leaves us wondering
whether it is really normal or indeed an exception to normality. The only way
to test and reinforce concepts and theories claims to normal regularity rather
than exceptionality is to look at a number of cases, not just one, and the more
COMPARATIVE CASE STUDIES 169
the better (knowing, as noted, that the generality can never be proved or
expected to be universal). Zartman (1994) and Hampson (1995) are the rst
works to attempt a theoretical analysis of multilateral negotiations, each using
cases somewhat differently; the rst subjects two cases (the Single Europe Act
and the GATT Uruguay Round) to a competition among six different theoret-
ical approaches (game, decision, small group, leadership, coalition, and orga-
nization theories) to try to draw out a common analytical perspective, whereas
the second examines nine cases from which to draw conceptual characteris-
tics and insights.3 Comparative case study exhibits the advantages of in-depth
analysis of reality while overcoming the weaknesses of focusing on one case
alone. But how and what to compare?
There is a large literature on case studies, starting with Mill (1843/1967)
and going on to the most authoritative recent statement made by Alexander
George (1982; George & McKeown 1985). With limited space and an argu-
ment to be made, cases are unlikely to be chosen at random, if indeed there
were a notion of randomness that was operationalizable and applicable. Most
likely, the analyst will begin by choosing a number of cases that are salient
and relevant. Saliency involves importance in the general discourse about
negotiation problems, including simple current events. Relevance concerns applic-
ability to the conceptual issues involved. The more cases can be chosen to
focus on variations relevant to the conceptual issues and hold other features
constant, the more explanatory factors can be isolated and identied, a con-
dition termed structured, focused comparisons. It would also be useful to
bring in negative cases as a control, rather than including only positive cases,
although comparing why it did not happen with why it happens signicantly
increases the difculty of holding constant the elements to be analyzed.
The simplest way to achieve comparison is to examine multiple instances
in the same case. Instances of failure can be compared with instances of suc-
cess in the same country in a comparative analysis that uses specic concepts
to provide an explanation of outcomes, as was well and explicitly done in
Touval (1982) examining nine attempts to mediate the Arab-Israeli conict,
and in Stedman (1991) comparing three failures and one success in mediat-
ing the Zimbabwean anti-colonial war. Interestingly but only coincidentally,
both works focus on ripeness as a major explanatory variable, making impor-
tant contributions to an understanding of the concept. Another comparative
single-case analysis can be provided by examining the role of various com-
ponents of a single set of negotiations, as is done in the rst study of
European integration as negotiation (Meerts & Cede 2004), which analyzes
negotiations in each of the major institutions of the European Union. If mul-
tiple instances or segments within a single case are not available or chosen,
170 I. WILLIAM ZARTMAN
then similar cases can be the corpus of analysis, following the same rules of
structured, focused comparison. As one moves away from multiple instances
within a single case, it becomes more difcult to hold elements constant in
order to focus on particular explanatory aspects of negotiation.
The problem with multiple case studies is that the more the cases, the big-
ger the book and the further the account gets from the important details of
reality. Some excellent multiple case studies guided by or testing concepts get
rather voluminous, such as Crocker, Hampson and Aall (1999, 21 cases in 735
pages) on mediation or Stedman, Rothchild, and Cousens (2002, 9 cases in
728 pages) on post-agreement settlement, leaving in the dust behind them
other studies such as Ali and Matthews (1999, 8 cases in 322 pages) on nego-
tiating African civil wars and Zartman (1995, 11 cases in 353 pages) on the
difculties of negotiating centralist and regionalist civil conict. Yet, the abil-
ity to make comparisons across a number of negotiations gives a rich harvest
of lessons and insights. In addition to their supporting case material, such
studies also contain varying amounts of conceptual or theoretical knowledge
drawn and tested from the cases. Obviously, support is not measured by the
number of pages, but more and longer case accounts can provide more data
in order to test and apply the theory and concepts.
It might be worthwhile examining the last-mentioned work (Zartman 1995)
in greater detail, as an example of the process and advantages (and disadvantages)
of a comparative set of case studies. Cases were chosen, not scientically,
but by the criteria of saliency and interest already mentioned. Most cases
were signicant and unresolved at the time the project started Sri Lanka,
Sudan, Eritrea, Lebanon, South Africa, Angola, Mozambique, Afghanistan,
Colombia plus two other cases, also unresolved but less well known
Euskadi and the Philippines. The challenge was to analyze the negotiations to
date and then provide conceptually-derived prescriptions on ways to bring
them to success. Interestingly, a number of cases Eritrea, Lebanon, South
Africa, Mozambique came to a conclusion while the project was under way,
while some others Angola and Afghanistan reached a successful conclu-
sion of the current phase of the conict, only to be followed by a new phase.
The cases further divided into conicts for control of the central government
and conicts for control of a region. A conceptual framework, built on con-
trasting notions of stalemate and on characteristics of internal wars asym-
metry, phases, agendas, escalation and mediator tactics, was set up at the
beginning and rened inductively in interaction with the case authors, pro-
viding the basis at the end for a review of factors in failure and success and
two parts of a bottom line weakness of both parties that hinders ripeness,
and weakness of mediators that impedes patronage or relationship to give
COMPARATIVE CASE STUDIES 171
a nal account for failure. It was above all the repeated interaction between
the inductive and deductive parts and people of the project that made the
study so effective. On the other hand, it is worth recalling that nothing was
proven; old and new propositions and concepts were supported and pro-
posed, open for further testing.
Multiple case accounts allow the analyst to develop a deeper understanding
of the details and idiosyncrasies of the case, so that the t between the gen-
eralizations and the data can be fully explored, explained, and understood.
This analytical formula has been an integral part of the studies of the
Processes of International negotiation (PIN) Program of the International Institute
of Applied Systems Analysis (IIASA) in Austria. These works include com-
parative case-and-concept studies of negotiations on such issues as civilian
and military use of nuclear material in order to test a number of propositions
regarding the impact of the worlds most dangerous substance on the process
of negotiation (Avenhaus, Kremenyuk & Sjstedt 2002), economic issues in
order to compare the strength of economic vs. negotiation explanations of out-
comes (Sjstedt and Kremenyuk 2000), symmetry and asymmetry in order to
derive strategies and employ a new concept of power (Zartman & Rubin
2000), environmental issues in order to develop analytical and strategic
concepts (Sjstedt 1993), and a broad range of historic and contemporary encoun-
ters in order to answer some major conundrums about resolution vs. transfor-
mation and peace vs. justice (Zartman & Kremenyuk 2004).
But events, unlike concepts, do not naturally come in boxes, with sharp
sides and square corners, and calling an event one thing or another has to be
done with extreme care and support. To evoke the concepts mentioned above,
was there really a formula in the negotiations? What was it, specically; and
when did negotiations pass from diagnosis to formulation? Or was there really
a mutually hurting stalemate (MHS)? How do we know and how did they
(the participants) know, and how long did that perception last (even if they
didnt say so)? Debates about the existence and effects of the concepts and
theories in reality can only be resolved by evidence from reality, and that can
only be supplied by detailed case studies.
If eight cases are better than one and 21 cases better than eight, what about
going to truly large sets containing many tens and even hundreds of cases?
How can we provide even larger collections of comparative data for more
conclusive application and testing? Ostensibly, the answer seems to be found
172 I. WILLIAM ZARTMAN
tion when the evidence is not given, and an assurance of intercoder reliabil-
ity is simply not adequate to relieve questions. Case study accounts may
be compressed, but at least one can spot missing data and questionable
judgments.
Third, direct data are often not available, only indicators, sometimes termed
proxies. Since the method can only handle comparable, quantiable data, and
so, because it has no feel for its subject, it has to rely on indicators or
proxies; subjective elements must be objectied to become data. As a result,
its indicators, such as per capita income or economic growth, are often far
away from the effect they are proxying, such as proneness to the breakdown
of order. Inequalities (in household incomes and in land ownership) are used
as indicators of grievance, as if there were some universal threshold of envy
or economic inequalities were the only and direct cause of protest and revolt
(Collier and Hoefer 2002; Collier et al. 2003: 66).
Fourth, what cannot be measured or proxied is not analyzed. Since it is
difcult to objectify intensity of feelings, such as nationalism or commitment,
or degree of satisfaction with outcomes, these phenomena become ignored
in analysis, even though they have been identied as crucial elements in
negotiation.
Fifth, as a result of the above, data become no longer data but are them-
selves events squeezed into generalizations. They become shorthand for the
event and so enter into a tangling tautology: they become theoretical gener-
alizations required to test theoretical generalizations. If household incomes are
used to proxy grievance in order to test its role in causing conict, they
contain the unproven theory that income inequality directly causes or relates
to grievance. But whether or how it does so or operates is outside the analy-
sis. Of course, the research is not as simple-minded as that; the analyst looks
for evidence or at least an indicator for the effect being evaluated, but the
notation that appears on the correlation chart sets the tautology trap.
Finally, there is remarkably little process in this analysis. Outcomes and
conditions are noted but they are static road signs, neither roads nor driving
skills. Getting there does not tell how, and so the dynamic of the process is
lost. That loss is serious indeed, and marks a step backward in the analysis
of negotiation. When the analysis was in the hands of the historians, attention
tended to be focused on results, with much less on the way in which they
were obtained. Process analysis in social science has been working to correct
the aim, building on the historical analysis. It would be a step backward to
focus simply on correlations between conditions and results.
174 I. WILLIAM ZARTMAN
Conclusion
This is a rich list of problems; an equally full list can be made for case stud-
ies as well. Case studies exchange feel for precision and thrive on it; their
strength is an understanding of the situations they analyze, even if it is
hard to place numbered pieces of those situations into columns in a chart.
Such studies are more interested in arguing and illustrating how perceptions,
processes, communications, and grievances operate in known instances of
negotiation than in correlating inputs and outcomes, and they spend little time
on absent effects, non-instances, or control cases. Case students may be
satised to understand one case well and produce some lessons for someone
else to test on other cases, rather than nding correlations in universes of
cases of varying importance. They are even happier when comparative case
studies can be undertaken, either through successive negotiation attempts in
the same conict or through several negotiations of several problems or
conicts. Yet their data suffer from loose formalization, necessarily small
numbers of cases, and deference to case idiosyncrasy.
At the other extreme of one or a few cases lie studies involving many,
many cases, summarized in aggregate data analyzed by statistical methods.
While useful for establishing correlations, this method has its own problems:
apples and oranges are often crammed into the same indicator, sensitive con-
cepts are crudely operationalized, the variables used for analysis are often so
distant from the phenomena named in the theory that it is hard to be sure the
theory is being tested, and process dynamics are almost invariably lost. All
this is not to say that such studies and their methodology are useless, as this
critique may imply. It does indicate that enormous renement is still awaited,
that conceptual links and assumptions still need analysis, that subjective data
still call for their place in the analysis, and that the statistical correlations can
well be used to provide hypotheses that closer analysis can test.
Yet the balance of advantages and weaknesses, inevitable in any method of
analysis, places case studies in the midst of a search for breadth and depth,
for data and theory. Much of the greatly expanded understanding of the nego-
tiation process made available over the past four decades involves case stud-
ies largely comparative case studies used either to generate or to test conceptual
and theoretical generalizations. Empirical soundness, including a feel for the
subject, harnessed to a concern for usefulness through accurate generalizations
and concepts, can be achieved perhaps even best achieved through com-
parative case studies.
COMPARATIVE CASE STUDIES 175
Notes
1. Too little work is done on historic possibilities at particular decision points, comparing the
possible against the actual and analyzing why a particular decision was made. For some
such case studies, see Tuchman (1984), Parker (1993), Jentleson (2000), and Zartman
(2005).
2. I have tried to cite signicant case studies to illustrate my analysis. There are many more
of them than those cited here and I apologize to their authors for the omissions.
3. Ten years later, the 1994 book was used as the basis for further case studies in Crump and
Zartman (2003) and Crump (2003), expanding the conceptual and empirical development.
4. For a serious effort to bridge this gap, see Sambanis 2004.
5. Walter (2002) combines both methods by examining two cases (Zimbabwe and Rwanda) in
addition to the 72 sets of data.
6. Although Walter (2002: 56) says it does not matter.
References
Ali, T.M. & Matthews, R.O., eds. (1999). Civil Wars in Africa. Toronto: McGill/Queens
University Press.
Avenhaus, R., Kremenyuk, V., & Sjstedt, G., eds. (2002). Containing the Atom: Nuclear
Negotiations for Safety and Security. Lanham, MD: Lexington Books.
Bunn, G. (1992). Arms Control by Committee. Stanford, CA: Stanford University Press.
Collier, John & Hoefer, Anke (2002). Greed and Grievance in Civil War. Working Paper
200201. Oxford University, Centre for the Study of African Economies.
Collier, John, et al. (2004). Breaking the Conict Trap. Washington: World Bank.
Crocker, C.A., Hampson, F.O., and Aall, P., eds. (1999). Herding Cats: Multiparty Mediation
in a Complex World. Washington: United States Institute of Peace Press.
Crump, L., editor (2003). Multiparty Negotiation and Complexity, special issue of International
Negotiation, 8, 2.
Crump, L., and Zartman, I.W. eds. (2003). Multilateral Negotiation and Complexity, special
issue of International Negotiation 8, 1.
Faure, G.O. & Rubin, J.Z., eds. (1993). Culture and Negotiation. Thousand Oaks, CA: Sage.
Fortna, V.P. (2004). Peacetime: Ceasere Agreements and the Durability of Peace. Princeton,
NJ: Princeton University Press.
Genesis 18: 2032.
George, A. (1982). Case Studies for Theory Development, paper presented to the Symposium
on Information Processing in Organizations, Carnegie-Mellon University.
George, A. & McKeown, T. (1985). Case Studies and Theories of Organizational Decision-
Making, Advances in Information Processing 2: 2158.
Hampson, F.O. (1999). Nurturing Peace. Washington: United States Institute of Peace Press.
Jentleson, B., editor (2000). Opportunities Missed, Opportunities Seized. Lanham, MD: Row-
man and Littleeld.
Meerts, P. & Cede, Franz, eds. (2004). Negotiating European Union. New York: Macmillan.
Mill, John Stuart, (1843/1967). A System of Logic: Ratiocinative and Inductive. Toronto: Uni-
versity of Toronto Press.
176 I. WILLIAM ZARTMAN
LINDA L. PUTNAM*
Clearly, I am the type of scholar who loves to muck around in the data. Qualitative
coding of communication processes allows me to get my ngers, hands, and
arms deep into the negotiation data. This mucking in the thick of things is the
key to discovering subtle nuances of not only what negotiators say but also
what they do not say. It provides opportunities to extend the functions of mes-
sages to their forms and structures, their symbolic features, and the broad
context of bargaining.
The coding of communication processes in negotiation encompasses a wide
array of qualitative and quantitative research methods. Qualitative methods
fall into the categories of textual and discourse analyses while quantitative
studies typically employ variations of content or interaction analyses (Wein-
gart, Olekalns and Smith 2004). In both instances, scholars examine language,
messages, talk, or written texts that emanate from a negotiation. Clearly, dis-
course units can be counted and converted to data amenable for quantitative
analysis. However, this particular paper focuses on qualitative methods that
are central to an in-depth examination of meaning and interpretations that
arise from the negotiation process. In this way, language and symbols are
keys to understanding how bargaining is constituted in a particular way, how
it is maintained or changed over time, and how participants make sense or
interpret this process (Putnam and Roloff 1992).
Research in negotiation typically centers on three different types of dis-
course analysis: conversational analysis, pragmatics, and rhetorical analyses.
To illustrate these methods and how they have been applied to negotiations,
I provide an overview of these approaches to discourse analysis. Then, I dis-
tinguish them from content and interactional analyses, methods that typically
employ quantitative coding. Finally, I focus on ways to conduct discourse
analyses, some strengths and weaknesses of this approach, and some insights
gleaned from personal experience in mucking around with the data. I draw
from negotiation research that employs discourse analysis to highlight the
strengths, difculties, and pitfalls of this approach.
A third type of discourse analysis draws from rhetoric or the study of per-
suasion and symbolic meaning (van Dijk 1997). Researchers who employ this
approach center their analyses on broad-based language patterns, such as
phrases that serve as evidence, claims, and reasons for arguments; metaphors
and symbolic language that invoke meanings in context; and patterns of talk
that construct narratives, such as references to plots, scenes, and story lines.
In studies of argumentation, discourse surfaces as persuasive attacks and defen-
sive reactions similar to bargaining strategies and tactics (Putnam and Jones
1992; Roloff, Tutzauer and Dailey 1989), as legitimating a bargainers claims
(Schuetz 1978), as ways of redening issues through policy deliberations (Putnam
and Wilson 1989), and as reason-giving for proposed changes in the status
quo (Putnam and Geist 1985; Putnam, Wilson and Turner 1990). Literary
analyses adopt a symbolic view of communication by focusing on the ways
words and phrases become short hand expressions for past discussions and
shared experiences. For example, Hamilton, (1997, 2000) research on talk
in pay negotiations between management and trade unions illustrates how
phrases such as stake in the organization and take out heads become
short-hand expressions that refer to a past sense of shared community and
current practices of company downsizing. Thus, researchers who focus on
gures of speech identify language patterns that have multiple and often con-
tradictory meanings within the larger bargaining context.
The nal type of rhetorical analysis stems from narrative or dramaturgical
texts. In narrative analyses, researchers focus on the way talk constructs sto-
ries, complete with villains and heroines, plot lines, time orientation, motives,
and values. Hence, negotiation surfaces as a ritualized performance in which
the parties develop narratives of their experiences within their bargaining
cultures (Friedman 1994). Narrative analysis reveals how administrators and
teachers build common ground through sharing stories of mutual enemies, and
yet hold very different understandings for what the bargaining ritual means
(Putnam, Van Hoeven and Bullis 1991).
These areas of discourse analysis represent only a sample of the many
ways that language forms a lens for understanding the complexities of nego-
tiation. The most informative analyses treat negotiation as more than simply
another type of conversation, but rather as a speech event in which the set-
ting, participants, norms, and history of the process inuence language
choices (Cicourel 1988) and the language choices, in turn, shape bargaining
activity.
Discourse analysis, then, is a qualitative approach to examining negotiation
texts. Researchers focus on the nuances of meaning within the broad context
of negotiation interaction and aim to untangle how the bargaining evolves as
180 LINDA L. PUTNAM
it does. Rather than rely on a priori categories to analyze texts, discourse ana-
lysts let the themes or patterns grow out of the transcripts as part of the
meanings that are co-developed during the negotiation. Researchers use lan-
guage patterns to demonstrate how bargaining concepts evolve and take shape
in the negotiation. For example, Putnam (2004) illustrates how symbolic uses
of the terms language and money shift in teacher-administrator negotia-
tions, moving from standing for sections of the contract itself to signaling a
bargaining formula for packaging items, and nally to establishing shared
control of the form and nature of the agreement (see Table 1). In this study,
shifts in the meanings of these words function as units of analysis to decipher
how negotiators come to a common understanding and how the bargaining
changes over time. Thus, qualitative analysis ts interpretative social science
that centers on how texts evoke multiple meanings. In this way it differs from
quantitative analyses that rely on unitizing a code, determining the frequency
of its use, and calculating reliability and validity of the codes. Instead,
discourse analysts seek to ascertain how language makes certain negotia-
tion practices possible, ones such as legitimating positions and identities,
managing role conicts with constituents, and reaching particular outcomes
(Donohue and Roberto 1994; Putnam, Van Hoeven and Turner 1991; Rogan
and Hammer 1995).
the corpus of texts to use in a study. For example, researchers could focus on
the interaction at the table or the talk in the caucus sessions (Donohue and
Roberto 1994; ODonnell 1990) or they could center on written proposals,
newspaper clippings, interviews, memos, or some combination of the above
that encompasses the context and culture of negotiation (Friedman 1992;
Putnam, Van Hoeven and Bullis 1991).
The choice among texts often hinges on the research questions that drive
the particular project. Usually, a researchers interests and philosophical posi-
tion guide the choice of texts and help a scholar work through the conundrum
of how broad or narrow, how inclusive or exclusive, and how ne grained or
contextual the study needs to be. A general guideline is that the researcher
should select a manageable subset of texts, make reference to the broader
discourses that impinge on these texts, and have sufcient data to justify
claims that emerge from analyses (Wood and Kroger 2000). For example, to
ascertain the way that trade negotiators rely on different types of strategies,
transcripts of the interaction might sufce, but to determine how cultural dif-
ferences and preferences for commodity trades guide these patterns, re-
searchers might need interview data and copies of trade policies that impinge
on the interactions (Grindsted 1989). Hence, beginning with an explicit re-
search question provides a frame for the study and aids in making decisions
about data collection and analysis.
Another factor that inuences choices about texts is the goal of the
researcher; that is, how descriptive, normative, or critical a researcher wants
to be. Often goals arise from the project itself, but researchers have their own
implicit biases as to the ultimate aim of their projects. Descriptive studies
focus on explaining and understanding how language use constitutes the nego-
tiation while normative studies center on ways to make the bargaining process
effective. Critical studies examine the practices that privilege or marginalize
certain players and the ways that power and control are managed through the
negotiation process (Hamilton 1997; Putnam 2004). In my experience, goals
typically emerge from a combination of the phenomenon under study,
the extant literature, and the puzzles that drive the project. In effect, most
researchers do not enter a discourse study with a blank template. They are
searching for patterns, insights, and new concepts for future work.
In summary, goals and research questions intertwine to shape the choices
that researchers make about the scope and depth of texts to use, the ne-
grained or broad-based nature of a study, and the type of discourse analysis
to employ. In the absence of one right way to proceed, researchers use
coherence as an overarching criterion, ask interesting questions, and make
their decisions as transparent and reasonable as possible. Researchers often under-
DISCOURSE ANALYSIS: MUCKING AROUND WITH NEGOTIATION DATA 183
particular ways. For instance, an analysis of words and phrases that signies
drawing closer together or distancing from each other (e.g., we versus
they, that versus this) shows how hostage negotiators protect the per-
petrators face, manage identity issues, and build trust with the hostage taker
(Rogan and Hammer 1994). The use of pronouns serves as the discourse units
and a comparison between the perpetrators use of these units and the hostage
negotiators response reveals the patterns or themes linked to distancing or
pulling together. The overall interpretations occur reexively by seeing how
these patterns evolve over time and how they interface with the context of the
negotiation, for example, type of hostage situation, training for the hostage
negotiator, length of the incident, and federal versus local level of the event.
Some discourse analyses are amenable to tracking through the use of com-
puter programs such as ATLAS and NVIVO. These programs consolidate dif-
ferent units across themes in a large data set and plot patterns pictorially.
Rather than consolidating data through statistical analyses, these programs
sort texts into patterns. However, they only help with sorting and organizing
the data, not with making interpretations or testing claims (Phillips and Hardy
2002).
Making interpretations from the themes and patterns that emerge in dis-
course studies requires sticking close to the data and then moving back and
forth between the broader context and the process. For example, a researcher
might discover that both sides in a labor-management negotiation make neg-
ative remarks about an accountant who handles nances for the organization.
Tracking these comments throughout the duration of the negotiation and in
the caucus meetings demonstrates that both parties view this person as a vil-
lain or as someone who makes bargaining difcult for all of them. Moreover,
when they mention this persons name at the table, it serves as a bonding
experience that leads them to sequential concessions and agreements on
difcult matters. A conclusion that could be drawn from these patterns is that
the accountant serves as a common enemy to unite the opposing sides.
This inference could then be veried through interview data with team
members about the role this person plays in the negotiations. Inferences, then,
are drawn from patterns of discourse, themes and content of talk, and are
related back to the larger context in which the parties work. They are then
tested through collecting additional data and comparing these ndings with
the extant literature. For example, Putnam (2004) draws inferences from shifts
in the meanings of the words language and money in a teachers negotiation
(see Table 1). This study reveals themes in the use of words that shift across
negotiation phases. These shifts show how both sides treat language and
money issues as commodities that become dualities to be traded in a win-lose
DISCOURSE ANALYSIS: MUCKING AROUND WITH NEGOTIATION DATA 185
fashion. Both parties were aware of this bargaining formula, lamented the
way it constrained them, but felt it was essential for an effective settlement.
Reecting back on the context and the bargaining relationship revealed a
duality of control in which board members used the formula to control the
length of the contract and teachers pressured the board for extra dollars in
raises and fringe benets. Both sides wanted to preserve their ways of con-
trolling the process while both felt highly constrained by their tacit norms and
bargaining formula.
Writing up a qualitative study is similar to, yet different from, quantitative
research. Consistent with quantitative studies, researchers make arguments
and provide evidence to support their claims. If they choose a social science
format, they follow the standard outline for preparing articles. Even with
qualitative discourse studies, a deductive model that lays out key themes
and conclusions to the study is easier to follow than inductive studies that
leave the interpretations to the end of the article (Wood and Kroger 2000).
Unlike quantitative coding, discourse studies need exemplars from the text to
illustrate and support claims about the patterns, themes, and conclusions of
the study. Scholars evaluate the quality of a study based on how well the evi-
dence supports the claims, the plausibility of the ndings, and the insights gleaned
from the analytic approach (Phillips and Hardy 2002). A studys overall
coherence and the way it rings true to the reader are also important crite-
ria for evaluating this research. For example, Friedmans (1995) observations
that negotiations are both constrained and enabled by the networks that bar-
gaining team members employ resonates both with the participants and with
other researchers. These insights not only challenge the assumptive ground of
dyadic negotiations, but they also parallel observations drawn from engaging
in actual bargaining experiences.
deliberations differed from past practices. Participants found rules and formu-
las to make the process work for them and in doing so became constrained
by their own norms for social interaction. Discourse analyses exposed the
rules and practices, exhibited how they functioned as constraints, and indi-
cated why they made it difcult for both sides to reach optimal settlements.
A second reason for using discourse analysis is that it provides a way to
link macro political, legal, and organizational processes to the micro behav-
iors in negotiation. These connections stem from the way that discourse
enacts the process while it refers back to the larger context in which the bargain-
ing occurs. Thus, a third strength of discourse analysis is its reexivity or
using language to reect back on the bargaining context, how a study was
conducted, and how processes evolve over time (Holland 1999). Reexivity
also addresses how discourse reveals instrumental, relational, and identity lev-
els of negotiation simultaneously. In this way discourse analyses demonstrate
how bargaining about issues and positions also includes negotiating ones
identity and the relationship among the parties, including ties to communities
and organizations.
The fourth reason to use discourse analysis is to uncover new concepts
and to extend the extant knowledge of negotiation in different directions.
Qualitative discourse processes can lead to the discovery of new concepts,
such as issue development, a concept that shows how reframing an issue
alters the way it is dened within the negotiation (Putnam and Holmer 1992).
This concept enriches the notion of a negotiation package by exhibiting how
splintering, combining, and dropping issues do more than create a bargaining
mix. Rather they lead to new understandings, ones that transform the way dis-
putants label a conict. Transformation of issues, in turn, opens up new
avenues for developing agreements through introducing topics created within
the negotiation. In effect, discourse analysis has the advantage of revealing
subtleties in the negotiation, providing methods of linking micro-behaviors to
macro-context, employing reexivity, and discovering new concepts to enrich
bargaining theory.
Discourse analysis also has its drawbacks and may be difcult to implement
in a number of settings. Researchers need texts, either transcripts of negotia-
tion interactions, documents, informal exchanges, interviews, eld notes, pro-
posals, or some form of verbal/written material to analyze. Its difculty as a
method stems in part from being labor intensive and from requiring trial and
error to select texts, decide on units, and link them to the larger negotiation
and social context. Researchers have to be willing to muck around in the data;
they have to follow a trail like a bloodhound or a detective to see where it
leads. To illustrate, a graduate student and I decided to perform a discourse
DISCOURSE ANALYSIS: MUCKING AROUND WITH NEGOTIATION DATA 187
study on the way that negotiators positioned themselves in the bargaining talk.
Relying on detailed methods of conversational analysis, we examined posi-
tioning patterns that revealed what we knew already about offensive and
defensive roles in labor-management negotiation. The lesson that I learned
from this experience was that discourse analysis is not just a method. As with
any research project, if you do not have an interesting problem, puzzle, or
dilemma, the study may lead nowhere. Thus, another shortcoming of this approach
is designing the study to address an important problem or puzzle, one that
arises from within the setting or from the extant negotiation literature.
Discourse analysis is often criticized for lacking rigor. It is a relatively new
and unproven method and not widely used in bargaining research. However,
as with most qualitative studies, rigor is a matter of precision, clarity, and sys-
tematic procedures. One way to conduct discourse analysis is to track patterns
through mapping language use in charts and spreadsheets. For example, in
studies of turning points in negotiations, Druckman and his colleagues (1991,
2001) tracked abrupt shifts in events through mapping internal and external
precipitants, subsequent departures, and the effects of these shifts on negoti-
ated outcomes. Through a systematic tracing of different paths, the investiga-
tors examined links between these negotiation shifts and conict escalation.
These detailed mappings enabled the researchers to uncover turning points, to
demonstrate how they evolved, and to distinguish between trade and security
negotiations.
One of the most difcult aspects of discourse analysis lies in drawing infer-
ences from the data. Researchers need to avoid inferences that are too gen-
eral or cannot be substantiated by detailed references to the text. For instance,
in our study of teachers bargaining, state law required some items to be
negotiated (e.g., salary, benets, reduction in force) and others (e.g., academic
freedom, teacher evaluation) were open to discussion but not negotiation
between teachers and administrators. However, since we did not collect data
at the state level, we could not make claims that this legal context inuenced
how issues were bargained in the two districts that we studied. We could
speculate about the effects of the law on the size and scope of issues included
in the contract, but we needed more information about how the law was inter-
preted throughout the state to make major claims about its effects.
A nal problem with discourse studies is keeping the research focused
on language analysis and not on a case study of the bargaining. In discourse
analysis, language forms the basis for the research and the context or se-
quence of events aids in interpreting these patterns. The discourse must be
analyzed through a particular method or approach. Discourse studies that get
lost in a play by play description of the event have turned into case studies.
188 LINDA L. PUTNAM
Thus, the study must focus on the way that patterns of discourse construct,
alter, and produce a negotiation. The bargaining events, the relationships
among players, the negotiation policies, and the history and cultural factors
are relevant only as they are produced and reproduced in the negotiation texts.
working back and forth from the case to concepts in the literature, we broad-
ened our focus to cover a complex communication system that merged well
with literature on back-channel negotiations. In like manner, Cobbs (1991)
study of neutrality as a discursive practice in mediation, illustrates how tying
particular language patterns to issues of legitimacy uncovered power moves
that mediators endorsed and used to marginalize some disputants and privi-
lege others. Similar to the rst guideline, working back and forth between the
text and the concepts keeps inferences about discourse patterns close to the
observations of the data.
A third guideline, look for inconsistencies and ironies in the text, focuses
on the discovery of puzzles and how researchers can engage in puzzle solv-
ing. Inconsistencies are contradictions that appear in the text and call for the
researcher to question the data. For example, illustrating how a particular pat-
tern of argument in negotiation both conceals and reveals information simul-
taneously leads the researcher to probe different questions about information
exchange (Putnam 1997). Friedmans (1992) discovery that intergroup bargaining
promotes exible stability aids in unpacking how negotiation groups both
perpetuate and counteract on-going struggles. These ironies lie at the roots of
new discoveries and provide puzzles for future studies.
A nal guideline for conducting discourse analysis is to dispute your own
interpretations. Qualitative data analysis involves discovering and assem-
bling pieces of a puzzle and drawing inferences to form a broad image. By
disputing interpretations of ndings, the researcher engages in a process of
eliminating explanations, particularly ones that are not plausible. Some
explanations do not ring true for the situation; others are too disconnected
from the data. Thus, discourse analysts need to make decisions based on
coherence among analytic schemes, evidence drawn from the texts, and inter-
pretations that resonate with the situation. In affect, several guidelines derived
from mucking around with the data aid in selecting discourse patterns by
letting the text talk to you, integrating patterns and negotiation concepts by
working back and forth from text to theory, discovering new concepts through
focusing on ironies and inconsistencies, and making appropriate claims
through a process of elimination among plausible interpretations.
Conclusion
Mucking around with the data, like making mud pies, is messy but fun. Just
as children create imaginative and often provocative images from this activ-
ity, researchers can uncover and construct original concepts from mucking around
190 LINDA L. PUTNAM
References
Boje, David (2001). Narrative Methods for Organizational & Communication Research.
London: Sage.
Cicourel, Aaron (1988). Text and Context: Cognitive, Linguistic, and Organizational Dimen-
sions of International Negotiations. Negotiation Journal, 4, 3: 257266.
Cobb, Sara and Rifkin, Janet (1991). Practice and Paradox: Deconstructing Neutrality in Mediation.
Law & Social Inquiry, 16: 3562.
Donohue, William A. and Diez, Mary (1985). Directive Use in Negotiation Interaction.
Communication Monographs, 52: 305318.
Donohue, William A. and Ramesh, Closepet (1992). Negotiator-Opponent Relationships, in
Linda L. Putnam and Michael E. Roloff, editors, Communication and Negotiation. Newbury
Park: Sage.
Donohue, William A., Ramesh, Closepet and Borchgrevink, Carl (1991). Crisis Bargaining:
Tracking Relational Paradox in Hostage Negotiation. International Journal of Conict
Management, 2, 4: 257274.
Donohue, William A. and Roberto Anthony (1994). Relational Development as Negotiated in
Hostage Negotiation. Human Communication Research, 20, 2: 175198.
Druckman, Daniel (2001). Turning Points in International Negotiations: A Comparative
Analysis. Journal of Conict Resolution, 45: 519544.
Druckman, Daniel, Husbands, Jo L. and Johnston, Karin (1991). Turning Points in the INF
Negotiations. Negotiation Journal, 7, 1: 5567.
Fant, Lars M. (1989). Cultural Mismatch in Conversation: Spanish and Scandinavian
Communicative Behavior in Negotiation Settings. Hermes, Journal of Linguistics, 3:
247265.
Firth, Alan (1994). Accounts in Negotiation Discourse: A Single-Case Analysis. Journal of
Pragmatics, 23: 199226.
Firth, Alan (1995). Talking for a Change: Commodity Negotiating by Telephone, in Alan
Firth, editor, The Discourse of Negotiation. Oxford: Pergamon.
Frances, David W. (1986). Some Structures of Negotiation Talk. Language in Society, 15:
5379.
DISCOURSE ANALYSIS: MUCKING AROUND WITH NEGOTIATION DATA 191
Putnam, Linda L. and Wilson, Steve R. (1989). Argumentation and Bargaining Strategies as
Discriminators of Integrative Outcomes, in M. Afzalur Rahim, editor, Managing Conict:
An Interdisciplinary Approach. New York: Praeger.
Putnam, Linda L., Wilson, Steve R. and Turner, Dudley B. (1990). The Evolution of Policy
Arguments in Teachers Bargaining. Argumentation 4: 129152.
Rogan, Randall G. and Hammer, Mitch R. (1994). Crisis Negotiations: A Preliminary Investiga-
tion of Facework in Naturalistic Conict Discourse. Journal of Applied Communication Research
22: 216231.
Rogan, Randall G. and Hammer, Mitch R. (1995). Assessing Message Affect in Crisis Negotiations:
An Exploratory Study. Human Communication Research 21: 553574.
Roloff, Michael E., Tutzauer, Frank E. and Dailey, William O. (1989). The Role of Argu-
mentation in Distributive and Integrative Bargaining Contexts: Seeking Relative Advantage
But at What Cost? in M. Afzalur Rahim, editor, Managing Conict: An Interdisciplinary
Approach. New York: Praeger.
Schuetz, Jan (1978). Argumentative Competence and the Negotiation of Henry Kissinger. Journal
of the American Forensic Association 15: 116.
Titscher, Stefan, Meyer, Michael, Wodak, Ruth and Vetter, Eva (2000). Methods of Text and
Discourse Analysis. London: Sage.
Van Dijk, Teun A. (1997). Discourse as Structure and Process: Volume 1. London: Sage.
Walker, Esther (1995). Making a Bid for Change: Formulations in Union/Management
Negotiations, in Alan Firth, editor, The Discourse of Negotiation. Oxford: Pergamon.
Weick, Karl E. (1979). The Social Psychology of Organizing. Reading, MA: Addison-Wesley
Publishing.
Weingart, Laurie R., Olekalns, Mara and Smith, Philip (2004). Quantitative Coding of
Negotiation Behavior. International Negotiation, 9: 441455.
Wodak, Ruth and Meyer, Michael (2001). Methods of Critical Discourse Analysis. London:
Sage.
Wood, Linda A. and Kroger, Rolf O. (2000). Doing Discourse Analysis: Methods for Studying
Action in Talk and Text. Thousand Oaks, CA: Sage.
Field Experiments on Social Conict
DEAN G. PRUITT
example, in an earlier study related to the one just described (Johnson and
Pruitt 1972), undergraduates negotiated about wage increases and hospitaliza-
tion benets in a simulated labor-management task. Three conditions were
imposed by instructions about what would happen if agreement were not
reached in 25 minutes. They were told in the rst condition that an arbitrator
would enter and make a binding decision; in the second condition that an
arbitrator would make a nonbinding decision; and in the third condition that
the negotiation would temporarily stop and then resume. The participants
made more concessions and were more likely to reach agreement under the
rst condition than the latter two.
In observational eld studies,2 the investigator measures all the variables
instead of manipulating some of them. For example, Kochan and Jick (1978)
conducted a study of archival records of mediated New York public employee
disputes, comparing outcomes before and after a med-arb procedure came into
law. The prior procedure was med-fact-nding, mediation followed by fact-
nding if mediation failed. This was an observational study rather than an
experiment because the investigator did not assign the cases to the conditions.
Rather assignment was done by the authorities who were following provisions
in the law.
The results of this study suggest that more cases were settled in mediation
under the med-arb law than under the med-fact-nding law. This nding is
consistent with the results of the two experiments described earlier, in that the
anticipation of binding arbitration produced conciliatory behavior, which was
presumably designed to escape the uncertainties of arbitration.
All three methods have their particular strengths and weaknesses. To under-
stand the arguments for and against eld experiments, we should rst exam-
ine the pros and cons of conducting experiments (in either the laboratory or
the eld) versus those of observational studies. Subsequently, we will consider
the respective pros and cons for doing eld and laboratory experiments.
often easier to sort out causation in such studies than in garden-variety obser-
vational studies. Still their name is a misnomer; they are not really experi-
ments because the investigator does not control the independent variable.
It is straightforward that experiments are not useful for studying all phenom-
ena. Some variables, such as the behavior of the heavenly bodies or of nation
states, cannot be manipulated. If we want to perform experiments on these
phenomena, we must employ simulations, which always run the risk of
198 DEAN G. PRUITT
differing in critical ways from the phenomena we wish to study. It is also not
possible to manipulate participant characteristics, such as sex, age, and race.
In addition, there are ethical objections to manipulating certain variables.
Even if we are able to create serious marital quarrels in a study of their
impact on the offspring, we should not try to do so. All of these variables
must be observed rather than measured.
Observational designs must also be used when one wants to look at the
interrelations among a large number of variables, since experiments are lim-
ited to manipulating only a few variables at a time. Surveys, in which dozens
or even hundreds of variables are measured, are a case in point. Hypotheses
about cause and effect among survey variables may be explored by means of
path analysis or causal modeling, but these methods seldom allow watertight
conclusions about causation. In addition, in-depth, qualitative studies of one
or a few cases can sometimes produce a rich theoretical tapestry concerning
the interrelations among a number of variables (Pruitt, in press). In such stud-
ies, causal sequences can be followed in still other ways, for example, by ask-
ing people why they behaved as they did. However, such methods only yield
speculative hypotheses, which should ultimately be tested by experimentation
(Cook and Payne 2002).
A third advantage of observational designs is that they allow us to estimate
the strength of relationships among naturally occurring variables. Experiments
are not so useful for this purpose, because good experimental design requires
augmenting the strength and potency of the manipulated variables and reduc-
ing variation in the measured variables. This allows us to reach statistical
signicance if our variables are related to each other, but unnaturally magnies
the strength of those relationships.
Both laboratory and eld settings are good places to observe new phenomena
and generate new hypotheses the laboratory because it isolates and mag-
nies relationships that might not be noticed in the eld, and the eld because
phenomena occur there that are unlikely to be programmed into the labora-
tory until they have been noticed. Thus, a mix of laboratory and eld research
is bound to be more heuristic than exclusive reliance on either approach.
Once new hypotheses have been developed, they should, if possible, be
tested experimentally to avoid confounding and establish cause and effect. But
should they be tested in the laboratory or the eld?
FIELD EXPERIMENTS ON SOCIAL CONFLICT 199
Control. Laboratory settings usually allow greater control over the elements
of research than do eld settings. This makes it easier to conduct good exper-
iments. For example, randomization is usually more possible and more secure
in laboratory settings. In many eld settings, random assignment to conditions
is not possible because society (e.g., mediation agencies) insists on having
this prerogative. We were lucky in the med-arb study to secure the cooperation
that allowed us to randomize. At the end of the study, however, the director
of the mediation center told us that she would never again allow a randomiza-
tion study because it was too disruptive for her staff.
Greater control also makes it possible to create more precise manipulations
in the laboratory, holding more variables constant between conditions and
thus reducing the number of confounds and alternative interpretations of the
results. Holding variables constant also reduces random error, making it eas-
ier to reach statistical signicance and hence to discover subtle effects with
smaller numbers of participants.
Mortality participants dropping out of the sample before the study is
done is also a larger problem in the less-controlled environment of eld
experiments, an issue we will return to toward the end of the article.
This kind of study was done by Cialdini and his colleagues (Cialdini,
Vincent, Lewis, Catalan, Wheeler and Darby 1975) to test the hypothesis that
concessions are reciprocated in everyday negotiation. Participants, who were
students walking alone on university walkways, were randomly assigned to
one of three conditions: a concession condition, in which they were asked a
large favor and, when they refused, asked a small favor; a no-concession con-
dition, in which they were only asked the small favor; and an exposure con-
dition, in which both favors were described and they were asked to make a
choice. The favors all had to do with volunteering for charitable activity. The
hypothesis was supported in that 50% agreed to do the small favor in the con-
cession condition, but only 17% agreed in the no-concession condition and
only 25% in the exposure condition.6 It was not possible to interview partic-
ipants about why they made the decisions they did because this would have
blown the investigators cover.
concern about the impression one is making on others will usually dimin-
ish as we move from the laboratory to the eld.
Our med-arb study (McGillicuddy et al., 1987) is an example of an effect
that was strengthened by moving into the eld. Before doing the eld exper-
iment, we ran some participants in a laboratory negotiation task involving the
prices of three products in a simulated wholesale market. A mediator, who
could become an arbitrator in the med-arb condition but not in the straight
mediation condition, helped them with the negotiation.7 This laboratory exper-
iment yielded only weak trends in the direction later taken by the results of
our eld experiment. In retrospect, it is quite certain that the issues faced in
the laboratory were less motivating and emotion producing than those faced
in the eld. Hence, the participants in the med-arb condition were probably
less anxious about the prospect of having the mediator take over their deci-
sion making, and hence less motivated to reach agreement.
A case in which a robust laboratory nding was weakened to the point of
disappearance can be seen in a pair of studies done by Druckman and Arai
(private communication). In a laboratory experiment, they found that media-
tion was less likely to lead to agreement if the disputants sat across a table
than if they faced each other across open air. But when they moved to a eld
setting and ran the same experiment in small claims court, this effect disap-
peared. This effect may have been partly drowned out by the greater error
variance that inevitable occurs in the eld. But it was probably also a victim
of the strong passions that were quite evident in the eld setting, where peo-
ple were dealing with frustrating and ego-involving issues from their own
lives. Heavy, emotion-laden involvement tends to block perception of back-
ground features of the environment. Hence, the conict and the adversary
probably stood out starkly while the rest of the scene including the pres-
ence or absence of a table probably receded into oblivion.
If I am right about the source of the difference between Druckman and
Arais two studies, it implies that their laboratory experiment was not a fail-
ure. The results of that study suggest that the absence of a table facilitates
agreement in low key, non-emotional conicts a subclass of all conicts.
What is needed next is a study that crosses presence or absence of a table
with high vs. low emotional involvement, to test the hypothesis that putting
a table between the disputants makes a difference when passions are weak but
not when they are strong.
Mortality
Why worry about this issue? Mortality poses a threat to random assignment
if participants drop out of one condition for different reasons than out of
another a phenomenon called differential mortality. For example, if the
med-arb conditions in our study had seemed more threatening than straight
mediation, braver people might have stayed in the former than the latter con-
dition, which could have biased the results.8 Mortality also poses a threat to
external validity if certain kinds of people drop out of all conditions, because
the results may only apply to the type of people who stay in the study
(Aronson et al., 1990).
A major problem of differential mortality was narrowly averted in our med-
arb study. Ofcers of the court became alarmed when some of the cases
assigned to straight mediation did not settle their disputes and came back to
the court. (This could not happen in the med-arb conditions, because the arbi-
trator always imposed a settlement.) They complained to the mediation cen-
ter employee who was doing the random assignment, and she changed her
procedure without notifying us. As each new case came into her ofce, she
was supposed to open the next envelope in a loose-leaf binder and read out
the instructions contained in that envelope. Instead, she began to open several
envelopes ahead of time and assign the easier cases to straight mediation so
that they would not go back to the court. We detected this problem during
one of our weekly inspections of the binder, because we found some folders
already opened. To avoid differential mortality, we were forced to drop from
our sample all of the cases that had gone through her ofce between the time
at which she said she had started this practice and the time we detected it.
If there is substantial mortality in an experiment, it is necessary to check
whether signicantly more participants have dropped out of one condition
than another, a sure sign of differential mortality. Mortality was a big prob-
lem in our med-arb study; indeed so many cases were dropped or withdrew
that we had to process 114 cases to get 12 cases in each of our three condi-
tions. Fortunately, the dropout rate was not signicantly greater in one condi-
tion than another. This was a comforting nding but it by no means ruled out
differential mortality.
If there is substantial mortality, it is also important to check whether dif-
ferent kinds of cases have dropped out of one condition than another, using
whatever information one has on the dropouts. Differential mortality exists if
the dropouts differ between conditions on any variable. It is a problem if that
variable is correlated with the dependent variable in a direction that could
produce ones ndings. An example where it was not a problem can be seen
in the Emery et al. (2001) twelve-year follow-up of divorce mediation. They
found evidence of differential mortality in that fathers from heavily conict-
laden relationships were less likely to drop out of the mediation sample than
FIELD EXPERIMENTS ON SOCIAL CONFLICT 205
the litigation sample. However, this did not endanger their conclusions
because conict should produce less contact with the children whereas they
found more contact after mediation than after litigation.
Such post-hoc analyses are worth doing in studies that involve heavy mor-
tality, but they are seldom denitive. Even if ones ndings are exonerated by
all the variables one has measured, it is still possible that differential mortal-
ity is a problem on some other unmeasured variable. Hence, it is always bet-
ter to conduct an experiment in which there is little or no mortality. Boruch
(1997) provides much useful advice about how to accomplish this goal (see
especially Chapter 6).
All experiments, whether run in the laboratory or the eld, require permission
from a human participants review board. But many eld experiments require
additional approval from the agency that supplies the cases. For example, in
the med-arb study, we had to persuade the mediation center to allow us access
to their cases and the right to randomize assignment of cases to conditions.
Center personnel also had to vouch for our study with the court from which
the cases were drawn; and, halfway through the study, they had to clamp
down on the employee who was disturbing our randomization.
To get agency cooperation especially at the high level we obtained
requires a courtship process whereby one does favors for the agency and/or
promises it something of value in exchange. Druckman and Arai were able to
get agency cooperation by promising that the agency would learn something
about its own procedures and make a contribution to science; but a larger
quid pro quo seemed necessary in our case. Before we proposed the med-arb
study, we evaluated one of the agencys programs for its annual report, and
we helped recruit mediators on our campus. Since the agency was making
heavy use of med-arb, we were also able to sell our study as an evaluation
of that procedure.9 Furthermore, we detected that sponsorship of our research
made the agency look progressive in the eyes of its sponsors and the broader
mediation profession. Hence, we made presentations of our research at the
yearly conferences of the New York state mediation community, often asking
the agency director to make preliminary remarks. Our approach to courtship
was essentially guided by a principle recently articulated by Gueron (2002),
Remember that you want them more than they want you (35).
It is especially hard to get agencies to agree to random assignment because
that procedure curtails some of an agencys freedom of action with its clients
and can provide an administrative burden to the agency. Because of this
difculty, Campbell and his associates (Campbell and Stanley 1963; Cook and
206 DEAN G. PRUITT
Condentiality
Disputants often want to keep the details of their conicts secret for fear
of embarrassment, benet to an adversary, or legal liability. Hence, eld
experiments on social conict require particular attention to condentiality.
Concerns about condentiality usually make it impossible to tape record
negotiation and mediation sessions. However, it is sometimes possible to do
on-site coding of negotiator or mediator behavior, using coding categories that
reect the process of what is happening but reveal little or nothing about the
content of the sessions. This approach, which was taken in our med-arb study
and the Druckman-Arai study, makes it difcult to do inter-coder reliability
checks and impossible to go back to the data with new coding categories.
In a later observational study (Pruitt, Peirce, McGillicuddy, Welton and
Castrianno 1993; Zubek, Pruitt, McGillicuddy, Peirce and Syna 1992), we
solved these problems by getting permission to take verbatim notes of what
was said in mediation. We argued successfully that these notes were tanta-
mount to hearsay and hence could not be admitted to a court of law. We also
put case numbers instead of names on these notes and other records; and
when we had assembled all the data for our cases, we destroyed the codebook
that linked these numbers to the participants names.
Conclusions
Writing this article has led me to conclude that eld experiments should be
employed more often then they are in the study of social conict. They are
more useful than observational studies for assessing the impact of novel con-
ditions, establishing cause and effect, and reducing confounding. And they are
more useful than laboratory experiments for examining long-term effects and
discovering externally valid effects that involve strong passions and complex
social interactions.
However, eld experiments clearly have their limitations. Some variables
cannot be practically or ethically manipulated and require the use of observa-
tional methods. Observational methods are also more useful for looking at the
FIELD EXPERIMENTS ON SOCIAL CONFLICT 207
Notes
1. Med-arb (diff) was intermediate between the other two conditions on these measures.
2. Sometimes called correlational eld studies.
3. The reader will recall that in the just prior section we said that the most plausible interpre-
tation of the Kochan and Jick nding is that the change in the law or some associated
change made agreement more likely. The confounded variable, time period, is an example
of an associated change.
4. To denitively pinpoint a causal variable, one must run several studies in each of which that
variable is confounded with a different set of alternative variables.
5. Randomization also removes confounds with other features of the situation such as the time
of day people enter the study, the room they are run in, the identity of the researcher, etc.
6. Nobody agreed to do the large favor.
7. The med-arb condition was quite similar to the med-arb (same) condition in the later eld
experiment. The straight mediation conditions were identical in the two experiments.
8. Differential mortality has an effect similar to self-selection, a topic discussed earlier. Both
produce confounds by placing different kinds of people in the conditions of the study.
9. The agency director was very pleased with the way the results turned out.
References
Aronson, E., Ellsworth, P.C., Carlsmith, J.M. and Gonzales, M.H. (1990). Methods of Research
in Social Psychology, 2nd edition. New York: McGraw-Hill.
Anderson, C.A. and Bushman, B.J. (1997). External validity of trivial experiments: The case
of laboratory aggression, Review of General Psychology, 1:1941.
Barry, B., Fulmer, I.S. and Van Kleef, G.A. (in press). I laughed, I cried, I settled: The role
208 DEAN G. PRUITT
of emotion in negotiation, in M.J. Gelfand and J.M. Brett, editors, Culture and Negotiation:
Integrative Approaches to Theory and Research.
Berkowitz, L. (1993). Aggression: Its Causes, Consequences, and Control. New York:
McGraw-Hill.
Boruch, R.F. (1997). Randomized Experiments for Planning and Evaluation. Thousand Oaks,
CA: Sage.
Campbell, D.T. and Stanley, J.C. (1963). Experimental and Quasi-Experimental Designs for
Research. Chicago: Rand McNally.
Cialdini, R.B., Vincent, J.E., Lewis, S.K., Catalan, J., Wheeler, D. and Darby, B.L. (1975).
Reciprocal concessions procedure for inducing compliance: The door-in-the-face technique.
Journal of Personality and Social Psychology, 31:206213.
Cook, T.D. and Campbell, D.T. (1979). Quasi-Experimentation: Design and Analysis Issues for
Field Settings. Chicago: Rand McNally.
Cook, T.D. and Payne, M.R. (2002) Objecting to the objections to using random assignment
in educational research, in F. Mosteller and R. Boruch, editors, Evidence Matters:
Randomized Trials in Education Research: 150178. Washington, DC: Brookings Institution
Press.
Emery, R.E., Laumann-Billings, L., Waldron, M.C., Sbarra, D.A. and Dillon, P. (2001). Child
custody mediation and litigation: Custody, contact, and coparenting 12 years after initial
dispute resolution. Journal of Consulting and Clinical Psychology, 69:323332.
Emery, R.E., Matthews, S.G. and Wyer, M.M. (1991). Child custody mediation and litigation:
Further evidence on the differing views of mothers and fathers. Journal of Consulting and
Clinical Psychology, 59:410418.
Geen, R.G. (2001). Human Aggression, 2nd edition. Buckingham, England: Open University
Press.
Gueron, J.M. (2002). The politics of random assignment: Implementing studies and affecting
policy, in F. Mosteller and R. Boruch, editors, Evidence Matters: Randomized Trials in Education
Research: 1549. Washington, DC: Brookings Institution Press.
Hersen, M. and Barlow, D.H. (1976). Single Case Experimental Designs. New York: Pergamon.
Johnson, D.F. and Pruitt, D.G. (1972). Pre-intervention effects of mediation vs. arbitration.
Journal of Applied Psychology, 56:110.
Kochan, T.A. and Jick, T. (1978). The public sector mediation process: A theory and
empirical examination. Journal of Conict Resolution, 22:209240.
McEwen, C.A. and Maiman, R.J. (1984). Mediation in small claims court: Achieving
compliance through consent. Law and Society Review, 18:1149.
McGillicuddy, N.B., Welton, G.L. and Pruitt, D.G. (1987). Third party intervention: A eld
experiment comparing three different models. Journal of Personality and Social Psychology,
53:104112.
Mikolic, J.M., Parker, J.C. and Pruitt, D.G. (1997). Escalation in response to persistent annoy-
ance: Groups vs. individuals and gender effects, Journal of Personality and Social
Psychology, 72:151163.
Mosteller, F. and Boruch, R., editors (2002). Evidence Matters: Randomized Trials in Education
Research. Washington, DC: Brookings Institution Press.
Nisbett, R.E. and Cohen, D. (1996). Culture of honor: The psychology of violence in the South.
Boulder, CO: Westview.
Pruitt, D.G. (in press). Escalation, readiness for negotiation, and third party functions in I.W.
FIELD EXPERIMENTS ON SOCIAL CONFLICT 209
Zartman and G.O. Faure, editors, Escalation and Negotiation. Cambridge, England: Cam-
bridge University Press.
Pruitt, D.G., Peirce, R.S., McGillicuddy, N.B., Welton, G.L. and Castrianno, L.B. (1993).
Long-term success in mediation. Law and Human Behavior, 17:313330.
Robinson, R.J., Keltner, D., Ward, A. and Ross, L. (1995). Actual versus assumed differences
in construal: Nave realism in intergroup perception and conict, Journal of Personality
and Social Psychology, 68:404414.
Sherman, L.W. and Berk, R.A. (1984). The specic deterrent effects of arrest for domestic
assault. American Sociological Review, 49:261272.
Vallone, R.P., Ross, L. and Lepper, M.R. (1985). The hostile media phenomenon: Biased per-
ception and perceptions of media bias in coverage of the Beirut massacre. Journal of Personality
and Social Psychology, 49:577585.
Zubek, J.M., Pruitt, D.G., McGillicuddy, N.B., Peirce, R.S. and Syna, H. (1992). Short-term
success in mediation: Its relationship to prior conditions and mediator and disputant behav-
iors. Journal of Conict Resolution, 36:546572.
Laboratory Experiments on Negotiation and
Social Conict
This article addresses the laboratory experiment from the perspective of the
experimental social psychologist interested in making statements about social
conict and negotiation. The main concern is the validity of these statements
whether the conclusions or inferences about the data are justied (Cook &
Campbell 1979; see also Borsboom, Mellenbergh, & van Heerden 2004). Note
that it is not that a particular method or study is valid, per se; rather, it is the
statements made about the data the propositions that have validity or not
(Brewer 2000). Such statements are best judged in light of a studys purpose.
One purpose might be to demonstrate an effect or relationship between
variables. For example, Thompson and Hastie (1990, Study 2) wanted to
demonstrate that when people face many issues of opposed interests, issues
of common interest go unseen (this happened in 85% of the negotiations!).
OConnor and Carnevale (1997) wanted to demonstrate that such an issue of
common interest would be exploited by people to gain advantage in an agree-
ment (in one situation, such exploitation occurred in 53% of the negotia-
tions!). Both studies were conducted by employing naive university students
in conducting an articial task in an articial setting. One wonders if these
patterns would occur in other settings, with other kinds of people, other peri-
ods of history, other cultures, or even with other levels of task incentive (see
Hertwig & Ortmann 2001). Such issues of generality are always present in
laboratory research or any research for that matter; indeed, any question about
boundary conditions is a question about underlying process. I observe X here;
over there, I do not observe X; since process Y does not operate over there,
maybe that is why I do not observe X over there. In other words, I believe
that process Y is critical for X to occur. Any situation without Y means no
X. The nice thing about psychological processes is that they occur not only
in the experimental psychologists laboratory, but they occur in the real
world as well (Mook 1983).
Background
thesizing after the results are known is not a problem. McGrath (1982) pro-
vides great wisdom on the many choices facing the researcher. Aronson and
Carlsmiths (1968) review of experimental method inspired many young social
psychologists when they noted, in the last line of the paper, that the enter-
prise is fun.
Many social psychology experiments involving negotiation and social
conict arise from observations gleaned in natural settings. Triplets (1897)
foundational laboratory experiment designed a task investigating the impact of
others presence on competition and performance. Having observed bicycle
racing, Triplet modeled the task to simulate performance on a bicycle race
track. Pruitt and Johnsons (1970) study of the impact mediators suggestions
have on negotiator concession-making built off Stevens (1963) real-world
mediation observations.
shown at the top of Table 2, and the union negotiator is assigned the issue
chart displayed at bottom. These charts list ve possible settlement points on
each issue. Again, each side sees only their own chart, and is typically
instructed that they can say anything to one another about it.
In Table 2, the issues are integrative in the sense that they can be traded
for one another. That is, each side can get what they want on their most val-
ued issue and gives in on their least valued issue. In this task, many possible
agreements exist and they vary in their value to the negotiators, both indi-
vidually and collectively. The two issues in the center offer equal value for
both parties and, in the parlance of negotiation research, are distributive
issues, following the unfortunate terminology used by Walton and McKersie
(1965) (unfortunate since the term distributive has other meanings in the expan-
sive social psychology of justice literature). They noted that distributive bar-
gaining is the process by which each party attempts to maximize his/her own
share in the context of xed-sum payoffs. Integrative bargaining is the process
by which the parties attempt to increase the size of joint gain without respect
to the division of the payoffs (8). Joint gain in the Table 2 task settles upon
70k on salary and 20% on medical. Such an outcome results in greater
value to each individual and the pair than an agreement that splits these
two issues down the middle ( joint gain of 800 versus 460 on these two
issues).
Table 3 presents a rough taxonomy of laboratory tasks; it indicates some
areas where tasks simply do not exist (see Bornstein 2003, for taxonomy of
intergroup tasks that also identify gaps). An organizing feature of the taxon-
omy is whether the task models single or multiple dimensions of value. The
former include one-issue negotiation tasks (as shown in Table 1); the latter
includes integrative bargaining tasks and their variations (as shown in Table 2,
including contingent agreement games and the integrative ultimatum game).
Another feature of the taxonomy concerns whether the task represents a game
of agreement or one of coordination. A game of agreement allows the behav-
ioral act of agreeing, reaching accord or developing consensus. Consider the
ultimatum game, where a proposer offers an exchange and the responder
either accepts or rejects it, and the game ends (Gth, Schmittberger, &
Schwarze 1982). In this game, non-agreement is allowed, as determined by
one or more participant saying no, withdrawing, or reaching a deadline that
terminates negotiation; no agreement implies that the status quo is maintained.
Any agreement usually represents a change in the status quo. Games of co-
ordination model individual decisions to distribute outcomes, usually over
time. These include, for example, the prisoners-dilemma and trust games.
Coordination games typically generate simple response variables (C or D
216 PETER J. CARNEVALE AND CARSTEN K.W. DE DREU
Table 1. The Singe Issue Negotiation: Issue Charts for Management and Union Negotiators
Management Issue Chart
Annual Raise (value)
15% (00)
12% (60)
9% (120)
6% (180)
3% (240)
Union Issue Chart
Annual Raise (value)
15% (240)
12% (180)
9% (120)
6% (60)
3% (00)
Note: All agreements sum to a constant for the dyad, 240, making the task xed-sum.
in a matrix choice; see Pruitt 1970). The notion of disagreement is not clearly
represented, and coordination games typically do not model more than one
dimension of value.
In negotiation experiments, the main measurement is the outcome, which
might be whether an agreement is reached or the quality of agreement, includ-
ing its joint value, degree of efciency, or its symmetry (the extent to which
it favors one side or the other). A useful discussion of coding negotiation
agreements can be found in Weingart, Hyder, and Prietula (1996). In more
complex simulations, the primary measure may be an escalatory sequence of
behaviors that reect ever-increasing levels of conict (see Mikolic, Parker, &
Pruitt 1997).
A laboratory experiment should be appreciated not only for its message, for
what it says about theory, but for the techniques used to produce the message.
This point became clear to this articles rst author while he was a graduate
student working in Dean Pruitts laboratory in Buffalo, New York in the early
LABORATORY EXPERIMENTS ON NEGOTIATION AND SOCIAL CONFLICT 217
Table 2. Issue Charts for Management (Top) and Union (Bottom) Negotiators
Management Issue Chart
1980s. One day, Pruitt walked into the project room with a sheet of paper in
hand and a gleam in his eye and announced that he had something and
it was something special about experimental design. The topic of the past
months had been the concept of accountability and how it might play out in
bilateral negotiation. We dened accountability in terms of social power and
as an aspect of the relationship between negotiator and constituent. One idea
posited that a mediators accountability to constituents might interfere with
the development of a trusting relationship between the opposing negotiators
and thereby interfere with problem solving and the development of integra-
tive agreements. Pruitt developed an idea for testing that hypothesis, and the
paper in his hand depicted a 2 2 experimental design, alongside some
path notation. The design employed a moderator variable to help make infer-
ences about mediating processes. The logic was straightforward: if the differ-
ence between high and low accountability was due to process X (in this case,
a trusting relationship between negotiators), then controlling X should mod-
erate the accountability effect. It soon became apparent that Pruitts design
was elegant and structurally reected Platts notion of strong inference (Platt
1964). To illustrate this type of design and others, what follows is a brief dis-
cussion of experimental design as it was developed, employed, and taught in
the Pruitt laboratory.
Path-diagram notation. Path-diagram notation is useful for presenting
causal propositions schematically. The basic building blocks of propositions
are variables and the links postulated between them. Variables are represented
in path diagrams by letters or words in open boxes, and causal links are rep-
resented by arrows pointing in the direction of assumed causation. Simple
path diagrams can be linked in chains. For example, if we know that A has
a direct effect on B and B has in inverse effect on C, we can write:
A B C
+
Other combinations of signs are also possible, + +, +, , and chains con-
sisting of four or more variables are possible. In such a chain, A is often
called the independent variable, B the mediating variable, and C the depen-
dent variable. If there are more than three variables in a chain, then there will
be more than one mediating variable. Pruitt pointed out that such chains
always permit theoretical derivations. Above, the relationship between A and
B and the relationship between B and C imply a relationship between A and
C, namely that increases in A lead to decreases in C. For example, if we
believe that frustration (A) leads to anger (B) and that anger (B) reduces lik-
LABORATORY EXPERIMENTS ON NEGOTIATION AND SOCIAL CONFLICT 219
ing (C), then we might conclude that frustration reduces liking. The concepts
in propositions are stated as variables rather than as conditions or states of
nature. Thus the parameter A might be time pressure, a variable ranging
from low to high (in other words, the A parameter would not be written as
high time pressure). More complex forms of path diagrams express the joint
effect of two variables on a third, for instance, additive and interactive effects.
Several basic experimental designs help assess the processes underlying a
research nding. These designs complement various statistical procedures
(e.g., Baron & Kenny 1986): conceptual replication, precise manipulation,
and the use of a moderator variable. Each of these fosters conceptual
internal validity, concerning the impact of one variable on another and the
quality of explanation. Conceptual internal validity provides a basis for
generalizing ndings and thus new research. It also fosters strong inference
(Platt 1964), a way of building cumulative knowledge.
Conceptual replication. We are on much stronger theoretical ground if
we manipulate variables in more than one way. Often this requires multiple
studies that each tests the variable differently. An example is negotiator
positive mood, demonstrated by Carnevale and Isen (1986) to produce more
cooperative negotiation and more integrative agreements; positive mood was
implemented via a small gift and reading cartoons just prior to negotiation.
Forgas (1998) also implemented positive mood and obtained a similar effect.
Since his procedure differed from that of Carnevale and Isen he used posi-
tive feedback on a test of verbal abilities just prior to negotiation we can
be more certain about the effect. The Forgas study was not just a replication
in part (he also examined negative mood), but it was a conceptual replication
for positive mood that is, important for ruling out alternative explanations
of the Carnevale and Isen results due to possible artifacts or confounds asso-
ciated with the gift/cartoon manipulation of positive mood.
In path diagram notation, the Carnevale and Isen study might be repre-
sented as A leads to B leads to C, where A is receiving a gift, B
is positive mood, and C is cooperation. However, B could be com-
posed of B1 and B2, where B1 is sense of obligation having received a gift,
offering an alternative explanation to positive mood. The Forgas study rules
that out as a possible explanation since it is hard to conceive of such oblig-
ation arising from receiving a positive test score. Of course, some unknown
common process (apart from mood) could be important in interpreting the
Carnevale and Isen mood procedure and the Forgas mood procedure, and this
might be an interesting basis of a future study.
Precise manipulation. Another approach is to try to rene the experimen-
tal manipulation. This is often done by including one or more control groups
220 PETER J. CARNEVALE AND CARSTEN K.W. DE DREU
time pressure affected perceived common ground, not concern. When per-
ceived common ground was free to vary, time pressure had an effect; but
when it was not free to vary, there was no relationship between time pressure
and the use of forceful mediation tactics. In the spirit of strong inference,
future studies should pursue perceived common ground, and not concern, as
the critical process that explains the impact of time pressure on mediator
behavior. This, of course, does not mean that mediator concern is unimpor-
tant, but rather that it is likely less valuable in understanding the impact of
time pressure on mediation.
Coda
Like a great work of art, a laboratory experiment can be appreciated not only
for its message, that is, what it says about theory, but for the technique used
to produce the message. The latter is especially relevant to its truthfulness and
its validity. Indeed, the latter is especially relevant to its beauty (see Overbye
2002). If we ask, What is a good experiment? the answer will often depend
on the purposes of the experiment (see Brewer 2000; Dawes 1996). If the
purpose is explanation, then the more the experiment controls for and rules
out alternative explanations, the better. And if the experiment helps with strong
inference, which some argue is the basis for cumulative knowledge in science,
perhaps this is best, although perhaps unrealistic (see Hafner & Presswood 1965).
Of course, the character of explanation, whether it is a neural mechanism,
a higher order mental process, the structure of an incentive, a property of
a group decision process, or a cultural process, will vary as a function of the
tastes of the researcher, and is very much a matter of the zeitgeist.
A key feature of laboratory experimentation is that it is programmatic. That
is, experiments are conducted in a series, with each experiment in the series
building on a prior experiment. This allows the researcher to narrow the set
of theoretical possibilities that might explain an effect. The main benet of
conducting a series of experiments, over a single experiment, is that we learn
far more. This is seen even when researchers with very different views agree to
collaborate on a followup study to resolve a theoretical controversy (e.g.,
Latham, Erez, & Locke, 1988). This occurred also in a program of research on
mediation of disputes (Carnevale, 1986, 1992; Murnighan, 1986; see Conlon,
Carnevale, & Murnighan, 1994). This helps strong inference (Platt 1964).
There are numerous other benets; for example, they allow us to test the-
ories in a safe place, and we can develop tests that are less expensive than in
natural settings. In experiments, we can allow behavior that approximates
222 PETER J. CARNEVALE AND CARSTEN K.W. DE DREU
intense behavior in natural settings, but host it in a relatively safe place where
it is less likely to produce harm. The wind tunnel in aviation research offers
a great analogy: differently-shaped wings can be built and own, but wind
tunnel testing saves lives and cost. Of course, the wind tunnel needs to rep-
resent processes that occur in nature or at least in that part of nature where a
plane with that wing will y. Another benet is that laboratory designs allow
measurement and tracking of processes that would be impossible or at least
extremely difcult to get at in natural settings, for example, measures of facial
electromyograph in conditions of cooperation and competition obtained in the
clever experiment by Lanzetta and Englis (1989) or fMRI data obtained by
Sanfey, Rilling, Aronson, Nystrom, and Cohen (2003) in the ultimatum game.
The main shortcoming of laboratory research is that it does not reveal the
relative importance of different variables as they affect negotiation. It only
tells what effects can occur (Mook 1983; and Locke 1986). Moreover, it is
often difcult to generalize results from laboratory settings to natural settings,
a problem shared with eld research since natural settings differ from one another.
There are two solutions: do parallel research in both laboratory and natural
settings, and conduct research derived from theory. With theory-based
research, the results may be generalized because the theoretical processes
occur in the laboratory as well as in natural settings.
References
Abelson, R.P. (1995). Statistics as principled argument. Hillsdale, NJ: Erlbaum.
Aronson, E. & Carlsmith, J.M. (1968). Experimentation in social psychology, in G. Lindzey
& E. Aronson (Editors), The handbook of social psychology (2nd ed.), Reading, MA:
Addison-Wesley: 179.
Aronson, E., Ellsworth, P.C., Carlsmith, J.M., & Gonzales, M.N. (1990). Methods of research
in social psychology. 2nd Edition. Reading, MA: Addison-Wesley.
Baron, R.M. & Kenny, D.A. (1986). The moderator-mediator variable distinction in social psy-
chological research: Conceptual, strategic, and statistical considerations. Journal of
Personality and Social Psychology, 51: 11731182.
Bem, D. (2002). Writing the empirical journal article. in Darley, J.M., Zanna, M.P., &
Roediger III, H.L. (Editors.), The complete academic: A career guide, Washington, DC: American
Psychological Association: 171201.
Bornstein, G. (2003). Intergroup conict: Individual, group, and collective interests. Per-
sonality and Social Psychology Review, 7: 129145.
Borsboom, D., Mellenbergh, G.J., & van Heerden, J. (2004). The concept of validity. Psychological
Review, 111: 10611071.
Brewer, M.B. (2000). in H.T. Reis & C.M. Judd (Editors) Handbook of research methods in
social and personality psychology. Cambridge, UK: Cambridge University Press.
Burnham, T., McCabe, K., & Smith, V.L. (2000). Friend-or-foe intentionality priming in an
extensive form trust game. Journal of Economic Behavior & Organization, 43: 5773.
LABORATORY EXPERIMENTS ON NEGOTIATION AND SOCIAL CONFLICT 223
Campbell, D.T., & Stanley, J.C. (1963). Experimental and quasi-experimental designs for
research. Chicago: Rand McNally.
Carnevale, P.J. (1986). An unnecessary neologism in two systems of mediation. Negotiation
Journal, 2: 357361. [reply to J.K. Murnighan (1986), The structure of mediation and intra-
vention: Comments on Carnevales strategic choice model. Negotiation Journal, 2:
351356.]
Carnevale, P.J. (1992). The usefulness of mediation theory. Negotiation Journal, 8: 387390.
[reply to E. van de Vliert (1992), Questions about the strategic choice theory of mediation.
Negotiation Journal, 8: 379386.]
Carnevale, P.J., & Conlon, D. (1988). Time pressure and strategic choice in mediation.
Organizational Behavior and Human Decision Processes, 42: 111133.
Carnevale, P.J., & Isen, A.M. (1986). The inuence of positive affect and visual access on the
discovery of integrative solutions in bilateral negotiation. Organizational Behavior and
Human Decision Processes, 37: 113.
Cha (2005). All that glitters is not gold: Examining the perils and obstacles in of collecting
data on the internet. International Negotiation.
Conlon, D.E., Carnevale, P.J., & Murnighan, K. (1994). Intravention: Third-party intervention
with clout. Organizational Behavior and Human Decision Processes, 57: 387410.
Cook, T.D., & Campbell, D.T. (1979). Quasi-experimentation: Design and analysis issues for
eld settings. Chicago: Rand McNally.
Dawes, R.M. (1996) The purpose of experiments: Ecological validity versus comparing
hypotheses. Behavioral and Brain Sciences, 19: 20.
Froman, L.A., & Cohen, M.D. (1970). Compromise and logroll: Comparing the efciency of
two bargaining processes. Behavioral Science, 30: 180183.
Forgas, J.P. (1998). On feeling good and getting your way: Mood effects on negotiator cog-
nition and behavior. Journal of Personality and Social Psychology 74: 565577.
Gth, W., Schmittberger, R. & Schwarze, B. (1982). An experimental analysis of ultimatum
bargaining. Journal of Economic Behavior and Organization 3: 367388.
Hafner, E.M., & Presswood, S. (1965). Strong inference and weak interactions. Science, 149:
503510.
Hertwig, H., & Ortmann, A (2001). Experimental practices in economics: A methodological
challenge for psychologists? Behavioral and Brain Sciences, 24: 383451.
Kelley, H.H. (1966). A classroom study of the dilemmas in interpersonal negotiations. In
K. Archibald. in Strategic interaction and conict: Original papers and discussion.
Berkeley, CA: Institute of International Studies.
Kelley, H.H., Beckman, L.L., & Fischer, C.S. (1967). Negotiating the division of reward under
incomplete information. Journal of Experimental Social Psychology, 3: 36198.
Kerlinger, F.N. (1986). Foundations of behavioral research. New York: Holt, Rinehart and Winston.
Kerr, N. (1998). Harking: hypothesizing after the results are known. Personality and Social
Psychology Review, 2: 196217.
Latham, G.P., Erez, M., & Locke, E.A. (1988). Resolving scientic disputes by the joint
design of crucial experiments by the antagonists: Application to the Erez-Latham dispute
regarding participation in goal setting. Journal of Applied Psychology, 73, 4: 753772.
Lanzetta, J.T., & Englis, B.G. (1989). Expectations of cooperation and competition and their
effects on observers vicarious emotional responses. Journal of Personality and Social
Psychology, 56: 543554.
Liebert, R.M., Smith, W.P., Hill, J.H., & Keiffer, M. (1968). The effects of information
224 PETER J. CARNEVALE AND CARSTEN K.W. DE DREU
Thompson, L.L. and Hastie, R. (1990). Social perception in negotiation. Organizational Behavior
and Human Decision Processes, 47: 98123.
Triplett, N. (1897). The dynamogenic factors in pacemaking and competition. American
Journal of Psychology, 9: 507533.
Walton, R.E., & McKersie, R. (1965). A behavioral theory of labor negotiations: An analysis
of a social interaction system. New York: McGraw-Hill.
Weingart, L.R., Hyder, E., & Prietula, M.J. (1996). Knowledge matters: The effect of tactical
descriptions on negotiation behavior and outcome. Journal of Personality and Social Psychology,
70: 12051217.
Weingart, L.R., Olekalns, M., & Smith, P.L. (2005). Quantitative coding of negotiation behav-
ior. International Negotiation.
Managing Conict in the Literature: Meta-analysis as a
Research Method
However, those drawn to meta-analysis for its objectivity soon realize that
the approach encompasses more than a straightforward recipe for integrating
research. While it has a standard methodology, meta-analysis takes on a life
of its own, dened by the creativity, planning, and decision-making of the
researchers who use it (Hedges 1990).
Development of Meta-analyses
Practical issues may also inuence formulating the problem. For example,
at some point, the researcher will want a sense of how many studies might
be included. If there are a limited number of studies, it might make sense
to broaden the denition of the independent or dependent variables. Or, it
might be feasible to paint a broader picture of the construct by including mul-
tiple dependent variables (e.g., effects on both cooperation and impasses;
Stuhlmacher et al. 1998). Invariantly, beginning researchers seek guidance or
directions for a magic number of studies. Rather than reiterating the con-
cerns of an ideal quantity of studies to sample (e.g., Glass, et al. 1981;
Rosenthal 1984), Hedges (1990) pointed out that researchers should examine
the representativeness of the participants and the treatments in the studies
relative to the area of interest. Thus, again, investigators are asked to use
judgment in determining the selection of the participants, treatments, and
studies to the research domain.
Even with clearly dened constructs, investigator judgment calls are neces-
sary. Hedges et al. (1989) warned, Treatments will not have been imple-
mented identically in all studies and different studies will not have measured
the outcome constructs in exactly the same way. Thus, you will have to judge
whether each operation is a legitimate representation of the corresponding
construct (5). Negotiation and conict research is marked by several research
traditions across multiple scholarly disciplines, thus creating some potential
misclassications. For example, Walters et al. (1998) questioned if research
on prisoner dilemma games should be included with explicit negotiation.
Ultimately, because the research literature has built on matrix games and
explicit negotiations, both of these operationalizations were combined. Studies
were coded for whether the task was a matrix game or an explicit negotiation
so that each type of task could be compared. In Walters et al. (1998), these
two tasks produced different results regarding gender differences. The ability
to run comparisons across treatments is an unmistakable advantage of syn-
thesizing the literature, and an example of how meta-analyses can clarify past
research efforts.
Similarly, because conict and negotiation are vast research literatures, a
common misperception is that large numbers of studies exist that can be com-
bined on a particular topic. We have been extremely surprised over the some-
times small number of studies that actually address a particular question and
report relevant data. The most startling nding was a search that resulted in
only 21 explicit negotiation studies (yielding 53 effect sizes) directly compar-
ing the prots men and women earned in explicit negotiation (Stuhlmacher &
Walters 1999). Despite several hundred articles being screened, we were puz-
zled that such a simple relationship was not reported in a way that allowed
MANAGING CONFLICT IN THE LITERATURE 231
effect size calculations. In this case, reviewers were equally puzzled, and
asked for validation of the search. It was certainly possible that gender com-
parisons may have been reported in a manuscript but were not retrieved by
database keywords or other searches. To locate more studies, a manual search
of frequent journal outlets of negotiation research was undertaken. When
missing studies did not materialize, attention was turned to explain why so
few studies were found and included. The smaller than expected sample size
in the meta-analysis could be due to a combination of factors. Undoubtedly,
some studies slipped past our search, particularly unpublished research. Ultimately,
although many researchers could have run the analyses for gender, gender
information was not incorporated in nal reports. In many cases, gender-prot
was not central to the reports, and thus probably omitted. Further, interest in
reporting gender differences in negotiation decreased after reviews (e.g.,
Rubin & Brown 1975) found limited gender differences. It was also possible
that researchers found gender differences that were mixed or non-signicant
(possible given the sample size to detect a small effect) and publication pres-
sures resulted in omitting gender comparisons from the report. While it is
unsatisfying to nd a small number of useable studies, clarifying the amount
and type of research on a topic is yet another important way that meta-analyses
can summarize past efforts and direct future research.
Collecting and Evaluating Data. The next stage, collecting data, involves
many of the same issues that arise in other research designs, such as choos-
ing participant characteristics. One of the main concerns with meta-analysis
has been the suitability of the studies selected for inclusion in a meta-analysis
and the implications for the inferences made from the results (Hedges
1990). Studies relevant to the research area may be scattered throughout aca-
demic libraries, presented at local conferences, or crammed into the dark
recesses of a laboratory le cabinet. Some of these studies are easily accessi-
ble (e.g., journal articles, books), while others require more diligent investi-
gation. Regardless, the comprehensive meta-analyst will search for pertinent
work by using a variety of tactics such as contacting prominent researchers
in the eld, poring over conference programs, and posting requests on Internet
sites.
Once assembled, the studies generally undergo scrutiny by more than one
researcher. At this juncture, certain characteristics of the research report are
coded. Coding may require simple distinctions that are easy to code reliably
(e.g., coding as a published vs. unpublished study) while some distinctions
require more training and judgment in coding (e.g., relative power between
parties, opponent strategy). Lipsey and Wilson (2001) clarify that variables are
coded for at least one of three reasons. Some variables relate to substantive
232 ALICE F. STUHLMACHER AND TREENA L. GILLESPIE
issues, or those variables that relate to the topic of study and may be hypoth-
esized to impact the results. A second group of variables is coded that relates
to the methods and procedures (e.g., task, participant population, use of con-
federates). These variables can examine if differences in procedures inuence
the meta-analytic results. Finally, source descriptors are coded for describing
the sample of reports and their context (e.g., published/unpublished, date of
study). Each type of coding plays a unique and important purpose in captur-
ing the studies and to assist in interpreting the ndings.
Analyses. After the data are collected, they often are analyzed using tech-
niques associated with a methodological tradition. Again, this is a judgment
call. Interestingly, the approach may be highly inuenced by the academic
discipline of the meta-analyst. In an analysis of meta-analytic trends (Mohr,
Zickar, & Russell 2000), Schmidt and Hunters methods were employed in
85% of the meta-analyses by industrial/organizational psychologists, Glasss
methods were frequently used by clinical (50%) and educational psychologists
(65.7%), while social psychologists were more likely to apply the Hedges
method (38.1%) or the Rosenthal method (23.8%).
The meta-analyses that have been published to date in negotiation have
used the Hedges method (i.e., Hedges & Olkin 1985). Our decision to use
Hedges procedures was driven by a desire to compare our results on gender
differences with seminal gender meta-analyses by Alice Eagly and colleagues
(e.g., Eagly & Johnson 1990; Eagly, Karau, & Makhijani 1995). Also, we
found that negotiation measures were not clearly amenable to the assumptions
and corrections for the study artifacts that Schmidt and Hunter propose.
Across traditions, assumptions and formulas differ somewhat; ultimately, the
choice of methods may inuence the reception of the meta-analytic research.
It is clear though that current developments require frequent re-evaluation and
consideration of any particular approach.
Also related to the analyses is the research design and choice of variables.
The interdependent nature of negotiation creates issues regarding the level of
analysis to address (Zetik & Stuhlmacher 2002). Many variables of interest in
negotiation can be considered across multiple levels. For example, when do
we want to look at joint or dyad level outcomes and when are we really inter-
ested in individual level variables? This creates yet more judgment calls for
the meta-analyst. Meta-analysts face other judgment calls when data in the
selected studies are missing or incomplete. Similar to sampling concerns, non-
randomly missing data impacts the validity of the inferences made from meta-
analytic results. The researcher then must determine how to deal with missing
data (Pigott 1994). In addition to the more traditional concerns of missing
MANAGING CONFLICT IN THE LITERATURE 233
data, negotiation research faces a dilemma of negotiations that may not come
to an agreement or ones that end early. For example, in the case of an
impasse, should the settlement be represented as a missing data point, a zero
prot, a best alternative to a negotiated agreement (BATNA) or some sort of
average? How should process issues be considered when the negotiation is
not resolved? It becomes more complex when the original researcher omits
how impasses were handled or how group statistics were calculated. Current
researchers can certainly assist future research by including complete method-
ological information and relevant descriptive statistics (particularly means,
standard deviations, sample sizes, and intercorrelations) across conditions and
variables. The quality of the meta-analysis results rests on the quality of the
data included in the review.
In short, investigators often nd themselves making judgment calls, some
arising very early in the process and others emerging after investigators have
logged extensive hours on the project. These junctures, while seemingly
objective, allow subjectivity to creep into the meta-analysis methodology, inuenc-
ing the project with the investigators biases. However, judgment calls are a
part of the process and can be managed professionally. One way of approach-
ing judgment calls is to construct a plan in advance for all the steps of the
project, detailing how judgments will be made, and reporting deviations from
the plan in the nal report (Hedges 1990). For example, a decision can be to
run the analysis in more than one way to incorporate a possible judgment call.
As mentioned earlier, Walters et al. (1999) ran results together and separately
for matrix games and explicit negotiations. Zetik and Stuhlmacher (2002)
compared negotiation performance across goal conditions in two ways (opti-
mal vs. suboptimal goals, and goal vs. no goal) to tease out if different oper-
ational denitions across studies made a difference in interpretation. In this
way, more information is brought to evaluating the analyses.
Interpretation and Reporting. After analyses are completed, the information
is interpreted and written up into a nal report. The nal report ideally
includes the judgments that were made throughout the course of the meta-
analysis; readers are thus alerted to judgment calls that limit the generaliz-
ability of the results or the inferences that can be made. For example, specic
information is needed on how the studies were selected, the descriptions of
the subjects in the studies, as well as a depiction of the treatments (Hedges
1990). Judgment continues as meta-analysts try to make sense of the data.
Given identical results, researchers may interpret it differently (Cooper
1990b), which inuences the resulting report and potentially future research.
As Cooper (1990b) explained, Thus, while a great deal of agreement might
234 ALICE F. STUHLMACHER AND TREENA L. GILLESPIE
Challenges of Meta-analysis
Anticipating and resolving the judgment calls within a meta-analysis are cen-
tral challenges of the meta-analysis process. There are resources devoted to
the judgment call issue in greater depth (e.g., Wanous, Sullivan, & Malinak
1989) and ideally these judgments should be addressed at the beginning of a
meta-analytic project. Anticipated in advance, the meta-analysts can agree on
the best process for navigating these issues. Glass et al. (1981) remarked that
many critics may be concerned about study quality, studies that include non-
independent data, and the problem of selection bias in the studies that are
more easily accessible to researchers. However, they countered these argu-
ments by showing the insignicance of the effects of these cited methodolog-
ical weaknesses in actual data analysis.
A further challenge is that meta-analyses require a substantial amount of
expertise and effort (Lipsey & Wilson 2001). The uninitiated may not appre-
ciate the amount of time, effort, and expertise involved in a completing a
high-quality meta-analysis. Initially, the researcher may encounter obstacles
convincing others about the value and difculty of such projects. As more sci-
entists become familiar with the technique, it will hopefully become less nec-
essary for the researcher to educate promotion or tenure committees, grant
agencies, and colleagues that meta-analyses involve signicant time, effort,
and resources but can yield a substantial contribution to the eld. Ultimately,
we nd that high-quality research requires time and a meta-analysis is no
different.
These challenges are less likely as meta-analyses are published which clar-
ify the status quo set by primary studies. The effects between variables can
appear far weaker or even inverted when a comprehensive meta-analysis is
conducted. One recent example is a meta-analysis by De Dreu and Weingart
(2003) where the general belief that task-conict can be benecial to team
performance was not supported. In fact, task-conict appeared to be as dys-
functional as relationship conict. Future research can clearly benet from the
research review that a meta-analysis can provide to a topic.
MANAGING CONFLICT IN THE LITERATURE 235
Future of Meta-analysis
The future for meta-analysis is bright, and advancing with increasing sophis-
tication. One way this can be seen is through an increase in software devel-
oped to assist the meta-analyst (e.g., Arthur, Bennett, & Huffcutt 2001; see
also Rothstein et al. 2002 for summary of software). It also appears clear that
one substantial advantage of the meta-analytic perspective is the opportunity
for researchers to handle large amounts of information (Lipsey & Wilson
2001). As a result of planned coding systems, a meta-analysis can deal with
large numbers of variables across many studies without the concern of stud-
ies being forgotten or receiving less attention from the investigator. If enough
literature exists, predictions from competing models can be tested (e.g., De
Dreu et al. 2000). In addition, moderators, mediators, or other relationships
between variables can be investigated that may have been less obvious in the
separate studies or if reviewed using more conventional methods of research
synthesis (Lipsey & Wilson 2001).
Meta-analyses have been criticized as emphasizing correlational rather than
causal relationships. However, this criticism hinges on the type of research
that has been conducted. Like other forms of modeling, causality assumptions
rely on the design of the research. Thus, in many ways, the meta-analyst is
constrained by the data that exists. One recent approach that expands the use
of existing data involves using meta-analyses to test mediators through build-
ing causal models (Becker 1992; Shadish 1996). For example, a correlation
matrix can be derived from several meta-analyses and can be used as the
input for path-analysis. This then has the potential to test not just moderators
but mediating pathways and competing models (e.g., Grant, Compas, Stuhlmacher,
Thurm, McMahon, & Halpert 2003).
fertile ground for meta-analytic work. Competition between ideas and tradi-
tions has always existed in negotiation research. These pressures may have
been intensied of late as negotiation has become more central to the cur-
riculum in many areas of study, as more researchers enter the eld, and as
pressures to publish research has increased. As studies continue to emerge and
results, however disparate from each other, continue to be reported, the role
of meta-analysis becomes greater in helping make sense of it all by exposing
and examining contradictory results. It is our belief that meta-analyses afford
opportunities for constructive dialogue and advances in understanding social
conict.
References
Arthur, Winfred. Jr., Bennett, Winston, Jr., & Huffcutt, Allen (2001). Conducting meta-analyses
using SAS. Mahwah, N.J.: Lawrence Erlbaum Associates.
Becker, Betsy Jane (1992). Models of science achievement: Forces affecting male and female
performance in school science, in Thomas D. Cook, Harris Cooper, David S. Cordray,
Heidi Hartmann, Larry V. Hedges, Richard J. Light, Thomas A. Louis, & Frederick
Mostellar, editors, Meta-analysis for explanation: A casebook. New York: Russell Sage
Foundation.
Coleman, Peter T. (2001). Concluding overview, in Morton Deutch & Peter T. Coleman, edi-
tors, The Handbook of Conict Resolution: Theory and Practice. San Francisco: Jossey-Bass.
Cooper, Harris M. (1990a). Moving beyond meta-analysis, in Kenneth W. Wachter & Miron
L. Straf, editors, The future of meta-analysis. New York: Russell Sage Foundation.
Cooper, Harris M. (1990b). On the social psychology of using research reviews, in Kenneth
W. Wachter & Miron L. Straf, editors, The future of meta-analysis. New York: Russell Sage
Foundation.
Cooper, Harris M., & Rosenthal, Robert (1980). Statistical versus traditional procedures for
summarizing research ndings. Psychological Bulletin, 87: 442449.
De Dreu, Carsten K., Weingart, Laurie R., & Kwon, Seungwoo (2000). Inuence of social
motives on integrative negotiation: A meta-analytic review and test of two theories. Journal
of Personality and Social Psychology, 78: 889905.
De Dreu, Carsten K., & Weingart, Laurie, R. (2003). Task versus relationship conict, team
performance, and team member satisfaction: A meta-analysis. Journal of Applied Psy-
chology, 88, 4: 741749.
Druckman, Daniel. (1994). Determinants of compromising behavior in negotiation: A meta-
analysis. Journal of Conict Resolution, 38: 507556.
Eagly, Alice H., & Johnson, Blair T. (1990). Gender and leadership style: A meta-analysis.
Psychological Bulletin, 108: 233256.
Eagly, Alice H., Karau, Steven J., & Makhijani, Mona (1995). Gender and the effectiveness
of leaders: A meta-analysis. Psychological Bulletin, 117: 125145.
Glass, Gene V. (1976). Primary, secondary and meta-analysis of research. Educational Researcher,
5: 38.
MANAGING CONFLICT IN THE LITERATURE 237
Glass, Gene V., McGaw, Barry, & Smith, Mary L. (1981). Meta-analysis in social research.
Beverly Hills, CA: Sage.
Grant, Kathyrn E., Compas, Bruce E., Stuhlmacher, Alice F., Thurm, Audrey E., McMahon,
Susan D., & Halpert, Jane A. (2003). Stressors and child/adolescent psychopathology:
Moving from markers to mechanisms of risk. Psychological Bulletin, 129: 447466.
Hedges, Larry V. (1990). Directions for future methodology, in Kenneth W. Wachter & Miron
L. Straf, editors, The future of meta-analysis. New York: Russell Sage Foundation.
Hedges, Larry V., & Olkin, Ingram (1985). Statistical methods for meta-analysis. Orlando, FL:
Academic Press.
Hedges, Larry V., Shymansky, James A., & Woodworth, George (1989). Modern methods of
meta-analysis. Washington, DC: National Science Teachers Association.
Huffcutt, Allen I. (2002). Research perspectives on meta-analysis. in Steven G. Rogelberg,
editor, Handbook of research methods in industrial and organizational psychology. Malden,
MA: Blackwell.
Hunter, John E., & Schmidt, Frank L. (1990). Methods of meta-analysis: Correcting error and
bias in research ndings. Newbury Park, CA: Sage.
Light, Richard J., & Smith, Paul V. (1971). Accumulating evidence: Procedures for resolving
contradictions among different research studies. Harvard Educational Review, 41: 429471.
Lipsey, Mark W., & Wilson, David B. (2001). Practical meta-analysis: Vol. 49. Applied social
research methods series. Thousand Oaks, CA: Sage.
Mohr, David C., Zickar, M.J., & Russell, S.S. (2000, April). An analysis of trends in meta-
analytic research. Poster presented at the Annual Meeting of the Society of Industrial/
Organizational Psychology. New Orleans, LA.
Pigott, Therese D. (1994). Methods for handling missing data in research synthesis, in Harris
Cooper & Larry V. Hedges, editors, The Handbook of Research Synthesis. New York: Russell
Sage Foundation.
Rosenthal, Robert (1984). Meta-analytic procedures for social research. Beverly Hills, CA:
Sage.
Rosenthal, Robert (1991). Meta-analytic procedures for social research: Vol. 6. Social research
methods series. Newbury Park, CA: Sage.
Rothstein, Hannah R., McDaniel, Michael A., & Borenstein, Michael (2002). Meta-analysis:
A review of quantitative cumulation methods, in Fritz Drasgow & Neal Schmitt, editors,
Measuring and analyzing behavior in organizations. San Francisco: Jossey-Bass.
Rubin, Jeffrey Z., & Brown, B.R. (1975). The social psychology of bargaining and negotiation.
New York: Academic Press.
Shadish, William R. (1996). Meta-analyses and the exploration of causal mediating processes:
A primer of examples, methods, and issues. Psychological Methods, 1: 4765.
Smith, Mary L., & Glass, Gene V. (1977). Meta-analysis of psychotherapy outcome studies.
American Psychologist, 32, 9: 752760.
Stuhlmacher, Alice F., & Citera, Maryalice. (in press). Negotiator hostility and prot in virtual
negotiations: A meta-analysis. Journal of Business and Psychology.
Stuhlmacher Alice F., Gillespie, Treena L., & Champagne, Matthew V. (1998). The impact of
time pressure in negotiation: A meta-analysis. The International Journal of Conict Management,
9: 97116.
Stuhlmacher, Alice F., & Walters, Amy E. (1999). Gender differences in negotiation outcome:
A meta-analysis. Personnel Psychology, 52: 653677.
238 ALICE F. STUHLMACHER AND TREENA L. GILLESPIE
Walters, Amy E., Stuhlmacher, Alice F., & Meyer, Lia L. (1998). Gender and negotiator
competitiveness: A meta-analysis. Organizational Behavior and Human Decision Processes,
76: 129.
Wanous, John P., Sullivan, Sherry E., & Malinak, Joyce (1989). The role of judgment calls in
meta-analysis. Journal of Applied Psychology, 74, 2: 259264.
Zetik, Deborah C., & Stuhlmacher, Alice F. (2002). Goal setting and negotiation performance:
A meta-analysis. Group Processes and Interpersonal Relations, 5: 3552.
When, Where and How: The Use of Multidimensional
Scaling Methods in the Study of Negotiation and
Social Conict
the elements of the matrix provide information about the relation among the
objects, events, or behaviors (which we will refer to as stimuli) that comprise
the rows and columns of the data matrix (Young & Hamer 1987). Unlike fac-
tor analysis, however, MDS does not require metric data, allowing for the use
of both metric (interval or ratio) and nonmetric (ordinal) data. Given that data
reecting attitudes and cognitions are nonmetric, MDS is ideally suited to the
study of conict and negotiation.
Although MDS has wide-ranging theoretical and applied appeal, it has been
highly underutilized in the conict and negotiation literature, which has
tended to rely on factor analysis to understand hidden data structures. In this
paper, we seek to illustrate the promise that MDS has in the study of conict
and negotiation. We begin with a discussion of how MDS can be differenti-
ated from other multivariate techniques, such as factor analysis, illustrating its
distinct advantages to conict scholars. Next we provide a brief overview of
multidimensional scaling techniques highlighting the various methods avail-
able for collecting proximity data and the MDS computer analysis programs
that can be used to analyze them. We further review the nature of the results
and the ways in which they are interpreted. We conclude with some examples
of the types of questions that have been addressed using MDS in the conict
and negotiation literature, highlighting the promise this technique holds for
future research.
There are two issues that scholars face when interpreting MDS congurations:
1) determining the number of dimensions that best represents the actual prox-
imities between the data, and 2) labeling each dimension or determining the
interpretability of each dimension. Each issue is discussed in turn below.
Determining Dimensionality. Three criteria are used to determine the opti-
mum number of dimensions needed to describe the stimulus space. While reliance
on these methods varies, experimenters typically use all three criteria for deter-
mining dimensionality.
The rst criterion used for determining dimensionality, is Kruskals (1964)
STRESS index or goodness of t, which indicates how the distances displayed
in the conguration reect the actual proximities in the similarities data.
Technically speaking, it is the square root of a normalized residual sum of
squares, which exhibits the amount of variance that remains unaccounted for
by the MDS model. Although measurements of STRESS vary from analysis
program to analysis program (such as M-D-SCAL, TORSCA, KYST, or
ALSCAL), the meaning of STRESS is always the same: Small STRESS indi-
cates good t, with good t nearing zero and poor t nearing one. As the
number of dimensions increase, STRESS becomes closer to zero. It should be
noted that a number of factors affect stress values. For example, when the
number of stimuli (I) and the number of dimensions (R) of stress are similar,
the STRESS index can be distorted, resulting in undue inuence of the inter-
pretation R. As a result, a good rule of thumb is to use at least four times as
many stimulus items as the number of dimensions likely to underlie the stim-
ulus space (i.e., I > 4R).
A second criterion for evaluating the number of dimensions necessary and
sufcient to adequately represent the stimulus space, is the RSQ index or
squared multiple correlation between the proximities in the similarities data
and the distances plotted by the MDS model. The RSQ index describes how
much of the variance in the proximity data is accounted for by the MDS
model. As with any squared correlation, a one indicates a perfect t and a
zero indicates no t at all.
To determine the appropriate number of dimensions, researchers typically
plot the rst two criteria (i.e., STRESS and RSQ) against the number of
dimensions to discover an elbow or bend, which is designated by a sudden
rise in RSQ and fall in STRESS. The number of dimensions that correspond
with the elbow represent a particularly good t of the MDS model to the
proximity data (Young & Hamer 1987). The elbow test is usually accompa-
nied by an assessment of how well the raw data ts the MDS model by
246 ROBIN L. PINKLEY, MICHELE J. GELFAND AND LILI DUAN
Several examples may further illustrate the inherent benets of using MDS
techniques in the study of conict and negotiation. Although we provide only
a couple examples here, a handful of scholars have used MDS techniques to
address such issues as negotiatior perceptions of conict situations (Pinkley
THE USE OF MULTIDIMENSIONAL SCALING METHODS 247
Example 1
Pinkley, Brittain, Neale, & Northcraft (1995) used MDS to conduct an induc-
tive analysis of managerial third-party dispute intervention strategies. The
objective of this study was to identify the dimensions that distinguish one class
or category of intervention strategies from others. In addition, the authors
examined the relationship between strategy selection and the nature of the
conict (i.e., managerial dispute intervention goals, dispute intensity, time pres-
sure, dispute importance, managerial power, and the relative power of the dis-
putants). To fulll this objective, the authors used a ve-step method to collect
and analyze the data. In step one, alumni from four universities lled out a
survey in which they provided a description of the last time they intervened in
a corporate conict, as well as, specics regarding the nature of the conict.
Of the 142 obtained descriptions, 40 were randomly chosen as step 2 stimulus
materials.
In step two, ten of the remaining 40 conict descriptions were randomly
selected as target descriptions. One hundred participants, were randomly sep-
arated into ten groups, each of which was assigned target description, such
that ten participants were given the rst target description, ten the second tar-
get description and so on. Each group was asked to rank-order the remaining
39 descriptions, in terms of how similar they were to their assigned target
description. Participants were also asked to specify the criteria they used for
making their similarity judgments.
A three-way MDS analysis using SASs alternative least squares scaling
(ALSCAL, see Takane, Young, & DeLeeuw 1977) implementation of the indi-
vidual differences scaling model (Carroll & Chang 1970). A new set of
allowed experimenters to test the hypothesized relationship between manage-
rial strategy selection (as dened by dimensionality) and the nature of
conict. Two criteria were used to determine the optimum number of dimen-
sions: 1) Kruskals (1964) STRESS index and 2) an elbow test accompanied
by an examination of the amount of variance accounted for by each dimen-
sional solution. Both procedures suggested that a six-dimensional solution did
not signicantly improve on the ve-dimensional solution, which accounted
for 95% of the variance.
248 ROBIN L. PINKLEY, MICHELE J. GELFAND AND LILI DUAN
In step three, one-hundred and forty potential labels were generated from
two sources: 1) the labels suggested by the criteria used by step two partici-
pants to rank-order the intervention strategy descriptions, and 2) the third-
party intervention categories used by past scholars such as Thibaut and
Walker (1975) and Carnevale (1986). A third group of participants rated the
degree to which each of the 40 intervention strategy descriptions reected
each of the 140 potential labels on a 9-point Likert-type scale. Two criteria
were used to label the dimensions: 1) multiple-correlations and F-tests
revealed that 22 of the 140 potential labels related to the dimensions at the
p-value level of .01 or better and 2) multiple regression produced direction
cosines (beta weights) to relate each potential label to each dimension. Ten
labels most closely related to each of the ve dimensions and were thus, used
to label the conguration.
In step four, the authors used conrmatory analysis to verify the appropri-
ateness of each of the ten labels by asking ve trained raters (unaware of the
step one three results) to rate the step one intervention strategy descriptions
in terms of each label. Cronbachs alpha was used to assess interrater relia-
bility (found to be quite high at .85). The F values and beta weights found
that all ten of the selected labels loaded onto the ve-dimensional solution
with reliability ranging from .77 to .97. As a result, the ve dimensions were
labeled: 1) Attention to the stated versus underlying problem, 2) Disputant
commitment forced versus encouraged, 3) Manager decision control versus
disputant decision control 4) Manager approaches conict versus avoids con-
ict, and 5) Dispute handled publicly versus privately.
Finally, step ve used multiple regression to relate the ve-dimensional solu-
tion to the intervenor goals and perceptions of conict specied by the step
one participant surveys. This step allowed the authors to evaluate when and
under what circumstances various intervention strategies are used by managers.
Example 2
Gelfand, Nishii, Holcombe, Dyer, Ohbuchi, and Fukuno, M. (2001) used MDS
to examine the dimensions that are used to construe conicts across cultures.
The purpose of the study was to discern if there are universal (or etic) dimen-
sions of conict construal and if there are culture-specic (emic) construals
that are consistent with prevailing cultural values and practices. This study
involved ve steps. In step one, students from the U.S. and Japan were asked
to write a description of a conict that they had experienced in the recent
past. Consistent with Pinkley (1990), they were told they could describe any
incident they chose, regardless of its nature, the type of relationship, or the
THE USE OF MULTIDIMENSIONAL SCALING METHODS 249
program was used to analyze the matrices. An elbow test of Kruskals mea-
sure of stress suggested that stress values did not decrease substantially from
the three to four dimensional solutions, yet did decrease substantially from the
two to three dimensional solutions for all of the MDS spaces. The three
dimensional solution was also chosen because it allowed for the most com-
prehensible interpretations for all of the MDS solutions.
In step six, the dimensions were labeled based on a) an examination of the
conicts in the MDS spaces as well as b) multiple regression analyses which
examined how well the location of each conict on these unidimensional
items was predicted by its location in the multidimensional space. Items with
a signicant multiple correlation and signicant Beta weights indicate that the
conguration explains the item well. Items that load on multiple dimen-
sions, however, are not as useful for labeling the dimensions. The results
demonstrated that Japanese and American participants construed both U.S.
and Japanese conicts through a Compromise versus Win frame (Pinkley
1990), providing evidence of a universal dimension of conict construal. The
results also illustrated that Japanese perceived both sets of conicts to be
more Compromise focused, as compared to Americans. In addition, there were
unique dimensions of conict construal among Americans and Japanese (e.g.,
Infringements to Self and Giri Violations, respectively), suggesting that iden-
tical conict episodes can be perceived differently across cultures.
Example 3
Druckman, Martin, Nan, & Yagcioglu (1999) used MDS to examine the struc-
ture of actual cases of international negotiation. The objective of the study
was to test whether Ikls (1964) typology of international negotiation could
account for similarities among the actual negotiations. This typology distinguished
among ve objectives in international negotiations, including extension, nor-
malization, redistribution, innovation, and side effects, which were construed
as distinct types of negotiations with particular processes and outcomes. Although
the taxonomy had been widely discussed in the literature, Druckman et al.
(1999) set forth to directly test the validity of Ikls notions using MDS.
This study involved six steps. In step one, 30 cases were randomly selected
from the Pew Case Studies in International Affairs (approximately 17% of
the cases available). Cases that were sampled differed along numerous char-
acteristics, including region and type of issue (e.g., economic cases, security
issues, environmental, hostage negotiations). All cases were approximately ten
to fteen pages and of a common format, consisting of background informa-
tion, a discussion of the unfolding of the negotiation, and an analysis of the
THE USE OF MULTIDIMENSIONAL SCALING METHODS 251
processes and the outcomes of the negotiation. After selecting the cases, the
authors categorized the conicts in terms of Ikls ve objectives. Final deci-
sions were reached by a consensus between at least two people (the rst
author and another student familiar with Ikls theory). The authors found that
the cases that were chosen were representative of the ve objectives identied
by Ikls theory (eight innovation cases, ve normalization cases, ten redistri-
bution cases, ve extension cases, and two side effect cases).
In step two, the 30 cases were divided randomly into approximately four
equal sets and who were assigned to one of four coders were unfamiliar with
the taxonomy being examined. The cases were coded for sixteen categories,
including characteristics of the parties (e.g., number, power, length and type
of relationship), number and type of issues, the process of negotiation (e.g.,
length and types of exchanges), negotiation outcomes (e.g., type of agree-
ment) and conditions surrounding the negotiation (e.g., time pressure and
media coverage). A smaller sample of cases was subject to inter-rater relia-
bility. There were high levels of agreement (generally over 90%) and dis-
agreements were resolved through a renement of the denitions of variables.
In step three, correlations were computed among the 30 cases across the 16
coded variables. These correlations were used as an indication of similarity
among the cases. Subsequently, a 30 30 matrix of correlations was subject
to MDS analysis. Based on stress values as well as interpretability, the authors
chose a two-dimension solution.
In step four, the authors examined whether the negotiation cases clustered
according to Ikls taxonomy. First, they visually inspected the clustering of
the cases and found initial support for the notion that negotiations cluster
according to their focus on innovation, redistribution, extension, side effects,
and normalization. They also found a new negotiation category, labeled multi-
lateral regimes, which they noted had not existed at the time of the publi-
cation of Ilks book. Next, they performed a Kruskal-Wallis ANOVA to test
for signicant differences among the clusters, and also performed a K-means
cluster analysis. This analysis revealed three clusters (multilateral and nor-
malization cases), innovation cases, and redistribution cases.
In step ve, the authors checked to see whether their initial categorization
of the cases could be veried. In order to do so, they performed discriminant
analyses that examined whether the sixteen coded variables could distinguish
among Ikls categories. These analyses illustrated that a high percentage of
cases were classied accurately.
Finally, in step six, the authors created proles of the clusters of types of
negotiations along the sixteen variables, providing a parsimonious under-
standing of the processes that characterize different types of international
252 ROBIN L. PINKLEY, MICHELE J. GELFAND AND LILI DUAN
Example 4
McLaughlin, Carnevale, and Lim (1991) used MDS to identify the dimensions
underlying professional mediators categorizations of mediation tactics. The
authors had four phases to achieve their research purpose: (a) data collecting,
(b) multidimensional scaling analyses, (c) regression analyses, and (d) cluster
analyses. In step 1, surveys were mailed to 230 mediators sampled randomly
from a list of members of the Society of Professionals Involved in Dispute
Resolution (SPIDR). The mediators were asked to sort 36 stimulus tactics,
taken from Carnevale and Pegnetters (1985), into as many mutually exclu-
sive categories as they wanted. After the sorting task, the mediators rated each
tactic on ve bipolar scales: friendly-unfriendly, assertive-passive, controlling-
uncontrolling, use frequently-use infrequently, and effective-ineffective. All mate-
rials were mailed back to the researchers.
In the next step, researchers created a tactics tactics (36 36) diagonal
matrix of similarities. The similarity of each pair of tactics was presented by
the number of times across mediators that both tactics were included in the
same category. The tactics tactics diagonal matrix was used in MDS analy-
ses and clustering analyses. A two-way nonmetric MDS program, KYST2A
(Kruskal, Young, & Seery 1977) scaled the similarities data. Kruskals stress
values were obtained for the one- through six-dimensional solution and the
elbow criterion was used to determine the number of dimensions. Because
stress values dropped greatly from the two- to three-dimensional solution
and decreased minimally beyond the three-dimensional solution, the three-
dimensional solution appeared to be closest to the true structure of mediation
tactics.
In the third step, unidimensional scales were regressed onto the MDS congura-
tion to label the dimensions. Those unidimensional scales, with (a) relatively
large squared multiple correlations and (b) a large weight on one of the three
dimensions but not on the other two, proved most useful for interpreting the
dimensions. Based on these analyses, researchers labeled the three dimensions
substantive versus reexive, affective versus cognitive, and forcing versus
facilitating.
In the last step, the tactics similarities matrix was clustered hierarchically
with the CLUSTER procedure in SPSS. The results of cluster analyses were
compared with MDS results. In addition, the authors found that the cluster
THE USE OF MULTIDIMENSIONAL SCALING METHODS 253
and MDS results were compatible with the results of factor analyses of medi-
ation tactics.
Conclusion
References
Beardsworth, A., & Keil, T. (1992). The vegetarian option: Varieties, conversions, motives and
careers. Sociological Review 40: 253293.
Bernard, H.R. (2000). Social Research Methods: Qualitative and Quantitative Approaches. Thousand
Oaks, California: Sage Publications, Inc.
Black, J.K. (1991). Development in theory and practice: Bridging the gap. Boulder, Colorodo:
Westview.
Busing, F.M.T.A., Commandeur, J.J.F., & Heiser, W.J. (1997). PROXSCAL: A multidimen-
sional scaling program for individual differences scaling with constraints, in W. Bandilla &
F. Faulbaum, editors, Softstat 97: Advances in Statistical Software, Stuttgart: Lucius &
Lucias, 6: 6774.
Carnevale, P.J.D., & Pegnetter, R. (1985). The selection of mediator tactics in public sector
disputes: A contingency analysis. Journal of Social Issues 41, 2: 6581.
Carroll, J.D., & Chang, J.J. (1970). A quasi-nonmetric version of INDSCAL, a procedure for
individual differences in multidimensional scaling. Paper presented at meetings of the Psycho-
metric Society, Stanford, CA.
Druckman, D. (1997). Dimensions of international negotiations: Structures, processes, and out-
comes. Group Decision & Negotiation 6, 5: 395420.
Druckman, D., Martin, J., Nanand, S.A., & Yagcioglu, D. (1999). Dimensions of international
negotiation: A test of Ikls typology. Group Decision & Negotiation 8: 89108.
254 ROBIN L. PINKLEY, MICHELE J. GELFAND AND LILI DUAN
Gelfand, M., Nishii, L., Holcombe, K., Dyer, N., Ohbuchi, K., & Fukuno, M. (2001). Cultural
Inuences on Cognitive Representations of Conict: Interpretations of Conict Episodes in
the U.S. and Japan. Journal of Applied Psychology 86, 6: 10591074.
Gelfand, M., Triandis, H., & Chan, D. (1996). Individualism versus collectivism or versus
authoritarianism? European Journal of Social Psychology 26, 3: 397410.
Green, P.E., & Carmone, F.J. (1970). Multidimensional Scaling and Related Techniques in Marketing
Analysis. Allyn & Bacon, Inc., Boston, Massachusetts.
Hair, J.F., Anderson, R.E., Tatham, R.L., & Black, W.C. (1998). Multivariate Data Analysis,
fth edition. Prentice-Hall, Inc., Upper Saddle River, New Jersey.
Hansen, J.C., Sarma, Z.M., & Collins, R.C. (1999). An evaluation of Hollands model of
vocational interests for Chicana(o) and Latina(o) college students. Measurement and Evalua-
tion in Counseling & Development 32: 214.
Hirschberg, N., Jones, L., & Haggerty, M. (1978). Whats in a face: Individual differences in
face perception? Journal of Research in Personality 12, 4: 488499.
Jehn, K.A. (1994). A multimethod examination of the benets and detriments of conict.
Administrative Science Quarterly 40: 256282.
Johnson, C.S. (1995). The Rokeach value survey: Underlying structure and multidimensional
scaling. Journal of Psychology 129: 583597.
Jones, L. & Young, F. (1972). Structure of a social environment: Longitudinal individual dif-
ferences scaling of an intact group. Journal of Personality & Social Psychology 24, 1:
108121.
Kruskal, J.B. (1964). Nonmetric multidimensional scaling: A numerical method. Psycho-
metrika 29: 115129.
Kruskal, J.B. (1968). How to use M-D-SCAL, a program to do multidimensional scaling and
multidimensional unfolding, (Version 4 and 4M of M-D-SCAL, all in FORTRAN IV), Bell
Telephone Laboratories, Murray Hill, New Jersey, March 1968 (mimeographed).
Kruskal, J.B., and Wish, M. (1978). Multidimensional scaling. Sage University Papers Series.
Quantitative Applications in the Social Sciences, no. 07011. Beverly Hills, California: Sage.
Kruskal, J.B., Young, F.W., & Seery, J.B. (1977). How to use KYST2A: A very exible program
to do multidimensional scaling and unfolding. Murray Hill, NJ: Bell Telephone Laboratories.
Liske, J. (1993). Regional subcultures of the United States. Journal of Politics 55: 888913.
McLaughlin, M., Carnevale, P., & Lim, R. (1991). Professional mediators judgments of medi-
ation tactics: Multidimensional scaling and cluster analyses. Journal of Applied Psychology
76, 3: 465472.
Pinkley, R.L. (1990). Dimensions of conict frame: Disputant interpretations of conict.
Journal of Applied Psychology 75, 2: 117126.
Pinkley, R.L., Brittain, J., Neale, M.A. & Northcraft, G.B. (1995). Managerial third party dis-
pute intervention: The impact of situational factors on strategic choice. Journal of Applied
Psychology 80: 363402.
Robinson, S.L. & Bennett, R.J. (1995). A typology of deviant workplace behaviors: a multi-
dimensional scaling study. Academy of Management Journal 38, 2: 555572.
Takane, Y., Young, F.W. & De Leeuw, J. (1977). Nonmetric individual differences in multi-
dimensional scaling: An alternating least squares method with optimal scaling features.
Psychometrika 42: 767.
Thibaut, J., & Walker, L. (1975). Procedural justice: A psychological analysis. Hillsdale, New
Jersey: Erlbaum.
Wish, M., Deutsch, M., & Biener, L. (1970). Differences in conceptual structures of nations:
an exploratory study. Journal of Personality & Social Psychology 16, 3: 361373.
THE USE OF MULTIDIMENSIONAL SCALING METHODS 255
Young, F.W. and Hamer, R.M. (1987). Multidimensional Scaling: History, Theory and Appli-
cation. Hillsday, New Jersey: Lawrence Erlbaum Associates.
Young, F.W. and Torgerson, W.S. (1968). A FORTRAN IV program for nonmetric multi-
dimensional scaling on the IBM 360/75 computer (obtained from Dr. Young at the University
of North Carolina).
Markov Chain Models of Communication Processes in
Negotiation
Markov chain analysis assumes that the set of tactics used in a negotiation
can be classied into a nite number of discrete tactics, or states, and that the
sequencing of tactics, which is represented as a transition between tactics, can
be described by a simple probability model. To carry out a Markov chain
analysis, we think of the tactic used by a particular member of a dyad at
speaking turn n as a random variable, denoted Xn. By this we mean that at
any given point in time, negotiators may choose between one of several tac-
tics, each with some probability. Which tactic they choose denotes the state
of the dyad at that time. The appendix shows an excerpt from a transcript in
which we have coded each strategy as either integrative or distributive. The
sequential dependencies among these 12 strategies can be represented schema-
tically as follows:
integrative1(Party A) distributive2(Party B) distributive3(Party A) . . .
. . . integrative10(Party B) distributive11(Party A) distributive12(Party B)
Put more formally, we can say that Xn is a dichotomous variable that takes
the values integrative or distributive, each with some probability. We can
therefore think of a negotiation consisting of N speaking turns as a sequence
of state transitions of the form:
X1(a) X2(b) X3(a) Xn-1(a) XN(b) (1)
In this representation, the random variables are subscripted with the speaking
turn and are superscripted a or b to show that they represent communication
between members of the dyad (Parties a and b) in alternation. The arrows
indicate a hypothesized casual, and hence statistical, dependency, between
consecutive values of Xn.
Within this framework, we may ask: What is the probability that a partic-
ular tactic is used at a given speaking turn, given the entire history of the
process up until that point? In terms of our example, this is like asking: What
MARKOV CHAIN MODELS OF COMMUNICATION PROCESSES IN NEGOTIATION 259
Time n-1 [ ]
0 .6 .4 .
1 .3 .7
(4)
integrative and distributive tactics, respectively, this matrix asserts that the
probability of reciprocating an integrative tactic at time n-1 (the row state)
with an integrative tactic at time n (the column state) is .6. The probability
of reciprocating a distributive tactic is .7. This transition matrix is equivalent
to the pair of probability statements:
P[X_Xn = 0|Xn-1 = 0] = .6
P[X_Xn = 0|Xn-1 = 1] = .3. (5)
Because the probabilities in each row of the transition matrix must sum to
1.0, these two equations sufce to determine the transition matrix uniquely.
The requirement that the rows sum to 1.0 expresses the fact that, at each time
step, the chain must be in exactly one of a nite set of states.
As discussed by Agresti (1990) and Bishop, Fienberg and Holland (1975), Markov
chain models may be analyzed using the same loglinear modeling techniques
that are used to analyze multi-way contingency tables. This approach is an
attractive one, because of the widespread availability of loglinear modeling
software. Unlike most psychological applications of loglinear models, how-
ever, the unit of analysis in Markov chain models is the tactic, or the speech
act, rather than the individual.
In loglinear models, statistical dependencies among variables are repre-
sented as interactions. An association between two variables is a two-way
interaction; an association among three variables is a three-way interaction,
and so on. In loglinear analyses of Markov chains, the order of the chain (i.e.,
whether it is rst-order or second-order, etc.) may be assessed by determin-
ing the highest order interaction needed to describe the dependencies in the
sequence of coded tactics. To do this, the sequence of tactics must be put into
contingency table form. The variables that form the margins of the contin-
gency table are the tactics at consecutive time steps. The number of dimen-
sions in the table depends on the hypothesized order of the underlying chain.
In general, a Markov chain of order m has dependencies among the tactics
at m + 1 time steps. The number of dimensions of the table must therefore
equal the hypothesized order of the chain plus one. Thus a rst-order chain
requires a two-dimensional table; a second-order chain requires a three-dimen-
sional table, and so on. Because the order of the chain is an empirical ques-
tion, it is necessary to test for the nonsignicance of interaction terms that
represent chains of higher order. Thus to conclude that a given set of data
MARKOV CHAIN MODELS OF COMMUNICATION PROCESSES IN NEGOTIATION 261
that a test that a chain is (at least) second order can be carried out by testing
the appropriate three-way interaction terms in a contingency table in which all
two-way interaction terms are present.
To demonstrate how we t rst- and second-order Markov chains, we gen-
erated two data sets (shown in Table 1) using Monte Carlo simulations of two
chains, each representing different underlying processes. The rst data set (left
side of Table 1) was based on the assumption that the underlying communi-
cation structure can be represented as a rst-order chain. This chain was gen-
erated using the probability model of Equation 5 (i.e., the transition matrix
(4)). The second data set (right side of Table 1) was based on the assumption
that the underlying communication structure can be represented as a second-
order chain. This chain was generated using the probability model below:
P[Xn = 0|Xn-1 = 0, Xn-2 = 0] = .7
P[Xn = 0|Xn-1 = 0, Xn-2 = 1] = .5
P[Xn = 0|Xn-1 = 1, Xn-2 = 0] = .4
P[Xn = 0|Xn-1 = 1, Xn-2 = 1] = .2 (6)
Note that the rst-order transition probabilities for this chain, averaged over
values of Xn-2, are .6 and .3, which are the same as those for the rst-order
chain. However, the statistical structure of the two chains is very different.
Table 2 shows the result of tting Markov chain models to these simulated
data. These models are specied using the notation of Fienberg (1980; see
also Agresti, 1990). In Fienbergs notation, a loglinear model is specied as
a set of bracketed model terms. This notation provides a succinct way of indi-
cating which variables in a model are independent of one another and which
variables interact. Variables that interact are grouped together inside a set of
brackets; variables that are independent of one another are placed in separate
brackets. Thus, for example, the models [X1][X2][X3] and [X1, X2][X2, X3] are
two possible representations of the structure of a three-dimensional contin-
gency table with margins X1, X2, and X3. The rst model states that the three
variables that form the margins of the table are independent of one another.
It may be thought of as a generalization of the usual chi-square test for inde-
pendence in two-way tables. Psychologically, it asserts that there are no sta-
tistical dependencies in the tactics used in consecutive speaking turns in the
negotiation.
The second model, [X1, X2][X2, X3] is a rst-order Markov chain. It states
that the only statistical dependencies in the table are the one-step transitions
between consecutive speaking turns: Xn-1 Xn. Both of these models can be
tested within a three-way table because neither of them is saturated. That is,
the table possesses sufcient residual degrees of freedom to allow failures of
MARKOV CHAIN MODELS OF COMMUNICATION PROCESSES IN NEGOTIATION 263
t to be detected. In the case of the rst-order Markov model, the table has
two residual degrees of freedom, one associated with the two-way interaction
X1.X3 and one associated with the three-way interaction X1.X2.X3. Failure of a
rst-order model to t implies that the chain is at least of second order. A
direct test of the hypothesis that the chain is second order can be carried out
in a four-dimensional table. To show this we need to show simultaneously
that the pair of three-way interactions, X1.X2.X3 and X2.X3.X4 that represent the
second-order effects in such a table are signicant, and the four-way inter-
action X1.X2.X3.X4 that represents the third-order effect is not.
Loglinear models like those described above may be tted in a generalized
linear modeling framework using a logarithmic link function and a Poisson
264 PHILIP L. SMITH, MARA OLEKALNS AND LAURIE R. WEINGART
distribution error. Generalized linear models generalize the classic linear (i.e.,
regression) model by permitting the error distribution to be non-normal and
by specifying a relationship between the linear predictor and error distribu-
tion in a more general form than that of classical regression models (see
McCullagh & Nelder, 1989; Venables & Ripley, 2002).
The rst part of Table 2 shows a t of independence and rst-order Markov
models to the rst-order data from Table 1. As can be seen, the data are best
tted by a rst-order order. The values of G2 demonstrate that the indepen-
dence model, [X1][X2][X3], fails to capture the structure of the data, whereas
the rst-order Markov model, [X1, X2][X2, X3], captures it accurately. Table 2
also shows conditional likelihood t statistics DG2. This statistic tests the improve-
ment in t (i.e., the change in G2 obtained by progressing from a simple to a
more complex model. In this case, it tests the improvement produced by mov-
ing from an independence model to a model with rst-order sequential depen-
dencies. The value of DG2 for the rst-order Markov model in the top part of
the table shows that it provides a signicantly better description of the data
than does the independence model.
The second part of Table 2 shows a t of a rst-order Markov model to
the second-order data from Table 1. The failure of a rst-order model to t
these data implies that the underlying chain is at least of second order. This
is tested directly in the third part of Table 2 which reports ts of the second-
order chain reclassied into a four-dimensional contingency table (i.e., into
sequences of length four). The rst-order Markov model, which in this table
is [X1, X2][X2, X3] [X3, X4], again fails to t. In contrast, the second-order
model [X1, X2, X3] [X2, X3, X4] accurately describes these data, in agreement
with the known statistical properties of the chain. This is demonstrated both
by the nonsignicance of the residual G2 statistic and the notable improve-
ment in t over the rst-order model, as indicated by the value of DG2.
spectively, at the end of the negotiation by the kind of outcome that negotia-
tors achieve. For example, Olekalns and Smith (2000, 2003) have considered
models in which outcomes from simulated negotiations are classied as dis-
tributive, suboptimal, optimal, or impasse, according to where the joint out-
come for the dyad fell in relation to the Pareto-optimum boundary for the
task. In experimental settings, the prospective and retrospective variables are
often the independent and dependent variables of the design and, conse-
quently, are viewed as qualitatively different from one another. However, in a
loglinear analysis framework, they are treated in an identical way. Specically,
hypotheses about any of the possible relationships among antecedent condi-
tions, communication behavior, or outcomes are tested via interaction terms
in a loglinear model.
Table 3 shows data from two simulated, rst-order, Markov chains from
subpopulations with different transition matrices. The transition matrix for the
rst subpopulation M = 0 was
0 1
[ ]
0 .7 .3 .
1 .4 .6
(7)
266 PHILIP L. SMITH, MARA OLEKALNS AND LAURIE R. WEINGART
[ ]
0 .5 .5 .
1 .2 .8
(8)
Markov chains. When testing such effects in empirical data, rather than test-
ing overall DG2 values for signicance, as we have done in Table 2, it is
preferable to base tests of improvement in t on only one of a set of redun-
dant terms to avoid inating the Type I error rate. Further discussions of these
issues may be found in Olekalns and Smith (1997, 1999, 2000) and Wasserman
and Iacobucci (1986).
Once signicant subpopulation effects have been identied in data, the next
step is to try to characterize them qualitatively. The most direct way to do
this is through the analysis of residuals. There are a number of different kinds
of residuals that have been considered in the categorical data analysis litera-
ture, but for our purposes the most relevant and easily interpretable is the
standardized residual. The standardized residual, ei, for any cell i of the con-
tingency table is dened as
= ni mi ,
ei
mi
where ni is the observed frequency of the cell and m i is the frequency pre-
dicted by the model.
The square of each standardized residual is distributed approximately as a
chi-square variable with one degree of freedom. This allows the residuals
themselves to be interpreted roughly as z-scores (i.e., the square root of a sin-
gle degree of freedom chi-square). As a rule of thumb, residuals of the order
of 2.0 can be viewed as making large or signicant contributions to the
t of a model. However, an analysis of residuals should take into account the
entire pattern of residuals, not just the large ones. Indeed, a strategy that is
often useful, especially with large tables, is to ignore the magnitude of the
residuals and to just look at their signs. This allows a researcher to discern
systematic patterns of overprediction and underprediction in the model that would
not be apparent from an exclusive focus on large residuals.
When dealing with sequences of nested models like those considered here,
one often has the problem of trying to understand why the addition of par-
ticular model terms, like subpopulation dependencies in sequential effects,
signicantly improves the t. Under these circumstances, a useful strategy is
to analyze the residuals from a model that does not include the term or terms
in question. The large residuals from this model identify those cells of the
table that make large contributions to the DG2 statistic when the term or terms
in question are added to the model.
268 PHILIP L. SMITH, MARA OLEKALNS AND LAURIE R. WEINGART
Sample Size
In practice, the complexity of the models that can be investigated using the
methods described here is limited by the size of the sample. Negotiation
experiments typically yield a corpus of a few thousand classiable responses,
but with models of even moderate complexity, this may result in sparse con-
tingency tables, in which the number of observations is small relative to the
number of cells. Goodness-of-t statistics like G 2 require samples of a
sufcient size to allow them to be tested for signicance against a chi-square
distribution. A typical recommendation is that not too many of the cells (e.g.,
not more than 20%) should have an expected frequency of less than 5. If the
majority of cells have expected frequencies of between 0.5 and 4, G2 tends to
provide too liberal a test, in which reported p values underestimate true ones
(Agresti, 1990; pp. 246250). This may result in overtting bias, in which
models that are more complex than necessary are selected as providing the
best description of the data. If the majority of expected frequencies are less
than 0.5, G2 tends to provide too conservative a test, and an undertting bias
can result.
For this reason, it is usually preferable to base inferences on the conditional
likelihood ratio tests, DG2, rather than on the t of the model to the table as
a whole. This is because the adequacy of the chi-square approximation to the
overall G2 statistic depends on the sparseness of the full table, whereas that
of the conditional likelihood ratio tests depends on the sparseness of the mar-
ginal tables on which they are computed. Because marginal tables are
obtained by collapsing the full table over one or more dimensions, they are
usually much less sparse than the original table and consequently, provide a
better basis for calculating t statistics.
Further Reading
Acknowledgements
integrative There are four aspects that actually go together. There is the
number of years of contract and the vacation, the removal can
be separated out from the package. So if, if we keep the pack-
age at six thousand for you and we have 60% removal costs,
that would be additional.
distributive No Id need more than 60% I was thinking 80% actually.
distributive What if we kept it at 60% and upped it to $8000 for the
package?
distributive I just want to get clear in my head about the vacation. Like I
said Im not prepared to relocate to Canberra unless I get more
than the two weeks.
integrative Well it depends how many years youre with us. The shorter the
time that you were with us, I would expect the less time you
would get in vacation. If you were going to be with us for six
years say youd get three weeks holiday.
distributive I would be prepared to go to Canberra for six years only for six
weeks vacation.
integrative So if we were to do say three weeks vacation, how many years
contract would you agree on?
integrative Perhaps if we get back to the package. Id be prepared to go to
Canberra for 6 years, if thats important to you. . . .
integrative Well thats related into vacation, and I know vacations impor-
tant to you because of family commitments. So we could even
look at four years and three weeks vacation each year for that
time and include the airfares in an $8000 package with 60% removal
costs on top of that.
integrative I think its a good starting point. It obviously needs some work.
My contract can be longer than four years; I think thats nego-
tiable. Canberra if its really important to you.
distributive The thing with Canberra is that weve got three positions there.
How open are you to challenge in new areas? Because we rec-
ognize that your strategic skills are applicable across any broad
range and you have demonstrated very high levels of achieve-
MARKOV CHAIN MODELS OF COMMUNICATION PROCESSES IN NEGOTIATION 271
References
Agresti, Alan (1990). Categorical data analysis. New York: Wiley.
Anderson, Theodore W., & Goodman, Leo A. (1957). Statistical inference about Markov
chains. Annals of Mathematical Statistics 28: 89109.
Bartlett, Maurice S. (1951). The frequency goodness of t test for Markov chains. Proceed-
ings of the Cambridge Philosophical Society 47: 8695.
Billingsley, Patrick (1961). Statistical methods in Markov chains. Annals of Mathematical
Statistics 32: 1240.
Bishop, Yvonne M., Fienberg, Stephen E., & Holland, Paul W. (1975). Discrete multivariate
analysis: Theory and practice. Cambridge, MA: MIT Press.
Burnham, Kenneth P., & Anderson, David R. (1998). Model selection and inference. New York:
Springer.
De Dreu, Carsten K.W., & Van Lange, Paul A.M. (1995). The impact of social value orien-
tations on negotiator cognition and behavior. Personality and Social Psychology Bulletin 21:
11781188.
De Dreu, Carsten K.W., Weingart, Laurie. R., & Kwon, Soon (2000). Inuence of social
motives on integrative negotiations: A meta-analytic review and test of two theories.
Journal of Personality and Social Psychology 78: 889905.
England, John L. (1973). Mathematical models of two-party negotiations. Behavioral Science
18: 189197.
Fienberg, Stephen E. (1980). The analysis of cross-classied, categorical data (2nd ed.) Cambridge,
MA: The MIT Press.
Hawes, Leonard C., & Foley, Joseph R. (1973). A Markov analysis of interview communica-
tion. Speech Monographs 40: 210219.
Hewes, Dean E. (1979). The sequential analysis of social interaction. The Quarterly Journal
of Speech 65: 5673.
Hoel, Paul G. (1954). A test for Markov chains. Biometrika 41: 430433.
McCullagh, Peter, & Nelder, John A. (1989). Generalized linear models (2nd ed.) London, UK:
Chapman and Hall.
Neale, Margaret A., & Bazerman, Max H. (1991). Cognition and rationality in negotiation.
New York: Free Press.
Norris, James R. (1998). Markov chains. New York: Cambridge University Press.
OConnor, Kathleen (1997). Motives and cognition in negotiation: A theoretical integration
and an empirical test. International Journal of Conict Management 8: 114131.
Olekalns, Mara, & Smith, Philip L. (2003). Testing a three-way relationship among negotia-
tors motivational orientations, strategy choices, and outcomes. Journal of Experimental
Social Psychology 39: 107117.
272 PHILIP L. SMITH, MARA OLEKALNS AND LAURIE R. WEINGART
Olekalns, Mara, & Smith, Philip L. (2003). Social motives in negotiation: The relationship
between dyad composition, strategy sequences and outcomes. International Journal of
Conict Management 14: 233254.
Olekalns, Mara, & Smith, Philip L. (2000). Negotiating optimal outcomes: The role of strate-
gic sequences in competitive negotiations. Human Communication Research 24: 528560.
Olekalns, Mara, & Smith, Philip L. (1999). Social value orientations and strategy choices in
competitive negotiations. Personality and Social Psychology Bulletin 25: 657668.
Olekalns, Mara, & Smith, Philip L. (1997). Understanding optimal outcomes: The role of
strategic sequences in competitive negotiations. Working Paper No. 110, Department of
Management & Industrial Relations, University of Melbourne.
Olekalns, Mara, Smith, Philip L. & Kibby, Rachel (1996). Social value orientations, negotia-
tor strategies and outcomes. European Journal of Social Psychology, 26: 299313.
Ross, Sheldon M. (2000). Introduction to probability models (7th ed.). San Diego, CA:
Academic Press.
Thomas, Andrew (1985). Conversational routines: A Markov chain analysis. Language and
Communication 5: 287296.
Thomas, Andrew, Roger, Derek, & Bull, Peter (1983). A sequential analysis of informal
dyadic conversation using Markov chains. British Journal of Social Psychology 22: 177188.
Ting-Toomey, Stella (1983). An analysis of verbal communication patterns in high and low
marital adjustment groups. Human Communication Research 9: 306319.
Wasserman, Stanley, & Iacobucci, Dawn (1986). Statistical analysis of discrete relational
data. British Journal of Mathematical and Statistical Psychology 39: 4164.
Wasserman, Staney, & Weaver. Sheila O. (1985). Statistical analysis of binary relational data:
Parameter estimation. Journal of Mathematical Psychology 29: 406427.
Weingart, Laurie R., Olekalns, Mara & Smith, Philip L. (2004) Quantitative coding of nego-
tiation behavior. International Negotiation 9 (3).
Weingart, Laurie R., Prietula, Michael J., Hyder, Elaine, & Genovese, Christopher (1999).
Knowledge and the sequential processes of negotiation: A Markov chain analysis of
response-in-kind. Journal of Experimental Social Psychology 35: 366393.
Venables, William N., & Ripley, Brian D. (2002). Modern applied statistics with S-Plus. (4th
ed.). New York: Springer.
All that Glitters is Not Gold:
Examining the Perils and Obstacles in Collecting Data
on the Internet
With the end of the semester looming ahead, I received notication that the
subject pool in the department has run dry again. Without other economical
avenues to recruit participants, I am resigned to the fact that my research has
to be shelved for the time being, until the new semester begins and until the
subject pool is open for use . . .
Many researchers in small and mid-sized psychology programs around the
world who do not have large research funding and who lament the difculty
of obtaining data are familiar with the above scenario. Unfortunately, nothing
is more disruptive and frustrating to an enthusiastic researcher when his
progress is constantly hindered by such data collection difculty. There is
intense competition for participants when the subject pool is open for use to
a large group of undergraduates, graduates and faculty members. In my
department, for instance, about 70 projects are vying for a limited pool of
about 1000 introductory psychology students in each academic year. Given
that each introductory student contributes ve hours of his/her time for
research participation, this averages to about 70 subject hours per project a
number that is hardly sufcient for a 2 2 between-subjects experimental
design!
Fortunately, there is now a real possibility that such woeful tales of in-
sufcient data and slow research progress will largely disappear as a result of
the internet revolution. The internet offers almost instantaneous access to a
world of participants. With a continual increase in connectivity and simultane-
ous decrease in connection cost due to advent in technology and economy of
scale, the number of users for the World Wide Web is growing at a rapid pace.
From a widely cited gure of about 30 million people world-wide connected
to the internet in 1995 (GVUs 7th WWW User Survey 1995), this estimation
has witnessed a steady increase. For instance, Couper (2000) cited a study by
the Strategies Group that reported the number of Americans connected to the
internet at the end of 1998 to be around 84 million and this gure further
increased to 106.1 million by December 1999. This large pool of internet users
makes the medium practically a goldmine for those involved in the data col-
lection endeavor, whether their aim is for entertainment (e.g., polling websites
such as misterpoll.com or news website such as CNN Quickvote), commercial
(e.g., marketing research companies or the on-line marketing research depart-
ments found in many companies), or serious academic research purposes.
In addition to the virtually unlimited pool of potential participants on the
internet, the obvious benets of using internet as the new medium for
research are many. At the top of the list, many lauded the ease in identifying
special population groups (e.g., homosexuals, identical twins, person with dis-
ability or special illnesses) through the help of newsgroups or listservs (auto-
mated mailing lists). This provides researchers the capability to conduct
research that they would otherwise nd too costly, time-consuming, and hence
impossible, to carry out a decade ago. Once the survey or experiment is
posted on a website, the data collection process runs itself like an automaton,
without the need to employ assistants to administer repeated data collection
sessions that stretch over weeks or months. There is also no need to cover for
the potentially large cost of papers, envelopes, postage, and for copying mate-
rials. Transcription errors are minimized, if not entirely eliminated, when one
can program for the transfer of data directly to a spreadsheet or statistical pro-
grams for analysis. Most attractive of all, imagine your data bank increases
continuously 24 hours a day even while you sleep. It is as exhilarating an
image that a researcher could dream of in terms of data collection! For
instance, Pettit (1999) collected over 800 responses in 21 days for a Computer
Anxiety Scale she puts on the Web. Similarly, Epstein and Klinkenberg
(2002) collected 1116 completed surveys in their rst month of the data col-
lection period. Nancarrow, Pallister and Brace (2001) succinctly summarized
the above obvious advantages as the speed, ease and cost appeals of con-
ducting an internet-based study.
Considering the appeals of internet-based study, the temptation for most
researchers would be to jump in the fray immediately. If the Web is akin to
a goldmine of easily available data, why wait? After all, an on-line study is
merely an electronic implementation of what used to be a paper-and-pencil
task a conversion that could be easily accomplished by someone literate in
programming language. Or is this not the case? What are the difculties of
implementing an internet-based study? What are the hidden threats to valid-
ity that may render data collected from such a medium unusable? What are
the related issues (e.g., ethics) that ought to be tackled with the introduction
of this new medium of research? Are there any ramications for the current
deluge of requests for research volunteers on the internet?
The major objective of this article is to answer the above questions, by
pointing out a few of the major problems encountered by researchers con-
ALL THAT GLITTERS IS NOT GOLD 275
The two most commonly employed research methods on the internet are sur-
veys, followed by experiments. The discussion of problems related to inter-
net-based study will be made relative to these two research methods.
One reason why surveys tend to be the more commonly employed method-
ology on the internet is due to the relative ease in adapting paper-and-pencil
forms of a survey instrument to the on-line format. On the other hand, exper-
iments conducted on the internet typically require additional interactivity (e.g.,
writing scripts for random assignment of participants to various conditions)
and hence additional demands on programming (refer to Birnbaum (2000) for
a discussion of the computer techniques used for internet-based study).
However, internet-based research has witnessed a steady increase in recent
years, as technological barriers are gradually lowered through the increased
availability of more user-friendly Web programming language and a concerted
276 CHA YEOW SIAH
growth of internet-access across all segments of the society. The sample ob-
tained by Gosling et al. (2004), for instance, compares favorably in terms of
gender and race composition to those obtained by Epstein and Klinkenberg
(2002) two years earlier.
Second, unlike telephone or mail surveys where we can construct an ade-
quate frame population for sampling based on phone or mail directories, there
are still tremendous difculties at the current stage to obtain an adequate
frame population for internet-based study from which to make statistical infer-
ences. Technological innovation, despite its far-reaching impact, will take time
to become a constant in all household. Given that the internet medium has
not achieved the same level of penetration as the telephone or the television,
we are therefore still missing a considerable proportion of people in the tar-
get population (to which one wants to make inference) from the frame popu-
lation, creating what is known as coverage error. This prevents researchers
from carrying out probability-based sampling (Couper 2000; Truell, Bartlett
and Alexander 2002), which in turn disallows any precise population esti-
mates to be made due to the unknown standard error of measurement.
Therefore, though the internet provides an efcient medium for a large
amount of data to be collected quickly and easily, the price to pay for such
conveniences may currently still be too high for survey researchers when they
are unable to make population estimates from the data. This leads researchers
such as Gonzalez (2002) to exercise extra care when interpreting results from
internet-based surveys.
For experimental research on the internet, the advantage of yielding a het-
erogeneous sample seems persuasive considering that the most common crit-
icism on psychological research is its over-reliance on college student samples
(McNemar 1942; Gosling et al. 2004). Thus, having a heterogeneous sample
is a commonly cited reason in favor of Web experimentation (Montgomery
and Ritchie 2002; Reips 2000), with the assumption that increased hetero-
geneity leads to increased external validity. However, one should refrain from
exaggerating the importance of external validity in experimental research. To
the extent that experimental research should rightfully be more concerned
with theory-testing and not with establishing population estimates, obtaining
a representative sample is secondary to the aim of ensuring internal validity
because the former is not necessary for causal hypothesis testing (Berkowitz
and Donnerstein 1982; Cook and Campbell 1979). In fact, the heterogeneous
nature of a Web sample opens itself to the criticism that it invariably increases
error variance (Greenberg 1987), and thus constitutes a weakness instead. In
response, Birnbaum (2004b) argued that the large increase in sample size
from an internet-based study is probably sufcient to counter the added error
278 CHA YEOW SIAH
Subject Fraud
While the internet empowers researchers with the ability to collect data from
all over the world, it is also a double-edged sword that handicaps researchers
in aspects that are otherwise easily controllable in face-to-face administration
of survey or experiment. A prime example would be subject fraud, which is
likely one of the most intractable problems in internet-based studies. Subject
fraud refers to participants lying on critical demographic information (e.g.,
sex, race, age, etc.) in a survey or making multiple submissions that compro-
mise the data integrity.
The consequence of lying on demographic information in a survey is that
it causes erroneous classication that results in inaccurate generalization.
Though the problem also exists in conventional research setting (e.g., a res-
pondent may still lie about his age in a telephone survey), the danger is con-
siderably greater in internet-based study because of the anonymous and
faceless setting. Similarly, multiple submissions, one of the most commonly
reported problems in the literature, create havoc during data analysis because
it violates the independent observation assumption of statistical techniques.
Multiple submissions are more likely to concern researchers who employ
personality inventories or intelligence tests on the internet. This is because
participants may be tempted to try again for curiositys sake, just to see how
changing one response would affect their overall score. Birnbaum (2004a)
suggested a list of solutions for multiple submissions though none of them is
foolproof. These solutions can be classied generally into proactive iden-
tication measures such as checking of Internet Protocol (IP) addresses, and
preventive ones which mainly focus on removing the incentive to participate
more than once.
Popular measures such as checking for repeated IP addresses have limita-
tions because the repetition could be due to different participants using the
ALL THAT GLITTERS IS NOT GOLD 279
Thus far, the picture that emerged from a comparison of results obtained
from internet-based versus laboratory studies has been very encouraging. For
instance, Krantz and Dalal (2000) concluded that the two methods yield very
similar results based on his review of nine studies that compared the two ver-
sions. Likewise, Gosling et al. (2004) reached the same conclusion when they
compared the score reliabilities and discriminant validities of scales derived
from their Web sample with those obtained by John and Srivastava (1999).
There are indeed good reasons to believe that mounting a study on the
Web enhances the reliability and validity of its results. In surveys, missing
responses and response errors are likely to become a thing of the past when
researchers capitalize on the power of programming to prevent respondents
from proceeding further until they provide a response. Programming to high-
light an improbable response that lies outside the reasonable range of a value
could also help to inform participants of their mistakes. In experiments con-
ducted on the internet, demand characteristics are minimized because there is
no longer direct contact between the experimenter and the participants (Piper
1998). This also prevents subtle experimenter bias from occurring because it
eliminates the possibility that participants may respond to certain attributes
(sex, age, race, physical appearances, and mannerism) of the experimenters
(Hewson, Laurent and Vogel 1996). Reips (2001) and Murray & Fisher (2002)
also argued that the increased social distance from the experimenter reduces
social desirability concerns and allows respondents to be more truthful in their
responses.
On the other hand, conducting a study on the internet introduces new con-
cerns. For instance, Montgomery and Ritchie (2002) found during the testing
of their experiment program that as many as one-fth of their respondents
either failed to read the directions on the screen or simply could not gure
out what they were supposed to do next. Whereas an experimenter present in
the laboratory could easily explain the task to make sure that any confusions
are cleared up, this is no longer possible for studies conducted on-line. A par-
ticipant who is unsure what he is supposed to do next will submit random
responses that are difcult for researchers to detect. This makes pilot testing
and performing additional checks before launching the study extra important
for researchers who wish to conduct internet-based study.
In addition, the hardware and software employed in internet-based study
also introduces extraneous inuences on the data. For instance, the use of
differently congured computer platforms, browsers and equipments (e.g.,
different sized computer monitors) lead to compatibility issues and varying
presentation (Buchanan and Smith 1999) that cannot be controlled by researchers.
Besides, the different connection speed among the participants may also lead
ALL THAT GLITTERS IS NOT GOLD 281
experiment. The results imply that participants may become more skeptical of
experimental procedures as a result of not seeing other real participants or
experimenters around them, thereby weakening treatments designed by the exper-
imenters. The possible attenuation of effect size means that internet-based
researchers should be even more attentive to the issue of design and power.
Finally, one other problem that plagues researchers of internet-based study
is the lack of control over the environmental factors in which participants take
part in the study (Buchanan and Smith 1999; Murray and Fisher 2002; ONeil
and Penrod 2001). It is no longer possible for researchers to provide a mini-
mally distractive environment for participants when the study is carried out
on-line. For instance, experimenters have little knowledge and control over
the mental state of the participants (e.g., whether the participant is alert,
intoxicated or overly tired), the time and day at which the data is collected,
the level of background noises (e.g., music may be blaring in the back-
ground), and the kind of activities that participants may be carrying out con-
currently during their participation (e.g., eating or watching television
programs). It is a serious problem, particularly for experiments, because the
inability to control such environmental factors introduces substantial error
variances that affect the results in an unknown and signicant manner. Can
the problem be solved currently? Unfortunately, it seems that the answer is
no. Can it be solved in the future? The answer is a hopeful yes when video
data in addition to text data could be transmitted to provide more information
to the experimenter. But until the technology is widely adopted, researchers
conducting internet-based study may have to live with larger random error
that has the possibility of masking the true treatment effects.
Thus there are potential advantages and disadvantages when conducting
an internet-based study. The possible cost of a weaker treatment effect and a
lack of control in an experiment and the possible benets of increasing self-
disclosure and decreasing social desirability have different implications for
researchers conducting experiments or survey on the internet. In sum, there
are strong reasons to believe that subtle differences between electronic and
paper-and-pencil implementation of surveys and experiments produce dynam-
ics and different experiences within individual participants that would interact
in an unintended manner with the investigated variables. However, such issues
have not been given sufcient attention to date. More empirical research in
this area is necessary for researchers to probe in-depth on the list of factors
mentioned.
284 CHA YEOW SIAH
Due to the recency of internet-based research, it is not surprising that the eth-
ical arena for such methodology resembles the Wild West with no clearly set
rules and guidelines for researchers. The latest revision of the American
Psychological Association (APA) Ethics Code (2002), for instance, provides
little help in guiding researchers who wish to make sure what they do falls
within the boundary of the Ethics Code. Thus, the approval of research con-
duct for internet-based studies is largely left to the discretion of individual
Institutional Review Board (IRB) which in some cases, may lead to different
adoption of standards. For instance, while the IRB from which Epstein &
Klinkenberg (2002) sought approval eventually allowed them to set cookies
in participants browsers to avoid subject fraud, the same measure was not
approved for reasons of intrusion to privacy by the Human Subjects Research
Review Committee to which Montgomery & Ritchie (2002) belongs.
The issue is an important one because there are many gray areas involved
in the conduct of internet-based research, due to the fact that the medium rep-
resents largely uncharted ground for researchers. Many of these gray areas are
found in domains such as subject recruitment, obtaining informed consent,
protecting the well-being and privacy of participants, and ensuring data
integrity. For instance, what are considered legitimate ways for researchers to
recruit participants from internet or specic newsgroups? What are the dos
and donts that researchers should take note of in order to protect participants
anonymity and privacy? These various concerns need to be addressed in order
for internet-based research to continue to ourish.
To this end, experiences shared by various researchers (e.g., Buchanan and
Smith 1999; Montgomery and Ritchie 2002; Pettit 1999) and related discus-
sions on ethical issues (e.g., Smith and Leigh 1997) have served as useful
guides in steering future researchers in the right direction. In particular, the
set of 30 guidelines provided by Michalak and Szabo (1998) lled a much
needed gap caused by the lack of code of conduct. In contrast to traditional
research methodology, there is an even greater need to continually update the
ethical codes for internet-based research because of the rapid advance of
technology.
Conclusion
from volunteering for studies conducted on the internet. When this happens,
we may nd ourselves staring into empty goldmines similar to the ones found
at the end of the Gold Rush Era.
Note
1. For more on email- and internet-based negotiation approaches, see International Negotiation
9(1) and 9(2), thematic issues that focus on the impact of technology on negotiation
processes.
References
Stanton, Jeffrey M. and Rogelberg, Steven G. (2002). Beyond Online Surveys: Internet
Research Opportunities for Industrial-Organizational Psychology, in Steven G. Rogelberg,
editor, Handbook of research methods in industrial and organizational psychology. Malden,
MA: Blackwell Publishers.
Truell, Allen D., Bartlett, James E., II and Alexander, Melody W. (2002). Response rate,
speed, and completeness: A comparison of internet-based and mail surveys. Behavior
Research Methods, Instruments & Computers, 34: 4649.
Turner, Charles F., Ku Leighton, Rogers, Susan M., Lindberg, Laura D., Pleck, Joseph H. and
Sonenstein, Freya L. (1998). Adolescent Sexual Behavior, Drug Use, and Violence:
Increased Reporting with Computer Technology. Science, New Series, 280, 5365: 867873.
The Method of Experimental Economics
RACHEL CROSON
Introduction
The 2002 Nobel Prize in Economics was awarded to Daniel Kahneman (an
experimental psychologist) and Vernon Smith (an experimental economist).
This award acknowledges an important trend in economics the growth of
experiments as a valid, accepted methodology and the inuence of psycho-
logical research in that growth.1
There are many similarities between experimental economics and psycho-
logical research. Researchers in both elds are concerned with similar sub-
stantive areas; bargaining in economics and negotiation in psychology, public
goods provision in economics and social dilemmas in psychology, just to
name a few. The two elds also share many methodological practices. Both
elds use convenient populations (like undergraduate students) as participants
in their experiments. Both elds elicit decisions or introspections from their
participants and use those to learn about the world. Both elds are concerned
with careful experimental design, avoiding demand effects, and appropriate
statistical analyses.
However, the objectives of psychology and economics experiments are
often different. Economics experiments are designed to address economic the-
ories; psychology experiments are designed to address psychological theories.
This distinction may seem obvious at rst, but it has important and often
unforeseen implications for methodological differences in the two elds.
These methodological differences are the subject of this article. I will discuss
ve areas where experimental economists and experimental psychologists dif-
fer. They are ordered from most to least important for economists, although
readers may disagree with my ordering.
This is by no means the rst article on experimental economics methodol-
ogy. However, most previous articles have been aimed at traditional (non-
experimental) economists to argue that the experimental methodology is a relevant
and valid one (e.g. Binmore 1987, 1999; Plott 1982, 1991a, 1991b, 1994;
Roth 1986, 1991, 1994; Smith 1976, 1989, 1994). Friedman and Sunder
practice and rationale (much less thoroughly). My hope is that this article will
lead not only to a deeper understanding of each elds choice of methodol-
ogy, but also to some practical advice to psychologists on how to have their
work read and accepted by economists. I begin with perhaps the largest gulf
between economic and psychological experiments: incentives.
Incentives
Economic theories describe and predict decisions individuals will make in the
presence of payoffs. Typically, theories specify the payoffs from taking one
action or another. For example, individuals contributing to a public good
(cooperating in a social dilemma) incur some private cost from doing so
which outweighs the private benet they receive from the public good. Since
earnings are higher when individuals defect than when they contribute, theory
predicts they will defect.
It is critical for theory testing that the participants actually face the payoffs
assumed by the theory. The fact that individuals cooperate in social dilemmas
when there are no payoff consequences from their actions is simply not infor-
mative. Economic theory makes no predictions of what individuals will say
they would do, only what they will actually do when faced with a given deci-
sion and the resulting payoffs. For example when studying ultimatum bar-
gaining it is important to have participants actually earn the amounts that their
decisions imply (e.g. Croson 1996, Forsythe et al. 1994).
This has led to the practice of induced valuation in experimental econom-
ics (Smith 1976), where participants compensation is not simply positive, but
importantly is responsive to the choices they make in a way that is consistent
with the theory being tested. This practice replaces the at-fee payments more
common in psychology experiments, where individuals are paid some amount
for their participation (perhaps zero, or perhaps earning extra credit in their
course) which is not contingent on the decisions they make.
For example, when two psychologists rst identied the disjunction effect
(nonconsequential reasoning), their papers included data from surveys and
hypothetical questions (Shar and Tversky 1992; Tversky and Shar 1992).
Participants were asked about gambles they might take after having won or
lost a previous gamble or vacations they might purchase after having passed
or failed their exams. When an economist wanted to further explore the
boundaries of the effect, participants played games (prisoner dilemmas, pub-
lic goods, and games of iterated dominance) and were paid based on their
decisions and the decisions of their counterparts (Croson 1999, Croson 2000).
292 RACHEL CROSON
This is not to say that all psychologists use unpaid participants many run
experiments with incentives.5 But all economics experiments involve salient
payments to participants. This practice is considered critical to the validity of
the experiment and the objective of testing the theory being addressed.
Typically, payments in economics experiments are made in cash directly
after the experiment. There are a few reasons for the use of cash. First, every-
one values it, in contrast with extra-credit points or other grade-related
rewards which may be valued only by students who are grade-conscious
and/or whose grade may be affected by the outcome. Second, cash is one
good that is non-satiable more is always better.
That said, some notable experiments use forms of payment other than cash.
All, however, use payments that are contingent on individuals decisions and
argue (successfully) that the payments involved are consistent with the theo-
ries being tested. For example, in Boyce et al. (1992) the authors want to
elicit willingness-to-pay and willingness-to-accept values for specic environ-
mental damages. They use dollars for the WTP and WTA measures, but for
the environmental damage they bring a baby tree into the lab that the exper-
imenter will chop down if the damage must be incurred. This tree-killing
action captures nicely the theorys assumptions of incurring environmental
damage and is an excellent example of non-nancial rewards that are none-
theless contingent on the participants actions.
A second issue in the area of incentives is the amount participants receive.
A number of papers have argued that participants must be paid enough to
compensate them for their time (thus, the average payoff should translate into
an hourly wage roughly equal to the salaries of on-campus jobs), and for the
thinking costs they incur during the experiment (Smith and Walker 1993).
One natural question is the extent to which paying participants contingent
on their actions affects outcomes. There are now a number of meta-analyses
on this question. Camerer and Hogarth (1999) review 74 studies and conclude
that nancial incentives have a large effect on judgment and decision tasks,
but the effect is smaller in games and markets. Hertwig and Ortmann (1998)
review 10 studies from the Journal of Behavioral Decision Making with and
without incentives. They nd that incentives decrease framing effects, bring
auction bids closer to optimality and eliminate preference reversals. They also
identify contexts when payment had no effect, specically condence judg-
ment and information acquisitions. For some tasks contingent payment does
not seem to affect the mean outcomes, while for others it does. There is gen-
eral agreement, however, that contingent payment reduces the variance of
responses one receives. Smith and Walker (1993) survey 31 experimental eco-
THE METHOD OF EXPERIMENTAL ECONOMICS 293
nomics studies varying incentives and conclude that in virtually all cases,
rewards reduce the variance of the data around the predicted outcome (245).
Paying contingent (and sufcient) payments can get expensive. There are a
number of tricks that economists play in order to stretch limited funding.
First, if participants in the experiment are playing the same game repeatedly,
researchers sometimes choose one round at random for real payment. Of
course, participants are told in advance that one round will be randomly cho-
sen (see the section on deception, below), but not which one. This payment
design has the advantage of avoiding wealth effects in which participants who
are accumulating money throughout the session begin making riskier and
riskier decisions. Similarly, for experiments with large numbers of partici-
pants, researchers sometimes choose one participant (or one dyad or one
group) at random for real payment. Again, participants are told in advance
that one will be chosen at random. While both these techniques have the
potential to reduce expenditures, experimental economists using these designs
typically make the expected earnings of any given participant equal to the
wage rate.
This is not to say that economists never use any data collected from a sur-
vey or hypothetical responses. Many experiments in economics use a post-
experimental questionnaire where participants are asked to introspect as to
why they made the decisions they made. But economists are very cautious
when reporting and interpreting this data, and always ask these questions
after they have collected the real-money decisions of interest. The perception
is that responses to questions like this are cheap talk and may have no rela-
tion to what is actually in the minds of experimental participants. Personally,
I consider this data like any other, taking into account the motivations of the
subjects including honest reporting, impression management and other factors
in interpreting these results. This data is often particularly useful in going
beyond documenting an outcome and digging deeper into the motivations
causing it.
My recommendation to psychologists who want economists to use their
work and to cite their results is simple pay your participants. And pay them
not a at fee, but an amount contingent on their decisions. This is particularly
important if the experiments results are inconsistent with economic theory. It
is far too easy (and common) for economists to simply disregard experimen-
tal results inconsistent with their theories because the participants werent
appropriately incentivized.
294 RACHEL CROSON
Context
need to argue that the results are not being driven by the particular context
chosen or, even better, demonstrate the same effect in multiple (sufciently
different) contexts.
Subject Pools
Deception
only enhances the likelihood of recruiting participants who enter the experi-
ment expecting deception and further weakens the link between actions and
earnings.
The norm against deception is quite strong in experimental economics an
economic journal will not publish an experiment in which deception was
used, while in psychology journals, deception is commonplace and accepted.
The closest the experimental economics eld has come to allowing deception
is surprising participants with additional decisions when they initially believed
the experiment had ended (e.g. Andreoni 1988; Croson 1996; Boles, Croson
and Murnighan 2000; Croson, Boles and Murnighan (2003)). Note that this
does not involve deception by commission; participants are never told any-
thing that is not true. Instead, it involves deception by omission; participants
are not told everything about the experiment when they begin.
For psychologists who want their work to be read and accepted by econo-
mists, deception will prove to be a difcult barrier to overcome. If its simply
a matter of saving money, economists react very skeptically to experiments
that use deception. If the deception is necessary for some other reason, an
explanation and justication will be needed in order to have the data taken
seriously.
Experimental Procedures
These last two topics differ from the previous four. Here, I describe a collec-
tion of practices in the implementation of economics experiments, rather than
globally-accepted wisdom about experimental design. Many of these proce-
dures are commonly used both psychology and economics.
Experimental projects begin with an application to ones human subject
committee. Experimental economists tend to have relatively little trouble with
human subjects committees, as many of the usual ags are not present in
their designs. The experiments do not involve deception, there are rarely any
consequences other than nancial, and participants earn money for their par-
ticipation. Little or no demographic data is collected, and participants are paid
privately (see below), reducing the impact of social comparisons on earnings.
Thus many projects receive expedited review and quick acceptance.
The most troubling constraint for experimental economists imposed by
human subject committees is guaranteeing positive earnings. Economists often
want to explore decision-making in the face of losses rather than gains, but
human subject committees (and other concerns) prevent us from taking money
from our participants. There are a few alternative responses which have been
THE METHOD OF EXPERIMENTAL ECONOMICS 299
different scales in the quiz and in the experiment, by making the quiz abstract
rather than concrete, or by allowing participants to ll in their own numbers
for their decisions. Then we ask participants to calculate their earnings (and
sometimes, their counterparts earnings). Experimental monitors circulate through-
out the room and check the answers, correcting and explaining for anyone
who completed it incorrectly.
A second issue that arises during experiments is how to deal with partici-
pant questions. Especially while reading instructions aloud, participants will
often have questions about the procedures. These questions you would like to
have asked and answered publicly. However, other participants ask leading or
contaminating questions (e.g. why dont we just all cooperate so we can earn
more money?). These are attempts to inuence the actions of others dis-
guised as questions, and these the experimenter doesnt want asked and
answered publicly. One tactic I nd useful is to ask those with questions to
raise their hand. The experimenter can then go to the participant and hear
the question privately. If the question is the type that should be publicly
addressed, the experimenter can repeat the question and answer it publicly. If
not, the question can be answered privately.
A nal issue that often arises during experiments involves randomization.
While it is now technologically very easy to use computers to generate ran-
dom numbers, participants are often skeptical about the unbiasedness of this
method. Thus economics experiments typically use actual randomizing de-
vices (e.g. dice as in Croson and Johnston 2000) to implement random out-
comes.
Once the experiment is over, participants receive their earnings (in cash)
and are asked to sign a receipt before they leave. In many economics exper-
iments it is important that these earnings be paid privately. Theories often
assume that individuals value money absolutely, without social inuence or
comparative preferences. If we want to test the implications of such a theory,
it is important to implement an experimental design where these other con-
siderations are not present. Of course, one can also test the assumptions of
such a theory, at which point knowing the earnings of others is an important
design considerations. When payments are private, participants often want to
know how well they did relative to others. I try hard to avoid answering that
question, reminding them that earnings are private information.
A nal end-of-experiment consideration involves debrieng. As with our
rst topic (human subject committees) extensive debrieng is not typically
needed when experiments involve no deception. Economists are concerned
about communicating the nature of the experiment to participants when
others will be coming to the lab later in the day or the week to participate
THE METHOD OF EXPERIMENTAL ECONOMICS 301
Data Analysis
A nal methodological dimension along which the elds differ is the data
analyses they employ. In my publications in both economics and psychol-
ogy/management journals, I have found referees and editors surprisingly parochial
about their favored statistical methods.
Both sides use nonparametric statistics, although economists favor Wil-
coxon tests (Mann-Whitney U tests) after an inuential paper demonstrating
its power in ultimatum and dictator game data (Forsythe et al. 1994).6 In con-
trast psychologists tend to use chi-squared and other tests that are appropriate
for discrete data. Once the decision is made to move to parametric analysis,
economists use regressions and psychologists ANOVAs. I have observed both
these choices even when not appropriate economists regularly use regres-
sions even when the independent variables are all discrete and psychologists
regularly discretize their continuous independent variables in order to be able
to use ANOVAs. These two techniques are, of course, quite related to each
other and I have never experienced a situation where they yield qualitatively
different responses. My interpretation of the source of these differences has
to do with the reference disciplines to which the various experimentalists are
trying to speak. For example, empirical economists use regressions; thus
experimental economists use regression techniques to convince them that
experimental results are valid.
A second methodological difference in analysis (and experimental design)
is the concept of interactions. Economists are often uncomfortable with
interaction effects (especially three-way or four-way interactions) while
for psychologists interactions are the bread-and-butter of publication. My
302 RACHEL CROSON
Conclusion
Notes
1. This article will be discussing the methodology of experimental economics and providing
comparisons with that of experimental psychology. I mean the latter term to apply quite
broadly and include research in management, dispute resolution and other related elds that
use experimental data and experimental psychological methods.
2. While many (perhaps almost all) economists acknowledge that the theory does not predict
and explain economic outcomes in all settings, the general consensus is that the theory does
exceptionally well in predicting and explaining in a large variety of settings. Furthermore,
there are costs and benets of making the theory more complex. The benets are clear, by
adding extra parameters you can increase the predictive ability of the theory (e.g. by adding
a parameter for the status-quo point around which the utility [value] function is asymmet-
ric you can capture different risk preferences in gains and losses, or by adding others pay-
offs into ones own utility function you can capture social factors like altruism, envy and
inequality-aversion). However, there are also costs to adding extra parameters. The parame-
ters need to be estimated in any given situation, and when this is not possible the expanded
theory makes no predictions. So when an expanded theory makes predictions they are more
often correct, but there are fewer cases where any predictions are possible. Furthermore, the
expanded theory is more complicated to work with and may not provide as many insights
or unexpected predictions. Psychologists and economists have made different choices in this
tradeoff between simplicity and descriptive ability.
3. The way to implement an innitely repeated game in the lab is to derive the theorys pre-
dictions under a discount rate of d, then implement the discount rate by setting the proba-
bility of the game ending in any given period equal to (1-d).
4. The way to implement a one-shot game is through a strangers or zipper design where each
participant meets each other participant at most once during the experiment. (Kamecke
(1997), Andreoni (1988), Croson (1996b), Andreoni and Croson (forthcoming)).
304 RACHEL CROSON
5. Shar and Tversky (1992) did include some studies on the prisoners dilemma game in
which participants were paid for their earnings. Morris, Sim and Girotto (1998) ran similar
experiments with a at fee.
6. This paper also demonstrated signicant differences between behavior in paid and unpaid
ultimatum and dictator games.
References
Andreoni, James (1988). Why Free Ride? Strategies and Learning in Public Goods Experi-
ments. Journal of Public Economics, 37: 291304.
Andreoni, James and Croson, Rachel (forthcoming). Partners versus Strangers: Random
Rematching in Public Goods Experiments, in Plott, C. and Smith V., editors, Handbook of
Experimental Results. New York: Elsevier.
Binmore, Kenneth (1987). Experimental Economics. European Economic Review, 31, 1/2:
257264.
Binmore, Kenneth (1999). Why Experiment in Economics? Economic Journal, 109, 453: F1624.
Boles, Terry; Croson, Rachel; Murnighan, Keith (2000). Deception, Retribution, and the Negotiation
Process. Organizational Behavior and Human Decision Processes, 83: 235259.
Bonetti, Shane (1998a). Experimental Economics and Deception. Journal of Economic Psychology,
19, 3: 377395.
Bonetti, Shane (1998b). Experimental Economics and Deception: Response. Journal of Economic
Psychology, 19, 3: 411414.
Boyce, Rebecca; Brown, Thomas, McClelland, Gary, Peterson, George and Schulze, William
(1992). An Experimental Examination of Intrinsic Values as a Source of the WTA-WTP
Disparity. American Economic Review, 82, 5: 13661373.
Buchan, Nancy; Croson, Rachel; Dawes, Robyn (2002). Swift Neighbors and Persistent Strangers:
A Cross-Cultural Investigation of Trust and Reciprocity in Social Exchange. American Journal
of Sociology, 108: 168206.
Camerer, Colin; Hogarth, Robin (1999). The effects of nancial incentives in experiments: A
review and capital-labor production theory. Journal of Risk and Uncertainty 19, 13: 742.
Carter, John; Irons, Michael (1991). Are Economists Different, and If So, Why? Journal of
Economic Perspectives, 5: 171177.
Cox, James; Isaac, Mark R. (1986). Experimental Economics and Experimental Psychology:
Ever the Twain Shall Meet? in Economic Psychology: Intersections in Theory and Applica-
tion, MacFadyen and MacFadyen, editors, North-Holland, Amsterdam: 647669.
Croson, Rachel (1996a). Information in Ultimatum Games: An Experimental Study. Journal
of Economic Behavior and Organization, 30: 197212.
Croson, Rachel (1996b). Partners and Strangers Revisited. Economics Letters, 53: 2532.
Croson, Rachel (1999). The Disjunction Effect and Reason-Based Choice in Games.
Organizational Behavior and Human Decision Processes, 80: 118133.
Croson, Rachel (2000). Thinking like a Game Theorist: Factors Affecting the Frequency of
Equilibrium Play. Journal of Economic Behavior and Organization, 41: 299314.
Croson, Rachel (2002). Why (and How To) Experiment: Methodologies from Experimental
Economics. University of Illinois Law Review, 2002: 921945.
Croson, Rachel; Boles, Terry; Murnighan Keith (2003). Cheap Talk in Bargaining Experi-
ments: Lying and Threats in Ultimatum Games. Journal of Economic Behavior and Organization,
51: 143159.
THE METHOD OF EXPERIMENTAL ECONOMICS 305
Croson, Rachel; Gneezy, Uri (2005). Gender Differences in Preferences. Working Paper, Wharton
School University of Pennsylvania.
Croson, Rachel; Johnston, Jason (2000). Experimental Results on Bargaining under Alternative
Property Rights Regimes. Journal of Law, Economics and Organization, 16: 5073.
Croson, Rachel; Mnookin, Robert. (1997). Does Disputing through Agents Enhance Coope-
ration? Experimental Evidence. Journal of Legal Studies, 26: 331345.
Cummings, Ronald; Holt, Charles; Laury, Susan (2002). The Georgia Irrigation Reduction
Auction. Working Paper, Georgia State University, Department of Economics.
Dyer, Douglas; Kagel, John; Levin, Dan (1989). Comparison of Naive and Experienced
Bidders in Common Value Offer Auctions: A Laboratory Analysis. Economic Journal, 99,
394: 108115.
Dyer, Douglas; Kagel, John (1996). Bidding in common value auctions: How the commer-
cial construction industry corrects for the winners curse. Management Science, 42, 10:
14631476.
Eckel, Catherine; Grossman, Phillip (forthcoming a). Volunteers and Pseudo-Volunteers: The
Effect of Recruitment Method on Subjects Behavior in Experiments, Experimental
Economics.
Eckel, Catherine; Grossman, Phillip (forthcoming b). Differences in the Economic Decisions
of Men and Women: Experimental Evidence, In Plott, C. and Smith, V. Handbook of
Experimental Results. New York: Elsevier.
Forsythe, Robert; Horowitz, Joel; Savin, N.E.; Sefton, Martin (1994). Fairness in Simple Bargaining
Experiments. Games and Economic Behavior, 6: 347369.
Frank, Robert; Gilovich, Thomas; Regan, Dennis (1993). Does studying economics inhibit
cooperation? Journal of Economic Perspectives 7: 159171.
Frank, Robert; Gilovich, Thomas; Regan, Dennis (1996). Do economists make bad citizens?
Journal of Economic Perspectives, 10: 187192.
Friedman, Daniel; Sunder Shyam (1994). Experimental Methods: A Primer for Economists.
New York: Cambridge University Press.
Gneezy, U., and A. Rustichini (2000). Pay Enough or Dont Pay At All. Quarterly Journal
of Economics, 791810.
Hertwig, Ralph; Andreas Ortmann (2001). Experimental Practices in Economics: A Methodological
Challenge for Psychologists? Behavior and Brain Sciences, 24, 383451.
Hey, John (1998). Experimental Economics and Deception: A Comment. Journal of
Economic Psychology, 19, 3: 397401.
Holt, Charles (1995). Psychology and Economics: A Discussion of Methodology. University
of Virginia, Discussion Paper.
Kamecke, Ulrich (1997). Rotations: Matching Schemes That Efciently Preserve the Best
Reply Structure of a One Shot Game: Note. International Journal of Game Theory, 26, 3:
409417.
Laband, David; Biel, Richard (1999). Are Economists More Selsh than Other Social Scientists?
Public Choice, 100, 12: 85101.
Loewenstein, George (1990). Experimental Economics from the Vantage-Point of Behavioral
Economics. Economic Journal, 109, F25-F34.
Marwell, Gerald; Ames, Ruth (1981). Economists Free Ride, Does Anyone Else?: Experiments
on the Provision of Public Goods IV. Journal of Public Economics, 15: 295310.
McDaniel, Tanga; Starmer, Chris (1998). Experimental Economics and Deception: A Com-
ment. Journal of Economic Psychology, 19, 3: 403409.
Morris, Michael; Sim, Damien; V. Girotto (1998). Distinguishing Sources of Cooperation in
306 RACHEL CROSON
the One-Round Prisoners Dilemma: Evidence for Cooperation Decisions Based on the
Illusion of Control. Journal of Experimental Social Psychology, 34, 494512.
Ortmann, Andreas; Hertwig, Ralph (1997). Is Deception Acceptable? American Psychologist,
52, 7: 746747.
Ortmann, Andreas; Hertwig, Ralph (1998). The Question Remains: Is Deception Acceptable?
American Psychologist, 53, 7: 806807.
Ortmann, Andreas; Hertwig, Ralph (2002). The Costs of Deception: Evidence from
Psychology. Experimental Economics, 5: 111131.
Plott, Charles (1982). Industrial Organization Theory and Experimental Economics. Journal
of Economic Literature, 20: 14851527.
Plott, Charles (1991a). Will Economics Become an Experimental Science? Southern
Economic Journal, 57, 4: 901919.
Plott, Charles (1991b). Economics in 2090: The Views of an Experimentalist. Economic
Journal, 101, 404: 8893.
Plott, Charles (1994). Market Architectures, Institutional Landscapes and Testbed Experi-
ments. Economic Theory, 4: 310.
Roth, Alvin (1986). Laboratory Experimentation in Economics. Economics & Philosophy, 2,
2: 245273.
Roth, Alvin (1988). Laboratory Experimentation in Economics: A Methodological Overview.
The Economic Journal, 98, 393: 9741031.
Roth, Alvin (1991). Game Theory as a Part of Empirical Economics. Economic Journal, 101,
404: 107114.
Roth, Alvin (1994). Lets Keep the Con out of Experimental Econ: A Methodological Note.
Empirical Economics, 19, 2: 279289.
Shar, Eldar and Tversky, Amos (1992). Thinking through uncertainty: Nonconsequential rea-
soning and choice. Cognitive Psychology, 24, 4: 449474.
Smith, Vernon (1976). Experimental Economics: Induced Value Theory. American Economic
Review, 66, 2: 274279.
Smith, Vernon (1989). Theory, Experiment and Economics. Journal of Economic Perspectives,
3, 1: 151169.
Smith, Vernon (1994). Economics in the Laboratory. Journal of Economic Perspectives, 8, 1:
113131.
Smith, Vernon and Walker, James (1993). Monetary Rewards and Decision Cost in Experi-
mental Economics. Economic Inquiry, 31, 2: 245261.
Tversky, Amos and Shar, Eldar (1992). The Disjunction Effect in Choice under Uncertainty.
Psychological Science, 3: 305309.
Yezer, Tony; Goldfarb, Robert; and Poppen, Paul (1996). Does Studying Economics Dis-
courage Cooperation?: Watch What We Do, Not What We Say or How We Play. Journal
of Economic Perspectives, 10, 1: 171186.
Legal Research on Negotiation
REBECCA HOLLANDER-BLUMOFF
Lawyers and conict are a classic combination. Lawyers are engaged in the
resolution of conict on a constant basis, whether they are negotiating settle-
ments during litigation or crafting deals during corporate transactions. The
study of legal conict resolution is a natural area for legal scholarship to
address. This article offers a sketch of the methodology used by legal schol-
ars studying negotiation, describing the range of approaches used and sug-
gesting some of the benets of and obstacles to empirical work on legal
negotiation.
The rst section provides a brief description of legal scholarship, including
the role of empirical research. Knowledge of this broader framework helps in
understanding the unique challenges to legal research on negotiation. The next
section offers an overview of the treatment of negotiation by legal academics.
The subsequent discussion highlights obstacles and challenges facing legal
scholars, and others, who want to conduct empirical studies of legal negotia-
tion. The conclusion discusses potential areas for future empirical exploration
in the legal negotiation context.
When law professors begin a traditional legal research project, they typically
start and nish by reviewing legal opinions written by judges, statutes
passed by Congress, and articles written by other law professors. Compiling
the materials to conduct such research is a relatively easy task to accomplish
using an online research engine like Lexis or Westlaw or using student
research assistants who are well-versed in online and library document
retrieval. The end goal of this kind of research is a well-written and thought-
fully argued law review article that offers the authors theoretical vision of a
legal issue to readers. The article may survey a vast collection of judicial
opinions, statutory codes, and administrative regulations, or it may just focus
on a handful of United States Supreme Court decisions, but it marshals legal
authority to make its central argument.
The use of social science conclusions is not restricted to one area of legal
scholarship. Areas as diverse as negotiation, litigation, general attorney advo-
cacy, legislative rule-making, and judicial rule-making have been examined
through the lens of empirical data. For example, Gary Blasi uses principles
from social psychology to animate his discussion of attorney advocacy in
light of human prejudice, citing the work of social psychologist John Bargh
to discuss the priming effects of stereotypes on behavior (Blasi 2002). Blasi
suggests that attorneys should take into account such effects when crafting
their courtroom strategies.
In recent years, a number of law review articles have begun to highlight
primary data collected for the purpose of the article. Such primary data takes
the form of empirical (both quantitative and qualitative) studies of things like
settlement agreements, legal opinions, monetary outcomes, particular disputes,
or the behavior of people engaged in negotiating. A number of legal scholars,
including Ayres (1991, 1995), Kim (1997), Korobkin (1998a), Guthrie
(Guthrie 1999) (Korobkin & Guthrie 1994, 1997), and Rachlinski (1996),
have collected original data and used it as the centerpiece of a law review
article in recent years. These writers offer useful illustrations of the potential
for legal scholarship that includes primary empirical research, especially in
the negotiation arena.
For example, Korobkin (1998a) argues that the status quo bias extends to
default contract terms. His argument is supported not just by a thorough dis-
cussion of previous social science research about the status quo bias but also
by a tailored experiment of his own design that specically addresses parties
preferences for default contract terms. His research suggests that contracting
parties view default contract terms as part of the status quo and therefore pre-
fer those terms. In a subsequent article, Korobkin (1998b) uses his data to
further develop a theory of contract negotiations called the inertia theory.
The burgeoning eld of behavioral law and economics, (Jolls, Sunstein &
Thaler 1998) which adds real human behavior into a rational economic
actor model, has helped such work gain popularity, prominence, and accep-
tance, but empirical work in the social science vein still accounts for a tiny
minority of all articles published.
Legal academias institutional biases in favor of normative theoretical
analysis over empirical research pose a challenge to the study of legal nego-
tiation. Very little case law or statutes concern legal negotiation, and thus nor-
mative approaches to negotiation that are wholly theoretical do not have the
weight of any traditional legal authority to support them. Yet clearly, with so
many legal disputes coming to a non-adjudicated resolution, negotiation is a
310 REBECCA HOLLANDER-BLUMOFF
critical component of the current legal system, and legal scholars have in-
creasingly turned their attention to the process of dispute resolution outside
the courts.
In studying how disputes are resolved through negotiation, then, legal aca-
demics face a core tension between the study of an important and pervasive
phenomenon in their eld and a dearth of traditionally favored authorities.
Because legal negotiation is largely extra-legal, that is, governed by few
hard-and-fast legal rules and occurring for the most part between individuals
in private settings, it makes sense that the legal study of negotiation would
rely in some part on empirical work. The traditionally marginal role of empir-
ical work in law scholarship, as described above, thus poses particular chal-
lenges to legal academics who want to study negotiation. An overview of the
legal scholarship on negotiation with an eye towards how legal academics
are approaching this challenge follows below.
The study of legal negotiation has been a growth industry over the past two
decades. During the past ten years there have been 211 articles published in
law reviews and legal journals with the word negotiation in their titles; in
the ten years before that, there were 70.2 The Negotiation Journal, the Journal
on Dispute Resolution (University of Missouri), and the Ohio State Journal
on Dispute Resolution were founded in the 1980s; I participated in the found-
ing of the Harvard Negotiation Law Review in 1995.3
Law review articles on negotiation take many forms. Typically, articles
have at least some descriptive elements (e.g., how legal negotiation does
work) and some prescriptive elements (e.g., how legal negotiation should
work that is, how practitioners should negotiate or how scholars ought
to conceptualize the legal negotiation process). The descriptive element may
be the end goal of the work or it might merely provide a backdrop. Descriptive,
however, need not mean empirical.
In line with classic legal scholarship, one might examine legal negotiation
from a particular theoretical framework. For example, Donald Gifford consid-
ers the implications of a client-centered model of lawyer behavior on legal
negotiation (Gifford 1987), offering a normative vision of how attorneys
should act towards clients during the negotiation process. Others have exam-
ined elements of negotiation from a law and economics perspective (Wachtler
& Cohen 1988). Some develop a coherent theory of negotiation to help guide
practitioners and scholars in thinking about the dynamics of negotiation. For
LEGAL RESEARCH ON NEGOTIATION 311
Empirical work about negotiation typically takes the form of surveys to legal
practitioners or experimental or quasi-experimental studies using law students
or other subjects.4 For example, a recent article by Andrea Kupfer Schneider
(2002) highlights her empirical survey work on the effectiveness of different
negotiation styles, concluding that hard bargaining is perceived as less effec-
tive than problem-solving negotiation tactics. Similarly, Janice Nadler (2004)
recently conducted empirical research on e-mail negotiations, concluding that
negotiators who engaged in small talk were more likely to reach agreement
rather than impasse, and were less angry and more respectful of their adver-
sary in negotiation, than those who did not engage in such small talk. Linda
Babcock and Greg Pogarsky (2001, 1999), too, have explored negotiation
empirically in their studies of the effects of damage cap rules on settlement
negotiation, nding that damage caps act as anchors for negotiated outcomes.
As noted earlier, a number of other legal scholars also employ empirical
research in their work on legal negotiation, and the eld of empirical research
into legal negotiation continues to grow. The work surveyed above describes
the state of largely mainstream law review and law journal literature on nego-
tiation; for reasons described both above in both the preceding and the fol-
lowing sections, some legal academics who research legal negotiation may choose
to publish their work in more specialized journals (Dickinson 2004), non-legal
journals (Babcock and Loewenstein 1997), or even non-academic publica-
tions. Still others who study legal negotiation empirically may include psy-
chologists, sociologists, or business school academics whose works are
published outside of law reviews and legal journals and are thus less acces-
sible to the legal community.
Law differs from psychology and other disciplines that study conict, like
sociology or anthropology, because law is not expressly about the study of
behavior. Law happens irrespective of whether anyone wants to do research
into the practice of law; that is, people sue each other, negotiate contracts and
settlements, and generally speaking engage in legal process all the time,
regardless of whether anyone bothers to study any of it. While human behav-
ior, too, happens whether or not psychologists bother to observe and study it,
psychology as a discipline is expressly about the study of human behavior.
The eld of law, in contrast, encompasses both legal practice itself (by prac-
titioners) and the study of the legal rules and legal practice (by legal acade-
mics). This makes it especially ironic that legal research on negotiation often
LEGAL RESEARCH ON NEGOTIATION 313
does not delve into the substance of real behavior. I attribute the dearth of
such research to two different sources: rst, obstacles within the legal acad-
emy, which I call institutional obstacles, and second, obstacles stemming from
studying the community of legal negotiators, which I term methodological
obstacles.
Institutional obstacles
Additionally, the way that negotiation may be taught as a subject in law schools
can sometimes send a message to students that rigorous intellectual study of
negotiation is unwarranted. Many law schools, for example, hire individuals
who are primarily practicing attorneys to teach negotiation classes. These
practitioners may not be familiar with academic literature on negotiation and
may rely largely on anecdotes to convey their negotiation lessons to students.
Students who receive the message from their law schools that negotiation is
a soft subject in which sophisticated intellectual analysis is not required
may be less likely to select negotiation articles to publish in their law
reviews.
There lingers, too, a basic mistrust of empirical data by the audience for
such articles typically, other legal academics. Legal scholars believe they
are students of social behavior but they are not scientists. (Moe 2002: 373).
They are not, as a rule, well-versed in statistical analysis or the interpretation
of data. Ironically, an untested theoretical idea can be more convincing to a
legal reader than the same idea supported by an empirical study. The study
can be picked apart, its methodology criticized, its nding limited and nar-
rowed until it only applies to the particular setting in which it occurred
while the theoretical idea need only hang together in the intellectual ether.6
Lawyers are, in fact, trained to use factual differences between cases to dis-
tinguish one case from another in order to develop the strongest argument for
the side they represent. Unsurprisingly, then, legal scholars use the same
process of analysis to dissect empirical work and are often left, after such
a process, with distinguishable details rather than generalizable principles.
Methodological obstacles
The structure of legal practice poses unique challenges for the researcher
interested in eld work. Practitioners are extremely busy and getting them to
agree to participate in a study, by allowing someone to interview, survey, or
observe them, can be difcult. Nonetheless, many lawyers are delighted to tell
an interested listener all about their experiences that is to regale an audi-
ence with what lawyers call war stories. These can be great fodder for
analysis but it takes lots of time and effort to deconstruct interviews into
useful explication for an audience.
In the war story vein, I interviewed eleven defense attorneys and prose-
cutors in 1996 for a project on negotiation in the plea bargaining context
(Hollander-Blumoff 1997). The goal of my project was to examine some of
the classic Getting to Yes negotiation principles to see whether they were
relevant and applicable to the setting of plea bargaining. Most of the lawyers
LEGAL RESEARCH ON NEGOTIATION 315
1. Condentiality/Privilege
The attorney-client privilege protects the relationship between lawyers and clients.
It is likely to be very difcult for researchers to gain access to the critical
interactions between principal and agent without waiving that privilege.
Lawyers are by nature a conservative and risk-averse lot, and there is not
much reason for most practicing attorneys to let anyone else get close to their
client. If anything goes awry, could they be opening themselves up to mal-
practice liability? Could a client claim that the presence of a researcher had
a negative impact on the representation provided by the attorney? And could
a researcher be made to testify about the lawyer-client interactions, either in
the context of a malpractice suit, or, perhaps worse, in the context of the orig-
inal legal matter? Not only might a researchers presence at a client meeting
be construed as a waiver of the attorney-client privilege, but sharing conden-
tial information outside of the relationship could be detrimental to the rela-
tionship itself.
Indeed, some attorneys are actually bound in what they can discuss by
their ofce policies. For example, I interviewed lawyers working in the
United States Attorneys Ofce; this ofce has strict rules on who can talk to
316 REBECCA HOLLANDER-BLUMOFF
whom, and what the substance of the conversation can cover. In my own research,
defense attorneys, bound only by their own clients and rules, rather than insti-
tutional guidelines, were far more voluble and far less circumspect in their
storytelling.
Given the restrictions on what attorneys may or will reveal about their
clients and their clients problems, one might reasonably worry about the
quality and completeness of any data collected. In order to uphold their
professional responsibility, attorneys might have omitted important details of
stories that could have compromised client condentiality but that might have
been important to an understanding of the problems they were describing. Lawyers
simply cannot always give as much contextual information about a negotia-
tion as would be helpful because of condentiality issues.
2. Selection bias
Lawyers who agree to discuss their cases with a researcher might practice law
very differently than lawyers who are not inclined to participate in the
research process. There was likely some selection bias among the lawyers
who agreed to see me, for example all were eager to help with my research
and interested in academic endeavors. Lawyers who are not interested in aca-
demic study of and reection on their behavior might somehow be different
kinds of people with different styles and experiences of negotiating. My sam-
ple, too, was by no means random or unbiased. I was given several names by
people that I already knew, and then used introductions by my original
sources and others to network with other lawyers in search of individuals who
would be willing to talk with me.
3. Internal Validity
Likewise, the lawyers that I spoke with most certainly brought their own
biases to their narratives and might have been less likely to tell me stories
that portrayed their negotiation skills in a negative light. This effect might be
exacerbated by the fact that, although the attorneys with whom I spoke
remained anonymous in the nal article, they were not anonymous to me as
a researcher. As an attorney myself, I am part of the same greater community
in which they work and this may have created incentives for how they pre-
sented themselves that could have skewed my results. As noted above, I net-
worked with other attorneys to nd my interviewees, and although I assured
them of condentiality, they may have been concerned about information get-
ting back to their peers.
LEGAL RESEARCH ON NEGOTIATION 317
4. External Validity
In my research described above, I sought to determine whether or not crimi-
nal plea bargaining could aptly be considered negotiation per se by explor-
ing both the extant literature about plea bargaining and the real experiences
of practitioners engaged in the process. I examined how well both what I read
about the plea bargaining process and my empirical ndings about plea bar-
gaining t within literature about the process of negotiation. My empirical
research was unquestionably anecdotal. I did not have a sufcient sample size
to draw any scientic conclusions about, say, how often prosecutors or
defense attorneys considered their BATNA (best alternative to a negotiated
agreement) or how relationships affected plea bargaining outcomes or
processes. I intended my research to be anecdotal and illustrative, rather than
statistically signicant and dispositive in any scientic manner, so a critique
of my work might fairly say that my results were not generalizable.
Even a large scale study, though, might be susceptible to concerns about
generalizability. For instance, Kupfer Schneiders (2002) exploration of the
effectiveness of negotiation style relied on a questionnaire sent to lawyers in
Chicago and Milwaukee; her response rate overall was 29 percent, but almost
40 percent of the Milwaukee recipients responded, while only 18 percent of
the Chicago recipients did. Proportionally, more women completed the survey.
Perhaps geographical or gender issues affected her ndings. Similarly, she
asked attorneys to comment on a self-selected matter or transaction there
could have been confounding variables related to the underlying nature of the
case selected that affected the results. As Korobkin (2001: 327) notes, The
great shortcoming of empirical observation is that it is difcult to generalize
the ndings of an empirical study to novel situations. This concern is one
that is shared by the legal academy.
Past research on negotiation in the law has stressed issues related to nego-
tiating style; the effects of legal endowments; and cognitive biases, such as
anchoring effects and reactive devaluation. My own research strikes out in a
different direction. One line of research applies the principles of procedural
justice to negotiation. Procedural justice literature has established that people
place great importance on the procedures by which authority gures make
decisions (Tyler & Lind 1992). When a third-party authority is involved, par-
ties have been shown to be more satised with and accepting of outcomes
when they are arrived at through the use of fair processes. Indeed, individu-
LEGAL RESEARCH ON NEGOTIATION 319
als may value features of the process above the actual outcome. The results
of my work show that there is a robust procedural justice effect even when
there is no third-party authority decision-maker.
Perhaps even more interestingly, my research challenges the assumption that
negotiators seeking to maximize gain ought to act adversarially in negotiation.
The data show that those who are perceived as fair, and who perceive them-
selves as acting fairly, do not get any better or worse monetary outcomes than
those perceived as less fair. There is, in fact, no correlation between the degree
of perceived fairness and the outcome. But the degree of perceived fairness has
powerful effects on how likely individuals are to accept, or to want to reject,
the outcome. This data has the potential for powerful normative implications
about appropriate and effective attorney behavior in negotiation.
An intriguing concern about this research is how it relates to the attorney
as an agent. Do the procedural justice effects transmit themselves to clients
through the way in which lawyers pitch a certain outcome? That is, if an
attorney feels a process is fair and is thus more ready to accept the resulting
outcome, does that translate into a perception on the part of the client that
the process is good or that the outcome is good? Does process between
attorneys matter even when clients are not present? In essence, do clients care
about process or not? This remains a largely unexplored area that is ripe for
future study.
Another line of my research explores how lawyers use the law in legal
negotiations that is, what differentiates the legal negotiation setting from
any other type of negotiation? How does the use of legal endowments work
practically during a negotiation? What effect does arguing the law, or not
arguing the law, have on negotiation process and outcome? I have developed
a coding scheme to analyze the data described above with respect to different
uses of legal argument in negotiations and its effect on process, numerical
outcome, and other outcome variables. Findings from such a study could be
useful to practitioners and to academics alike, as we think prescriptively about
what effective legal negotiation and able advocacy look like.
It remains the great challenge for legal academics interested in negotiation
research: how to extend social science methodology into a remarkably con-
servative, risk-averse eld, so that we are able to amass more research on real
attorneys and enable our empirical ndings to become more generalizable.
Empiricism has made inroads towards overcoming the institutional obstacles.
As momentum favoring empirical work builds in the legal academy, scholars
interested in legal negotiation can hope and push for similar receptivity
by practitioners themselves that will help to erode some of the methodological
barriers.
320 REBECCA HOLLANDER-BLUMOFF
Notes
1. I leave aside for this article the debate over what empirical means. First, a wide gulf sep-
arates qualitative from quantitative empiricism in the law; and then, there is hearty dis-
agreement over what may be considered qualitative empirical data. For the purpose of this
discussion, I dene empirical work as that involving a systematic collection and analysis of
data using social science methodology.
2. Lexis/Nexis search conducted April 30, 2004.
3. A number of articles about legal negotiation appear in journals that are not law reviews per
se. For example, publications including Law and Social Inquiry, Law and Human Behavior,
and the Journal of Legal Studies have included important work on legal negotiation. These
publications, however, tend to be outside the mainstream of the most prestigious law
reviews, and many legal academics are not familiar with the work of these journals; indeed,
some do not even have access in any easy way to journals that are not within Lexis or
Westlaws law review database.
4. Russell Korobkin (2002: 1038) identied four potential sources of empirical data on con-
tract law: judicial opinions, studies of actual contracting practices of contracting parties,
experimental studies of contracting behavior, and studies of contracting parties opinions.
These are useful categories to consider in thinking about sources of data for studying legal
negotiation. His rst category is not relevant to the negotiation context, as very few opin-
ions address the negotiation process. His second and fourth sources roughly correspond to
survey research that asks real attorneys what they do as well as what they think about
what they do, while the third source roughly corresponds with experimental work involving
simulations.
5. There are several peer-reviewed law journals in specialized areas of the law; many do not
emphasize empirical work, and in any event these journals generally have a smaller circu-
lation and audience than more mainstream law reviews.
6. Korobkin raises a related point when he notes that a single empirical study might be less
fullling to the author than an article laying out a theory that seamlessly and completely
resolves a legal problem on its own, with no questions left open (Korobkin 2002: 1055).
References
Adler, Robert S. & Silverstein, Elliot M. (2000). When David meets Goliath: dealing with
power differentials in negotiations. Harvard Negotiation Law Review, 5: 1112.
Ayres, Ian (1991). Fair driving: gender and race discrimination in retail car negotiations.
Harvard Law Review, 104: 817872.
Ayres, Ian (1995). Further evidence of discrimination in new car negotiations and estimates
of its cause. Michigan Law Review, 94: 109147.
Babcock, Linda & Loewenstein, George (1997). Explaining bargaining impasse: the role of
self-serving biases. Journal of Economic Perspectives, 11: 10926.
Babcock, Linda & Pogarsky, Greg (1999). Damage caps and settlement: a behavioral
approach. Journal of Legal Studies, 28: 341370.
Bartels, William K. (2000). Recent developments: the stormy sea of apologies: California Evidence
Code Section 1160 provides a safe harbor for apologies made after accidents. Western State
University Law Review, 28: 141157.
LEGAL RESEARCH ON NEGOTIATION 321
Birke, Richard & Fox, Craig R. (1999). Psychological principles in negotiating civil settle-
ments. Harvard Negotiation Law Review, 4: 157.
Blasi, Gary (2002). Advocacy against the stereotype: lessons from cognitive social psychol-
ogy. UCLA Law Review, 49: 12411281.
Bodie, Matthew T. (2004). Questions about the efciency of employment arbitration agree-
ments. Georgia Law Review, 39: 182.
Busch, Marc L. & Reinhardt, Eric (2000). Bargaining in the shadow of the law: early settle-
ment in GATT/WTO Disputes. Fordham International Law Journal, 24: 158172.
Cohen, Jonathan R. (1999). Advising clients to apologize. Southern California Law Review,
72: 10091069.
Dallas, Lynne L. (2003). The multiple roles of corporate boards of directors. San Diego Law
Review, 40: 781820.
Dau-Schmidt, Kenneth G. (1992). A bargaining analysis of American labor law and the search
for bargaining equity and industrial peace. Michigan Law Review, 91: 419514.
Dickinson, David L. (2004). A comparison of conventional, nal-offer, and combined arbi-
tration for dispute resolution. Industrial and Labor Relations Review, 57: 288301.
Ehrenreich Brooks, Rosa (2003). The new imperialism: violence, norms, and the rule of
law. Michigan Law Review, 101: 22752340, 2308.
Epstein, Lee & King, Gary (2002). Empirical research and the goals of legal scholarship: The
rules of inference. University of Chicago Law Review, 69: 1133.
Fisher, Roger & Jackson, William (1993). Teaching the skills of settlement. Southern
Methodist University Law Review, 46: 19851993.
Fisher, Roger, Ury, William, & Patton, Bruce (1991). Getting to yes. New York, NY: Penguin
Books.
Gifford, Donald G. (1987). The synthesis of legal counseling and negotiation models: pre-
serving client-centered advocacy in the negotiation context. UCLA Law Review, 34: 811862.
Guthrie, Chris (1999). Better settle than sorry: The regret aversion theory of litigation behav-
ior. University of Illinois Law Review, 1999: 4390.
Hoffmann, Joseph L. & Kahn, Marcy L. (2001). Plea bargaining in the shadow of death.
Fordham Law Review, 69: 23132391.
Hollander-Blumoff, Rebecca (1997). Getting to guilty: Plea bargaining as negotiation.
Harvard Negotiation Law Review, 2: 115148.
Jolls, Christine, Sunstein, Cass R., & Thaler, Richard (1998). A behavioral approach to law
and economics. Stanford Law Review, 50: 14711550.
Katsh, Ethan, Rifkin, Janet & Gaitenby, Alan (2000). E-Commerce, e-disputes, and e-dispute
resolution: In the shadow of Ebay law. Ohio State Journal on Dispute Resolution, 15: 705734.
Kim, Pauline T. (1997). Bargaining with imperfect information: a study of worker perceptions
of legal protection in an at-will world. Cornell Law Review 83: 105156.
Korobkin, Russell (1998a). Inertia and preference in contract negotiation: The psychological
power of default rules and form terms. Vanderbilt Law Review 51: 15831651.
Korobkin, Russell (1998b). The Status quo bias and contract default rules. Cornell Law
Review 83: 608687.
Korobkin, Russell (2000). A Positive Theory of Legal Negotiation. Georgetown Law Journal,
88: 17891831.
Korobkin, Russell (2001). A multi-disciplinary approach to legal scholarship: economics,
behavioral economics, and evolutionary psychology. Jurimetrics Journal, 41: 319336, 327.
Korobkin, Russell (2002). Empirical scholarship in contract law: possibilities and pitfalls.
University of Illinois Law Review, 2002: 10331066.
322 REBECCA HOLLANDER-BLUMOFF
Korobkin, Russell & Guthrie, Chris (1994). Psychological barriers to litigation settlement: an
experimental approach. Michigan Law Review, 93: 107192.
Korobkin, Russell & Guthrie, Chris (1997). Psychology, economics, and settlement: a new
look at the role of the lawyer. Texas Law Review, 76: 77141.
Korobkin, Russell & Guthrie, Chris (2004). Heuristics and biases at the bargaining table.
Marquette Law Review, 87: 795808.
Latif, Elizabeth (2001). Apologetic justice: evaluating apologies tailored toward legal solu-
tions. Boston University Law Review, 81: 289320.
Mnookin, Robert H. & Kornhauser, Lewis (1979). Bargaining in the shadow of the law: the
case of divorce. Yale Law Journal 88: 950997.
Mnookin, Robert H., Peppet, Scott R., & Tulumello, Andrew (2000). Beyond Winning.
Cambridge, MA: Belknap Press.
Moe, Terry M. (2002). Cynicism and political theory. Cornell Law Review, 87: 362383.
Nadler, Janice (2004). Rapport in legal negotiation: how small talk can facilitate e-mail deal
making. Harvard Negotiation Law Review, 9: 223251.
Neumann, Richard & Krieger, Stefan (2003). Empirical inquiry twenty-ve years after the
lawyering process. Clinical Law Review, 10: 349397.
Oesterle, Dale A. (1986). The negotiation model of tender offer defenses and the Delaware
Supreme Court. Cornell Law Review, 72: 117157.
Pogarsky, Greg & Babcock, Linda (2001). Damage caps, motivated anchoring, and bargaining
impasse. Journal of Legal Studies, 30: 143159.
Rachlinski, Jeffrey J. (1996). Gains, losses, and the psychology of litigation. Southern
California Law Review, 70: 113184.
Rossi, Jim (2001). Bargaining in the shadow of administrative procedure: the public interest
in rulemaking settlement. Duke Law Journal, 51: 10151058.
Schneider, Andrea Kupfer (2002). Shattering negotiation myths: empirical evidence on the
effectiveness of negotiation style. Harvard Negotiation Law Review 7: 143233.
Subramanian, Guhan (2003). Bargaining in the shadow of takeover defenses. Yale Law
Journal, 113: 621686.
Tyler, Tom R., & Lind, E. Allan (1992). A relational model of authority in groups, in Mark
Zanna, editor, Advances in Experimental Social Psychology, 25: 115191, New York:
Academic Press.
Wachter, Michael L. & Cohen, George M. (1988). The law and economics of collective bar-
gaining: an introduction and application to the problems of subcontracting, partial closure,
and relocation. University of Pennsylvania Law Review, 136: 13491417.
Methodologies for Studying Personality Processes in
Interpersonal Conict
Even recent conict theories have focused mainly on the surrounding contexts
involved in conict (e.g., Shantz & Hartup 1992).
Situational context provides only part of the conict story. Given that inter-
personal conict is dened as two or more individuals opposing one another,
personality differences should also contribute signicantly to the dyadic rela-
tion of individuals. Here is one major source of individual differences in atti-
tudes, beliefs, and life experiences discussed in the previous denition of
conict. It is too simplistic and atheoretical, however, to treat personality as
a non-interacting set of individual differences. Instead, personality may be
more usefully dened as structured system of individual differences organized
to assist an individual person and his/her adaptation to the environment
(Graziano 2003). For each individual, personality is a process summarizing
the continuing negotiated compromises among competing motives and ten-
dencies. One way to conceptualize these personality processes is in terms of
levels of analysis. For example, McAdams (1995) theorized that individual
differences in personality might be described at three different levels. Two of
these levels may be particularly important for understanding personality and
its inuence within conict situations. His rst level consists of comparative
dimensions of personality like the Big Five personality traits that are not con-
ditional or dependent on the situation. In other words, Level I traits involve
what a person has and brings to the conict situation. The second level
involves context-dependent strategies, plans, and concerns that enable individuals
to resolve conicts. In other words, Level II speaks to what a person does
within a given conict situation. These levels of analysis are not necessarily
derivative of one another (McAdams 1995). Nonetheless, we suggest that
there will be at least a loose linkage, between Level I traits and the organized
systems of personal concerns within a conict at Level II.
When two people are involved in conict, there is interdependence among
effects due to the person, effects due to the partner, and effects due to the rela-
tionship and situation (Kenny & LaVoie 1984; Lewin 1935). Thus, under-
standing the inuence of personality on social conict must look past simple
main effects models that focus on either situations or personality, per se. Here
we start our paper by discussing various research methods that can be used
to assess personalitys contribution to conict behavior. We then focus on sta-
tistical advances that recognize the complex nature of interactions between
individual, their partners, and the situation when studying conict. The meth-
ods reported here are not intended as an exhaustive list for studying person-
alitys contribution to conict, but will provide the reader with a starting
point.
METHODOLOGIES FOR STUDYING PERSONALITY PROCESSES 325
et al. 2001 for actual vignettes). The adolescents task was to tell us how
good or bad each choice was in solving conicts in the different situations/
relationships. As predicted, the personality dimension of agreeableness
uniquely predicted the endorsement of both constructive and destructive
conict resolution tactics.
In terms of strengths, vignettes are direct, and focus participants attention
on the variables of interest. In addition, the vignette method allows the
researcher to control the conict situation presented and to avoid some ethi-
cal problems associated with deception and manipulation of participants per-
ceptions of conict situations. There are also several potential weaknesses
with using these role-playing methodologies (Graziano 1987; Greenberg & Folger
1988: 3960). The primary argument against role-playing methodologies,
however, is that social desirability processes may bias responding toward the
ideal response rather than what participants would actually do in the situation.
In his literature review, Laursen (1993) found that this data collection tech-
nique biased empirical results in the conict literature. Specically, he found
that adolescents responded to hypothetical conicts with ideal resolution
choices (i.e., negotiation). In actual conicts, however, power assertion and
disengagement were used more frequently than negotiation. As a general rule,
responses to hypothetical conicts may be biased toward responses perceived
to be desirable, but social desirability processes themselves are complex, elu-
sive phenomena (Graziano & Tobin 2002). Furthermore, some studies report
convergence. Using a multi-method format that collected data for both hypo-
thetical vignettes and daily diary records across a two-week period, Jensen-
Campbell and Graziano (2001) found convergence between hypothetical conicts
and diary records of actual daily interpersonal conicts. Even so, our out-
comes suggest that reactions to vignettes may have an underlying structure
different from accounts of in-vivo conict behaviors (Graziano 1987: 284
287). In other words, the hypothetical vignette method relies heavily on
simplistic, singular, deterministic views of personality on conict behavior
and fails to attend to contextual factors (e.g., conict partner) that probably
inuence the behavior. (For a related discussion of personality and social
desirability artifacts in conict research, see Graziano & Tobin 2002).
Self-Report Methodologies
Event-by-Event Recordings
Behavioral Observation
Another approach that has been used in conict research is structured behav-
ioral observation in the laboratory. For example, Jensen-Campbell, Gleason,
Adams, and Malcolm (2003) observed the relations between personality and
conict in late childhood by using a board game (that created conict) and
then unobtrusively videotaping and transcribing the events that occurred
within the xed situation. Laboratory observations can also involve less struc-
tured observations. For example, Ickes and his colleagues (1986, 1990) have
developed a method to study unobtrusively naturally occurring interactions among
dyads. Recently, Schweinle, Ickes, and Bernstein (2002) used a modied ver-
sion of this paradigm to assess personalitys contribution (i.e., empathic accu-
racy) to husband-to-wife aggression.
These methods are important and provide a rich body of empirical evidence
in psychology. The data collected with these approaches, however, are usually
constrained by the environment (e.g., laboratory) and often involve xed activ-
ities. They do not allow for probing the full extent of everyday interpersonal
conicts in adolescence. In other words, the conict behavior is mandated into
existence by denition. Similarly, the obtrusiveness of the observation may
inuence the behaviors displayed. In addition, task demands constrain conict
resolution strategies (Laursen & Collins 1994). Disengagement is usually not
possible in laboratory settings. For example, participants in the Jensen-
Campbell et al. (2003) study were told that if they had a disagreement they
must work it out. Their choices for conict resolution were constrained; they
could not seek the researchers help or leave the experimental session. Thus,
the children could not withdraw from the situation before the experiment was
nished and could not seek out any third-party interventions.
METHODOLOGIES FOR STUDYING PERSONALITY PROCESSES 331
agreeable persons generate more positive responses than their peers to inter-
personal conict. Discussing the specics of moderated regression is beyond
the scope of this paper, but an excellent review of these procedures can be
found in both Aiken and West (1994) and Cohen et al. (2003).
Multi-level Modeling
Often data that assesses personalitys inuence on daily conicts is also struc-
tured hierarchically. In other words, there are clusters or levels within the data
set. Consider the conicts reported by an individual participant using the
event-by-event recording method. We would expect some association between
the resolutions of conicts across relationship classes and situations within the
same individual. For example, individuals high in agreeableness will consis-
tently resolve their conict episodes more constructively with parents, peers,
and teachers than will their peers. Moreover, conicts that are reported for
members of the same or similar conict dyad(s) (e.g., mother/child; High/
High Agreeable dyads) should also be more highly correlated than other pair-
ings. When data are collected at multiple levels, the data should be treated as
a multi-level or hierarchical model and the use of random coefcient regres-
sion should be used to retain the datas complexity (Cohen, et al. 2003).
In Jensen-Campbell and Graziano (2001), both individuals and their inter-
actions were treated as units of measurement because individual conicts
were collected from a participant across a two-week period. For example, we
rst calculated regression equations that estimated how much each participant
reported systematic ratings with his/her partners based on the sex of the part-
ner. In the second step, the rst-step regression coefcients were regressed on
participant-agreeableness, sex of participant, and the cross product of Sex X
Agreeableness (see Bosker & Snijders 1999; Bryk & Raudenbush 2002;
Cohen et al. 2003; Kenny et al. 1998; Kenny, Bolger, & Kashy 2002).
By using multi-level modeling instead of aggregating our data by partici-
pant, we could assess the within-individual information instead of assuming
that each conict or conict partner for the individual were equivalent. In
other words, we could assess more accurately the contribution of an individ-
uals personality while assessing effects due to the interaction and interaction
partner. We specically constructed a two-level model, with variations in
change parameters among interactions within an individual at level one (e.g.,
partners traits) and variation among participants as level two (e.g., partici-
pants traits). It is possible, however, to have more than two levels in a given
data set. In sum, multilevel procedures treat both lower- or micro-levels (e.g.,
METHODOLOGIES FOR STUDYING PERSONALITY PROCESSES 333
Mediational Analyses
Another statistical method that can be used to study how personality inuences
perceptions of conict in groups (e.g., triads) involves decomposing self- and
other data into perceiver and target effects using Kennys (1994) Social
Relation Model (SRM). Groups can be formed and asked to perform a
conict-inducing task or group decision-making task that creates conict. Because
each member in the group interacts and rates all other members in the group
(i.e., a round-robin design), SRM analyses allow for personality effects of the
individual to be partitioned from partner and relationship effects.
334 LAURI A. JENSEN-CAMPBELL AND WILLIAM G. GRAZIANO
Dyadic Analysis
Contrary views of personality such as Shadel and Cervone (1993) argue that
it is inappropriate to study personality as a basic unit of analysis because
interpersonal behavior needs to be dened within context. To address these
concerns, we used a multivariate approach for statistical analysis of dyadic
METHODOLOGIES FOR STUDYING PERSONALITY PROCESSES 335
Acknowledgements
References
Ahadi, S.A. & Rothbart, M.K. (1994). Temperament, development, and the Big Five, in
C.F. Halverson, Jr., G.A. Kohnstamm, & R.P. Martin, editors, The developing structure of
temperament and personality from infancy to adulthood. Hillsdale, NJ: Lawrence Erlbaum
Associates.
Aiken, L.S., & West, S.G. (1991). Multiple regression: Testing and interpreting interactions.
Newbury Park, NJ: SAGE Publications.
Aronson, E., & Carlsmith, J.M. (1968). Experimentation in social psychology, in G. Lindzey
& E. Aronson, editors, Handbook of social psychology, 2: 179. Reading, MA: Addison-
Wesley.
Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable distinction in social
psychological research: Conceptual, strategic, and statistical considerations. Journal of
Personality and Social Psychology, 51: 11731182.
Barrick, M.R., & Mount, M.K. (1991). The big ve dimensions and job performance: A meta
analysis. Personnel Psychology, 44, 1: 126.
Barrick, M.R., Mount, M.K. & Judge, T.A. (2001). Personality and performance at the begin-
ning of the new millennium: What do we know and where do we go next? International
Journal of Selection & Assessment, 9, 12: 930.
Barrick, M.R., Stewart, G.I., Neubert, M.J., & Mount, M.K. (1998). Relating member ability
and personality to work-team processes and team effectiveness. Journal of Applied
Psychology, 83: 377391.
Barry, B., & Friedman, R.A. (1998). Bargainer characteristics in distributive and integrative
negotiation. Journal of Personality and Social Psychology, 74: 345359.
Bentler, P.M. (1990). Comparative t indexes in structural models. Psychological Bulletin,
107: 238246.
Biesanz, J., West, S.G., & Kwok, O.M. (2003). Personality over time: Methodological
approaches top the study of short-term and long-term development and change. Journal of
Personality, 71: 905941.
Blaney, P.H. (1986). Affect and Memory: A review. Psychological Bulletin, 99: 229246.
Bosker, R.J., & Snijders, T.A.B. (1999). Multilevel analysis: An introduction to basic and
advanced multilevel modeling. Newbury Park, NJ: SAGE Publications.
Bryk, A.S., & Raudenbush, S.W. (1992). Hierarchical Linear Models: Applications and data
analysis methods. Newbury Park, CA: SAGE.
Campbell, L., & Kashy, D.A. (2002). Estimating actor, partner, and interaction effects for
dyadic data using PROC MIXED and HLM: A user-friendly guide. Personal Relationships,
9: 327342.
Cantor, N. & Kihlstrom, J.F. (1987). Personality and social intelligence. Englewood Cliffs, NJ:
Prentice Hall.
Cohen, J., Cohen, P. West, S.G., & Aiken, L.S. (2003). Applied multiple regression/correlation
analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Lawrence Erlbaum Associates.
METHODOLOGIES FOR STUDYING PERSONALITY PROCESSES 337
Csikszentmihalyi, M., & Larsen, R. (1984). Being adolescent: Conict and growth in the
teenage years. New York: Basic Books.
De Dreu, C.K.W., & Carnevale, P.J. (2003). Motivational bases of information processing and
strategy in negotiation and social conict, in M.P. Zanna, editor, Advances in Experimental
Social Psychology, 35, New York: Academic Press.
DeShon, R.P. (1998). A cautionary note on measurement error corrections in structural equa-
tion models. Psychological Methods, 3: 412423.
Farrell, A.D. (1994). Structural equation modeling with longitudinal data: Strategies for exam-
ining group differences and reciprocal relationships. Journal of Consulting and Clinical
Psychology, 62: 477487.
Finch, J.F. & Graziano, W.G. (2001). Predicting depression from temperament, personality,
and patterns of social relations. Journal of Personality, 69: 2755.
Furman, W. & McQuaid, E.L. (1992). Intervention programs for the management of conict.
in C.U. Shantz & W.W. Hartup, editors.) Conict in child and adolescent development:
402429. New York: Cambridge University Press.
Gilligan, S.G. & Bower, G.H. (1984). Cognitive consequences of emotional arousal. in
C.E. Izard, J. Kagan, & R.B. Zajonc, editors, Emotions, cognition, and behavior: 547588.
Hillsdale, NJ: Erlbaum.
Gleason, K.A., Jensen-Campbell, L.A., & Richardson, D. (2004). Agreeableness and aggres-
sion in adolescence. Aggressive Behavior, 30: 4361.
Goldberg, L.R. (1992). The development of markers of the Big Five factor structure. Psychological
Assessment, 4: 2642.
Graziano, W.G. (1987). Lost in thought at the choice point: Cognition, context, and equity,
in J.C. Masters & W.P. Smith, editors, Social comparison, social justice, and relative depri-
vation: Theoretical, empirical and policy perspectives: 265294. Hillsdale, NJ: Erlbaum.
Graziano, W.G. (2003). Personality development: An introduction. Toward process approaches
to stability and change in persons. Journal of Personality, 71, 6: 893903.
Graziano, W.G., Hair, E.C., & Finch, J.F. (1997). Competitiveness mediates the link between
personality and group performance. Journal of Personality and Social Psychology, 73:
13941408.
Graziano, W.G., Jensen-Campbell, L.A., & Finch, J.F. (1997). The self as a mediator between
personality and adjustment. Journal of Personality and Social Psychology, 73: 392404.
Graziano, W.G., Jensen-Campbell, L.A., Hair, E.C. (1996). Perceiving interpersonal conict
and reacting to it: The case for agreeableness. Journal of Personality and Social
Psychology, 70: 820835.
Graziano, W.G., Jensen-Campbell, L.A., Steele, R.G., & Hair, E.C. (1998). Unknown words
in self-reported personality: Lethargic and provincial in Texas. Personality and Social Psychology
Bulletin, 24: 893905.
Graziano, W.G., & Tobin, R.M. (2002). Agreeableness: Dimension of personality or social
desirability artifact? Journal of Personality, 70: 695727.
Greenberg, J., & Folger, R. (1988). Controversial issues in social research methods. New York:
Springer Verlag.
Halverson, C.F. (1988). Remembering your parents: Reections on the retrospective method.
Journal of Personality, 56: 435444.
Henry, B., Moftt, T.E., Caspi, A., Langley, J., Silva, P.A. (1994). On the remembrance of
things past: A longitudinal evaluation of the retrospective method. Psychological Assess-
ment, 6: 92101.
Hodges, E.V.E., Malone, M.J., Jr., & Perry, D.G. (1995, March). Behavioral and social
338 LAURI A. JENSEN-CAMPBELL AND WILLIAM G. GRAZIANO
antecedents and consequences of victimization by peers, in N.R. Crick, Recent trends in the
study of peer victimization: Who is at risk and what are the consequences? Symposium con-
ducted at the biennial meeting of the Society for Research in Child Development, India-
napolis, IN.
Ickes, W., Robertson, E., Tooke, W., & Teng, G. (1986). Naturalistic social cognition: Methodology,
assessment, and validation. Journal of Personality and Social Psychology, 51: 6682.
Ickes, W., Stinson, L., Bissonnette, V. (1990). Naturalistic social cognition: Empathic accuracy
in mixed-sex dyads. Journal of Personality and Social Psychology, 59: 730742.
Isen, A.M. (1984). Toward understanding the role of affect in cognition, in R.S. Wyer, Jr., &
T.K. Srull, editors, Handbook of social cognition: 179236. Hillsdale, NJ: Erlbaum.
Jensen-Campbell, L.A., Adams, R., Perry, D., Furdella, J.Q., Workman, K.A., & Egan, S.
(2002). Agreeableness, Extraversion, and Peer Relations in Early Adolescence: Winning
Friends and Deecting Aggression. Journal of Research in Personality, 36: 224251.
Jensen-Campbell, L.A., Gleason, K.A., Adams, R., & Malcolm, K.T. (2003). Interpersonal
conict, agreeableness, and personality development. Journal of Personality, 71: 10591085.
(Special Issue: Personality Development).
Jensen-Campbell, L.A. & Graziano, W.G. (2000). Beyond the schoolyard: Relationships as
moderators of daily interpersonal conict. Personality and Social Psychology Bulletin, 26:
923935.
Jensen-Campbell, L.A., & Graziano, W.G. (2001). Agreeableness as a Moderator of Inter-
personal Conict. Journal of Personality, 69: 323362.
Jensen-Campbell, L.A., & Graziano, W.G. (in press). Resisting Temptation: Differing Motives
for Self-Control. Merrill-Palmer Quarterly.
Jensen-Campbell, L.A., Graziano, W.G., & Hair, E.C. (1996). Personality and relationships as
moderators of interpersonal conict in adolescence. Merrill-Palmer Quarterly, 42: 148164.
Kashy, D.A., & Kenny, D.A. (2000). The analysis of data from dyads and groups. in H.T.
Reis & C.M. Judd, editors, Handbook of research methods in social and personality psy-
chology: 451477. New York: Cambridge University Press.
Kenny, D.A. (1994). Interpersonal perception: A social relations analysis. New York: Guilford.
Kenny, D.A. (1996). The design and analysis of social-interaction research. Annual Review of
Psychology, 47: 5986.
Kenny, D.A., & Kashy, D.A. (1992). Analysis of the multi-trait-multi-method matrix by
conrmatory factor analysis. Psychological Bulletin, 112: 165172.
Kenny, D.A., Kashy, D.A., & Bolger, N. (1998). Data analysis in social psychology, in D.T.
Gilbert and S.T. Fiske, editors, The handbook of social psychology, 1: 233265. New York:
McGraw-Hill.
Kenny, D.A., Bolger, N., & Kashy, D.A. (2002). Traditional methods for estimating multilevel
models, in D.S. Moskowitz & S.L. Hershberger, editors, Modeling intraindividual variabil-
ity with repeated measures data: Methods and applications. Multivariate applications book
series: 124. Mahwah, NJ: Lawrence Erlbaum Associates.
Kenny, D.A. & La Voie, L. (1985). Separating group and individual level effects. Journal of
Personality and Social Psychology, 48: 339348.
Laursen, B. (1993). The perceived impact of conict on adolescent relationships. Merrill-
Palmer Quarterly, 39: 535550.
Laursen, B., & Collins, W.A. (1994). Interpersonal conict during adolescence. Psychological
Bulletin, 115: 177209.
Lewin, K. (1935). A dynamic theory of personality. New York: McGraw-Hill.
METHODOLOGIES FOR STUDYING PERSONALITY PROCESSES 339
Mehrabian, A. (1969). Methods and designs: Some referents and measures of nonverbal
behaviors. Behavioral Research Methods and Instruments, 1: 203207.
McAdams, D.P. (1995). What do we know when we know a person? Journal of Personality,
63: 365396.
McDonald, R.P. (2002). Principles and practice in reporting structural equation analyses.
Psychological Methods, 7: 6482.
Mendoza, J., & Graziano, W. (1982). The statistical analysis of dyadic social behavior: A mul-
tivariate approach. Psychological Bulletin, 92: 532540.
Mischel, W. (1973). Toward a cognitive social learning reconceptualization of personality.
Psychological Review, 80: 252283.
Patterson, G.R. (1982). Coercive family process. Eugene, OR: Castalia.
Raudenbush, S.W., & Bryk, A.S. (2002). Hierarchical Linear Models: Applications and Data
Analysis Methods (2nd ed.). Newbury Park, NJ: SAGE Publications.
Reis, H.T., & Wheeler, L. (1991). Studying social interaction with the Rochester Interaction
Record, in M.P. Zanna, editor, Advances in experimental social psychology, 24: 269318.
New York: Academic Press.
Shadel, W.G., & Cervone, D. (1993). The Big Five versus nobody? American Psychologist,
48: 13001302.
Shadish, W.R., Cook, T.D., & Campbell, D.T. (2001). Experimental and quasi-experimental designs
for generalized causal inference. Boston: Houghton Mifin.
Shantz, C.U., & Hartup, W.W., editors, (1992). Conict in child and adolescent development.
New York: Cambridge University Press.
Sherif, M. (1956). Experiments in group conict. Scientic American, 195: 5358.
Schwarz, N., & Clore, G.L. (1983). Mood, misattribution, and judgments of well-being: Informative
and directive functions of affective states. Journal of Personality and Social Psychology,
34: 513523.
Schwartz, D., Dodge, K.A., & Coie, J.D. (1993). The emergence of chronic peer victimiza-
tion in boys play groups. Child Development, 6: 17551772.
Schweinle, W.E., Ickes, W., & Bernstein, I.H. (2002). Empathic inaccuracy in husband to wife
aggression: The over attribution bias. Personal Relationships, 9, 2: 141158.
Sternberg, R., & Soriano, (1984). Styles of conict resolution. Journal of Personality and
Social Psychology, 47: 115126.
Suls, J., Martin, R., David, J. (1998). Person-Environment t and its limits: Agreeableness,
neuroticism, and emotional reactivity to interpersonal conict. Personality and Social Psychology
Bulletin, 24: 8898.
Terhune, K.W. (1970). The effects of personality in cooperation and conict, in P. Swingle,
editor, The Structure of Conict. New York: Academic Press.
Tobin, R.M., Graziano, W.G., Vanman, E., & Tassinary, L.G. (2000). Personality, emotional
experience, and efforts to control emotions. Journal of Personality and Social Psychology,
79: 656669.
West, S.G. & Aiken, L.S. (1997). Toward understanding individual effects in multiple com-
ponent prevention programs: Design and analysis strategies, in K. Bryant, M. Windle, &
S.G. West, editors, The science of prevention: Methodological advances from alcohol and
substance abuse research: 167209. Washington, DC: American Psychological Association.
The Heart of Darkness: Advice on Navigating
Cross-Cultural Research
CATHERINE H. TINSLEY
My apologies of course to Joseph Conrad for any implication that this arti-
cle is as entertaining or as well written as The Heart of Darkness. Conrads
novel tracks a British trader who descends deep into the jungle of Central
Africa, where the lack of rules (or at least any structure that he recognizes)
drives him crazy. Freedom obsesses him with power, and he becomes a psy-
chotic despot. The story parallels ones descent into cross-cultural research. In
both instances, the path is murky, there are too many degrees of freedom for
one to be truly comfortable, and the lack of structure leads researchers to
establish methodological efdoms without oversight from others. This lack of
structure or unifying paradigm for cross-cultural research explains in large
measure why mainstream researchers have difculty accepting cross-cultural
studies or results, which are often perceived as descriptive or epiphenomenal
ndings with limited rigor or merit.
Several years ago, a group of colleagues, including myself, wrote two chap-
ters exploring cross-cultural research methods issues (Lytle, Brett, Barsness,
Tinsley, & Janssens 1995; Brett, Tinsley, Janssens, Barsness, and Lytle 1997).
We were beginning our own forays into cross-cultural research, and our goal
was to put forth a methodology for conducting rigorous, conrmatory cross-
cultural research. These chapters have been generally accepted by other schol-
ars doing cross-cultural research, and many of the specic suggestions for
doing cross-cultural research remain useful, such as: modeling culture as a set
of overlapping dimensions, studying both moderating and main effects of cul-
ture, assessing conceptual, structural, and functional equivalence, and constructing
research designs that rule out alternative explanations for conrmatory results.
However, as I reect back on the overall gestalt of these chapters, I notice a
certain innocence and optimism that is borne out of a theoretical rather than
practical understanding of the issues surrounding cross-cultural research.
The chapters lay out for the reader a rich array of choices in which to con-
duct rigorous cross-cultural research. For example, in the second chapter
(Brett et al. 1997) we discuss the vast multi-dimensional space in which a
researcher can position his or her cross-cultural endeavors. It is a space where
type I error or nding a relationship where one does not truly exist (Cook &
Campbell 1979). This seems to be a more dangerous risk than the risk of type
II error, which would characterize the underspecied intersection model,
where a researcher fails to nd a relationship that truly does exist (Cook &
Campbell 1979). Since science progresses as a body of knowledge that is modied
over time, scientists generally adopt a conservative approach, so that type II
error is considered more benign (Cook & Campbell 1979). Similarly, it may
be less dangerous to design cross-cultural comparisons using the intersec-
tion model of all cultures emic models, rather than the union model. The researcher
risks losing nuances of a particular culture (and hence risks missing a rela-
tionship that truly does exist), but minimizes the risks of concluding a rela-
tionship with spurious variables.
A similar logic applies to the measurement issues, when the same construct
may have a different domain across the various cultures of interest. The
approach has been to capture this domain using culturally sensitive (emic)
indicators when necessary (Brett et al. 1997; Hui & Triandis 1985; Triandis
1992). Thus a construct may be operationalized with measures A, B, and C
in one culture and with measures A, B, C, and D in another culture. The bias
seems to be towards cultural sensitivity, hence being as true to a cultures
emic knowledge as possible. Yet, this raises obvious issues of conceptual and
measurement equivalence. If the domains are substantially different, are they
really the same construct across cultures? And if they are measured with dif-
ferent indicators, then there are likely to be differences in correlations of this
construct with others that have more to do with measurement differences than
with cultural differences. Thus, I now think it is best to keep emic indicators
to a minimum. My advice would be for the researcher to use the intersection
set of all emic indicators, that is, to use only those items that are valid in all
the cultures. Although this may fail to capture the entire domain of a con-
struct in a particular culture, the researcher is more assured of equivalent
comparisons. Moreover, the discussion section can explain to the reader the
broader domain of a construct in a particular culture and speculate as to how
that limits ndings.
References
Adams, J.S. (1963). Toward an Understanding of Inequity. Journal of Abnormal and Social
Psychology, 67: 42236.
Bazerman, M.H. (2002). Judgment in Managerial Decision Making, 5th edition. New York:
John Wiley & Sons.
Berry, J.W. (1980). Introduction to Methodology, in H.C. Triandis & J.W. Berry, editors,
Handbook of Cross-Cultural Psychology, volume 2. Needham Heights, MA: Allyn & Bacon.
Berry, J.W. (1989). Imposed Etics-Emics- and Derived Etics: The Operationalizations of a
Compelling Idea. International Journal of Psychology, 24: 721735.
Brett, J.M., C.H. Tinsley, M. Janssens, Z.I. Barnsness & A.L. Lytle. (1997). New Approaches
to the Study of Culture in I/O Psychology, in P. Christopher Earley and Miriam Erez,
editors, New Perspectives on I/O Psychology: 75129. San Francisco, CA: Jossey-Bass, Inc.
Cattell, R.B. (1988). Multivariate Method and Theory Construction, in J.R. Nesselroade.
Y.R.B. Cattell, editors, Handbook of Multivariate Experimental Psychology (2nd edition):
120. New York: Plenum.
Cook, T.D. & Campbell, D.T. (1979). Quasi Experimentation: Design and Analysis Issues for
Field Settings. Boston: Hougton-Mifin.
Duhem, P. (1962). The Aim and Structure of Physical Theory. New York: Atheneum.
Guba, E.G. & Lincoln, Y.S. (1994). Competing Paradigms in Qualitative Research, in N.K.
Denzin & Y.S. Lincoln, editors, Handbook of Qualitative Research, 105117. Thousand
Oaks, CA: Sage.
Hanson, N.R. (1958). Patterns of Discovery: An Inquiry into Conceptual Foundations of
Science. Cambridge, England: Cambridge University Press.
Hui, C.H. & Triandis, H.C. (1985). Measurement in Cross-Cultural Psychology: A Review and
Comparison of Strategies. Journal of Cross-Cultural Psychology, 16 (2): 131152.
James, L.R., Mulaik, S., & Brett, J.M. (1982). Causal Analysis: Assumptions, Models, and
Data. Thousand Oaks, CA: Sage.
350 CATHERINE H. TINSLEY
Kuhn, T.S. (1962). The Structure of Scientic Revolutions. Chicago: University of Chicago Press.
Moore, J.F., Johns, G., & Pinder, C.C. (1980). Towards Middle Range Theory: An Overview
and Perspective, in C.C. Pinder & J.F. Moore, editors, Middle Range Theory and the Study
of Organizations, 116. Boston: Nijhoff.
Lytle, A.L., Brett, J.M., Barnsness, Z.I., Tinsley, C.H. & Janssens, M. (1995). A Paradigm for
Quantitative Cross-Cultural Research in Organization Behavior, in B.M. Staw and L.L.
Cummings, editors, Research in Organizational Behavior, 17: 167214.
Popper, K.R. (1959). The Logic of Scientic Discovery. New York: Basic Books.
Spradley, J.P. (1979). The Ethnographic Interview. Chicago: Holt, Rinehart & Winston, Co.
Spradley, J.P. (1980). Participant Observation. New York: Holt, Rinehart, & Winston, Co.
Triandis, H.C. (1972). Major Theoretical and Methodological Issues in Cross-Cultural Psychology,
in W.J. Lonner Walters, editor, Readings in Cross-Cultural Psychology, pp. 2638.
Triandis, H.C. (1992). Cross-Cultural Research in Social Psychology, in D. Granberg &
G. Sarup (Eds.), Social Judgments and Intergroup Relations: Essays in Honor of Muzafer
Sherif. New York: Springer-Verlag.
Weingart, L.R. (1997). How did they do that? The Ways and Means of Studying Group
Processes, in L.L. Cummings & B.M. Staw, editors, Research in Organizational Behavior,
Vol. 19: 189239. Greenwich, CT: JAI Press.
Disparate Methods and Common Findings in the
Study of Negotiation
Conict and negotiation lie at the very heart of many different types of social
interaction, be it among chimpanzees, businessmen, diplomats, spouses, or
drug dealers and their junkies. Accordingly, conict and negotiation is studied
in many disciplines within the social sciences, including social and personal-
ity psychology, organizational behavior, communication sciences, economics,
and political science. Yet at the same time, it seems that within each of these
academic disciplines conict and negotiation is studied in very different ways,
that some methods and techniques are commonplace in some areas while unknown
or even frowned upon in adjacent areas. An example is the debate about the
usefulness of laboratory experiments in the study of conict and negotiation.
Proponents of the laboratory experiment argue that it provides high internal
validity and allows one to detect causal relationships (Pruitt 1981). Opponents
counter that laboratory experiments lead to a science of 18-year old psychol-
ogy students performing abstract tasks of short duration and only because
they have to. How could this provide any information about, for example,
peace negotiations between Israel and the PLO, or an ongoing conict between
a manager and her subordinates?
The problem with this type of debate is that it may lead scholars to dis-
count research ndings obtained with the wrong method. When some meth-
ods are more popular and more accepted in some academic disciplines
than in others, this could reduce the extent to which insights about conict
and negotiation transcend academic boundaries. An alternative would be to
assume that no method or technique is perfect and able to answer all ques-
tions at once. Triangulation the use of different methods to study the same
phenomenon offers a solution to deal with deciencies of each method and
technique (Campbell & Stanley 1966).
In this article, we address two tasks. First, we analyze the conict and
negotiation literature to provide an empirical prole of the variety of methods
and techniques in the various disciplines; we suspect that some methods are
more popular in some disciplines than others. Second, we selectively review
effects in conict and negotiation that have been studied with different meth-
ods to see whether triangulation indeed works and provides useful insight.
Table 2. Methods and Techniques in the Study of Negotiation and Social Conict
Observed Number Percentage
Use of Mathematical Modeling 124 35.9
Computer Simulation 2 0.6
Use of Games for Model Testing 82 23.8
Laboratory Experiment 161 46.7
Use of Outcome Measures 14 140.9
Coding of Communication Processes 41 11.9
Quantitative Coding of Behavioral Sequences 3 0.9
Use of Surveys and Questionnaires 134 38.8
Markov-Chain Analysis 3 0.9
Methods of Culture Comparison 18 5.2
Multi-dimensional Scaling 1 0.3
Multi-Level Analysis 1 0.3
Meta-Analysis 4 1.2
True Field Experiment 1 0.3
Field Research 35 10.1
Analysis of Archival Data 52 15.1
Time Series Analysis 3 0.9
Workshop as Research Tool 5 1.4
Participant Observation 1 0.3
Comparative Case Analysis 14 4.1
Internet as a Method 1 0.3
Methodology for Personality Processes 19 5.5
Note: Articles can be coded in multiple categories when multiple methods and techniques are
used (e.g., laboratory experiment and surveys and questionnaires). Total number of articles
coded = 345, with 1 article not tting into any of the above categories.
DISPARATE METHODS AND COMMON FINDINGS IN THE STUDY OF NEGOTIATION 355
Method X 2(4)
Common Findings
The above analysis supports the idea that some methods and techniques
(e.g., laboratory experiments, surveys) are more frequently used than others
(e.g., time series analysis, meta-analysis, coding of behavioral sequences).
Substantial differences between social science disciplines emerge as well.
Organizational behavior, social and personality psychology, and the commu-
nication sciences share a heavy reliance on laboratory experiments, the use of
surveys, and the coding of communication processes. These methods and
techniques are far less popular in economics and political science, where
mathematical modeling, the use of games for model testing, and the use of
archival data tend to dominate.
To some extent the differences in methods and techniques reect different
research questions scholars in different areas pursue. Questions with regard to
the economic correlates of actual war decisions cannot be answered with lab-
oratory experiments, whereas questions with regard to neurological correlates
of facing a competitive opponent require at this time a laboratory design.
Sometimes, however, scholars within different areas of the social sciences
pursue similar research questions, but with different methods and techniques.
As mentioned, the use of different methods and techniques to study similar
processes and phenomena may result in more condence in the validity and
robustness of these processes and phenomena (Campbell & Stanley 1966).
Furthermore, using different methods and techniques allows one to view a
particular phenomenon from different angles. To illustrate this, we discuss research
on gain-loss framing conducted in organizational behavior and social and per-
sonality psychology on the one hand, and in political science on the other,
and research on interaction sequences conducted in social psychology on the
one hand, and the communication sciences on the other.
way value, and the other half were presented with payoffs presented as less
than their highest possible outcome. Thus, although negotiators objective out-
comes were the same in both conditions, half were led to view these as gains,
and half were led to view these as losses. Consistent with the idea that loss-
framing makes people more risk-seeking and that losses are more painful than
gains are pleasant, results showed that dyads with a loss frame more often
reached an impasse and less often reached integrative, win-win solutions than
dyads with a gain frame. Since this original experiment, multiple studies have
replicated and extended these results, uncovering both underlying psycholog-
ical processes for the effects (Bottom 1997; De Dreu, Carnevale, Emans &
Van de Vliert 1994), as well as boundary conditions under which the framing
effect is mitigated or even reversed (e.g., De Dreu & McCusker 1997).
Jervis (1989), a political scientist, observed that states seem to make greater
efforts to preserve the status quo against a threatened loss than to improve
their position by a comparable amount. A state might be willing, for exam-
ple, to ght and defend the same territory that it would not have been will-
ing to ght to acquire, or to accept greater costs in order to maintain an
international regime than to create it in the rst place. Loss aversion losses
are more painful than equivalent gains are pleasurable could explain these
tendencies as much as it explains the laboratory ndings on gain-loss fram-
ing reviewed above (Levy 1992). A series of in-depth case studies substanti-
ates these claims. For example, Farnham (1992) demonstrates that loss
aversion may well explain Roosevelts decision-making behavior during the
Munich crisis in September 1938. McDermott uses prospect theory to explain
President Carters decision to rescue the American hostages in Iran in 1980.
He argues that Carter found himself in a loss-frame and took considerable risk
to regain the status quo.
Conclusion
Our data on the use of various methods in general and across disciplines sug-
gest two immediate observations: rst, the study of negotiation and social
conict is a very small part of some domains (about 2% of published articles
in social psychology as well as organizational behavior; about 8% in com-
munication), and closer to the core of other domains (political science and
economics). Of course, one reason for this may be the existence and attrac-
tion of the specialty outlets such as International Journal of Conict Management,
Group Decision and Negotiation, Journal of Conict Resolution, Negotiation
Journal, and of course, International Negotiation.
DISPARATE METHODS AND COMMON FINDINGS IN THE STUDY OF NEGOTIATION 359
Second, disciplines tend to maintain their own way of doing things: labo-
ratory experiments that entail coding of behavior and self-report data using
surveys are especially popular in psychology, organizational behavior, and
communication sciences; mathematical modeling, the use of experimental games,
and the use of archival data, are especially popular in economics, and politi-
cal science.
Diverse methods can be a source of convergent insights, yet they also can
be a barrier to understanding if effects observed in one domain are not easily
accessible to scholars in another. Such a barrier is less an issue when schol-
ars working in different areas are aware of the various methods and tech-
niques that exist and have some basic understanding of their virtues and limitations.
The consistency in research ndings with regard to gain-loss framing, and
reciprocity, suggests that different methods and techniques allow one to study
similar phenomena and to build a body of knowledge that would have been
impossible had the research community limited itself to a narrow set of set-
tings, populations, and research methods.
If our effort to organize these two special issues of International Nego-
tiation on method tells us anything, it is that researchers should adopt, or
should continue to employ, triangulation as an approach to validity. In navi-
gation, triangulation is the technique where two visible points are used to
determine the location of a third point. Applied to validity and reliability of
measurement, triangulation is the use of multiple, different indicators in such
a way that errors can be excluded and underlying constructs can be identied
(Campbell & Fiske 1959). The concept applies quite well not only to mea-
surement, but to all aspects of method. When two or more methods or data
sources converge on a construct, we have greater assurance that our conclu-
sions are not driven by an error or artifact of any one procedure. Each method
offers strengths and weaknesses, and to the extent they do not overlap, we can
stand on more solid ground with conclusions about theory. An exciting direc-
tion is the development of collaborative enterprises such as the multi-disci-
plinary research team where scholars with complementary expertise work together
on common problems, a successful model in the natural sciences (Hopmann
2002).
References
Aureli, F., & De Waal, F.B.M. (2000). (Editors). Natural conict resolution. Berkeley: Univer-
sity of California Press.
Bazerman, M.H., Magliozzi, T., & Neale, M.A. (1985). Integrative bargaining in a competi-
tive market. Organizational Behavior and Human Decision Processes, 35: 294313.
360 CARSTEN K.W. DE DREU AND PETER J. CARNEVALE
Bottom, W.P. (1998). Negotiator risk: Sources of uncertainty and the impact of reference
points on negotiated agreements. Organizational Behavior and Human Decision Processes,
76: 89112.
Campbell, D.T., & Fiske, D.W. (1959). Convergent and discriminant validation by the multi-
trait-multimethod matrix. Psychological Bulletin, 56: 81105.
De Dreu, C.K.W., & McCusker, C. (1997). Effects of gain-loss frames on cooperation in two-
person social dilemmas: A transformational analysis. Journal of Personality and Social
Psychology, 72: 10931106.
De Dreu, C.K.W., Carnevale, P.J.D., Emans, B.J.M., & Van de Vliert, E. (1994). Effects of
gain-loss frames in negotiation: Loss aversion, mismatching, and frame adoption. Organizational
Behavior and Human Decision Processes, 60: 90107.
Deutsch, M. (1949). A theory of cooperation and competition. Human Relations, 2: 199231.
Farnham, B. (1992). Roosevelt and the Munich Crises: Insights from Prospect Theory.
Political Psychology, 13: 205235.
Jervis, R. (1989). The meaning of the nuclear revolution. Ithaca, NY: Cornell University Press.
Gouldner, A.W. (1960). The norm of reciprocity: A preliminary statement. American
Sociological Review, 25: 161178.
Hopmann, P.T. (2002). Negotiating data: Reections on the qualitative and quantitative analy-
sis of negotiation processes. International Negotiation, 7: 6785.
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk.
Econometrica, 47: 263291.
Komorita, S.S., Parks, C., & Hulbert, L. (1992). Reciprocity and the induction of cooperation
in social dilemmas. Journal of Personality and Social Psychology, 62: 607617.
Leng, R.J. (1993). Interstate crisis behavior, 18161980: Realism versus reciprocity. New York,
NY: Cambridge University Press.
Levinger, G., & Rubin, J.Z. (1994). Bridges and barriers to a more general theory of conict.
Negotiation Journal, (July): 201215.
Levy, J.S. (1992). Prospect theory and international relations: theoretical applications and ana-
lytical problems. Political Psychology, 13: 283311.
McDermott, R. (1992). The failed rescue mission in Iran: An application of prospect theory.
Political Psychology, 13: 1629.
Olekalns, M. & Smith, P. (2000). Understanding optimal outcomes: The role of strategy
sequences in competitive negotiations. Human Communication Research, 26: 527557.
Ostrom, E. (1998). A behavioral approach to the rational choice theory of collective action.
The American Political Science Review, 92: 122.
Pruitt, D.G. (1981). Negotiation. New York: Academic Press.
Pruitt, D.G. (1998). Social conict, in D. Gilbert, S.T. Fiske, & G. Lindzey (Editors).
Handbook of social psychology (4th ed.), 2. New York: McGraw-Hill: 89150.
Weingart, L.R., Prietula, M.J., Hyder, E.B., & Genovese, C.R. (1999). Knowledge and the
sequential processes of negotiation: A Markov Chain analysis of response-in-kind. Journal
of Experimental Social Psychology, 35: 366393.
INDEX