Journal of Personality and Social Psychology Copyright 2005 by the American Psychological Association

2005, Vol. 89, No. 3, 345–357 0022-3514/05/$12.00 DOI: 10.1037/0022-3514.89.3.345

The Possessive Self as a Barrier to Conflict Resolution: Effects of Mere

Ownership, Process Accountability, and Self-Concept Clarity on
Competitive Cognitions and Behavior

Carsten K. W. De Dreu Daan van Knippenberg

University of Amsterdam Erasmus University Rotterdam

The authors propose that people have difficulty managing conflict because they quickly develop
ownership of arguments and positions they use in the dispute, that these arguments and positions become
part of their (extended) self-concept, and that any opposition or counterargumentation therefore becomes
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

an ego-threat. Four studies reveal that individuals value arguments and beliefs more when these are
This document is copyrighted by the American Psychological Association or one of its allied publishers.

associated with the self and that anticipated or real opposition triggers ego-defensive cognition and
behavior, including competitive communication, retaliatory responses, negative perceptions of the
partner, and attitude polarization. These effects were weaker when epistemic needs were raised through
process accountability or when individuals had high rather than low self-concept clarity. The authors
conclude that because people develop ownership of arguments and make these part of their self-concept,
conflict is difficult to manage and bound to escalate.

Keywords: conflict resolution, self, aggression, cooperation, social interaction

Conflict is part and parcel of all forms of social life and quite arguments and positions, that these arguments and positions be-
difficult to manage. Often conflict even escalates into competitive come part of their self-concept, and that any threat or opposition to
cycles of exceedingly hostile exchanges (Rubin, Pruitt, & Kim, these arguments thus represents a threat to the self. This ego-threat
1994). For example, marital disputes about division of household (Baumeister, 1998) then triggers hostility, competitive cognitions,
chores may end in wife battering (Baumeister, Smart, & Boden, and exceedingly difficult and tenacious exchanges. In other words,
1996); parliamentary debates sometimes end in shouting games we suggest that conflict in and by itself produces ego-threat and is,
and fist fighting; and work teams manage their task-related con- therefore, destined to become difficult to manage and bound to
flicts so that their effects are as detrimental as the more self- escalate.
relevant conflicts about deeply held political values and beliefs Our take on the role of the self in conflict breaks down into three
(De Dreu & Weingart, 2003). interrelated positions that we develop and test in the four experi-
But why is managing conflict so difficult? Why do mild disputes ments reported in this article. First, we argue and show that people
about seemingly unimportant matters so readily escalate into quickly and effortlessly develop ownership of arguments and
heated conflict? Why do businesslike exchanges of arguments so attitude positions, that these arguments and positions become part
often give way to anger and frustration, and why do we respond to of the extended self, and that because people desire a positive view
minor misunderstandings much more emotionally than we would of the self, arguments and positions associated with the self are
like to? Why is it that our counterparts so often take our comments valued more positively than otherwise identical arguments and
much more personally than we intended them to be? Why, in short, positions not associated with the self (Studies 1– 4). Second, we
is it that conflicts so often and so easily become personal and show that anticipating conflict and opposition amplifies these mere
emotional? ownership tendencies (Study 1). Third, we demonstrate that be-
To answer these questions, we developed the idea that in any cause arguments become part of the extended self, any (antici-
type of conflict or dispute, people instantly identify with their pated) opposition to these arguments represents an ego-threat,
which in turn triggers competitive communication, retaliatory re-
sponses, negative perceptions of one’s conflict partner, and atti-
Carsten K. W. De Dreu, Department of Psychology, University of tude polarization (Studies 2– 4).
Amsterdam, Amsterdam, the Netherlands; Daan van Knippenberg, Rot- By deepening our understanding of the fundamental role of the
terdam School of Management, Erasmus University Rotterdam; the self in predicting conflict escalation, we make four contributions.
Netherlands. First, we expand research on mere ownership (e.g., Beggan, 1992)
This research was financially supported by a Royal Netherlands Acad- by showing that the extended self not only involves material items
emy of Sciences fellowship awarded to Carsten K. W. De Dreu. We thank
but also nonmaterial items such as arguments and attitude posi-
Gerben van Kleef and Bram Duyx for their help in collecting the data and
Paul Van Lange and Craig Parks for their comments and suggestions.
tions. Second, we expand work on ego-threat, ego-defensive bol-
Correspondence concerning this article should be addressed to Carsten stering, and hostility (e.g., Brehm, 1966; Bushman & Baumeister,
K. W. De Dreu, Department of Psychology, University of Amsterdam, 1998; Carver & Scheier, 1981; Exline, Baumeister, Bushman,
Roetersstraat 15, 1018 WB Amsterdam, the Netherlands. E-mail: Campbell, & Finkel, 2004; Greenwald, 1968; Mackie, Devos, &
[email protected] Smith, 2000; Mo’az, Ward, Katz, & Ross, 2002; Paese & Yonker,

345
 346 DE DREU AND VAN KNIPPENBERG

2001) by showing that it is the extended self as much as the core exposure, because they were involved in generating them, or
self that may be prone to threat and serves as a trigger of hostility because they already publicly defended them. The mere ownership
and aggression. Third, whereas this earlier work demonstrated that effect suggests that people identify with arguments they are asso-
conflict escalates when the self is threatened, we show why ego- ciated with and develop an egocentric and positive evaluation of
threat emerges in the first place, even when the conflict is the most these arguments regardless of any other reason.
simplistic and people dispute totally irrelevant issues. Fourth, and To determine whether the association with the self adds above
finally, although we maintain that conflict is bound to escalate and beyond other grounds to favor arguments and beliefs, we
because in and by itself it raises ego-threat, we acknowledge that resort to the methodology advanced by research on mere owner-
boundary conditions exist. In Study 2, we examine the moderating ship and reluctance to trade (e.g., Beggan, 1992; Kahneman et al.,
influence of process accountability, a situation variable known to 1990). If we assess an individual’s perception of an argument or
reduce self-enhancement tendencies (Sedikides, Herbst, Hardin, & belief associated with the self and compare this with the individ-
Dardis, 2002). In Study 4, we examine the moderating influence of ual’s perception of an otherwise identical argument or belief not
self-concept clarity, a personality variable known to make people associated with the self, we would expect that the same argument
more or less defensive in response to ego-threat (Stucke & Spore, or belief is valued more when it is associated with the self. The
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

2002). Whereas invoking mere ownership explains why conflict is first goal of Study 1 was to test the idea that, everything else kept
difficult to manage, identifying these moderator variables high- constant, people value arguments they are associated with (i.e.,
lights when, and for whom, conflict management is particularly that they own) more than otherwise identical arguments they are
challenging. not associated with (i.e., that they do not own). Support for this
Below, we first develop the notion of mere ownership and the idea is important because it would indicate that even when the
extended self, and we explain why we think that conflict and issues at stake bear no direct link to the self, through mere
opposition amplifies this tendency. We test hypotheses in Study 1 ownership the self-concept gets implicated.
that form the basis of subsequent experiments in which we exam-
ined the implications of mere ownership and the extended self for Conflict as Self-Threat
competitive cognition and behavior in the anticipation and man-
agement of conflict and opposition. The second goal of Study 1 was to examine the influence of
anticipating conflict on the strength of mere ownership effects.
Mere ownership studies thus far have considered settings void of
Study 1 conflict, but there is reason to expect mere ownership effects to be
Mere Ownership of Arguments stronger in conflict situations. Conflict involves threat, and threats
to the (extended) self produce a momentary drop in positive affect
Evaluations of the self are positively biased: People ascribe (e.g., Jussim, Yen, & Aiello, 1995). To escape this uncomfortable
positive traits to themselves, see negative traits as more descriptive state, people engage in self-enhancement (W. K. Campbell &
of others, and evaluate themselves as more moral and honest than Sedikides, 1999). Conflict also involves competition and social
the average other (Baumeister, 1998). This egocentric, inflated comparison processes, and self-enhancement tends to be stronger
view of the self extends to objects associated with the self. People in competitive settings. W. K. Campbell and Sedikides (1999)
have greater liking for objects they possess, compared with oth- explain this by arguing that “when a participant has a competitive
erwise identical objects they do not own (Heider, 1958). They also orientation, the participant is concerned with the magnitude of his
ask more money to give up something they own (e.g., a coffee or her own contribution and seeks to differentiate his or her
mug) compared with what they are willing to pay to acquire the performance from the competitor’s performance. These concerns
same object (Carnevale, 1995; Kahneman, Knetch, & Thaler, will instigate social comparison processes and will result in ele-
1990). vated levels of self-threat” (p. 27).
Although part of this “ownership effect” may be explained in The work summarized by W. K. Campbell and Sedikides (1999)
terms of loss aversion— giving up an object induces greater pain considered competitive settings and self-serving attributions for
than receiving that same object induces pleasure (Tversky & success and failure, not mere ownership effects in conflict. How-
Kahneman, 1979)— egocentric tendencies to overvalue objects ever, because threat and a competitive orientation are part and
associated with the self cannot be ruled out. Mere ownership parcel of conflict situations, their work implies Hypothesis 1:
effects occur without the possibility of loss (e.g., Nesselroade, People value arguments they own more than otherwise identical
Beggan, & Allison, 1999), and they are stronger for people in high arguments they do not own, especially when they anticipate
need of self-enhancement (Beggan, 1992). conflict.
We suspect that ownership effects also extend to arguments and
beliefs in debate and conflict situations. In fact, ownership of Method
arguments and beliefs may be quite common. “That idea was
mine” is a feeling many of us can relate to (Pierce, Kostova, & Overview. To test Hypothesis 1, we used the methodology used in
previous mere ownership studies (Beggan, 1992; Kahneman et al., 1990).
Dirks, 2003), and ownership of arguments and beliefs can be seen
Participants were provided with an attitude topic and were told that they
in expressions such as “to have a belief,” “to acquire a belief,” and would receive one of two sets of arguments (Set A and Set B), which were
“to surrender your principles” (Abelson, 1986, p. 230). However, pretested to be equally valid and persuasive. After they were shown one
valuing one’s own arguments and beliefs more than the other’s set, participants were asked to set a price for which they would be willing
arguments and beliefs does not necessarily reflect mere ownership. to sell the arguments to another participant, under the explicit understand-
People may favor their arguments and beliefs because of mere ing that the deal (i.e., arguments for money) would be closed if the other
 SELF IN SOCIAL CONFLICT 347

participant was willing to pay that price (cf. Kahneman et al., 1990). To facing the same task as they did, but instead of having received arguments,
contrast valuation of arguments owned and not owned, participants were they had money available to buy a set of arguments.
asked to set a price for both sets of arguments. Before they set their prices, Following the procedure developed by Kahneman et al. (1990), participants
participants were led to anticipate the writing of an essay about the attitude were told that the other persons would be asked to indicate the maximum
topic or a debate about this topic. According to Hypothesis 1, mere amount of money they would be willing to spend to buy a set of arguments.
ownership effects should be stronger in the debate condition. Participants were asked to indicate the minimum amount of money they would
Design and participants. We used a Task (essay vs. debate) ⫻ Own- ask for a set of arguments to be sold. They were told that they were randomly
ership (Set A vs. Set B) ⫻ Selling Order (Set A first vs. Set B first) ⫻ Price linked with 1 participant who had money instead of arguments and that if their
(Set A vs. Set B) factorial design, with the last factor within-participant. minimum price was equal to or below the maximum offered by this other
Dependent variables were the minimum price asked for Set A and for Set participant, a transaction would take place such that they would “receive the
B and manipulation checks. Participants were 95 students at the University money to take home, and the other participant receives the arguments.” It was
of Amsterdam. About half of them (46%) were men (exploratory analyses emphasized that whether or not a transaction would take place, participants
revealed no effects for gender). Participants received 10 guilders (approx- could always continue to use the set of arguments they owned. Thus, our
imately 4 U.S. dollars) for participation and were randomly assigned to procedure differed from the one used in prior research only in the conse-
conditions. quences of setting a price. In past research, matching prices led to an exchange
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Procedure and independent variables. For each session, 4 to 8 partic- of money for property. Here, matching prices led to an exchange of money for
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ipants were invited to the laboratory. On arrival, they were seated in arguments—participants would still be able to use the arguments (if they
individual cubicles in front of a computer on which all instructions and owned them), but they would lose unique possession. This procedural
tasks were presented; responses could be given by means of the computer deviation from earlier mere ownership studies reflects what happens in
keyboard. Participants were told that the study was concerned with the many cases in which intellectual property is sold (e.g., with transfer of
ways people think about the general dilemma between economic growth copyright, the author can continue to use his or her work, and selling patent
and preservation of the natural environment.1 rights often includes a clause that the seller can continue to produce). We
Manipulation of task. Participants in the essay condition were then told return to this in Study 4.
that they would be asked to write a brief essay in which they would defend Participants were asked to indicate “the minimum price for which [they]
the position that “we all should accept reduced economic growth to would sell Set A (or Set B, depending on selling order), which is the set
preserve the natural environment.” Participants in the debate condition that [they] (do not) own.” A table listed prices (Hfl 5.00; 4.50; 4.00, . . . ,
were told that they would be asked to defend the position that “we all 0.00), and with each price a sentence read, “For this price I would sell the
should accept reduced economic growth to preserve the natural environ- arguments YES/NO.” For each price, participants were asked to indicate
ment” in a debate with another participant. They were told that the debate their decision. The higher the number of “no” answers, the higher the price
would take place over the computer network and that they would not see participants asked for this set of arguments (A or B, depending on selling
order). The procedure was repeated to have participants indicate “the
the other person.
minimum price for which [they] would sell Set B (or Set A, depending on
Manipulation of argument ownership. Participants were told, “Be-
selling order), which is the set that [they] (do not) own.” Of importance, we
cause the dilemma between economic growth and preservation of the
emphasized that they were linked to another participant than the one they
natural environment may be a topic you haven’t thought about a great deal,
were dealing with in the prior task by explaining, “This person is not the
we decided to prepare several sets of arguments supporting the view you
same as the one you have been dealing with before.” The persons they were
have to defend.” Participants learned further that we had “developed two
dealing with were not competing with them in any way (i.e., we would not
sets of arguments, A and B, that are highly similar and that have been
be comparing the essays they wrote in the essay condition or the way they
pretested extensively to ensure they are identical in length, validity, and
debated in the debate condition).
persuasiveness.” After a brief delay, participants in the Set A (or Set B)
On completion of the pricing tasks, participants responded to manipu-
condition were told that the computer had determined, on a random basis,
lation checks (see the Results section), were informed that the experiment
that Argument Set A (or Set B) would be theirs to use and that Argument
was over, and were debriefed. Participants received Hfl 2.50 extra because
Set B (or Set A) would not be theirs. To emphasize this, we told them,
we did not continue with the selling task. Analyses of the responses to our
“You own Argument Set A (or Set B), which is highly similar to Argument
checks indicated that participants correctly indicated that the transactions
Set B (or Set A), which you do not own.”
were with different persons and that there was no competition of any kind
Hereafter, participants were told that the computer had randomly se-
between them. Participants also correctly indicated that they had to write an
lected one set of arguments to be shown to them first. Participants were essay or enter a debate with another participant and that whether or not they
shown Set A, which contained four arguments supporting the view that one established an exchange had consequences for the amount of money they
should sacrifice economic prosperity to save the environment. Participants would take home, but would not affect their chance to use the arguments.
were reminded that they did or did not own Set A and were given a few
minutes to read the arguments. We showed Set A for two reasons. First, we
wanted to examine whether mere ownership effects develop in the absence Results
of actual knowledge of the content of the arguments. If valuation of an
argument set depends only on whether participants owned the set and is not Selling prices. We submitted the price asked for Set A and Set
influenced by whether they saw the set, we can conclude that content B to a 2 (task) ⫻ 2 (ownership) ⫻ 2 (selling order) ⫻ 2 (price for
knowledge is no prerequisite for mere ownership effects to emerge. Sec- Set A vs. Set B) multivariate analysis of variance (MANOVA),
ond, we wanted to verify possible demand characteristics in argument with the last factor within-participant. Overall price was higher in
valuation. We reasoned that participants should be equally likely to con- the debate compared with the essay condition (M ⫽ 6.00 vs. M ⫽
clude that we expect them to value the set shown more than the argument 4.75), F(1, 87) ⫽ 6.14, p ⬍ .015, and an interaction between
set not shown. If we find differential valuation as a function of ownership
but not as a function of whether the argument set had been shown, demand
1
characteristics would become a less parsimonious explanation of our In all experiments, we measured prior attitudes and issue involvement.
results (see also Conclusions and General Discussion). Ratings never differed as a function of experimental manipulations and
Selling prices and selling order. Once participants had received a set were never correlated with other measures. Further details can be obtained
of arguments, they were told that some other participants present were from De Dreu.
 348 DE DREU AND VAN KNIPPENBERG

selling order and price, F(1, 87) ⫽ 34.84, p ⬍ .001, showed that ownership of arguments is due to motivated construal, conflict
higher prices were set for arguments sold first (MA ⫽ 6.44 and parties under process accountability should display mere own-
MB ⫽ 5.95) rather than last (MA ⫽ 4.70 and MB ⫽ 4.46), both ership to a lesser degree than parties not held accountable
ts(92) ⬎ 2.50, ps ⬍ .025. (Hypothesis 2).
Hypothesis 1 predicted higher prices for arguments owned, Whereas in Study 1 we provided first-time evidence for the idea
especially when individuals anticipate conflict. Results indeed that conflict amplifies mere ownership effects, in Study 2 we
revealed an interaction between ownership and price, F(1, 87) ⫽ examined whether and how mere ownership influences people’s
28.61, p ⬍ .001, which was qualified by an interaction among task, behavior toward their conflict partner. In conflict, two types of
ownership, and price, F(1, 87) ⫽ 3.86, p ⬍ .06 (marginal). Simple behavioral moves are important. Initial moves such as opening
effects analysis showed that in the essay condition, the interaction statements and initial demands are critical because they provide
between price and ownership only approached significance, F(1, boundaries to the counterpart’s repertoire of responses and set the
87) ⫽ 3.81, p ⬍ .10, whereas in the debate condition the interac- stage for future conflict interaction (Rubin et al., 1994). Responses
tion was significant, F(1, 87) ⫽ 20.73, p ⬍ .001 (see also Table 1). to the counterpart’s behavior are critical because they punish or
A Duncan test ( p ⬍ .05) showed that participants anticipating a
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

reward the other for his or her prior behavior (e.g., Axelrod &
This document is copyrighted by the American Psychological Association or one of its allied publishers.

debate set a higher price for arguments owned compared with Hamilton, 1981). Study 2 involved initial moves, and Studies 3 and
arguments not owned (see Table 1). 4 involved responses to the other’s behavior.
There is much evidence to suggest that ego-threat predicts
Discussion and Introduction to Study 2 hostility. Threats to inflated self-views link to a preference for
advantageous inequity (Diekman, Samuelson, Ross, & Bazerman,
Study 1 revealed a mere ownership effect with regard to argu-
ments, which is consistent with past research looking at ownership 1997), to escalatory behavior and impasse in negotiation (e.g.,
of coffee mugs (Kahneman et al., 1990) or bottles of wine (Van Paese & Yonker, 2001), to verbal aggression (Bushman &
Dijk & van Knippenberg, 1998). This means that the method used Baumeister, 1998), to refusal to forgive (Exline et al., 2004), and
here yields results consistent with those found in other studies to derogation of out-group members (e.g., Mackie et al., 2000;
using similar or different methods. Mo’az et al., 2002). Although this work is all about threats to the
The mere ownership effects observed in Study 1 may reflect core self, social influence research shows that anticipating coun-
transfer of affect, in that people feel good about themselves and terargumentation and opposition produces defensive bolstering
thus, by definition, about anything that is associated with the especially when the self is temporarily or chronically accessible
self. Alternatively, it may be that mere ownership of arguments (e.g., Brehm, 1966; Carver & Scheier, 1981; Greenwald, 1968).
is the result of motivated construal—people want to feel good Thus, it seems reasonable to assume that threats to excessively
about themselves, and thus they are motivated to perceive positive views of arguments associated with the self are linked to
anything associated with the self as good, valuable, and positive competitive cognition and behavior.
(Cohen, 2003). Study 1 provided indirect support for such Negotiation studies have shown that the extent to which judg-
motivated construal, in that anticipating opposition amplified mental biases influence behavior is moderated by process account-
mere ownership effects (cf. Hypothesis 1). In Study 2, we tested ability (De Dreu, Koole, & Steinel, 2000). Epistemic motivation
this idea in another way. If mere ownership of arguments and preemptive self-criticism do not directly affect competitive-
reflects motivated construal, factors known to reduce such ness but instead make people reluctant to apply their knowledge
motivated construal should (also) reduce mere ownership ef- and to make inferences on the basis of available information (De
fects. One such factor is process accountability. Under process Dreu & Carnevale, 2003). In other words, process accountability
accountability, individuals expect to be observed and evaluated not only reduces bias (cf. Hypothesis 2) but also the translation of
by others with unknown views about the process of judgment bias into behavior. In Study 2, after participants had set their prices
and decision making (Lerner & Tetlock, 1999). Process ac- for the arguments, we allowed them to send competitive messages
countability produces preemptive self-criticism, more even- to their counterparts. We predicted a positive relationship between
handed evaluation of decision alternatives, and reduced self- mere ownership and competitive messages, but less so under
enhancement tendencies (Sedikides et al., 2002). If mere process accountability (Hypothesis 3).

Table 1
Selling Price as a Function of Ownership, Task, and Argument Set (Study 1)

Task

Writing essay Debate

Selling price asked Own A Own B Own A Own B

Price for A 4.52a (3.02) 4.66a (3.13) 6.62b (2.92) 5.33a (2.91)
Price for B 3.95a (3.12) 5.87b (3.13) 4.70a,b (2.85) 7.41c (2.80)

Note. Cell means in one row not sharing the same superscript differ at p ⬍ .05, according to Duncan tests.
Standard deviations are given in parentheses.
 SELF IN SOCIAL CONFLICT 349

Method Selling prices. Prices for Set A and for Set B were submitted
to a 2 ⫻ 2 ⫻ 2 ⫻ 2 MANOVA, with process accountability
Design and participants. The design was a 2 (process accountability:
(present vs. absent); ownership (Set A vs. Set B); and selling order
present vs. absent) ⫻ 2 (ownership: Set A vs. Set B) ⫻ 2 (selling order: Set
A first vs. Set B first) between-participants factorial. Dependent variables (Set A first vs. Set B first) as between-participants variables and
were the minimum price asked for Set A and for Set B, the number of price (Set A vs. Set B) as the within-participant variable. Results
competitive messages sent to the partner, and manipulation checks. Par- revealed a Selling Order ⫻ Price interaction, F(1, 291) ⫽ 55.88,
ticipants were 299 students at the University of Amsterdam. About one p ⬍ .001, showing higher prices for arguments sold first (MA ⫽
third (36%) were men (exploratory analyses revealed no effects involving 5.48 and MB ⫽ 5.99) than for those sold second (MA ⫽ 4.46 and
gender). Participants received 10 Dutch guilders for participation and were MB ⫽ 4.57, ts ⬎ 3.0, ps ⬍ .01).
randomly assigned to conditions. Hypothesis 2 predicted that mere ownership effects would be
Procedure and independent variables. The task, procedures, and ma-
weaker when participants were under process accountability. In-
nipulation of argument ownership and selling order were the same as in
Study 1, except that all participants were in the debate condition. Process deed, a significant interaction between ownership and price, F(1,
accountability was manipulated as in past research (De Dreu et al., 2000). 291) ⫽ 19.38, p ⬍ .001, was qualified by a significant interaction
between ownership, price, and process accountability, F(1, 291) ⫽
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

We gave participants in the process accountability present condition the

This document is copyrighted by the American Psychological Association or one of its allied publishers.

following instructions just before they would set their price for Argument 3.97, p ⬍ .05. Simple effects analyses revealed that under process
Set A (or Set B, depending on selling order): accountability, the interaction between ownership and price was
not significant, F(1, 291) ⫽ 2.87, p ⬍ .12, whereas it was strong
Before you decide the price for the set of arguments, please keep in
and significant when process accountability was absent, F(1,
mind why you are doing so. After the experiment, we will interview
291) ⫽ 16.11, p ⬍ .001. Duncan tests ( p ⬍ .05) showed that when
you in detail about the reasons and considerations you had for setting
a particular price. We are not so much interested in the actual price, process accountability was absent, arguments owned were valued
but more in the underlying reasons you had for setting it at that level. more than arguments not owned (see Table 2). These results
If you want, you can make some notes using the scrap paper placed support Hypothesis 2.
beside your computer, and you can bring these notes to the interview. Competitive messages. Hypothesis 3 predicted that mere own-
ership would affect competitiveness but less so under process
Participants in the process accountability absent condition received no such accountability. To test this hypothesis, we conducted a series of
instructions and proceeded immediately with setting the price for the set of moderated regressions in which ownership (the price set for owned
arguments.
arguments minus the price set for arguments not owned), process
Dependent variables. Pricing and manipulation checks were the same
as in Study 1, the only difference being that after setting their prices,
accountability, and the interaction between ownership and process
participants were allowed “to send messages to the other party in order to accountability served as predictor variables.3 Main effects were
set the stage for the debate that will start in a few minutes.” Participants entered in the first step, and the interaction term was always
were presented with a list of nine prewritten messages, and they were asked entered in the second step. The dependent variable was the amount
to select one message to send to the partner. The list contained three of competitive messages sent.
competitive messages (“I’m interested in winning, not in what you think”; Results revealed a significant model, F(3, 295) ⫽ 3.38, p ⬍
“Don’t count on me giving in one bit,” and “I wouldn’t wait for me to .015. Ownership predicted the number of competitive messages,
become angry, if I were you”), three cooperative messages (e.g., “Hi, let’s
␤ ⫽ .45, t(296) ⫽ 2.31, p ⬍ .025 (constant b [SEb] ⫽ 0.18 [0.09]).
work together on this topic and see where we agree”), and three learning-
The hypothesized interaction term was significant also, ␤ ⫽ ⫺.31,
oriented messages (e.g., “I’m curious to learn about your position in this
debate”). After participants had selected a message, we repeated the t(296) ⫽ ⫺1.65, p ⬍ .05 (one-tailed). The effect is not very strong,
procedure another three times. From the four messages participants sent, but regression is notoriously likely to result in Type I error when
we computed a competitive messages index (ranging between 0 and 4).2 testing for interaction effects (McClelland & Judd, 1993). Indeed,
Hereafter, we proceeded as in Study 1. As in Study 1, at the very end of the as predicted in Hypothesis 3, the regression of competitive mes-
experiment, participants correctly indicated that the transactions were with
different persons and that there was no competition of any kind between
2
them. In addition, all participants correctly indicated they were about to We also computed a cooperative messages index (ranging between 0
enter a debate with another participant. Finally, all participants correctly and 4) and a learning oriented messages index (ranging between 0 and 4).
indicated that whether or not they would establish a transaction had Because these indices are interdependent (sending more competitive mes-
consequences for the amount of money they would take home, but it would sages implies fewer learning-oriented or cooperative messages), and be-
not affect their chance to use the arguments. cause cooperative messages can be seen as the flip side of competitive
messages, only competitive and learning-oriented messages were analyzed.
The number of competitive messages is the present focus, and we report
Results results only for this index. Analysis of cooperative or learning-oriented
Manipulation checks. To check the adequacy of the process messages can be obtained from De Dreu.
3
accountability manipulation, we asked participants at the end of The use of difference scores in regression is problematic when (a)
the experiment whether they felt accountable for the judgment and internal reliability of the measurement differs across variables and/or (b)
the variance in one variable deviates from the variance in the other.
decision-making process (1 ⫽ not at all to 5 ⫽ very much). An
Differences in internal reliability are not an issue in the current case, and
analysis of variance (ANOVA) only revealed a main effect for differences in variance were not significant, Cochran’s C(149, 2) ⫽ 0.54,
process accountability, showing that participants in the account- p ⬍ .34. Furthermore, hierarchical regression analyses in which price for
ability present condition felt more accountable than those in the argument set not owned was entered in Step 1, and price for argument set
accountability absent condition (M ⫽ 2.34 vs. M ⫽ 3.80), F(1, owned was entered in Step 2, yielded similar results and identical conclu-
291) ⫽ 40.65, p ⬍ .001. sions as the regression analyses that were based on difference scores.
 350 DE DREU AND VAN KNIPPENBERG

Table 2
Selling Price as a Function of Ownership, Process Accountability, and Argument Set (Study 2)

Process accountability

Present Absent

Selling price asked Own A Own B Own A Own B

b a c
Price for A 5.25 (3.02) 4.87 (3.13) 6.07 (2.92) 4.63a (2.91)
Price for B 4.42a (3.12) 4.65a,b (3.13) 5.15b (2.85) 5.87c (2.80)

Note. Cell means in one row not sharing the same superscript differ at p ⬍ .05, according to Duncan tests.
Standard deviations are given in parentheses.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

sages on ownership was positive and significant when process 3 were shown Set A, and all were sold Set A first. The design was a 2
This document is copyrighted by the American Psychological Association or one of its allied publishers.

accountability was absent, b ⫽ .25, t(296) ⫽ 2.89, p ⬍ .005, but (ownership: Set A or Set B) ⫻ 2 (other’s feedback on Set A: positive or
not when process accountability was present, b ⫽ .07, t(296) ⬍ negative) between-participants factorial, with the same dependent variables
1.00, ns. as in Study 2. Participants were 44 students at the University of Amster-
dam. About one third (32%) were men (exploratory analyses revealed no
effects involving gender). Participants received 10 Dutch guilders for
Discussion and Introduction to Study 3 participation and were randomly assigned to conditions.
Procedure and independent variables. The task, procedures, and ma-
Study 2 showed that mere ownership effects are reduced when
nipulation of argument ownership were the same as in Study 2, except that
people are held accountable, which supports the idea that epistemic all participants were placed in the no accountability condition. After
motivation and preemptive self-criticism reduce self-enhancement participants had set their price for Set A and Set B, they were told that the
tendencies (e.g., Sedikides et al., 2002) and that mere ownership of researchers were interested in how debates develop when opponents have
arguments is due to motivated construal. In addition, and central to versus do not have insight into their counterpart’s initial arguments. There-
the current thesis, results showed that when process accountability fore, some participants would face an opponent who had been shown
is absent, mere ownership is associated with competitive behavior. Argument Set A as well and had been asked to state his or her opinion
This finding lends further support for the idea that the (extended) about it. After a few seconds, participants in the positive feedback condi-
self provides a foundation for competitive and escalatory conflict tion received the following statement by the other party: “This set of
behavior. arguments is really good. This debate is going to be interesting, as these are
clever arguments.” Participants in the negative feedback condition received
Whereas Study 2 was concerned with initial moves, Study 3 was
the following statement by the other party: “This set of arguments is really
designed to gain insight into the consequences of ego-threat on
stupid. This debate is going to be interesting, as these are bad arguments.”
responses to the partner’s hostile behavior. In general, people tend Although a main effect of this manipulation may be attributed to a number
to reciprocate the other’s behavior, especially the other’s nonco- of differences between the positive and the negative comment, the critical
operative behavior. Reciprocating the other’s noncooperative be- test in our study is that there is a differential reaction to these messages
havior serves one’s own greed, and it also protects against being depending on ownership. Hereafter, the experiment proceeded as in Study
exploited. When the counterpart is (expected to be) cooperative, 2. As before, comprehension checks revealed that participants understood
people may be tempted to exploit the other out of greed and the task correctly.
mismatch rather than match the other’s cooperativeness (Erev &
Roth, 2001; Komorita, Parks, & Hulbert, 1992; Ostrom, 1998; Results
Pruitt, 1998).
Selling prices. The price for Set A and for Set B were sub-
In Studies 3 and 4, we focus on the partner’s noncooperative
mitted to a 2 (ownership) ⫻ 2 (other’s feedback) ⫻ 2 (price)
behavior and generally expected reciprocity. The tendency to
ANOVA, with the last factor within-participant. We did not expect
reciprocate noncooperative behavior has been argued to fuel con-
any effects of the other’s feedback because this variable was
flict escalation (Rubin et al., 1994). However, it has also been
manipulated after prices had been set. Indeed, the only significant
shown that individuals react with more ego-defensive bolstering
effect was the predicted interaction between price and ownership,
when the self is chronically or temporarily salient (e.g., Carver &
F(1, 40) ⫽ 9.63, p ⬍ .004, showing a trend for Set A to be valued
Scheier, 1981). For example, Wood, Pool, Leck, and Purvis (1996)
more when it was owned (M ⫽ 6.91) than not owned (M ⫽ 5.68),
showed that opposition by persuasive sources had an impact only
t(42) ⫽ 1.65, p ⬍ .10, whereas Set B was valued more when it was
when these sources were judged self-relevant. We thus predicted
owned (M ⫽ 7.26) than not owned (M ⫽ 5.37), t(42) ⫽ 1.98, p ⬍
that the counterpart’s competitive remarks are reciprocated to a
.05. This effect is similar to the one found in the debate condition
higher degree when the remarks concern arguments associated
of Study 1 and the debate–no process accountability condition of
with the self, rather than identical ones not associated with the self
Study 2.
(Hypothesis 4).
Competitive messages. The numbers of competitive messages
were submitted to a 2 (set owned) ⫻ 2 (other’s feedback) analysis
Method of variance. Results revealed an interaction between set owned and
Design and participants. Because the first two experiments showed no other’s feedback, F(1, 40) ⫽ 6.41, p ⬍ .015. Simple effects
interaction between selling order and ownership, all participants in Study analysis showed that when participants owned Set B, the other’s
 SELF IN SOCIAL CONFLICT 351

feedback about Set A had no effects, F(1, 40) ⬍ 1. When they individuals in whom explicit and implicit self-esteem correspond
owned Set A, the other’s derogative remarks about Set A resulted (Jordan, Spencer, Zanna, Hoshino-Browne, & Correll, 2003).
in more competitive messages than the other’s appreciative com- These studies suggest that mere ownership will be stronger among
ments, F(1, 40) ⫽ 7.33, p ⬍ .01 (see Figure 1). This supports individuals with low rather than high self-concept clarity (Hypoth-
Hypothesis 4. esis 5). We further predicted more competitive responses (Hypoth-
esis 6), less attitude change toward the counterpart’s position
Discussion and Introduction to Study 4 (Hypothesis 7a), and more negative impressions of the conflict
partner (Hypothesis 7b) when the other’s derogative remarks are
Results thus far are consistent with the idea that involvement of about arguments owned rather than not owned, especially when the
the (extended) self and concomitant ego-defensive tendencies con- individual has low self-concept clarity.
tributes to conflict escalation. Although we limited ourselves to The final issue considered in Study 4 is that mere ownership
competitive messages, research has shown less attitude change effects have been attributed to self-enhancement tendencies (e.g.,
toward the partner’s position, and more negative views of the Beggan, 1992) and to loss aversion (Kahneman et al., 1990). The
conflict partner, when the issue under attack is self-relevant predicted moderating role of self-concept clarity addresses the role
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

(Brehm, 1966; Wood et al., 1996), or when the self is activated and
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of self-enhancement tendencies (i.e., self-enhancement tendencies

made accessible (e.g., Carver & Scheier, 1981). Together with are stronger for low self-concept clarity individuals). To address
Study 3, this suggests that parties change their attitude less toward loss aversion, we asked participants in Study 4 whether fear of
the partner’s position and hold more negative views of their losing the arguments played a role when determining a price for
partner, when the other’s derogative remarks are about arguments Set A and for Set B. Evidence for loss aversion would reveal itself
owned rather than not owned. To examine this, in Study 4 we in mediation analyses if loss aversion is partly responsible for the
measured attitude polarization and source derogation. valuation of arguments owned. In other words, we predicted that
In Study 4, we also sought to further develop our self- the effect of ownership on price for Set A and Set B is partly
enhancement– ego-defensiveness argument. If the mere ownership mediated by loss aversion (Hypothesis 8).
effect and its consequences are due to self-enhancement and ego-
defensiveness needs, one would expect stronger effects when
individuals have a higher need to defend and restore the self- Method
concept. In Study 4, therefore, we included a measure of the
Design and participants. The design was a 2 (owned: Set A vs. Set
individual’s sensitivity to self-threat, namely, self-concept B) ⫻ 2 (other derogates: Set A vs. Set B) ⫻ 2 (self-concept clarity: high
clarity—the extent to which the contents of an individual’s vs. low) between-participants factorial. In addition to the dependent vari-
self-concept are clearly and confidently defined, internally consis- ables of Study 3, we measured attitudes regarding environmental policies
tent, and temporally stable (J. D. Campbell et al., 1996). Hostility both before and after the experiment, assessed impressions of the counter-
following self-threat is particularly strong among individuals with part, and, to assess loss aversion, measured reasons for setting a price.
low rather than high self-concept clarity (Bushman & Baumeister, Participants were 82 students at the University of Amsterdam. About one
1998; Exline et al., 2004; Kernis, Granneman, & Barclay, 1989). third (34%) were men (exploratory analyses revealed no effects involving
For example, people respond with more verbal aggression after gender). Participants received 5 Euro (approximately 5.5 U.S. dollars) for
negative feedback or after failure when they had low rather than participation and were randomly assigned to conditions.
Procedure and independent variables. On arrival, participants were
high self-concept clarity (Stucke & Spore, 2002). Similarly, indi-
told that they would participate in two unrelated studies—a scale validation
viduals with high explicit self-esteem but low implicit self-esteem study and a study on decision making. The scale validation study involved
tend to be more hostile and prone to prejudice compared with the measurement of self-concept clarity, using a Dutch translation of the
12-item Self-Concept Clarity scale developed by J. D. Campbell et al.
(1996; Stucke & Spore, 2002). Sample items from this scale, to be rated on
a 5-point scale (1 ⫽ not at all to 5 ⫽ very much), include “my beliefs about
myself often conflict with one another” and “on one day I might have one
opinion about myself, and on another day I might have another opinion.”
Ratings were averaged into one index (Cronbach’s ␣ ⫽ .85), and partici-
pants were allocated to the high self-concept clarity group when their score
was below the median (Md ⫽ 2.60) or to the low self-concept clarity group
when their score was equal to or above the median.
Participants then proceeded with “an unrelated study about decision-
making.” The task, procedures, and the manipulation of argument owner-
ship (Set A or Set B) were largely the same as in Study 3, except that all
participants were in the negative feedback condition (i.e., they received
derogative remarks about Set A or Set B). A few other changes in the
procedures were made. In the first three experiments, we excluded mere
exposure as an explanation for mere ownership effects by showing Argu-
ment Set A but not Set B. However, to also rule out an “unequal exposure”
explanation, we exposed participants in Study 4 to both sets of arguments
after they had been randomly assigned to own either Set A or Set B. We
created two sets of arguments and exposed all participants to both sets after
Figure 1. Competitive messages as a function of ownership and other’s they had been randomly assigned to own either Set A or Set B. To ensure
appreciative or derogative remarks about Set A (Study 3). that Set A and Set B appeared as highly similar, we created eight arguments
 352 DE DREU AND VAN KNIPPENBERG

Table 3
Selling Price as a Function of Ownership, Self-Concept Clarity, and Argument Set (Study 4)

Self-concept clarity

High Low

Selling price asked Own A Own B Own A Own B

b a b
Price for A 5.05 (3.02) 3.70 (3.13) 5.44 (2.92) 3.35a (2.91)
Price for B 5.09b (3.12) 5.10b (3.13) 4.76a,b (2.85) 6.91c (2.80)

Note. Cell means in one row not sharing the same superscript differ at p ⬍ .05, according to Duncan tests.
Standard deviations are given in parentheses.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

(the four from Set A and four new ones) and asked 22 psychology students clarity) ⫻ 2 (price) MANOVA, with the last factor within-
to rate all arguments on three items relevant to persuasiveness (how good, participant. Results revealed a main effect for price, F(1, 74) ⫽
valid, and persuasive is this argument; all ratings ranged from 1 ⫽ not at 22.77, p ⬍ .001, and an interaction between price and ownership,
all to 5 ⫽ very much) and on one item related to comprehensibility (this
F(1, 74) ⫽ 12.67, p ⬍ .001. Consistent with Hypothesis 5, these
argument is difficult to understand; 1 ⫽ not at all to 5 ⫽ very much).
Ratings for persuasiveness were averaged into one index (Cronbach’s ␣ ⫽
effects were qualified by an interaction among price, ownership,
.83), and paired-sample t tests revealed no differences between the argu- and self-concept clarity, F(1, 74) ⫽ 3.94, p ⬍ .05. Simple effects
ments from Set A and the new ones created for this study: persuasiveness, analyses showed that the interaction between price and ownership
t(21) ⬍ 1 (M ⫽ 3.78, SD ⫽ 0.43); comprehensibility, t(21) ⫽ 1.02, ns was significant when participants had low self-concept clarity,
(M ⫽ 2.21, SD ⫽ 0.55). Set A and Set B in this study each contained two F(1, 74) ⫽ 15.90, p ⬍ .001, but not when they had high self-
“old” arguments used in the previous experiments and two new ones. concept clarity, F(1, 74) ⫽ 1.16, ns. Table 3 shows that partici-
Second, the selling task was made less abstract by telling participants pants with low self-concept clarity set higher prices for Set A when
that ownership of nonmaterial items, and trading ownership for money, are
they owned Set A rather than Set B, whereas they set higher prices
quite common and that many of them may have been involved in such
transactions. We gave two examples of such situations (copyright transfer
for Set B when they owned Set B rather than Set A.
and patent rights). Third, in the previous experiments it was unclear what Competitive messages. Hypothesis 6 predicted more compet-
would happen with a set of arguments when it was sold. We now explicitly itive messages when the partner derogated the argument set owned
stated that if a transaction would take place, the participant would lose the rather than not owned, especially when self-concept clarity is low.
set of arguments and could “no longer use it in the upcoming debate.” We The number of competitive messages was submitted to a 2 ⫻ 2
further explained that during the debate, the computer would track the (Set Owned ⫻ Other’s Feedback) ANOVA. Results revealed an
arguments participants used, and if arguments that were sold were used, interaction between ownership and other’s negative feedback, F(1,
they would not be forwarded to the other party. This change in procedure
74) ⫽ 3.97, p ⬍ .05, and a three-way interaction among owner-
was made to establish a closer parallel with previous studies on ownership,
in which transactions led to losing the items owned (Kahneman et al., ship, other’s negative feedback, and self-concept clarity, F(1,
1990). As in the previous experiments, comprehension checks and suspi- 74) ⫽ 4.28, p ⬍ .05 (see Figure 2). Simple effects analyses showed
cion probes revealed the task was understood correctly, and no single that when participants had high self-concept clarity, neither the
participant was able to guess our hypotheses. main effects for ownership and other’s negative feedback nor the
Dependent variables. The setting of price and the assessment of com- interaction between these two was significant, all F(1, 74) ⬍ 1, ns.
petitive messages were done as in Study 3. To assess attitude change, we When participants had low self-concept clarity, the interaction
asked participants (both before any manipulation took place, and after they between ownership and other’s negative feedback was significant,
had sent messages to their counterpart) to rate their opinions about pre-
F(1, 74) ⫽ 9.33, p ⬍ .01. Participants owning Set A sent more
serving the natural environment on four scales, including “by cutting down
economic growth we will save the environment from deteriorating further,” competitive messages when their counterpart derogated Set A
and “preserving the environment should be our top priority” (1 ⫽ totally rather than Set B, t(75) ⫽ 2.23, p ⬍ .05, whereas participants
disagree to 7 ⫽ totally agree). Ratings were recoded so that higher scores owning Set B sent more competitive messages when their coun-
reflected more agreement with the counterpart’s position and averaged into terpart derogated Set B rather than Set A, t(75) ⫽ 1.99, p ⬍ .05
pre- and postattitude indices (␣ ⫽ .73 and .79, respectively). We measured (see Figure 2). This supports Hypothesis 6.4
source derogation by asking participants to rate their counterpart in terms
of likeability, competitiveness, and trustworthiness (all ratings ranged from
1 ⫽ very much to 5 ⫽ not at all). Ratings were recoded so that higher 4
A regression analysis with the number of competitive messages as the
scores reflected a more negative view of the other and were averaged into dependent variable, and as independent variables ownership, other’s feed-
one index (␣ ⫽ .76). To assess loss aversion, we asked participants to rate back (both dummy coded), self-concept clarity (continuous), and all two-
to what extent their price setting for Set A and Set B was influenced by fear and three-way interactions, produced similar results: The overall model
of losing the set of arguments (both 1 ⫽ not at all to 5 ⫽ very much). explained 19% of the variance, F(7, 74) ⫽ 2.42, p ⬍ .027. Significant
effects were obtained for other’s feedback, b ⫽ ⫺.48, t(75) ⫽ ⫺2.25, p ⬍
Results .027; for the Ownership ⫻ Other’s Feedback interaction, b ⫽ ⫺.69,
t(75) ⫽ ⫺2.48, p ⬍ .015; and for the three-way interaction among
Selling prices. Price for Set A and for Set B were submitted to ownership, other’s feedback, and self-concept clarity, b ⫽ 1.19, t(75) ⫽
a 2 (ownership) ⫻ 2 (other’s negative feedback) ⫻ 2 (self-concept 2.82, p ⬍ .006.
 SELF IN SOCIAL CONFLICT 353

interaction between ownership and set, F(1, 74) ⫽ 9.03, p ⬍ .004,

showing that loss aversion regarding Set A was stronger when Set
A rather than Set B was owned (M ⫽ 2.67 vs. M ⫽ 2.20), whereas
loss aversion regarding Set B was stronger when Set B rather than
Set A was owned (M ⫽ 2.20 vs. M ⫽ 1.69). We also computed a
loss aversion index by subtracting loss aversion for the set not
owned from the set owned. This index was strongly correlated with
the price set for arguments owned minus the price set for argu-
ments not owned, r(80) ⫽ .60, p ⬍ .001.
In Hypothesis 8, we predicted that loss aversion would mediate
the effect of ownership on price for Set A and Set B. Because
self-concept clarity did influence price but not loss aversion, loss
aversion does not qualify as a mediator for the interaction between
ownership, price, and self-concept clarity. However, the strong
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

correlation between loss aversion and price qualifies loss aversion

Figure 2. Competitive messages as a function of ownership and other’s
as a potential mediator underlying the effects of ownership on
derogation of Set A or Set B for individuals with high (left four bars) and
price. Indeed, an analysis of covariance, with price for Set A and
low (right four bars) self-concept clarity (Study 4).
for Set B as a within-participant variable, ownership (Set A vs. Set
B) and self-concept clarity (high vs. low) as between-participants
Attitude change. Hypothesis 7a predicted less attitude change variables, and loss aversion for Set A and for Set B as covariates,
toward the partner’s position when the partner derogated the revealed (a) a significant regression of the dependent variables on
argument set owned rather than not owned, especially when self- the covariates, F(1, 78) ⫽ 32.79, p ⬍ .001, and (b) a substantial
concept clarity is low. To test this, we computed a change index by and significant reduction in variance explained by the Owner-
subtracting preattitudes from postattitudes and submitted this in- ship ⫻ Price interaction, that is, from F(1, 78) ⫽ 12.67, p ⬍ .001,
dex to a 2 ⫻ 2 ⫻ 2 (Ownership ⫻ Other’s Negative Feedback ⫻ to F(1, 78) ⫽ 4.53, p ⬍ .05 (Sobel test: z ⫽ 2.65, p ⬍ .008).6 All
Self-Concept Clarity) MANOVA (using pre- and postattitudes as a in all, these results support Hypothesis 8: Loss aversion is partially
repeated measures variable produced identical results). Results responsible for the effects of ownership on selling price. Because
revealed a significant three-way interaction, F(1, 74) ⫽ 5.27, p ⬍ including loss aversion as a mediator left the influence on self-
.025. Simple effects analyses showed that when participants had concept clarity unaffected, it appears that both loss aversion and
high self-concept clarity, no effects were significant, all F(1, 74) ⬍ ego-defensiveness are responsible for the mere ownership ob-
1.26, ns. When participants had low self-concept clarity, however, served here.
the interaction between ownership and other’s feedback was sig-
nificant, F(1, 74) ⫽ 5.30, p ⬍ .05. Table 4 shows that counterparts
who derogated Set A produced less attitude change toward their Discussion
position when participants owned Set A rather than Set B, whereas
Study 4 replicated the finding that people respond with more
counterparts who derogated Set B produced less attitude change
hostility when their partner derogates arguments owned rather than
toward their position when participants owned Set B rather than
identical arguments not owned. Study 4 expanded previous exper-
Set A. This supports Hypothesis 7a.
iments by going beyond competitive messages to include attitude
Source derogation. Hypothesis 7b predicted more negative
views of the partner when the partner derogated the argument set polarization and source derogation. It also showed these effects to
owned rather than not owned, especially when self-concept clarity be present for people with low self-concept clarity and absent for
is low. A 2 ⫻ 2 ⫻ 2 (Ownership ⫻ Other’s Negative Feedback ⫻ people with high self-concept clarity. This lends support for the
Self-Concept Clarity) ANOVA only revealed a significant three-
way interaction, F(1, 74) ⫽ 6.00, p ⬍ .025. Simple effects anal- 5
Regression analysis with source derogation as dependent variable, and
yses showed that when participants had high self-concept clarity, as independent variables ownership, other’s feedback (both dummy
no effects were significant, all F(1, 74) ⬍ 2.61, p ⬍ .15. When coded), self-concept clarity (continuous), and all two- and three-way in-
participants had low self-concept clarity, however, the interaction teractions, produced similar results: The overall model explained 15% of
between ownership and negative feedback was significant, F(1, the variance, F(7, 74) ⫽ 1.91, p ⬍ .10. Significant effects were obtained
74) ⫽ 4.38, p ⬍ .05. Figure 3 shows that among participants with for the Ownership ⫻ Other’s Feedback interaction, b ⫽ ⫺.51, t(79) ⫽
low self-concept clarity, opponents who derogated Set A were ⫺2.16, p ⬍ .034; and for the Ownership ⫻ Other’s Feedback ⫻ Self-
viewed more negatively when participants owned Set A rather than Concept Clarity interaction, b ⫽ .84, t(79) ⫽ 2.34, p ⬍ .022.
6
Set B, t(21) ⫽ 2.93, p ⬍ .008. Likewise, partners who derogated A regression analysis with an ownership effect computed by subtract-
Set B were viewed more negatively when participants owned Set ing price for arguments not owned from price for arguments owned, and a
loss aversion effect computed by subtracting loss aversion for argument set
B rather than Set A, t(18) ⫽ 1.81, p ⬍ .05 (one-tailed). This
not owned from loss aversion for argument set owned, yielded almost
supports Hypothesis 7b.5 identical results: Ownership predicted loss aversion, b ⫽ ⫺0.98, t(79) ⫽
Loss aversion. Ratings for fear of losing Set A and of losing ⫺2.99, p ⬍ .001, and price predicts loss aversion, b ⫽ 1.02, t(79) ⫽ 5.73,
Set B were submitted to a 2 ⫻ 2 ⫻ 2 ⫻ 2 (Ownership ⫻ Other’s p ⬍ .001. Controlling for loss aversion led to a drop in variance explained
Negative Feedback ⫻ Self-Concept Clarity ⫻ Set) ANOVA, with from b ⫽ ⫺2.92, t(79) ⫽ ⫺4.76 to b ⫽ ⫺1.92, t(79) ⫽ ⫺3.53 (Sobel test,
set as the within-participant variable. Results only revealed an z ⫽ 2.64, p ⬍ .008).
 354 DE DREU AND VAN KNIPPENBERG

Table 4
Attitude Change as a Function of Ownership, Other’s Negative Feedback, and Self-Concept
Clarity (Study 4)

Self-concept clarity

High Low

Other derogates Own A Own B Own A Own B

Set A 0.56a (1.06) 0.38a (1.18) ⫺0.75b (1.02) 0.36a (1.67)

Set B 0.60a (1.23) 0.09a (0.96) 0.25a (0.76) ⫺0.44b (1.46)

Note. Positive numbers indicate a change toward the counterpart’s position; negative numbers reflect a
polarization away from the counterpart’s position. Cell means in one row or column not sharing the same
superscript differ at p ⬍ .05, according to Duncan tests. Standard deviations are given in parentheses.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

idea that mere ownership effects are, in addition to loss aversion, when one-tailed tests were used, and in a few cases, asymmetries
in part due to self-enhancement needs. emerge. In Study 4, for instance, effects appeared to be stronger
Whereas in the first three studies we found a general tendency when participants owned Set A rather than Set B, whereas in the
toward mere ownership, Study 4 seems to suggest that mere first three experiments, effects appeared stronger for Set B. Across
ownership effects emerge only when participants have low self- experiments, these discrepancies may be due to selling order,
concept clarity. It is important to emphasize that such a conclusion exposure, or unintended differences in “true” argument quality.
would be based on the median-split analyses, which we conducted These discrepancies were unrelated to mere ownership effects, in
for ease of interpretation. When self-concept clarity is viewed as a that ownership never interacted with selling order, and ownership
continuous variable, both the Ownership ⫻ Price and the Owner- effects occurred regardless of whether one set of arguments or both
ship ⫻ Price ⫻ Self-Concept Clarity interactions were significant sets were shown. Nevertheless, because of these asymmetries, we
(see Footnotes 4 and 5). The three-way interaction is an ordinal provide a quantitative summary of the results in order to give
(rather than disordinal) qualification of the two-way interaction, insight into the overall strength and robustness of the key results.
thus indicating that effects are generally stronger the lower self- We conducted a meta-analysis of the four experiments using
concept clarity becomes. This is consistent with our general argu-
Hedges and Olkin’s (1985) approach. Prices set for Set A when Set
ment that the greater the self-enhancement needs, the stronger the
A was owned versus not owned, and for Set B when Set B was
ego-threat experienced, and the more hostile reactions will be.
owned versus not owned, provided the input for the meta-analysis.
This provided a total of 14 independent comparisons (i.e., 4 in
Meta-Analysis of Studies 1– 4 Study 1, 4 in Study 2, 2 in Study 3, and 4 in Study 4). One
Across the four experiments, some clear consistencies in results moderator was examined here, namely, whether valuation took
emerge. At the same time, it also appears that some specific place under a low bias condition (expecting to write an essay in
replications within or between experiments were significant only Study 1, presence of process accountability in Study 2, and high
self-concept clarity in Study 4) or under a high bias condition
(expecting debate in Study 1, no accountability condition in Study
2, all conditions in Study 3, and low self-concept clarity in Study
4). We used DSTAT 1.11 (Johnson, 1989) to calculate effect sizes
(Cohen’s d), 95% confidence intervals (CIs), and the effect of the
a priori specified moderator.
The average effect size was 0.29 (number of effect sizes, k ⫽
14). Although small, it is greater than zero in that the 95% CI
ranges from 0.17 to 0.41. As expected, the moderator analysis
yielded a significant effect, QB(1) ⫽ 4.72, p ⬍ .03, showing
greater effect sizes in the high (k ⫽ 8; d ⫽ 0.42; 95% CI ⫽ 0.25,
0.58) rather than low bias condition (k ⫽ 6; d ⫽ 0.15; 95% CI ⫽
⫺0.03, 0.33). Thus, when individuals anticipate conflict, have low
epistemic motivation, or have low self-concept clarity, mere own-
ership is rather pronounced.
Experiments 2– 4 revealed a correlation between mere owner-
ship (price for set owned minus price for set not owned) and
competitive messages. The correlation obtained in the no account-
Figure 3. Hostile views of the counterpart as a function of ownership and ability condition of Study 2, r(150) ⫽ 0.22; in Study 3, r(44) ⫽
other’s derogation of Set A or Set B for individuals with high (left four 0.17; and in the low self-concept clarity group in Study 4, r(44) ⫽
bars) and low (right four bars) self-concept clarity (Study 4). 0.39, were combined using DSTAT. The effect is significant and
 SELF IN SOCIAL CONFLICT 355

of moderate size (d ⫽ .50, total N ⫽ 238; 95% CI ⫽ 0.31, 0.68), Third, our work contributes to our understanding of the conflict
showing that mere ownership of arguments triggers hostility. process by identifying core moderators of the general tendencies
outlined above. Consistent with a motivated information process-
Conclusions and General Discussion ing model of negotiation (De Dreu & Carnevale, 2003), Study 2
showed that when individuals are motivated to process information
To understand why people have difficulty managing conflict thoroughly, their self-enhancement tendencies are reduced, and
and why conflict so easily escalates into exceedingly hostile and inflated self-views less readily translate into hostile opening
competitive exchanges, we proposed that even in relatively minor moves. Study 4 identified self-concept clarity as a key moderator,
disputes, people quickly identify with their arguments and posi- suggesting that the general tendency for inflated self-views to lead
tions and that any threat or opposition to these arguments is taken to hostility holds especially for individuals with unstable views of
as a threat to the self. This ego-threat triggers competitive cogni- themselves.
tions and hostile behavior, thus making the conflict process ex- Fourth, and finally, the present studies provide first-time evi-
ceedingly difficult and tenacious. The results of four experiments dence that the extended self plays a vital role in rendering conflict
were consistent with this line of reasoning. Study 1 showed that a difficult process that easily escalates into exceedingly hostile
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

people develop positive views of arguments and beliefs that are exchanges. Past work on the self in conflict and influence situa-
associated with the self, compared with identical arguments and tions has considered the core self and examined threat to basic,
beliefs not associated with the self. Study 1 also showed that this deeply held values and beliefs, to core skills, and to defining
mere ownership effect was stronger when conflict was anticipated, characteristics. Our work pushes these lines of inquiry further by
suggesting that conflict in itself produces self-enhancement. Stud- focusing on the extended rather than the core self and by showing
ies 2– 4 showed that inflated evaluations of arguments associated that threat to recently acquired components of the extended self
with the self triggered competitive opening moves and hostile suffices to make conflict management difficult and escalation
reactions to the opponent’s derogative remarks. Study 2 showed likely. Whereas several lines of inquiry indicated that conflicts
that this general tendency was reduced when people were process about personal issues and deeply held values and beliefs easily
accountable, and Study 4 showed that the tendency was reduced escalate, our work indicates that even when people are randomly
when people have a stable and secure self-concept. assigned to attitude positions and have been given arguments that
Taken together, the present work shows that because of mere they have not seen (before), they tend to identify with these
ownership conflict is difficult to manage, and that this is particu- positions and arguments and make them part of their extended self.
larly true when conflict parties have low epistemic needs and high Any opposition to these positions or arguments then becomes an
self-enhancement needs. In the remainder of this section, we ego-threat and thus triggers hostility.
highlight several contributions to conflict theory, and we discuss Given that the (extended) self gets implicated so quickly and
some limitations and methodological issues. renders the conflict process so difficult, why do conflict parties
sometimes behave cooperatively (Rubin et al., 1994)? Also, why
Contributions to Conflict Theory do people often say (and perhaps believe) that self-interest is the
primary driver of their behavior, when in fact it is not (Miller &
Our findings highlight, first of all, the crucial role of the self in Ratner, 1998)? A first answer to these questions is that although
rendering conflict a difficult enterprise. From past research, we the self gets easily implicated, this pattern of results was most
knew that when their positive self-view is threatened, people pronounced among individuals with low self-concept clarity. The
become defensive and hostile (see Baumeister et al., 1996; Elle- level, stability, and clarity of the self-concept may be important
mers, Spears, & Doosje, 2002; Ross & Ward, 1995). We did not moderators of the extent to which people are competitive with their
yet know that merely anticipating conflict and opposition increases counterpart. Second, people in conflict and negotiation have or
the tendency to see oneself as positive and good, and that this adopt a selfish or a prosocial goal (e.g., De Dreu & Carnevale,
tendency in itself triggers competitive cognition and hostile be- 2003; Penner, Dovidio, Piliavin, & Schroeder, 2005). A prosocial
havior. Thus, a first contribution is that we showed that conflict in goal may be genuine (one likes the conflict partner and wishes him
and by itself produces ego-threat, and therefore, it is destined to or her well) or instrumental (one needs the conflict partner again in
become difficult to manage and is bound to escalate. Whether the near future). Regardless, a prosocial goal reduces competitive-
anticipated or enacted, whether about deeply held values or irrel- ness toward the partner (Rubin et al., 1994). Put differently, we
evant nonissues, conflict produces ego-threat, and the need to suggest that conflict almost inevitably triggers ego-threat and
defend and restore the self inevitably leads to escalatory cognition thereby sets the stage for a competitive exchange but that prosocial
and behavior. goals reduce ego-defensive tendencies. Although this hypothesis is
Our experiments make a second contribution. Past work on consistent with literature reviews (De Dreu & Carnevale, 2003),
ego-threat and hostility confined itself to showing that when pos- research is needed to provide an empirical foundation.
itive self-view is threatened, people respond to the other’s actions
with more hostility. We replicated this general finding in Studies Limitations
3 and 4. But in Study 2, we showed that ego-threat also triggers
hostile opening moves. That is, in absence of any knowledge of the Our conclusions and contributions are based on four experi-
other’s intentions or behavior, disputants act with greater hostility ments in which participants performed a number of tasks, one
the more inflated their self-views are. This is first-time evidence being the setting of prices for arguments they owned and argu-
that inflated self-view in and by itself makes the conflict process ments they did not own. Our meta-analytic integration of Studies
difficult and likely to escalate. 1– 4 already revealed that participants set a higher price for argu-
 356 DE DREU AND VAN KNIPPENBERG

ments owned, and this effect was independent of procedural vari- we showed that anticipating conflict increases self-enhancement
ations across experiments. However, setting prices for sets of needs. Further evidence comes from Study 2, where we showed
arguments is not a task in which our participants often engage. that process accountability reduces not only self-enhancement
This lack of mundane realism is, however, countered by experi- (Sedikides et al., 2002) but also mere ownership of arguments.
mental realism: Many participants were disappointed when learn- Finally, in Study 4, mere ownership effects were more pronounced
ing that they would not really debate with the other person, and among individuals with low self-concept clarity and concomitant
many persisted in their desire to learn whether an arguments-for- high self-enhancement needs. We therefore conclude that both loss
money transaction was established. We thus believe that although aversion and self-enhancement account for the mere ownership
the tasks lack mundane realism, the behaviors we observed were effects observed here and for the hostility that we observed in
real. people’s opening moves, reactions to their counterpart, attitudes,
Another issue is that participants may have inferred that the and perceptions.
quality of the sets of arguments was good (why otherwise would
the experimenter give them these arguments). The question, then,
Concluding Thoughts
is whether our results extend to situations in which participants
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

believe the quality of the sets of arguments is equivalent but poor. The present work shows that it is not so much when but because
Research on mere ownership and the endowment effect, always conflict elicits ego-threat that it is difficult to manage and bound to
considered positive features and likable objects, and future re- escalate. People quickly identify with arguments and positions in
search are needed to examine how object valence moderates feel- the debate even when these arguments and positions have been
ings of mere ownership. assigned to them on a random basis and even when they have not
Demonstrating mere ownership of arguments was not our only seen the arguments. Still, the arguments become part of the ex-
goal. At least as important was to examine whether mere owner- tended self, and when in conflict we oppose and reject each other’s
ship predicts competitiveness in conflict situations. As predicted, arguments, we thus oppose and reject each other’s self-view.
Studies 2– 4 revealed a correlation between mere ownership (price That conflict by default triggers ego-threat and is bound to
for set owned minus price for set not owned) and competitive escalate may seem pessimistic. However, many societies appear to
messages. Thus, mere ownership was positively related to com- have some implicit understanding of the dangers the self brings to
petitive communication both in responding to and approaching the conflict situations. Religious writings provide moral rules about
counterpart. In Study 4, we further showed that mere ownership how to behave in conflict situations, and people embellish debates
not only predicted competitive messages but also source deroga- and conflict in regulations of all kinds and all sorts. This research
tion and attitude polarization. These measures are quite removed shows that these actions may be needed because, in fact, the
from the selling task, and it is difficult to see how the artificiality (extended) self gets involved pretty quickly and accounts for a
of the selling task affected the degree to which mere ownership difficult and tedious process that easily escalates into excessive
brings out hostility. levels of hostility.
At the conceptual level, we need to consider the possibility that
the mere ownership effects observed here not (only) reflect in-
flated self-views. First, it has been argued that mere ownership References
effects can be explained in terms of mere exposure—people have Abelson, R. P. (1986). Beliefs are like possessions. Journal for the Theory
been exposed to items owned more, and more intensely, than items of Social Behavior, 16, 223–250.
not owned (e.g., Beggan, 1992). This explanation can be excluded Axelrod, R., & Hamilton, W. (1981, March 7). The evolution of cooper-
because mere ownership of arguments occurred whether or not ation. Science, 211, 1390 –1396.
participants had seen the arguments (i.e., in Studies 1–3) and when Baumeister, R. F. (1998). The self. In D. Gilbert, S. T. Fiske, & G. Lindzey
both argument sets had been shown the same amount of time (i.e., (Eds.), Handbook of social psychology (4th ed., Vol. 1, pp. 680 –740).
in Study 4). Second, one may argue that mere ownership effects New York: McGraw-Hill.
were the result of demand characteristics, but this explanation can Baumeister, R. F., Smart, L., & Boden, J. M. (1996). Relation of threatened
egotism to violence and aggression: The dark side of self-esteem. Psy-
be excluded as well. Because participants could earn money with
chological Review, 103, 5–33.
setting their prices, a demand characteristics account has to com- Beggan, J. K. (1992). On the social nature of nonsocial perceptions: The
pete with the participants’ desire to earn some extra money. Also, mere ownership effect. Journal of Personality and Social Psychology,
as mentioned in the Method section of Study 1, demand charac- 62, 229 –237.
teristics should have produced higher valuation of both arguments Brehm, J. W. (1966). A theory of psychological reactance. San Diego, CA:
owned versus not owned and of arguments shown versus not Academic Press.
shown. The latter was never observed. Finally, the interactions Bushman, B. J., & Baumeister, R. F. (1998). Threatened egotism, narcis-
between ownership and other variable(s) found in Studies 2– 4 are sism, self-esteem, and direct and displaced aggression: Does self-love or
consistent with our theory yet difficult to explain in terms of self-hate lead to violence? Journal of Personality and Social Psychol-
demand characteristics. ogy, 75, 219 –229.
Campbell, J. D., Trapnell, P. D., Heine, S. J., Katz, I. M., Lavallee, L. F.,
Third, mere ownership effects have been explained in terms of
& Lehman, D. R. (1996). Self-concept clarity: Measurement, personality
loss aversion (Kahneman et al., 1990). In Study 4, we found that correlates, and cultural boundaries. Journal of Personality and Social
self-reported loss aversion partially mediated the effects of own- Psychology, 70, 141–156.
ership on price setting. However, loss aversion does not exclude Campbell, W. K., & Sedikides, C. (1999). Self-threat magnifies the self-
inflated self-views. In fact, the favorable view of the “extended serving bias: A meta-analytic integration. Review of General Psychol-
self” (Beggan, 1992) reasoning is consistent with Study 1, where ogy, 3, 23– 43.
 SELF IN SOCIAL CONFLICT 357

Carnevale, P. J. (1995). Property, culture, and negotiation. In R. Kramer & Lerner, J. S., & Tetlock, P. E. (1999). Accounting for the effects of
D. Messick (Eds.), Negotiation as a social process: New trends in theory accountability. Psychological Bulletin, 125, 255–275.
and research (pp. 309 –323), London: Sage. Mackie, D. M., Devos, T., & Smith, E. R. (2000). Intergroup emotions:
Carver, C. S., & Scheier, M. F. (1981). Attention and self-regulation: A Explaining offensive action tendencies in an intergroup context. Journal
control-theory approach to human behavior. New York: Springer- of Personality and Social Psychology, 79, 602– 616.
Verlag. McClelland, G. H., & Judd, C. M. (1993). Statistical difficulties of detect-
Cohen, G. L. (2003). Party over policy: The dominating impact of group ing interaction and moderator effects. Psychological Bulletin, 114, 376 –
influence on political beliefs. Journal of Personality and Social Psy- 390.
chology, 85, 808 – 822. Miller, D. T., & Ratner, R. K. (1998). The disparity between the actual and
De Dreu, C. K. W., & Carnevale, P. J. (2003). Motivational bases for assumed power of self-interest. Journal of Personality and Social Psy-
information processing and strategic choice in conflict and negotiation. chology, 74, 53– 62.
Advances in Experimental Social Psychology, 35, 235–291. Mo’az, I., Ward, A., Katz, M., & Ross, L. (2002). Reactive devaluation of
De Dreu, C. K. W., Koole, S., & Steinel, W. (2000). Unfixing the fixed-pie: an “Israeli” vs. “Palestinian” peace proposal. Journal of Conflict Reso-
A motivated information processing of integrative negotiation. Journal lution, 46, 515–546.
of Personality and Social Psychology, 79, 975–987. Nesselroade, K., Beggan, J., & Allison, S. (1999). Possession enhancement
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

De Dreu, C. K. W., & Weingart, L. R. (2003). Task versus relationship in an interpersonal context: An extension of mere ownership effect.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

conflict, team performance and team member satisfaction: A meta- Psychology & Marketing, 16, 21–34.
analysis. Journal of Applied Psychology, 88, 741–749. Ostrom, E. (1998). A behavioral approach to the rational choice theory of
Diekman, K., Samuelson, S., Ross, L., & Bazerman, M. (1997). Self- collective action. The American Political Science Review, 92, 1–22.
interest and fairness in problems of resource allocation: Allocators Paese, P. W., & Yonker, R. D. (2001). Toward a better understanding of
versus recipients. Journal of Personality and Social Psychology, 72, egocentric fairness judgments in negotiation. International Journal of
1061–1074. Conflict Management, 12, 97–186.
Ellemers, N. E., Spears, R., & Doosje, B. (2002). Self and social identity. Penner, L. A., Dovidio, J. F., Piliavin, J. A., & Schroeder, D. A. (2005).
Annual Review of Psychology, 53, 161–186. Prosocial behavior: Multilevel perspective. Annual Review of Psychol-
Erev, I., & Roth, A. E. (2001). Simple reinforcement learning models and
ogy, 56, 337–363.
reciprocation in the prisoner’s dilemma game. In R. Selten & G. Gig-
Pierce, J., Kostova, T., & Dirks, K. (2003). The state of psychological
erenzer (Eds.), Bounded rationality: The adaptive toolbox (pp. 215–
ownership: Integrating and extending a century of research. Review of
231). Cambridge, MA: MIT Press.
General Psychology, 7, 84 –107.
Exline, J. J., Baumeister, R. F., Bushman, B. J., Campbell, W. K., & Finkel,
Pruitt, D. G. (1998). Social conflict. In D. Gilbert, S. T. Fiske, & G.
E. J. (2004). Too proud to let go: Narcissistic entitlement as a barrier to
Lindzey (Eds.), Handbook of social psychology (4th ed., Vol. 2, pp.
forgiveness. Journal of Personality and Social Psychology, 87, 894 –
89 –150). New York: McGraw-Hill.
912.
Ross, L., & Ward, A. (1995). Psychological barriers to dispute resolution.
Greenwald, A. G. (1968). Cognitive learning, cognitive response to per-
Advances in Experimental Social Psychology, 27, 255–304.
suasion, and attitude change. In A. G. Greenwald, T. C. Brock, & T. M.
Rubin, J. Z., Pruitt, D. G., & Kim, S. H. (1994). Social conflict: Escalation,
Ostrom (Eds.), Psychological foundations of attitudes (pp. 147–170).
stalemate, and settlement. New York: McGraw-Hill.
San Diego, CA: Academic Press.
Sedikides, C., Herbst, K. C., Hardin, D. P., & Dardis, G. J. (2002).
Hedges, L., & Olkin, I. (1985). Statistical methods meta-analysis. San
Diego, CA: Academic Press. Accountability as a deterent to self-enhancement: The search for mech-
Heider, F. (1958). The psychology of interpersonal relations. New York: anisms. Journal of Personality and Social Psychology, 83, 592– 605.
Erlbaum. Stucke, T. S., & Spore, S. L. (2002). When a grandiose self-image is
Johnson, B. T. (1989). DSTAT: Software for the meta-analytic review of threatened: Narcissism and self-concept clarity as predictors of negative
research literatures. Hillsdale, NJ: Erlbaum. emotions and aggression following ego-threat. Journal of Personality,
Jordan, C. H., Spencer, S. J., Zanna, M. P., Hoshino-Browne, E., & Correll, 70, 509 –532.
J. (2003). Secure and defensive self-esteem. Journal of Personality and Tversky, A., & Kahneman, D. (1979). Prospect Theory: An analysis of
Social Psychology, 85, 969 –978. decision under risk. Econometrica, 47, 263–291.
Jussim, L., Yen, H., & Aiello, J. R. (1995). Self-consistency, self- Van Dijk, E., & van Knippenberg, D. (1998). Trading wine: On the
enhancement, and accuracy in reactions to feedback. Journal of Exper- endowment effect, loss aversion, and the comparability of consumer
imental Social Psychology, 31, 322–356. goods. Journal of Economic Psychology, 19, 485– 495.
Kahneman, D., Knetch, J. L., & Thaler, R. H. (1990). Experimental tests of Wood, W., Pool, G. J., Leck, K., & Purvis, D. (1996). Self-definition,
the endowment effect and the Coase-theorem. Journal of Political Econ- defensive processing, and influence: The normative impact of majority
omy, 98, 1325–1348. and minority groups. Journal of Personality and Social Psychology, 71,
Kernis, M. H., Granneman, B. D., & Barclay, L. C. (1989). Stability and 1181–1193.
level of self-esteem as predictors of anger arousal and hostility. Journal
of Personality and Social Psychology, 56, 1013–1022.
Komorita, S. S., Parks, C., & Hulbert, L. (1992). Reciprocity and the Received February 12, 2004
induction of cooperation in social dilemmas. Journal of Personality and Revision received January 3, 2005
Social Psychology, 62, 607– 617. Accepted January 21, 2005 䡲