Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

www.elsevier.com/locate/obhdp

ARTICLE IN PRESS

Powerful perceivers, powerless objects: Flexibility of powerholders’ social attention �

Jennifer R. Overbeck a,¤, Bernadette Park b

a Marshall School of Business, University of Southern California, Los Angeles, CA 90089-0808, USAb Department of Psychology, 345 UCB, University of Colorado, Boulder, CO 80309-0345, USA

Received 9 October 2004

Abstract

We argue that the eVect of power on social attention is a function of Xexible, instrumental information processing that allows thehigh power perceiver to attain situation speciWc goals using whatever means are available, including attention. Study 1 assigned pow-erful participants to more “people-centered” or more “product-centered” goals, and found that people-centered powerholders betterindividuated low-power targets. Study 2 examined responses by both high- and low-power organization members, and found power-ful judges more responsive to organizational goals in setting priorities and using information about the organization than powerlessjudges. Together, these results suggest that powerholders use social attention, like other resources, in order to advance their ability tofulWll organizational goals. 2005 Elsevier Inc. All rights reserved.

Keywords: Power; Social attention; Individuation; Stereotyping; Subject; Object; Instrumentality; Goals

In the American version of the sitcom The OYce, theboss arranges a recognition party for his staV, tobolster morale. Through an endless-seeming evening,he doles out awards based on stereotypes (the Indian-American employee gets an otherwise-unexplained“Curry Award”) and limited, outdated informationabout employees (he debates giving his receptionist a“Longest Engagement” award for the third or fourthconsecutive year). The employees groan and suVer, butclearly none are surprised. Despite his eVorts to deliver

� This research was supported in part by National Institute of MentalHealth Grant R01 MH45049. Study 1 was part of Jennifer Overbeck’s dis-sertation thesis, completed at the University of Colorado at Boulder. Weare grateful to Charles Bentley for his untiring assistance with data collec-tion, to Ashley Prosper and Cathleen Cho for data coding, and to the CUStereotyping and Prejudice lab, the Tiedens lab at the Stanford GSB, andthe USC/UCLA OB lab for valuable feedback on drafts of the paper.

* Corresponding author. Fax: +1 213 740 3582.E-mail addresses: overbeck@marshall.usc.edu (J.R. Overbeck),

bpark@psych.colorado.edu (B. Park).

0749-5978/$ - see front matter 2005 Elsevier Inc. All rights reserved.doi:10.1016/j.obhdp.2005.10.003

individual recognition, the boss is unaware of who hisemployees are as individuals. He has little insight intotheir skills and contributions, let alone their concerns ortheir personalities. Like a Dilbert cartoon, the episodederives humor from a sadly common experience: that theboss is oblivious to those around him.

How do the powerful pay attention to others?According to the cynical view depicted in The OYce, notwell. On the other hand, perhaps in real organizationspowerholders do pay attention to others well, knowingthat such attention can yield beneWts that might help tomaintain or even increase their power.

The basic research on this topic has leaned stronglytoward the former perspective: that power leads peopleto stereotype others and to pay careless social attentioncharacterized by cognitive laziness and shortcuts(DeDreu & Van Kleef, 2004; Ebenbach & Keltner, 1998;Fiske, 1993; Goodwin, Gubin, Fiske, & Yzerbyt, 2000;Keltner, Gruenfeld, & Anderson, 2003; Keltner & Rob-inson, 1997; Rodriguez-Bailon, Moya, & Yzerbyt, 2000).

2 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

This research, whose basic arguments and implicationsare outlined in Table 1, is most often conducted by socialpsychologists, and published in psychology journals. Itsstrongest proponents are Fiske and colleagues (Fiske,1993; Fiske & Dépret, 1996; Goodwin et al., 2000) andKeltner et al. (2003), who argue that power should beassociated with careless processing for a number of rea-sons.

Fiske (1993) points out that there is usually a one-to-many ratio of powerholders to subordinates, and thusthe powerholders’ cognitive load is much higher, andprecludes careful attention. Powerholders’ outcomes arenot dependent on the powerless, and so there is littleexternal inducement to attend. Finally, people with dom-inant personalities are simultaneously more likely to paycareless attention to others, and drawn to positions ofstructural power. As an example, Fiske cites the Jack-sonville Shipyards case, in which a traditionally maleworkplace was integrated by gender. The enteringwomen were confronted by a hostile work environmentin which they were stereotyped and otherwise ignored,Fiske argues, because the men did not feel they neededanything from the women’s presence. Managers, whocould potentially have remedied the situation, wereattentionally overloaded and found it easy to dismissindividual women’s complaints as unimportant annoy-ances. Because of these factors, the hostile environmentpersisted at least until court decisions mandated other-wise.

Keltner et al. (2003) echo this argument, and extend itby saying that power instills a tendency to approachbehavioral rewards. Powerholders will attend to objectsthat promise the potential for reward, and because pow-erless people in the environment oVer little reward, theyget little attention. Further, powerful people tend to havemore positive aVect—a psychological state associatedwith more careless cognition—and so may be even morelikely to follow the motivational pattern.

A few psychologists have worked to establish bound-ary conditions for these Wndings (Overbeck & Park,2001; Vescio, Snyder, & Butz, 2003), identifying situa-tional variables that seem to moderate whether powerfulpeople’s attention is careful. In general, this work hasfound that power is associated with the active use ofattention, and that powerholders can be careful attend-ers or more stereotype-bound attenders, depending ontheir expectations and responsibilities. For example,Overbeck and Park (2001) argued that powerholders inorganizations feel a greater sense of responsibility forgood performance. To the extent that the targets ofresponsibility are altered, attention will follow.

In the current work, we hope to challenge what seemsto be a strongly prevailing notion from the social psy-chological literature, by integrating work from that liter-ature with organizational behavior theories andobservations. Theorists of management and organiza-

tional behavior have crafted models of power use inorganizations (generally under the heading of leadership)that would seem to require eVective use of attention (e.g.,Wilemon & Cicero, 1970; Yukl, 1989). Organizationalpowerholders must solve problems and make decisionsregarding people (Hollander, 1978); form coalitions, inpart by responding to other people’s concerns and devel-oping relationships with them (Kanter, 1983; Kaplan,1986); display not only task-oriented behavior, but alsorelationship-oriented behavior, according to situationaldemands (Bass, 1985; Burns, 1978; Fleishman, 1953;Halpin & Winer, 1957; Hill, 1969; Kahn, Wolfe, Quinn,& Snoelk, 1964); respond to human as well as task con-cerns (Blake & Mouton, 1982; Conger & Kanungo,1987); and interpret performance information aboutsubordinates (Green & Mitchell, 1979) All of these pro-cesses, it would seem, should demand the competent useof attention—in particular, Xexible attention directedtoward objects that help fulWll the organization’s goals.

Although the OB arguments are compelling, empiri-cal testing has generally been restricted to evaluation ofgeneral leadership models and broad managerial activi-ties. As such, speciWc implications of these theories forsocial attention have not been tested to the same extentas the powerholder-as-poor-attender perspective. Thecurrent paper seeks to integrate across the two litera-tures by testing a more inclusive, integrative view ofpower and social attention.

Keltner et al. (2003) have suggested that power isassociated with a set of approach-related aVects andbehaviors; in short, power reXects an “action orienta-tion” in which the powerful are thought to pursueopportunities and beneWts in their immediate environ-ments. On the other hand, low power is associated withgeneral aVective and behavioral inhibition; the powerlessare thought to be vigilant for threats in the environment,and to strive to avoid these threats. This approach-inhi-bition model posits that powerholders will tend to favoraction over inaction in virtually all cases. For example,Galinsky, Gruenfeld, and Magee (2003) demonstratedconvincingly that powerholders will act to serve theirown comfort, to advance their own concerns, and to ben-eWt the group at large. In short, powerholders in theirstudy pursued whatever behavioral option was the mostaction-oriented, shunning the passive.

As stated, Keltner et al. (2003) argue that power isassociated with heuristic processing of informationabout other people. In particular, they argue that poweris associated with inattention to low-power people andtheir concerns. Viewed more broadly, however, Keltneret al. can be seen as arguing for a more inclusive storyabout how power might aVect social attention. The the-ory of action orientation suggests that powerholders,more than others, are predisposed to act and toapproach. Likewise, Deschamps (1982) demonstratedthat people with power assume and are granted the role

J.R. O

verbeck, B. P

ark / Organizational B

ehavior and Hum

an Decision P

rocesses xxx (2006) xxx–xxx3

AR

TIC

LE

IN P

RE

SS

Table 1Overview

Source Implications for OB

GoodwiFiske,Dépre

Powerful members of newly-integrated workplaces are likely to stereotype new entrants; Power fosters sexism, racism in the workplace; Power is not associated with careful attention to subordinates or their concerns

rty providedorrectly

Keltner Robin

pponents’ Powerful partisans in ideological conXicts tend to see more conXict than is actually present, and fail to recognize areas of common ground

EbenbacKeltn

ntage in Powerful partisans in conXict over desired resources do not recognize the views of the other side

Rodrigu(2000)

s illegitimate, ve their position eotypes

Powerful members of organizations will denigrate others (through direction of attention) in order to maintain their own power

Copelan ttention could ore likely to

Supervisors and managers may be unlikely to go beyond their own expectations in gathering information about subordinates

De Dreu(2004)

ttention could secure a

Powerful negotiators are unlikely to go beyond their own expectations and, as a consequence, may deal with incomplete information

Vescio e ttention tner more

Supervisors and managers may be unlikely to go beyond their expectations in gathering information and are thus vulnerable to stereotypingothers; however, they can also pay good attention when the structure encourages it

Overbec(2001)

ttention was only ailable to HPs

Powerful members of organizations are likely to use attention as a resource to accomplish tasks well

of social psychological views of power and social attention

Power & social cognition Postulated mechanisms Utility of social attention

n et al. (2000) (also 1993; Fiske and t, 1996)

Power !Stereotyping(careless social attention)

1. Outcome independence Little:2. Cognitive load3. Dominant personality

1. Selecting unknown 3rd-paapplicants for a job: no goal2. Possible desire to decide cwho should get a prize

and son (1997)

Power !Lack of awareness of others’ perspectives, exaggerating intergroup diVerences

Outcome independence fosters lack of motivation to attend accurately

Little: Potential to counter oarguments more eVectively

h and er (1998)

Power !Inability to recognize others’ attitudes

1. Lack of motivation to attend accurately

Little: Potential to gain advadebate over resources

2. More positive aVect fosters poor attention

ez-Bailon et al. Power !More attention to negative stereotypes

1. Motivation to explainone’s own higher position

Moderate: When hierarchy iHPs are motivated to preserby attending to negative sterabout LPs

2. Motivation to justify one’s behavior

d (1994) Power !ExpectancyconWrmation biases

Power provides the meansto attain goals; Here, the goal is to form a predictable impression

Moderate to high: Careful alead to choosing a partner mhelp the HP win a prize

and Van Kleef Power !Expectancy conWrmation, leading questions

Outcome independencefosters lack of motivation to attend accurately

Moderate to high: Careful aenable the HP negotiator to better outcome

t al. (2003) Power !goal- and context-dependent attention to both stereotype andindividuatinginformation

Attention is guided by expectations and the goal of forming impressions to maximize outcomes

Moderate to high: Careful acould lead to choosing a parlikely to perform well

k and Park Power !More careful, individuating attention to others

Power fosters desire to excel in given tasks, sense of responsibility for performing

Moderate to high: Careful aperformance opportunity av

4 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

of “subjects” who act upon the world, as opposed tolow-power others, who play the role of object in thesocial world.

Just as the powerful tend to orient their actionstoward objects that promise fulWllment of desired goals,they should similarly orient their attention to social andperceptual stimuli that appear instrumental. Rather thanbeing careless or superWcial in their processing of socialinformation, powerholders should attend carefully toany stimulus—even a subordinate— that facilitates theirown goal attainment. This argument concords with theOB literature’s predictions that powerholders (in theform of organizational leaders) should use attentionaccording to their management goals and chronic styles.

To return to the opening question of this paper—howdo the powerful pay attention to others?—we suggestthat the answer is, “Xexibly.” Sometimes that attentionwill be careful and systematic, provided that the otherpeople in question are of some use in attaining the pow-erholder’s goals. If people are irrelevant to these goals,then the powerholder is likely to direct his or her atten-tion elsewhere, to other objects that are more immedi-ately instrumental. The studies that have shown heuristicprocessing by the powerful have tended to use para-digms in which the targets are of little practical use tothe powerful perceivers. Those that have shown moresystematic processing have used designs in which carefulsocial attention was important for discerning how well atarget might perform on a valued task, or for the power-holder’s ability to complete his or her own task. Keltneret al. acknowledge that power is likely to yield generalattention focused on attainment of one’s personal goals.We simply extend this perspective to social attention—attention to others and their concerns—and to goalsbeyond the personal.

In this work, we follow Thibaut and Kelley (1959) inusing the term “power” to refer to the potential or abil-ity to control others’ outcomes or even behaviors; assuch, power is a property of relationships between peo-ple, rather than a property of an individual. The power-holder can make decisions about the other’s resources,opportunities, and rewards, and can evaluate the other’sperformance.

Note that in an organization, any member’s formalpower is constrained by the organization itself. Theorganization provides and legitimates the individual’spower because it wants him or her to act on its behalfand to achieve its goals. The legitimacy provided by theorganization may be the only base of power, or may bethe foundation for additional bases of power (French &Raven, 1959; Raven, 1965). In any case, the legitimacyconferred by the organization is a necessary componentof the person’s power. Because of this, individuals hold-ing formal power in organizations should feel height-ened responsibility to produce desired results. In thefollowing studies, then, we manipulate not personal goals

of powerful participants, but rather the goals espousedby the organizations in which they interact. We arguethat powerholders will allocate their attention in aninstrumental manner to satisfy the organization’s goals.

For example, consider what happens when an oYcemember is promoted to a management position.Whereas previously she may have ridiculed managementinitiatives and rolled her eyes at cost-cutting eVorts, insolidarity with her co-workers, now as a manager sheWnds herself charged with creating and enforcing thoseinitiatives and cutting costs. It is likely that she will cometo accept and even endorse what she now needs to pro-mote. This could be because her incentives are alignedwith these behaviors, because of conformity pressuresfrom fellow managers, or because she fears for her ownjob if she fails to meet expectations. However, we suggestthat the more important cause is that she genuinelycomes to internalize the organization’s goals, and there-fore uses her available resources—including attentionand ability to act—in service of those goals. By doing so,she ensures that she will remain in power, and she enjoyscontinued legitimacy as an organizational powerholder.In our work, we will isolate the eVects of power andgoals from eVects of incentives, conformity pressures, orexpectations.

To integrate the social psych and OB literatures, weset out to accomplish the following. The dominant argu-ment in the psychological literature is that power leadsto stereotyping or other forms of careless perception,leading to failure to individuate targets (i.e., to see themas distinct individuals with unique characteristics andnot just the conferred characteristics of their socialgroups). Thus, we set out in Study 1 to use measures anda design that closely echo previous work on power andsocial attention, and to establish a very simple eVect:that at equivalent levels of power, we can Wnd variancein the degree to which perceivers individuate targets as afunction of diVerences in the perceivers goals. We lookonly at perceivers with high power, because it is thisgroup that has been the focus of most of the theorizingregarding allocation of attention as a function of power.Like past literature, Study 1 examined attention to onlyone attentional object, and we looked for goal diVer-ences in attention to that object.

In Study 2, we explore our central assertion: thatpower should lead to Xexible use of attention, wherebypowerholders allocate their attentional resources towardobjects that will help further the organizational goalsthat are the raison d’etre for the powerholders’ power.We predicted that, when examining the responses ofboth high- and low-power participants, we would Wndthis kind of goal-sensitive Xexibility only among thepowerful—a Wnding that would resonate with the atten-tion-relevant aspects of leadership theories. Further,Study 2 provides more than one attentional object. Weexpected that powerful perceivers would make more

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 5

ARTICLE IN PRESS

Xexible shifts between those objects, according to organi-zational goals.

In Study 1 we assigned diVerent goals—or, more spe-ciWcally, an equivalent set of goals with diVerent empha-sis—to powerholders in two conditions. Theseconditions reXect two distinct goal contexts: people-cen-tered and product-centered. Participants in both envi-ronments conducted the same tasks, and were assignedthe same two goals, but the emphasis on these goalsdiVered such that people-centered powerholders weretold that issues of employee engagement were para-mount, and product-centered powerholders were told toemphasize task productivity.

Also, in Study 1, we stayed close to the measures andmethods used in recent work (particularly Overbeck &Park, 2001) by focusing on individuation processes. Wepredicted that the people-centered supervisors wouldindividuate their workers well, and better than product-centered supervisors.

Study 1

Method

OverviewParticipants role-played supervisors in “a simulated

publishing company whose workers telecommute.” Allparticipants were assigned to the high-power supervisorrole, and were told that they would interact with Wveworkers (actually pre-scripted) who were completing aproofreading task. Instructions emphasized diVerentgoals for participants in two conditions: one, a people-centered context, stressed making workers feel engagedand included. The other, a product-centered context,emphasized productivity and eYciency. Note that eventhough we only collected data from participants in ahigh-power role, these participants believed that theywere interacting in a reciprocal hierarchy with low-power counterparts. Participants spent approximatelyan hour communicating, then answered questions aboutthe scripted targets.

ParticipantsNinety-Wve undergraduates at the University of Colo-

rado at Boulder participated in the study, in partial ful-Wllment of course requirements. Of these, three expressedsuspicion about the presence of the workers, Wve couldnot complete the study within the allotted time, and oneparticipant’s data were inadvertently deleted. Afteromitting these, 86 participants’ responses were analyzed.The group included 49 women and 37 men.

Stimulus materialsThe materials for the study were presented using a

HyperCard program on Macintosh computers. Each

participant/supervisor received 20 “e-mails”—four fromeach of Wve scripted worker targets. Each e-mail took theform of a report from a worker to the supervisor on oneof four proofreading tasks. The supervisor received aform on which the worker ostensibly provided numbersof errors found, reactions to the particular proofreadingtask, and any additional comments.

All written comments included two pieces of informa-tion for which memory was later tested. Every partici-pant dealt with the same 20 messages, presented in thesame order, and including the same scripted informa-tion. For example, in the Wrst message they received(shown in Fig. 1), supervisors learned that Joe (worker#1) found it easy to miss repeated-word errors and hadbeen to the town of Cluny, France.1 This informationwas identical across conditions.

Each target could be individuated through boththe information presented and the language and contentof the e-mails, which were manipulated to give eachtarget a “personality.” The Wve targets displayed thefollowing personalities, respectively: “good-natured/frien-dly,” “unintelligent/irresponsible,” “negative/argumenta-tive,” “competitive/aggressive,” and “smart/eYcient.” Joe,for example, consistently used friendly language, greetedthe participant aVably, and talked about his friends. Thepersonality traits were selected on the basis of their clarityand distinctiveness from one another (see Norman, 1963),and because they oVered the opportunity to script clearbehavioral manifestations into the e-mail exchanges. Theyalso reXect most of the dimensions used in our prior work(Overbeck & Park, 2001).

In addition, the proofreading performance of theworker targets was manipulated. Although the group oftargets improved somewhat, on average, from the begin-ning of the work session to the end, some targets consis-tently performed better than others. In particular, the“smart” target performed very well and the “unintelli-gent” target very poorly; the others performed betweenthese two extremes. Although the participant/supervi-sors were told to strive for a Wnal average of 95% accu-racy, the workers ultimately attained only about 85%accuracy, in order to maintain the participants’ motiva-tion to aVect worker behavior through the Wnal task.

Except for condition-speciWc instructions, all materi-als presented on the computer were exactly the same forparticipants in both conditions.

1 The scripted information, as shown here, included items both rele-vant and irrelevant to the interaction at hand. In past work (Overbeck& Park, 2001), powerholders’ memory for individuating informationwas particularly strong with respect to task-relevant information.Here, we did not observe such a diVerence, and item relevance will notbe discussed further. Participants were notiWed that their counterpartshad been told to share personal information, and despite substantialprobes for suspicion we did not Wnd that they found this informationto lack credibility.

6 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

ProcedureParticipants were seated at computers in separate

rooms and told that they were taking part in a study ofworkgroup eVectiveness in telecommuting contexts.Written instructions said that Wve of the current partici-pants would be assigned to a “worker” role and wouldbe proofreading several text passages, and that onewould have the “supervisor” role, to oversee the Wveproofreading workers. In fact, all were assigned to besupervisors. They were told that they would interact overe-mail; in fact, all the e-mails were pre-scripted. Partici-pants received a set of numeric goals (numbers of errorsand time for proofreading each passage) that eachworker was supposed to achieve.

After reading instructions, participants were assignedto their role. Assignment to goal condition was random,and was accomplished through the instructions given toparticipants. In the product-centered condition, partici-pants read,

Your task is to get the workers performing at the goallevel by the end of the work session. In order for tele-commuting to be a viable option, it is critical that peopleperform as productively and eYciently as possible.Therefore, we need to discover whether we can assurethe necessary levels of production. We do not wantworkers sabotaging the company’s goals by sitting athome watching TV when they should be working. Usewhatever means you can derive to accomplish this goal,including the kinds of recommendations and feedbackyou provide, and decisions about workers’ pay.

As a secondary aim, we Wnd that when workers feel posi-tive, engaged and with a sense of belonging to the organi-

zation, they may be more able to focus on doing what isbest for the company. You should also consider fosteringsuch feelings in your workers, but remember, your primarygoal is to get them to produce what the company needs.

In the people-centered condition, participants read,

Your task is to establish an atmosphere in which work-ers feel positive, engaged, and with a sense of belongingto the organization. It has been demonstrated that thisatmosphere causes workers to be more enthusiastic andmotivated to focus on doing what is best for the com-pany. If the right environment is established, the rest ofthe company’s goals can be met with much less eVortand diYculty. Workers are more likely to channel theireVort to the company’s beneWt, by doing their workrather than (for example) watching TV while they are athome, which is basically taking pay without working.Use whatever means you can derive to accomplish thisgoal, including the kinds of recommendations and feed-back you provide, and decisions about workers’ pay.

Of course, any company cares about productivity andmaking money, so you should also consider how wellyour workers are meeting productivity and accuracygoals. But remember, your primary goal is to foster theright kind of “company culture” for your workers.

Note that both conditions received both kinds ofgoals. It was merely the emphasis placed on each goal—like the hierarchy of goals found in actual organiza-tions—that distinguished the two conditions. Tostrengthen and check the manipulation, participantswere asked to list speciWc strategies they might use inorder to meet their assigned goal.

Fig. 1. Sample e-mail screen for people-centered participants in Study 1. Note the highlighted position diagram in the upper left, the tone and scriptedcontent of the e-mail, and elements of the feedback form that support manipulation.

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 7

ARTICLE IN PRESS

Participants next learned the names of the Wve(scripted) targets with whom they would interact, alongwith a spatial diagram labeled with the name of eachperson, to maximize the number of cues for participantsto keep track of targets’ identities. (This diagram, unla-beled but shaded to indicate the current target, isdepicted at the top of Fig. 1.) Participants were quizzedto ensure that they learned all names and spatial loca-tions: they saw a blank diagram with text boxes intowhich they had to type the name of each target. Partici-pants had to complete the quiz perfectly in order to pro-ceed.

The proofreading task was presented next. Partici-pants were given a sample proofreading task to try,along with feedback on the actual numbers of errors inthe task. They then received e-mailed work reports fromthe workers, one at a time. The spatial diagram appearedon each e-mail, and each worker’s position on the dia-gram was highlighted when his or her reports appeared.The supervisor completed a form on the computer, pro-viding feedback to each worker after each task: he or sheindicated the number correct out of the number possible(supervisors were told what the correct numbers were foreach task) and gave written feedback on the worker’saccuracy. Participants were also told that workers wouldautomatically be paid 75 ! for each task unless the super-visors changed this amount. A box was provided forsupervisors to assign a higher or lower pay. In order tosupport the goal manipulation, people-centered supervi-sors were asked to evaluate workers’ enthusiasm andmotivation, whereas product-centered supervisors evalu-ated worker’s productivity. Finally, participants wereasked to keep notes about their group’s progress on a“tracking sheet”; people-centered supervisors wroteideas to improve workplace inclusiveness, and product-centered supervisors tracked worker accuracy and speed.This was designed to reinforce the salience of the goalcondition.

After the work session, participants completed depen-dent measures. As each person left the lab, the experi-menter probed for suspicion and then fully debriefedeach participant.

Dependent measuresFour measures assessed the degree to which partici-

pants individuated the Wve targets. The Wrst was a confu-sions task (Taylor, Fiske, EtcoV, & Ruderman, 1978).The 40 pieces of information from the work reports wererandomly ordered and presented, and participants indi-cated which target provided each item. Correctly identi-fying the worker paired with each bit of informationreXects greater individuation of targets.

Next was a trait-rating task. Each target was rated on20 trait adjectives, including 10 traits characteristic ofthe various targets, two per target (good-natured,friendly, unintelligent, irresponsible, negative, argumen-

tative, competitive, aggressive, smart, eYcient) and 10words of opposite meaning (harsh, hostile, conscien-tious, dependable, easy-going, agreeable, passive, weak,slow, sloppy). Ratings were made on a seven-point scale(1 D “not at all descriptive” to 7 D “highly descriptive”).

Participants then completed a paired similarity task,in which they saw each possible dyadic combination ofthe Wve targets, and rated the similarity between the twomembers of each pair (1 D “not at all similar” to7 D “extremely similar”). In addition, two independentjudges rated the degree of actual similarity between eachpair of targets, based on both the scripted behaviors andpersonality information and on the targets’ work perfor-mance. The judges agreed on 86% of ratings. Bothjudges’ ratings were averaged together, and the meanratings were then used as the criterion to identify pairs oftargets who objectively were of low, medium, and highsimilarity. These objective categories were then used inevaluating how well participants perceived actual diVer-ences in the degree of similarity from pair to pair.

Finally, participants evaluated their power in thework session and their motivation to achieve time andaccuracy goals (1 D “not at all,” 7 D “a great deal”).Experimenters recorded the gender of each participant.2

Results

Manipulation checks

Participants’ strategy lists were coded for mention ofcontextual goals and values (1 D strongly product-cen-tered, 5 D strongly people-centered) by a single coder,blind to condition, as a check of the goal manipulation.3

Analysis revealed that, though the means do not fall atthe extreme ends of the rating scale, product-centeredsupervisors planned to use a more product-centeredapproach to their task (M D 2.56) and people-centeredsupervisors’ planned approach was signiWcantly morepeople-centered (M D 3.15), F(1, 85) D 6.42, p < .02,R2 D .07, .06 6 b 6 .52 (see Table 2 for all manipulationchecks). The quality of comments also indicated that

2 No eVects of gender were found, and gender did not moderate con-text for any results. Therefore, it is not discussed further.

3 For example, a people-centered participant wrote, “I want to createan environment where the workers feel their ideas are heard and imple-mented. I will do this by listening and compiling all their info to mosteVectively run the company.” A statement like this received a code of 5.A product-centered supervisor wrote, “I will let them know the incen-tives for working harder and more eYciently, which include an in-crease in pay for superior work. I will also let them know that thebetter the company does as a whole, the more successful it will be,therefore increasing their pay.” Such statements were coded as 1. A listcontaining no strong people- or product-centered content was assigneda code of 3.

8 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

participants were involved in the study and that psycho-logical realism was high.

It is important to conWrm that our values manipula-tion did not create diVerences in felt power among par-ticipants in the two conditions. Indeed, there was nodiVerence in how powerful members of each conditionfelt, Mproduct D 4.24 vs. Mpeople D 4.47, F(1,85) D 0.40, ns.There was also no diVerence in motivation to achievetime and accuracy goals, F(1, 85) D 0.09, ns.

Confusions task

Performance on this task indicates whether the goalvariables aVected participants’ ability to attribute infor-mation to the correct source and to distinguish the tar-gets from each other. The dependent variable is a countof the number of items correctly attributed. As pre-dicted, people-centered supervisors correctly matched agreater number of items to the target who provided theinformation (M D 25.65) than did product-centeredsupervisors (M D 21.52), F(1, 85) D 8.65, p < .005, R2 D .09,.47 6 b 6 2.44.

Trait ratings

Individuation involves successfully identifying eachcharacteristic trait with the target correctly described bythat trait (e.g., rating Joe high on “good-natured” and“friendly”) and rating that target low on the oppositetraits (rating Joe low on “negative” and “hostile”). Cor-rect targets (i.e., those who were in fact described bythese traits) should receive high ratings on characteristictraits and low ratings on uncharacteristic traits. Amongother targets, there should be less of a diVerence betweencharacteristic and uncharacteristic traits.

The signiWcant 2-way interaction of trait type £ targettype showed that, overall, there was a much greaterdiVerence in ratings of characteristic relative to unchar-acteristic traits for the correct targets than for the othertargets, F(1,85) D 212.87, p < .001, R2 D .70, 1.64 6

b0 6 2.15 (see Table 3). Further, the three-way interac-tion of trait type £ target type £ goal condition con-Wrmed that this pattern diVered for product-centered

Table 2Means and standard deviations of manipulation check responses(Study 1)

Note. Listed strategies were assigned a code on a 1–5 scale with1 D highly product-centered and 5 D highly person-centered. Powerand motivation ratings were made on a 1–7 point scale with highernumbers indicating higher levels of the measured construct. Means inthe same row with diVerent subscripts diVer at p < .05.

People-centered Product-centered

M SD M SD

Strategy to meet assigned goal 3.15a 1.07 2.56b 1.14Power felt 4.47a 1.71 4.24a 1.66Motivation 4.14a 1.39 4.22a 1.13

and people-centered participants, F(1, 85) D 5.29, p <.001, R2 D .06, .04 6 b 6 .56. Importantly, for the correcttargets only, people-centered participants were evenmore likely to correctly rate the characteristic traits highfor these targets and the uncharacteristic traits low, rela-tive to the product-centered participants, F(1, 85) D 6.55,p < .02, R2 D .07, .08 6 b 6 .69. It is clear from ratings ofthe “other” targets that participants did not simply over-generalize the presence of the traits. The diVerences forthe other targets were very small and negative, and didnot depend on condition, F(1,85) D 0.68, ns.

Paired-similarity task

Given that the targets were in fact quite distinct fromone another in their behavior and personality character-istics, lower average similarity ratings (across all targets)indicates greater individuation in that the raters haveresisted lumping all targets together. A test of the linearcontrast (low vs. medium vs. high similarity) can revealwhether participants pick up on the Wner distinctions inpair similarity.

This test showed that, although both groups did iden-tify a linear diVerence among low, medium, and highsimilarity pairs, F(1, 85) D 21.71, p < .001, R2 D .20, .44 6

b 6 1.11, people-centered supervisors made much stron-ger distinctions than product centered supervisors,F(1, 85) D 8.59, p < .005, R2 D .09, .16 6 b 6 .82, for thetest of the linear by goal condition interaction (seeTable 4).4 Collapsing across the linear contrast, meanswere consistent with greater overall similarity judgmentsby product-centered supervisors, though the diVerencewas not signiWcant.

A Proxscal multidimensional scaling analysis (Carroll& Chang, 1970) was carried out on the similarity judg-ments. Proxscal uses distance data to plot targets in atwo-dimensional space, allowing interpretations of how

4 The quadratic contrast was nonsigniWcant, although the quadraticcontrast £ organizational values manipulation was also marginally sig-niWcant, F(1, 85) D 3.39, p < .07.

Table 3Results of trait ratings (Study 1)

Note. Ratings on 1–7 scale with 7 D “extremely characteristic.” Means the same row with diVerent subscripts diVer at p < .05.

People-centered Product-centered

M SD M SD

Correct targetsCharacteristic traits (C) 5.16a 0.88 4.54b 0.95Uncharacteristic traits (U) 2.33a 0.62 2.47a 0.78DiVerence (C–U traits) 2.83a 1.28 2.07b 1.52

“Other” targetsCharacteristic traits (C) 3.39a 0.37 3.31a 0.45Uncharacteristic traits (U) 3.66a 0.37 3.51a 0.45DiVerence (C–U traits) ¡0.27a 0.37 ¡0.20a 0.47

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 9

ARTICLE IN PRESS

the targets are seen to diVer. It also calculates a score foreach participant that reXects the degree to which eachdimension was used in judging similarity.

Across all participants, targets were judged using twoprimary dimensions. Dimension 1 (the X-axis in Fig. 2)appears to involve the personality traits “uptight” versus“lazy”: The four targets grouped at the left end of theaxis were more motivated (or tense) about doing well,and the target at the far right was exceptionally lazy.Dimension 2 (the Y-axis) appears to capture the topicfocus of targets’ e-mails. That is, the targets toward thebottom of the axis used more personal references andstories. The other targets’ comments were more task-focused.

Interestingly, the two conditions weighted thesedimensions diVerently when judging target similarity,F(1,85) D 13.14, p < .001, R2 D .13, .03 6 b 6 .09. Whileproduct-centered supervisors used both dimensionsabout equally, people-centered supervisors gave moreweight to the personality trait dimension and much lessto the topic focus dimension. This diVerence suggeststhat people-centered supervisors may be more attuned topersonality traits than to focus, and product-centeredsupervisors attend equally to both.

Discussion

The results from Study 1 support the argument thatpowerful perceivers can be Xexible with their attention,

Table 4Paired similarity ratings (Study 1)

Note. Similarity ratings on 1–7 scale with 7 D “extremely similar.”

Actual (scripted) similarity

Low Moderate High

M SD M SD M SDPeople-centered 2.07 1.02 4.37 0.96 4.71 1.01Product-centered 2.82 1.21 4.25 0.78 4.39 1.00

Fig. 2. Paired similarity judgments, Study 1: Two-dimensional spacerepresentation of how supervisors formed judgments of similarityamong targets.

-1.5

-1

-0.5

0

0.5

1

1.5

-1.5 -1 -0.5 0 0.5 1 1.5 2

Uptight - Lazy

Fri

end

lines

s

Joe

Chris

Amy

Matt

Katie

using perception as an instrumental resource to achievetheir ends. Two perceivers who were equally powerful,when operating in environments that elicited diVerentobjectives, displayed diVerent patterns of social percep-tion. Powerholders with the primary objective of foster-ing a sense of inclusion attended more to individuals,kept straight information about them, learned abouttheir personalities, and saw them as distinct from oneanother. Powerholders with the primary objective ofassuring output were less attuned, more confused, andless able to discern the unique characteristics of socialtargets. In addition, the Proxscal analysis shows thatpowerholders’ goals apparently created an informationexpectancy that guided the kinds of impression informa-tion used in judging similarity. This study conWrms thatpower does not lead Wxedly to heuristic thinking; rather,powerholders can attend Xexibly, even engaging in con-trolled and systematic interpersonal perception if thatfacilitates the goals they’re charged with meeting.

In one sense, it is not terribly surprising that diVerentsocial or cognitive goals produce diVerent patterns ofperson perception—this has been argued convincingly inearlier research (Neuberg, 1989; Sanitioso, Freud, & Lee,1996; Seta & Hayes, 1994; Trzebinski & Richards, 1986).However, the present Wndings contribute in importantways to the power literature, going beyond the demon-stration of simple goal eVects. First, the goal that partici-pants were given was to achieve a particular outcome.The presentation of this goal did not explicitly mentionattention. It was up to the participants to decide how toallocate their attention in order to achieve that outcome.Thus, it is unlikely that participants would have paidattention diVerently when trying to, for example, enact aproduct-centered environment, unless they took theintermediary step of deciding that this attention was ameans to achieve that goal.

Second, much of the psychological power literaturehas developed under the notion that eVects of power areinvariant and predictable, as long as level of power isidentiWed. Thus, it is important that one can observevariations in perception within a single power level. Asargued above, it is important that the judgment patternsdisplayed by participants correspond to what was mostinstrumental for them, given their assigned goals. ThisreXects powerholder Xexibility, as perceivers withequally high power paid attention diVerently accordingto these goals.

Though the results of Study 1 support our contentionthat powerholders can direct their attention in an instru-mental manner, it leaves some important questionsunanswered. First, if indeed attention is Xexible, weshould be able to observe powerholders allocating atten-tion diVerently to diVerent kinds of information, depend-ing on what oVers immediate utility. In other words, notonly do we want to show that powerholders pay more orless attention to one attentional object, but also that they

10 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

Xexibly transfer their attention from one object toanother depending on the instrumental value of thoseobjects. Keltner et al. (2003), and previous researchers(Goodwin et al., 2000; Overbeck & Park, 2001), havealso tended to focus on arguments about the amount ofattention paid by powerful and powerless perceivers.Our current focus, however, is on how that attention isdirected. We assume that powerholders generally paysubstantial attention to the external environment, asthey monitor it for opportunities (Keltner et al., 2003). Itmay sometimes appear that little attention is being paidto people and their concerns because the powerholder’sattention, though substantial, is being directed towardsomething more instrumental. Study 1 did not allow usto assess this, because only one attentional object wasprovided. Therefore, we designed Study 2 to include twokinds of attentional objects—information about bothpeople-oriented issues and productivity issues—to moredirectly examine how attention is directed.

In addition, we broadened the kind of informationprovided. Having already examined attention to indi-viduating information about the targets to provide alink to past research, we now wanted to examine a mea-sure that is perhaps more directly related to the day today functioning of an organization—namely, the issuesand concerns that are judged to be of greatest impor-tance in running the organization. If a powerholder ispresented with some issues and concerns that focus onthe people in the company and some that focus on pro-ductivity and Wnancial performance, our Xexibilityargument predicts that the type of concerns judged tobe more important will vary as a function of the goalgiven to the powerholder. To this end, we providedinformation about people-oriented policies, achieve-ments, and challenges faced by the organization, ratherthan simply the personal characteristics of individualworkers.

Study 1 is not informative about whether Xexible,instrumental attention is unique to powerholders,given that powerless actors were not included in thedesign. It might be that all organizational actors allo-cate their attention according to the goals emphasizedby the organization. Study 2 addressed this issue byusing a full design in which both powerful and power-less actors were presented with the same organizationalgoals and the same information about the organiza-tion. We predict that attention should be allocated inXexible, instrumental way by powerholders, but not bythe powerless.

Finally, as our earlier examples made clear, usingattention Xexibly could be a result of other variables thatmight covary with power in organizations, particularlyincentives, pressures toward conformity, and the expec-tations of peers and superiors. We designed our condi-tion manipulations to dismiss these alternativeexplanations for attentional diVerences.

Study 2

Method

OverviewParticipants were told that they would be role-playing

a company work team, interacting live with one managerand many subordinates, and were assigned randomly toone of these roles. They were further told that their com-pany held either people-oriented values or production-oriented values, to provide a goal context to guide theirattention. Next, participants were presented with infor-mation about the company’s recent accomplishmentsand challenges, including both people-oriented and pro-duction-oriented items. Measures assessed participants’use of this information and how it reXected their sense ofpriorities.

ParticipantsEighty-four undergraduates at the University of

Southern California (51 men and 33 women) partici-pated in the study, in partial fulWllment of courserequirements for an introductory course in organiza-tional behavior.

Materials and procedureParticipants were seated in small conference rooms in

groups of 8–10 and told that they would be conductingan organizational simulation. They Wrst received a briefquestionnaire containing items that assessed extraver-sion and surgency (cf. Norman, 1963), and that mightappear on their face to be validly related to leadershipaptitude. This questionnaire was ostensibly scored onthe spot, and the results used to assign participants to arole that they would play in the simulation: either man-ager (high power) or worker (low power).5 After theapparent scoring, participants received a packet thatcontained the power manipulation and instructions forthe study. In the high-power (HP) condition, partici-pants read:

You are going to be the MANAGER of this workgroup.Everyone else present will be your subordinates, and youwill supervise them.

5 In the literature, social power is manipulated in various ways, in-cluding semantic priming (Bargh & Raymond, 1995), experientialpriming (Galinsky et al., 2003), measurement of naturally-occurringpower diVerences (Ebenbach & Keltner, 1998), or providing diVerentresources to high- and low-power participants (Kim & Fragale, inpress). Here, we follow past work that has instantiated a more involv-ing and psychologically real experience of power by placing partici-pants into positions of formal power, in interactive situations with realor simulated others (Copeland, 1994; Overbeck & Park, 2001; Vescioet al., 2003). We believe that our manipulation is practical and exter-nally valid, and is most likely to evoke the cognitions and responsesthat ordinary people feel when interacting in a real-world hierarchy.

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 11

ARTICLE IN PRESS

In the low-power (LP) condition, participants read:

You are going to be a WORKER in this workgroup.One of the people present will be the Manager and willsupervise you and the rest of the group, who will be yourfellow workers.

Though they were led to believe that each group con-tained only one manager and that all others were work-ers, this was not true. In fact, roles had been randomlypre-assigned and the questionnaire responses were notused in the assignment.

The same packet of information contained a manipu-lation of the primary values held by the organizationbeing simulated. In the “people-oriented” condition,participants read:

Your company has a history of particularly good work-ing relationships with employees. The company’s mis-sion statement emphasizes that it cares very much foremployees and is concerned with making its people feelvalued and included. As you act as manager to promotethe interests of the company, please keep in mind thatthe company wants all its employees to experience highsatisfaction and a sense of personal involvement in theirwork. In order to accomplish this, your decisions shouldbe consistent with the high value that the companyplaces on its people and their concerns.

In the “production-oriented” condition, participantsread:

Your company has a history of pursuing optimal pro-ductivity and running a lean operation. The company’smission statement emphasizes that it wants to be theleader in its industry, to maximize proWts and share-holder returns, and to ensure the highest possible pro-ductivity. As you do your work to promote the interestsof the company, please keep in mind that the companywants eYciency and productivity to be reXectedthroughout all activities. In order to accomplish this,your actions should be consistent with the high valuethat the company places on Wnancial performance.

Participants were asked to summarize these instruc-tions before continuing, both to heighten the salience ofthe manipulation and to verify that they understood it.

Following the delivery of these two manipulations,participants viewed a sample of the workgroup task thatwould ostensibly be the main activity of the simulation.They were told that they would create a company bud-get; each worker would request funding for a projectrelated to their functional area, and try to win fundingfor that project. The manager would have Wnal authorityover the budget, but could not exceed $1,000,000. Partic-ipants were told that the workers would have to makedetailed proposals and defend those proposals by per-suading the manager and answering questions, and thatthe manager had sole discretion in Wnal budget decisions.

Several features of this design deserve special note.First, this task was highly relevant to the students’ class-room work, and they were accustomed to participatingin organizational simulations for course credit; thus, webelieve that this constituted an engaging task in whichparticipants felt a genuine desire to perform well, andfelt that the power diVerences were salient and conse-quential. Second, each ostensible workgroup (which par-ticipants believed included one manager and multipleworkers) was seated in a separate room, and did notexpect to interact with any other participants. Thus, nopressures toward conformity among managers werepresent. Third, there were no additional incentives pro-vided to managers, relative to workers, for meeting orga-nizational goals. The workers had reason to want tosucceed at an individual level, insofar as they would beproud to get their own projects funded and mightimpress the experimenter; similarly, the manager’s onlyincentive for good performance was potential for prideand the experimenter’s approval. Finally, all participantswere told that they were personally expected to meetorganizational goals.

After this preview of the workgroup task, participantswere told that they must Wrst become familiar with thecurrent state of the company, so that their proposals anddecisions would be more informed. To this end, all par-ticipants saw the same list of 20 statements (hereaftercalled “target items”) about “major events at the com-pany right now: problems, accomplishments, and state-ments of current conditions.” Embedded within this listwere 10 items that focused on employees’ personalachievements and concerns (“Employees of the companyhave expressed concern that vacation policies are toostrict”; “One of your researchers has just won a majorinternational scientiWc prize”) and 10 that focused onproductivity and Wnancial health (“The industry ishighly competitive and this company has been losingmarket share”; “A recent retooling of the productionline has speeded production by 8%”). The full set of tar-get items is presented as Appendix A.

After reading all the items, participants were directedto write a concise “State of the Company” report, sum-marizing the few target items of greatest importance tothe company. If the participant was a manager, thisreport was to be written to the Board of Directors. If theparticipant was a worker, the report was written forinclusion in the employee newsletter. Though they werefree to choose the items themselves, we recommendedthat they limit themselves to approximately six itemsthat they found most important. Next, we presented thesame list of 20 target items and asked participants torank the importance of each item for the functioning ofthe company, from 1 (the most important item on thelist) to 20 (the least important item on the list).

At this point, participants handed in their study materi-als and believed that they would go on to the workgroup

12 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

(budgeting) task. However, the experimenter informedthem that they would not be completing the workgrouptask. They were debriefed, thanked, and released.

Dependent measures

Two measures assessed participants’ attention andpriorities. The Wrst was the State of the Company reportwritten by each participant. These reports were coded bytwo independent coders, blind to condition and hypothe-ses, for the presence of the 20 target items. Agreement onthese codes was 97%; disagreements were resolvedthrough discussion. The measure of interest was thenumber of people-oriented versus production-orienteditems included in the report.

The second measure was the priority ranking of the 20target items. To focus precisely on the participants’ toppriorities, we examined those items that each participantranked in their own top 5. Again, we counted the numberof people-oriented versus production-oriented itemsincluded in these top 5 priorities. The measure analyzed isthe number of people-oriented items that participantsincluded in the top 5, given that the number of people plusproduction-oriented items necessarily summed to 5.

Pilot study

To examine the eYcacy of the power manipulationwithout contaminating responses on the DVs, we admin-istered this procedure (minus the presentation of targetitems and associated DVs) to a group of 49 undergradu-ates at the University of Southern California. After read-ing about their assigned role, organizational goals, andthe upcoming workgroup task, participants were askedabout their feelings of power and expectations placedupon them. Then they were debriefed and dismissed.

In particular, we were concerned with demonstratingthat our manipulation resulted in a greater sense of poweramong HP relative to LP participants, but that it did notproduce diVerences in perceptions of the extent to whichparticipants felt they were expected to act on behalf of thecompany’s best interests. Results of the pilot study con-Wrm that our manipulation led to signiWcant diVerences infeelings of power. We averaged across responses to fouritems (To what extent do you feel that you have the abilityto obtain the outcomes that the company prefers in thistask; To what extent do you feel that, in this exercise, youare given authority and control by virtue of your positionin the simulated organization; To what extent do you feelthat you have the ability to aVect the outcomes of othersin your workgroup in this task; How much power do youfeel that you’ll have in the workgroup task?) to create afelt power scale. HPs felt signiWcantly more powerful(MD6.11) than LPs (MD 4.79), F(1,44)D28.94, p < .001,R2 D .40, 0.416b60.89.

In contrast, when we asked participants about thedegree to which they believed that they were expected toact in accordance with the organizational goals assignedto them (To what extent do you feel that you areexpected to behave in ways that maximize proWts andshareholder returns for the company; ƒin ways thatensure high satisfaction and personal involvementwithin the company?), there were no diVerences betweenhigh- and low-power participants on either proWt expec-tations, MHP D 6.32 vs. MLP D 6.00, F(1,44) D 0.67, ns,or people expectations, MHP D 6.14 vs. MLP D 5.65,F(1,44) D 2.34, ns, conWrming that our power manipula-tion did not create diVerent senses of expectations forperformance. That is, supervisors and workers alike feltthat they were expected to be as good an employee aspossible in achieving company goals.

Results

This study used a 2 (participant power: high vs.low) £ 2 (company values: people-oriented vs. produc-tion-oriented) between-subjects design. Because of non-responses on various measures, the number ofobservations used in the following analyses also varies.

Comprehension of values instructions

We checked to ensure that participants understoodthe company values instructions. Indeed, 100% reportedthem correctly.

Reports

We predicted that powerful participants shouldreXect the assigned organizational goals in their reports;that is, powerholders in the company that valued peopleshould include more people-oriented items in theirreports than should those in the company valuing pro-duction. Those in the production-valuing companyshould include more production-oriented items thanthose in the people-valuing company. We did not expectlow-power participants to show this Xexibility accordingto company values condition.

All participants tended to write more about productionitems than people items, F(1,74)D9.67, p<.003, R2 D .08,¡1.046b6¡.23. This was especially true for those in theproduction-valuing company, F(1,74)D7.02, p< .01,R2 D .09, ¡.946b6¡.13, and, separately, for powerfulparticipants, F(1,74)D5.20, p<.03, R2 D .07, ¡.876b 6

¡.06. The 3-way interaction of power, values condition,and item type was marginally signiWcant, F(1, 74)D2.79,p<.10, R2 D .04, ¡.746b6 .07. More important, as pre-dicted, the simple 2-way interaction of values conditionand item type was signiWcant only for powerful partici-pants, F(1,74)D7.78, p<.01, R2 D .10, ¡1.506b6¡.25.

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 13

ARTICLE IN PRESS

Simple eVects tests of the means presented in Table 5revealed that, consistent with predictions, powerholderswrote about signiWcantly more production items in theproduction than in the people condition, F(1,74)D5.95,p< .02, R2D .08, ¡1.126b6¡.15, and wrote about sig-niWcantly more people items in the people than in the pro-duction condition, F(1,74)D6.71, p<.02, R2 D .08,.116b61.10. The 2-way interaction of values conditionand item type for low-power participants was not signiW-

cant, F(1,74)D0.61, ns, nor were either of the simple eVectsof item type as a function of the two values conditions,people F(1,74)D0.01, ns, product F(1,74)D 1.02, ns. Thisindicates that powerful participants used company goalsinformation to guide their attention to and selection ofinformation for inclusion in the company report, whereasthe powerless did not.

PrioritiesAgain, we predicted that powerholders in the people-

valuing company would include more people-orienteditems in their top 5 rankings than would those in theproduction-valuing company. We expected no suchdiVerence among powerless participants.

Overall, all participants tended to rank a greater numberof production items (MD3.39) in their top 5 than peopleitems (MD1.61), F(1,80)D29.05, p<.001, R2D .25,¡2.116b06¡.97. However, as expected, this eVect wasmoderated by the 2-way interaction of power level andcompany values, F(1,80)D4.45, p<.04, R2D .05,¡.596b6¡.02. As shown in Table 5, powerholdersincluded more people-oriented items in their top 5 rankingsin the people-oriented company than in the production-ori-ented company, F(1,80)D6.35, p<.02, R2D .07,¡.986b6¡.12. For low-power participants, there was nodiVerence as a function of the company-values manipula-tion, F(1,80)D0.53, ns, R2D .001. This Wnding indicates thatpowerholders indeed allocated their attention in an instru-mental manner. Low-power participants, though providedwith the same information, goals, and activities, did not.

Discussion

The results of Study 2 conform nicely to our predic-tions. People assigned to powerful roles in a workgroup

were more responsive to the stated values and goals ofthe organization that empowered them: they used value-consistent information more in their reports on the stateof the organization, and they ranked such informationas more important to the continued functioning of thecompany. People assigned to powerless roles in theworkgroup did not follow this pattern; instead, theyappeared insensitive to the stated values and goals of theorganization. It would be easy to expect all participantsto simply follow the goals that they were given. Thatpower aVects the likelihood of doing so is, we think, anovel and dramatic Wnding.

It is noteworthy that Study 2 goes beyond the narrowrealm of interpersonal cognitive measures of stereotyp-ing and individuation emphasized in previous work onpower and cognition (Fiske, 1993; Keltner et al., 2003;Overbeck & Park, 2001; Vescio et al., 2003). The mea-sures used here assess attention to matters relevant tothe human side of the organization, as well as mattersinvolving the organization’s task functions. Further, theresponses required of participants are transparent intheir relevance to the kind of instrumental use of infor-mation that takes place in real organizations.

General discussion

Study 1 provides a bridge from past work on powerand individuation, by using similar methods and mea-sures to reveal an eVect that the social psychology litera-ture would generally not predict: powerful perceiversshowed variance in individuating social targets, based onthe organizational goal that the perceiver was trying tomeet. Powerholders who received people-centered goalstended to show better individuation of powerless targetsthan did those who received product-centered goals,across personality and memory measures, suggestingthat the former summoned more of their attentionalresources for the task.

Study 2 extended this result by using diVerent mea-sures and domains of attention. In addition, Study 2conWrmed that instrumental, Xexible use of attentionwas unique to powerholders in this context; the power-less did not show the same Xexibility. The manipulationused in Study 2 helps us to discount alternative explana-

Table 5Means on priority and memo tasks (Study 2)

Note. In the top two rows, means with diVerent subscripts diVer at p < .05 when they occur within the same power-condition heading. Across all of thethird row, means with diVerent subscripts diVer at p < .05.

High power Low power

Production condition People condition Production condition People condition

M SD M SD M SD M SD

Production items in memos 4.37a 1.12 3.15b 1.57 2.91a 1.59 2.78a 1.24People items in memos 1.58c 0.96 2.85b 1.57 2.39a 1.53 2.83a 1.34People items in top 5 priorities 1.10a 1.09 2.19b 1.28 1.91b 1.41 1.79b 1.38

14 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

tions that our diVerences were based on diVerentunderstanding of the goals, on diVerent incentives, ondiVerent pressures for high- versus low-power partici-pants, or on diVerent expectations about how theyshould behave. Instead, we argue that power helps peo-ple to direct attentional resources more Xuently andeVectively toward goal attainment.

The whole point of power, one might argue, is to getsomething done. Whether a political leader striving tocontrol territory and policy, a corporate executive seek-ing proWt or protecting turf, or even a parent trying toquiet a child in a restaurant, people see power as a meansto an end. Indeed, the approach-inhibition model ofsocial power (Keltner et al., 2003; see also Galinskyet al., 2003) focuses strongly on power as action. As men-tioned earlier in the paper, this model implies that pow-erful people will orient toward opportunities in theenvironment and will use resources—such as attention—that help them obtain rewards and fulWll goals. It makeslittle sense, then, to expect that power might invariantlylead to heuristic, careless, or deindividuating socialattention. Instead, as OB theorists have long argued,diVerent circumstances call for diVerent degrees of atten-tion and interpersonal sensitivity. This paper supportstheir work by showing that, indeed, leaders may beuniquely suited to meeting these demands.

Power and goals

The key variable determining how power will aVectsocial attention is, we argue, the goals of the power-holder. In the laboratory, it is easy to see why power-holders will adopt the goals they are given: there is alimited, narrow set of activities and attentional objectsprovided to participants, and in this stimulus-poor envi-ronment the goal manipulation looms large and com-manding. (What is truly remarkable about our Wndings,of course, is that it did not loom similarly large for thepowerless participants.) In a real organization, however,powerholders do not just passively receive goals, but areactively involved in creating them. Nonetheless, webelieve that, as part of their creating goals for themselvesand the organization, a process of internalization willoccur whereby powerholders adopt the goals of theorganization as their own.

There are two reasons that powerholders shouldtend to internalize their organizations’ goals. The Wrstis constraint. Although it is tempting to think of power-holders as unconstrained and free (Brauer, Chekroun,& Judd, 2004; Keltner et al., 2003), in fact powerhold-ers too operate under constraints, particularly in orga-nizational contexts. Hamilton and Biggart (1985)argued that the expectations associated with their rolesconstrain powerholders, who must produce results, actpowerful, and behave so as to ensure both the coopera-tion of those below them and the approval of those

above. Organizations grant power to individuals tomake them agents who can act on the organization’sbehalf, accomplishing desired ends. Though power-holders may have considerable latitude in terms of howthey accomplish these ends, they also nonetheless bearthe obligation to do so (also see Overbeck, Tiedens, &Brion, in press). The powerholder needs not only to getthings done, but to do them on behalf of the organiza-tion, and thus the goals guiding action should be thoseof the organization.

The second and, we believe, more compelling reasonfor the powerful to internalize organizational goals isthat adopting these goals allows the powerholder tomaximize legitimacy. In an organization, most formalpowerholders are not innately powerful—rather, theorganization provides and legitimates their power. It canbe argued that a meta-goal of the powerful is to main-tain, if not increase, their power (ConniV, 2002; PfeVer,1992). Bachrach and Baratz (1962) stated that one of theprincipal forms and purposes of power is the manipula-tion of attention in order to perpetuate the power statusquo. Powerholders are more likely to maintain power ifthey stay in the good graces of the organization, andadopting the organization’s goals as their own—thenpaying attention to the things that fulWll those goals—isobviously a promising strategy.

Cognitive eVectiveness

It seems likely that power improves eVectiveness notonly through the direct sense of control that powerhold-ers enjoy, but also through their potential for more Xexi-ble cognition. In the book Black Hawk Down, journalistMark Bowden discusses the power and team dynamicscharacterizing two groups of soldiers involved in a cata-strophic incident in Mogadishu, Somalia, in the 1990s.The elite, high-status Delta Force soldiers are given thepower to make decisions for themselves. Their tacticalmeetings have the Xavor of group consensus sessions,and each individual is able to control critical aspects of amission. The somewhat lower-status Rangers follow tra-ditional chain-of-command procedures and enjoy farfewer freedoms and less autonomy.

Bowden’s account touches on attention—the youngRangers appear to be monitoring how “cool” the Deltasare, rather than on mission-critical objects. Deltasattended to the environment, monitoring cues that couldtell them how to plan, act, respond, in order to keepthemselves and their buddies alive. The Rangers, theycomplained, were perpetually unready and unaware,even to the extent of Wring on their own team during anoperation because they did not pay enough attention towhere the other team members were. In combat in Mog-adishu, these diVerences were painfully consequential,with Rangers tending more to hang back and wait forDeltas to make decisions about how to proceed, in some

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 15

ARTICLE IN PRESS

cases causing problems because squads remained sepa-rated and thus harder to rescue.

Though the kind of power held by Delta Force maylook somewhat diVerent from that held by an organiza-tion manager, nonetheless this illustrates how power canlead people to use attention as one of their availableresources, and to move among goals with Xexibility andlack of tunnel vision. Using attention eVectively, in thisway, can be an enormous advantage for the power-holder.

What about the powerless?

Although this example suggests that lower-powerindividuals show poorer attention, our data oVer littleinsight into how the powerless use social attentionresources. There is some evidence that the powerless seepowerholders in an undiVerentiated, unindividuated wayin at least some settings (cf. Overbeck & Park, 2001; Ric,1997) and they may, as a consequence, attend to a Wxed(and perhaps erroneous) set of objects that they believewill be useful, failing to attend Xexibly or instrumentallywith respect to the goals of the organization.

A more interesting possibility is that the powerlessare not using social attention as a resource at all. Someevidence suggests that the powerless are primarily con-cerned with projecting a positive impression, and withevaluating the impressions they’re making on others.Copeland (1994) found that, when powerless partici-pants interacted with powerful prospective partners,they were most concerned with how the powerful per-son was seeing them—that is, they experienced theinteraction as objects rather than as subjects or per-ceivers. Similarly, Earle, Giuliano, and Archer (1983)found that, in dyads interacting in an explicit hierar-chy, the low-power partner tried to reveal informationabout themselves—to make themselves more know-able and more perceptible—in a way that reinforcedtheir status as an object and not an agent ofperception.

The powerless, then, may tend to be self-focused insocial interactions with the powerful. Rather than look-ing outward for opportunity (the perspective of anagent), powerless people may look inward. If so, theirself-orientation could supersede the outward direction ofattention. In Study 2, though we did not measure thispossibility, the powerless participants may well havebeen inattentive to company goals and informationwhile they thought anxiously about how they would beperceived when the workgroup budgeting task occurred.In a similar vein, Bowden (2000) suggests that the lower-power Rangers used attention largely to compare them-selves with the elite Delta Force, in part assessing theirown competence and toughness. Certainly, this echoeswork on performance decrements caused by self-focusdue to social anxiety and stereotype threat (Mansell,

Clark, & Ehlers, 2003; Steele & Aronson, 1995). This is acompelling possibility that we plan to examine in futureresearch.

Limitations of the current work

We have argued, and found conWrmation, that power-holders adopt the goals of their organization; however,our work is silent on the issue of how powerholdersmight respond to competing attentional demands. Forexample, will the powerful always privilege the values ofthe organization, or will they put their personal concernsWrst when given the opportunity to do so? What if thereare sanctions or punishments for privileging personalgoals? It would be useful for future research to explorepowerholders’ allocation of social attention givenaligned and misaligned personal and organizationalgoals. Further, we have argued that goal-directed atten-tion is “Xexible”; however, our design did not allow us totest whether powerholders change attentional directionin response to changing situations. To truly reXect Xexi-bility, that should occur. Thus, additional work is neededto explore whether attentional variability occurs justbetween powerholders, or within the individual power-holder, as well.

The studies are also limited in their test of instrumen-tality. Limited goals were oVered to participants, and theattentional objects that might help meet those goals werequite circumscribed. It would be fruitful in futureresearch to explore how powerholders act on goals thatcan be carried out in multiple ways, and whose implica-tions for the interpersonal environment are less clear (cf.Galinsky et al., 2003).

Finally, the studies are of course limited by the factthat such a vast and complex construct as power ismeasured in a laboratory setting, using power-unfamil-iar undergraduates, and in role-playing simulations.Although these are real limitations, we argue that thesestudies constitute a strong and valid test of our hypoth-eses. First, power in the real world is so vast and com-plex as to be downright messy; the laboratory setting isessential for isolating eVects of the variable of poweritself, and not dominance, status, incentives, expecta-tions, or the myriad other constructs that covary withpower in more natural settings. Second, althoughundergraduates do not have extensive experience withbeing in power, our design creates relative diVerences inpower, which are suYcient to test our hypotheses.Third, through years of testing various power manipu-lations, we have found that role-playing provides themost absorbing and psychologically real experience ofpower possible in the laboratory. Participants’ perfor-mance within the studies, as well as their anecdotalcomments after the studies, indicate that they take thesimulations very seriously and Wnd them convincingand engrossing.

16 J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx

ARTICLE IN PRESS

Conclusion

Rather than being heuristic and automatic, power-holders can be Xexible and variable. The more researchilluminates the responses of powerful individuals to theirsocial worlds, the more it appears that there is little utilityin analyzing power as an individual diVerence with invari-ant consequences for those who have it. Power is embed-ded in a context of relationships, roles, and resources, andwhether resulting actions or cognitions appear corrupt orbenevolent is determined by the interaction of these con-textual elements rather than by power itself. Theapproach-inhibition model oVers a useful framework forconsidering how power may inXuence outcomes whilebeing sensitive to variability due to context. It is hopedthat this work extends that Xexible perspective on powerto the domain of social attention, as well.

Appendix A

Twenty target items presented to participants inStudy 2. People-oriented items are designated with a (P)following the item.

1. The building’s infrastructure is aging and requirescostly updating, which will cut into this year’s proWts.

2. The industry is highly competitive and this com-pany has been losing market share.

3. Employees of the company have expressed concernthat vacation policies are too strict. (P)

4. A group of employees hopes to start a mentorshiptraining program, to make new hires’experiencesmore positive. (P)

5. An independent auditor has found that the com-pany is wasting money through ineYcient purchas-ing and poor accounting practices.

6. An independent review board has found thatincreasing demands from the company on the timeand energy of employees are making it diYcult toachieve a good work-life balance. (P)

7. The company is trying to negotiate several newcontracts that will increase its sales substantially.

8. A recent retooling of the production line hasspeeded production by 8%.

9. New government rules mean that many employeeswill stop being eligible for overtime pay in the nextfew months. (P)

10. A group of employees has formed a sailing team,under the company name, that has won severallocal championships. (P)

11. Workers are being injured on the production lineat a higher rate than average. (P)

12. Several managers have been criticized by employ-ees for creating unpleasant work environments intheir departments. (P)

13. The company has established new productivitystandards, but most production lines are operatingat only 85% of the new standards.

14. An expensive new software package, if imple-mented throughout the company, might providebetter cost containment and Wnancial planning.

15. The company conducted a layoV in the past yearthat saved a substantial amount of money, primar-ily through savings on salaries and health care.

16. Employees have expressed a great deal of concernabout the negative eVects of the layoV on team-building and trust. (P)

17. The company is spending a great deal of money onlegal fees to Wle several new patents and protect itsintellectual property rights on its existing patents.

18. The COO has proposed a plan that could nearlydouble production with only 10% additionalinvestment.

19. One of your researchers has just won a major inter-national scientiWc prize. (P)

20. The company has inspired such loyalty from work-ers that many have been there 15 years or more. (P)

References

Bachrach, P., & Baratz, M. S. (1962). The two faces of power. AmericanPolitical Science Review, 56, 947–952.

Bargh, J. A., & Raymond, P. (1995). The naive misuse of power: Non-conscious sources of sexual harassment. Journal of Social Issues, 51,85–96.

Bass, B. M. (1985). Leadership and performance beyond expectation.New York: Free Press.

Blake, R. R., & Mouton, J. S. (1982). Management by grid principlesor situationalism: Which? Group and Organization Studies, 7,207–210.

Bowden, M. (2000). Black Hawk Down. New York: Penguin Books.Brauer, M., Chekroun, & Judd, C. M. (2004). Power and social control:

Reactions to deviance in high and low power groups. Unpublishedmanuscript.

Burns, J. M. (1978). Leadership. New York: Harper & Row.Carroll, J. D., & Chang, J. J. (1970). Analysis of individual diVerences

in multidimensional scaling via an N-way generalization of “Eck-art-Young” decomposition. Psychometrika, 35, 283–319.

Conger, J. A., & Kanungo, R. (1987). Toward a behavioral theory ofcharismatic leadership in organizational settings. Academy of Man-agement Review, 12, 637–647.

ConniV, R. (2002). The natural history of the rich: A Weld guide. NewYork: W.W. Norton & Company.

Copeland, J. T. (1994). Prophecies of power: Motivational implicationsof social power for behavioral conWrmation. Journal of Personalityand Social Psychology, 67, 264–277.

DeDreu, C. K. W., & Van Kleef, G. A. (2004). The inXuence of poweron the information search, impression formation, and demands innegotiation. Journal of Experimental Social Psychology, 40, 303–319.

Deschamps, J. C. (1982). Social identity and relations of power betweengroups. In H. Tajfel (Ed.), Social identity and intergroup relations(pp. 85–98). Cambridge: Cambridge University Press.

Earle, W. B., Giuliano, T., & Archer, R. L. (1983). Lonely at the top:The eVect of power on information Xow in the dyad. Personalityand Social Psychology Bulletin, 9, 629–637.

J.R. Overbeck, B. Park / Organizational Behavior and Human Decision Processes xxx (2006) xxx–xxx 17

ARTICLE IN PRESS

Ebenbach, D. H., & Keltner, D. (1998). Power, emotion, and judgmen-tal accuracy in social conXict: Motivating the cognitive miser. Basicand Applied Social Psychology, 20, 7–21.

Fiske, S. T. (1993). Controlling other people: The impact of power onstereotyping. American Psychologist, 48, 621–628.

Fiske, S. T., & Dépret, E. (1996). Control, interdependence, and power:Understanding social cognition in its social context. EuropeanReview of Social Psychology, 7, 31–61.

Fleishman, E. A. (1953). The description of supervisory behavior. Per-sonnel Psychology, 37, 1–6.

French, J. R. P., & Raven, B. H. (1959). The bases of social power. In D.Cartwright (Ed.), Studies of social power (pp. 118–149). Ann ArborMI: Institute for Social Research.

Galinsky, A. D., Gruenfeld, D. H., & Magee, J. C. (2003). From powerto action. Journal of Personality and Social Psychology, 85, 453–466.

Goodwin, S. A., Gubin, A., Fiske, S. T., & Yzerbyt, V. Y. (2000). Powercan bias impression formation: Stereotyping subordinates bydefault and by design. Group Processes and Intergroup Relations, 3,227–256.

Green, S. G., & Mitchell, T. R. (1979). Attributional processes of lead-ers in leader-member exchanges. Organizational Behavior andHuman Performance, 23, 429–458.

Halpin, A. W., & Winer, B. J. (1957). A factorial study of the leaderbehavior descriptions. In R. M. Stogdill, & A. E. Coons (Eds.),Leader behavior: Its description and measurement (pp. 39–51).

Hamilton, G. G., & Biggart, N. W. (1985). Why people obey: Theoreti-cal observations on power and obedience in complex organizations.Sociological Perspectives, 28, 3–28.

Hill, W. (1969). The validation and extension of Fiedler’s theory ofleadership eVectiveness. Academy of Management Journal, 12, 33–47.

Hollander, E. P. (1978). Leadership dynamics: A practical guide to eVec-tive relationships. New York: Free Press.

Kahn, R. L., Wolfe, D. M., Quinn, R. P., & Snoelk, J. D. (1964). Organi-zational stress: Studies in role conXict and ambiguity. New York:Wiley.

Kanter, R. M. (1983). The change masters. New York: Simon &Schuster.

Kaplan, R. E. (1986). Trade routes: The manager’s network of relation-ships. Organizational Dynamics, 37–52.

Keltner, D., Gruenfeld, D. H., & Anderson, C. (2003). Power,approach, and inhibition. Psychological Review, 110, 265–284.

Keltner, D., & Robinson, R. J. (1997). Defending the status quo: powerand bias in social conXict. Personality and Social Psychology Bulle-tin, 23, 1066–1077.

Kim, P. H., & Fragale, A. R. (in press). Choosing the path to bargain-ing power: An empirical comparison of BATNAs and contribu-tions in negotiation. Journal of Applied Psychology.

Mansell, W., Clark, D. M., & Ehlers, A. (2003). Internal versus externalattention in social anxiety: an investigation using a novel paradigm.Behaviour Research and Therapy, 41, 555–572.

Neuberg, S. L. (1989). The goal of forming accurate impressions duringsocial interactions: Attenuating the impact of negative expectan-cies. Journal of Personality and Social Psychology, 56, 374–386.

Norman, W. T. (1963). Toward an adequate taxonomy of personalityattributes: Replicated factor structure in peer nomination personal-ity ratings. Journal of Abnormal and Social Psychology, 66, 574–583.

Overbeck, J. R., & Park, B. (2001). When power does not corrupt:Superior individuation processes among powerful perceivers. Jour-nal of Personality and Social Psychology, 81, 549–565.

Overbeck, J. R., Tiedens, L. Z., & Brion, S. (in press). The powerfulwant to, the powerless have to: Perceived constraint moderatescausal attributions. European Journal of Social Psychology.

PfeVer, J. (1992). Managing with power. Boston: Harvard BusinessSchool Press.

Raven, B. H. (1965). Social inXuence and power. In I. D. Steiner & M.Fishbein (Eds.), Current studies in social psychology (pp. 371–381).New York: Holt Rinehart and Winston.

Ric, F. (1997). EVects of control deprivation on subsequent use of ste-reotypes. Journal of Social Psychology, 137, 333–342.

Rodriguez-Bailon, R., Moya, M., & Yzerbyt, V. (2000). Why do superi-ors attend to negative stereotypic information about their subordi-nates? EVects of power legitimacy on social perception. EuropeanJournal of Social Psychology, 30, 651–671.

Sanitioso, R., Freud, K., & Lee, J. (1996). The inXuence of self-relatedgoals on the use of stereotypical and individuating information.European Journal of Social Psychology, 26, 751–761.

Seta, C. E., & Hayes, N. (1994). The inXuence of impression formationgoals on the accuracy of social memory. Personality and Social Psy-chology Bulletin, 20, 93–101.

Steele, C. M., & Aronson, J. (1995). Stereotype threat and the intellec-tual test performance of African Americans. Journal of Personalityand Social Psychology, 69, 797–811.

Taylor, S. E., Fiske, S. T., EtcoV, N. L., & Ruderman, A. J. (1978). Cate-gorical and contextual bases of person memory and stereotyping.Journal of Personality and Social Psychology, 36, 778–793.

Thibaut, J. W., & Kelley, H. H. (1959). The social psychology of groups.New York: Wiley.

Trzebinski, J., & Richards, K. (1986). The role of goal categories in personimpression. Journal of Experimental Social Psychology, 22, 216–227.

Vescio, T. K., Snyder, M., & Butz, D. (2003). Power in stereotypicallymasculine domains: A social inXuence strategy £ stereotype matchmodel. Journal of Personality and Social Psychology, 85, 1062–1078.

Wilemon, D. L., & Cicero, J. P. (1970). The project manager—Anoma-lies and ambiguities. Academy of Management Journal, 13, 269–287.

Yukl, G. (1989). Managerial leadership: a review of theory andresearch. Journal of Management, 15, 251–289.