SEEING THE FOREST FOR THE TREES: EXPLORATORY LEARNING ... · SEEING THE FOREST FOR THE TREES:...

24
r Academy of Management Journal 2015, Vol. 58, No. 3, 739762. http://dx.doi.org/10.5465/amj.2013.0991 SEEING THE FOREST FOR THE TREES: EXPLORATORY LEARNING, MOBILE TECHNOLOGY, AND KNOWLEDGE WORKERSROLE INTEGRATION BEHAVIORS JEAN-NICOLAS REYT McGill University BATIA M. WIESENFELD New York University Role integration is the new workplace reality for many employees. The prevalence of mobile technologies (e.g., laptops, smartphones, tablets) that are increasingly wearable and nearly always onmakes it difficult to keep role boundaries separate and distinct. We draw upon boundary theory and construal level theory to hypothesize that role integration behaviors shift people from thinking concretely to thinking more abstractly about their work. The results of an archival study of Enron executivesemails, two experiments, and a multi-wave field study of knowledge workers provide evidence of positive associations between role integration behaviors, higher construal level, and more exploratory learning activities. Social roles, and how they relate to one another, are shifting in the modern workplace. Ubiquitous and rapidly evolving mobile technologies such as smartphones and tablets (and increasingly wearable mobile devices such as wrist technologies) allow people to work any time, anywhere, and on any task. Knowledge work, which involves skilled mental labor that is generally information- intensive, has been most acutely transformed by these capabilities because they increase access to knowledge and information. Mobile work practices and tools have altered worker and organizational demands, norms, preferences and abilities, but generally in the direction of greater role integration. Increasingly, people work all the time, everywhere, and on everything (e.g., Mazmanian, Orlikowski, & Yates, 2013). Moreover, due to pervasive bring your own devicepolicies, knowledge work is done on personal devices that mix business and non- business use (Cisco, 2013). In short, these technol- ogies have the effect of diminishing time, space and social role boundaries, rendering roles increasingly integrated, decontextualized, overlapping and combined (e.g., Mazmanian, Orlikowski, & Yates, 2005; Middleton, 2008; Schlosser, 2002; Towers, Duxbury, Higgins, & Thomas, 2006). Knowledge workers, who are likely to experience these role integration pressures most acutely, are also the employees whose mental mindsets and in- tellective abilities are especially relevant to their work. They tend to be the workers primarily re- sponsible for steering organizational learning efforts. Organizational learning is critically important to enable adaptation, especially in highly dynamic contexts. How does role integration shape the way knowledge workers think and behave on the job? A key aspect of role integration is how it taxes psy- chological resources and demands adaptation. We draw upon boundary theory, construal level theory, and organizational learning research, which share a common conceptual foundation in bounded ratio- nality (Simon, 1955) and notions of limited psycho- logical resources, to suggest that role integration behaviors may influence knowledge workersmental mindsets. In particular, these behaviors may raise workersconstrual level, leading to more abstract thinking as people try to adapt to over- lapping roles. We suggest that higher construal, in turn, may promote exploratory learning behavior (i.e., the development of new skills and capabilities that enable adaptation to changing contexts). We empirically explore the relationships between role integration, construal level, and organizational We wish to thank Martine Haas and our three anony- mous reviewers, the members of the Organizational Be- havior work-in-progress seminar at Stern and the Trope lab at New York University, and Yaacov Trope, David Kalkstein, and Catherine Cramton for their valuable comments and suggestions on earlier versions of this manuscript. 739 Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holders express written permission. Users may print, download, or email articles for individual use only.

Transcript of SEEING THE FOREST FOR THE TREES: EXPLORATORY LEARNING ... · SEEING THE FOREST FOR THE TREES:...

r Academy of Management Journal2015, Vol. 58, No. 3, 739–762.http://dx.doi.org/10.5465/amj.2013.0991

SEEING THE FOREST FOR THE TREES: EXPLORATORYLEARNING, MOBILE TECHNOLOGY, AND KNOWLEDGE

WORKERS’ ROLE INTEGRATION BEHAVIORS

JEAN-NICOLAS REYTMcGill University

BATIA M. WIESENFELDNew York University

Role integration is the new workplace reality for many employees. The prevalence ofmobile technologies (e.g., laptops, smartphones, tablets) that are increasingly wearableand nearly always “on”makes it difficult to keep role boundaries separate and distinct.We draw upon boundary theory and construal level theory to hypothesize that roleintegration behaviors shift people from thinking concretely to thinking more abstractlyabout their work. The results of an archival study of Enron executives’ emails, twoexperiments, and a multi-wave field study of knowledge workers provide evidence ofpositive associations between role integration behaviors, higher construal level, andmore exploratory learning activities.

Social roles, and how they relate to one another,are shifting in the modern workplace. Ubiquitousand rapidly evolving mobile technologies such assmartphones and tablets (and increasingly wearablemobile devices such as wrist technologies) allowpeople to work any time, anywhere, and on anytask. Knowledge work, which involves skilledmental labor that is generally information-intensive, has been most acutely transformed bythese capabilities because they increase access toknowledge and information. Mobile work practicesand tools have altered worker and organizationaldemands, norms, preferences and abilities, butgenerally in the direction of greater role integration.Increasingly, people work all the time, everywhere,and on everything (e.g., Mazmanian, Orlikowski, &Yates, 2013). Moreover, due to pervasive “bringyour own device” policies, knowledge work is doneon personal devices that mix business and non-business use (Cisco, 2013). In short, these technol-ogies have the effect of diminishing time, space andsocial role boundaries, rendering roles increasinglyintegrated, decontextualized, overlapping and

combined (e.g., Mazmanian, Orlikowski, & Yates,2005; Middleton, 2008; Schlosser, 2002; Towers,Duxbury, Higgins, & Thomas, 2006).

Knowledge workers, who are likely to experiencethese role integration pressuresmost acutely, are alsothe employees whose mental mindsets and in-tellective abilities are especially relevant to theirwork. They tend to be the workers primarily re-sponsible for steering organizational learning efforts.Organizational learning is critically important toenable adaptation, especially in highly dynamiccontexts. How does role integration shape the wayknowledge workers think and behave on the job? Akey aspect of role integration is how it taxes psy-chological resources and demands adaptation. Wedraw upon boundary theory, construal level theory,and organizational learning research, which sharea common conceptual foundation in bounded ratio-nality (Simon, 1955) and notions of limited psycho-logical resources, to suggest that role integrationbehaviors may influence knowledge workers’mental mindsets. In particular, these behaviorsmay raise workers’ construal level, leading to moreabstract thinking as people try to adapt to over-lapping roles. We suggest that higher construal, inturn, may promote exploratory learning behavior(i.e., the development of new skills and capabilitiesthat enable adaptation to changing contexts).

We empirically explore the relationships betweenrole integration, construal level, and organizational

We wish to thank Martine Haas and our three anony-mous reviewers, the members of the Organizational Be-havior work-in-progress seminar at Stern and the Tropelab at New York University, and Yaacov Trope, DavidKalkstein, and Catherine Cramton for their valuablecomments and suggestions on earlier versions of thismanuscript.

739

Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s expresswritten permission. Users may print, download, or email articles for individual use only.

learning activities in four studies. First, we estab-lish the relationship between role integrationbehaviors and construal level in an archival studyutilizing a large email corpus. We then explore thecausal links between role integration and construallevel, and between construal level and exploratorylearning, in two experiments. Finally, we test therelationship between role integration, construallevel, and exploratory learning behaviors ina multi-wave field study of knowledge workersengaging in role integration using mobile technol-ogies that was conducted over four months (seeFigure 1 for an overview of our hypotheses andstudies).

This research contributes to the existing litera-ture in a number of ways. We extend boundarytheory research by exploring a key cognitivemechanism underlying the effects of role in-tegration activities (i.e., construal level), and byexpanding the scope of role integration outcomes.In particular, whereas prior work has focused onattitudinal outcomes of role integration, such as jobsatisfaction, commitment and work-family con-flict, we argue that role integration may haveheretofore unexplored cognitive and behavioralimplications. We also contribute to research onexploratory learning. Exploratory learning enablesadaptation through cognitive processes but exist-ing literature seldom examines the cognition the-orized in that work. We advance the literature byevaluating the content of knowledge workers’mental mindsets (with respect to level of abstrac-tion versus concreteness) and link this to their

learning activities. We also extend construal leveltheory beyond its predominant focus on mentalrepresentations of discrete objects and events toconsider mental representations of key social roles.In so doing, we provide a platform for theorizingabout how construal level shapes organizationalbehavior.

BOUNDARIES AND MENTAL MINDSETS

Individuals enact various role identities in theirdaily lives. Roles are contextualized with respectto time, space, and social interaction partners. Forexample, people may enact their organizationmember role at their office during work hours, butenact their family member role at home. Roleidentities are characterized by specific attitudesand behavior patterns and they are designed to fiteach domain’s rules and expectations. For exam-ple, at home family members may be expectedto be sensitive and affectionate with each other,but employees may have to be competitive andself-reliant at the office (Greenhaus & Beutell,1985). A manager may be submissive with de-manding clients or bosses, but dominant withsubordinates.

Boundary theory is a growing stream of researchin management describing the demarcations thatdelimit the role identities that people enact acrossvarious social domains, and how people transitionfrom one domain to another (Ashforth, Kreiner, &Fugate, 2000; Clark, 2002; Kossek, Noe, & DeMarr,1999; Michaelsen & Johnson, 1997; Nippert-Eng,

FIGURE 1Conceptual Framework

Studies 1 & 2 Study 3

Role integration/segmentation

Work-based construallevel

Emploratory learning

Study 4

740 JuneAcademy of Management Journal

1996; Rau & Hyland, 2006; Rothbard, Phillips, &Dumas, 2005). Boundary theory refers to how peo-ple make “micro” role transitions: those involved ineveryday movement between roles such as “work”and “home” roles, or at-work transitions such asfrom supervisor to subordinate or from Team A toTeam B (Ashforth et al., 2000). Traditionally, whenpeople transition from one social domain to an-other, contextual cues bring relevant role identitiesinto salience (Turner, 1982). Thus, an employeewho goes back home after a day at work, or entersa conference room tomake a presentation to a client,must switch “cognitive gears” (Louis & Sutton,1991) by adapting to expectations in the new do-main in order to avoid behavioral discrepancies.Out-of-place behaviors (e.g., being affectionate witha coworker, being bossy with a spouse or supervi-sor) occur when expectations associated witha specific role are misinterpreted or violated, whichis often costly with respect to reputation and socialrelationships.

According to boundary theory, roles mayvary along a continuum from highly segmented(i.e., distinct and separate) to highly integrated(i.e., overlapping and combined) both psycholog-ically and behaviorally (e.g., Ashforth et al., 2000;Edwards & Rothbard, 1999; Hartmann, 1997;Nippert-Eng, 1996). When roles are more seg-mented, the various responsibilities that occupya person’s life are more distinct and separate. Forexample, workers may dedicate specific time,space and equipment to use for each task. Con-versely, when roles are more integrated, the vari-ous responsibilities that are part of someone’slife are relatively more overlapping or combined.For example, the same smartphone is used forwork and personal activities (sometimes simulta-neously), and workers may compose a work emailwhile watching a child’s soccer game or text col-leagues working on one team project while at-tending a client meeting for a different teamproject. In sum, roles are highly segmented whenrole-associations (e.g., behaviors, attitudes) aremetaphorically “left at the door”when disengagingfrom one social domain and engaging in anotherone (Olson-Buchanan & Boswell, 2006), while rolesare highly integrated when observers and workersthemselves find it difficult to contextually or be-haviorally distinguish the social domains a workeroperates in.

Boundary theory recognizes that psychologicalresources are limited and therefore people mustenact different types of role boundaries to manage

the different kinds of relationships between theirroles (Ashforth et al., 2000). Role segmentation andintegration behaviors have benefits and costs asso-ciated with their effect on role boundaries (Ashforthet al., 2000). Greater role segmentation requireserecting relatively impermeable (“thick”) bound-aries with respect to time, space, and social in-teraction partners, and crossing such boundariesis quite effortful (Nippert-Eng, 1996). Therefore,interruptions and blurring of roles are less likely,and movement between relatively segmented rolesis more infrequent and salient (to role occupants butalso to observers) than movement between moreintegrated roles (Ashforth et al., 2000). On the otherhand, more integration makes role boundariesthinner and more permeable. This allows people tofluidly move from one role to another but theresulting overlaps and blurring of roles increases thelikelihood of confusion, interruptions, role “con-tamination”, and competing demands (like thoseassociated with taking a cell phone call from one’ssupervisor while enjoying dinner out with one’sspouse on a Saturday night; Ashforth et al., 2000;Rau & Hyland, 2006).

Mobile devices (e.g., smartphones, tablets, net-books) and mobile computing (e.g., “cloud” stor-age) are key enablers of role integration especiallyfor knowledge workers. They do this by consider-ably broadening the range of job-related tasks thatcan be performed anytime and anywhere. Quanti-tative (e.g., Leung, 2011; Richardson & Thompson,2012) and qualitative (e.g., Jarvenpaa, Lang, &Tuunainen, 2005; Mazmanian et al., 2005; Towerset al., 2006) studies have described the boundary-blurring behavior associated with mobile technol-ogy use. These include overlapping conversations(e.g., Cameron & Webster, 2005; Schlosser, 2002;Tarafdar, Tu, Ragu-Nathan, & Ragu-Nathan, 2007),interruptions (e.g., Fenner & Renn, 2010; Ren-neker & Godwin, 2005), and multi-tasking (e.g.,Benbunan-Fich & Truman, 2009). However, mo-bile work research has yet to explore the cognitiveand organizationally-relevant behavioral outcomesof such integration activities.

Most of the attention in the boundary theoryliterature is focused on describing boundarymanagement strategies (e.g., Ashforth et al., 2000;Nippert-Eng, 1996), their antecedents (e.g.,Edwards & Rothbard, 1999; Rau & Hyland, 2006),employee preferences for segmentation or in-tegration, and how such preferences interact withorganization-level policies (e.g., Kossek, Lautsch, &Eaton, 2006; Kossek et al., 1999; Kreiner, 2006;

2015 741Reyt and Wiesenfeld

Rothbard et al., 2005).1 Work on boundary manage-ment strategies and preferences depicts workers’efforts to control their role integration behaviors.However, recent evidence suggests that mobile tech-nologies produce an “autonomy paradox” in whicha worker’s role integration behaviors are constrainedby social norms and the preferences and behavior oftheir interaction partners, limiting workers’ ability toact on their strategies and preferences (Mazmanianet al., 2013). Most likely, a worker’s role integrationbehaviors are a function of a complex interactionbetween the person and the context, with strongsituations (such as social norms, the expectationsand behavior of social interaction partners, andincentives) attenuating the effect of individualdesires and boundary preferences while weak orambiguous situations allow individual preferencesto dominate (Weiss & Adler, 1984).

Prior research has shown that role boundaries areenacted—indeed “boundary work” refers to the ef-fortful initiatives taken by people to enact roleboundaries in more segmented or more integratedways (Kreiner, Hollensbe, & Sheep, 2009; Nippert-Eng,1996). This research assumes that role boundaries areimportant psychologically (Ashforth et al., 2000), yetwe know little about how enacting more integrated ormore segmented roles shapes the way people think—their mental representations. We suggest that in orderto understand whether and how role integrationbehaviors influence knowledge workers’ activities onthe job, we must explore how role enactment shapesthe way people conceptualize their roles.

Construal level theory (e.g., Trope & Liberman,2010) is a highly influential theory in social psy-chology that offers a vocabulary and powerfulconceptual foundation for understanding indi-viduals’ mental representations. The theory sug-gests that mental representations are structuredin a hierarchy, and that domain-specific mental

representations can be characterized along a con-tinuum from more abstract (higher construal) tomore concrete (lower construal). Low-level con-struals are specific and contextualized, capturingperipheral and therefore subordinate features oftargets. Conversely, high-level construals are ab-stract, schematic, and decontextualized, capturingsuperordinate characteristics of targets and thus ly-ing higher on a conceptual hierarchy defined withrespect to centrality, general meaning, and valence.The old adage that “it is hard to see the forest for thetrees” reflects this distinction; seeing the forestreflects higher construals while seeing the treesreflects lower construals. The saying also highlightsthe trade-offs in construals associated with boundedrationality: in order to see the forest we must sur-render our capacity to see each individual tree, andwhen examining one tree we lose the ability to payattention to the whole forest.

Activities are fundamental to roles and behavioralexpectations define them in large part (Biddle, 1979).Thus, applying construal level theory to roles requiresconsideration of how actions also form hierarchiesthat are defined by their level of abstraction (Vallacher& Wegner, 1989). Each action can be thought of withrespect to the superordinate and more abstract con-ceptualization of why that action is performed or thesubordinate details of how that action is performed,with more abstract representations of actions beingmore general, central and incorporating value whilemore concrete representations contain more detailed,specific and peripheral features (Liberman & Trope,1998; Vallacher & Wegner, 1989). For instance,when thinking about a common work role activitysuch as “attending a meeting,” a concrete and con-textualized construal might be “convening ina conference room with colleagues” while a moreabstract and decontextualized one might be “be-coming well-informed”. The latter results in a lossof information about the fact that a meeting isinvolved—people could also become well-informedby reading a book or attending a class—but it retainsmore central information about the general meaningand valence of the action. In sum, abstract con-struals reflect individuals’ implicit choices re-garding which features of an object or activity arecentral and which are peripheral, with the abstrac-tion process enabling goal-relevant features (why) totake center stage while more practical features (how)may fade from salience (Trope & Liberman, 2010).

Construal level theory has generally been appliedto people’s mental representation of targets that aretangible, such as specific objects, activities, or

1 For example, Kossek et al. (1999) explored the con-sequences of the fit (or misfit) between individualemployees’ boundary management preferences and theirorganizational context. Edwards and Rothbard (1999)empirically investigated the attitudinal outcomes of“person–environment fit” on well-being, while Kreiner(2006) investigated its effects on role conflict, stress, andjob satisfaction. In an effort to clarify the mechanismlinking individual-level integration preferences andorganizational-level policies, Rothbard and Dumas (2005)explored the moderating effect of employees’ desire forsegmentation on the relationship between organizationalwork–family policies (e.g., onsite childcare, flextime), jobsatisfaction, and organizational commitment.

742 JuneAcademy of Management Journal

people (Liberman & Trope, 1998, 2008; Trope &Liberman, 2003). While roles are less tangible, bound-ary theory suggests that roles also have boundarieswith regard to time, space, and social interactionpartners (Ashforth et al., 2000). We suggest that thesepsychological boundaries enable people to mentallyrepresent their roles more abstractly or more con-cretely just as they do events, objects, and people.Moreover, boundary theory and related notions ofboundary work and boundary preferences suggestthat role boundaries are relatively stable. Therefore,mental representations of roles can be thought of asdomain-specific—associated with the role and trig-gered by psychological engagement with the role,rather than reflecting either a dispositional orien-tation that people apply consistently to every situ-ation or a momentary orientation that changescontinually within role.

We suggest that greater role integration behaviorswill elicit more abstract mental representations; inother words, higher construal level. Two assump-tions underlie our assertion: bounded rationalityand functional adaptation. First, construal leveltheory is based on the notion that cognitive resourcesare bounded. When people have broader mentalhorizons, they do not have the resources to processas much detail and complexity (Trope & Liberman,2010). Indeed, this dynamic may originate in thebrain, where brain activity moves along an axis inthe medial pre-frontal cortex as psychological pro-cesses become more or less abstract (Amodio &Frith, 2006). Role integration behaviors enablebridging and overlapping of roles rather than sharpfocus on a single role domain, and thus necessarilybroaden the individuals’ scope to take a wider set ofcontextual factors into account. This taxes cognitiveresources, forcing people to relinquish attention todetail. Therefore, the broader scope of role in-tegration necessitates greater abstraction in the formof higher construal levels.

When roles are highly segmented they are enactedsequentially rather than simultaneously, so thecontext and one’s social interaction partners pro-vide clear cues to direct behavior. Thick roleboundaries associated with enacting segmentedroles limit conflicting cues, buffer people frominterruptions, and diminish the salience of com-peting demands, pressures and priorities when theyare within any one role (Ashforth et al., 2000). Thus,role segmentation narrows role occupants’ focuson their immediate, specific, and actionable roledemands. This prompts them to move down thehierarchy of action; by eliminating overlaps and

thus uncertainty regarding what should be done andwhy, role occupants can focus on how to accom-plish their goals within a single role domain. Thus,the thick boundaries associated with role segmen-tation behaviors enable more concrete, feasibility-oriented lower-level construal by reducing uncertaintyand confusion.

However, when roles are more integrated thenmultiple roles and all of their demands are si-multaneously salient, leaving the role occupant tomake the difficult choices about how to directtheir energy and attention. People who must copewith these competing pressures are likely tofunctionally adapt by relying upon mental strate-gies that enable them to prioritize, make appro-priate trade-offs, and identify opportunities forsynergy. Higher construals are essential to thesecapabilities (Trope & Liberman, 2010). For exam-ple, higher construals increase the likelihood thatpeople will focus on central, goal-related features(Trope & Liberman, 2010), distinguish signal fromnoise (Nussbaum, Liberman, & Trope, 2006), and becapable of making comparisons between unalign-able features (Wakslak & Trope, 2009). These arethe capabilities that enable people to set priorities,make trade-offs and identify synergies. Withoutthese capabilities, more integrated roles will be ex-perienced as more stressful and depleting. Thus,higher construals are more functionally adaptivefor people who engage in greater role integrationbehavior.

Hypothesis 1. Role integration behaviors will bepositively associated with construal level (i.e.,more abstract mental representations of work).

Theory and research on construal level effectssuggest that causal relationships are typicallyreciprocal—for example, distance raises level ofconstrual but higher construals shape psychologicaldistance (Trope & Liberman, 2010). The relation-ship between behavior and cognition is also oftenreciprocal; thus, just as workers under normativeand social pressure to integrate roles are likely tocope by mentally representing their roles more ab-stractly, it is also possible that workers who men-tally represent their roles abstractly may be morelikely to integrate their roles, such as by using mo-bile technologies in an integrative way. However,based on role theory we expect that one direction ofcausality, in which role behavior shapes construal,will be primary. In particular, roles are composed ofpatterned behaviors characteristically performed bypeople in a given context or situation (Biddle, 1979).

2015 743Reyt and Wiesenfeld

Role theory suggests that people must be socializedinto roles and their behavior is patterned by theexpectations and norms associated with those roles(Mead, 1934). Likewise, role integration behaviorsmust be socially negotiated with role alters to follownorms and conform to others’ expectations (Kreineret al., 2009). In sum, roles are socially and con-textually constrained and thus less individuallycontrollable than cognition. Therefore, we expectcontextually-defined role integration behaviors toinfluence construal level more powerfully than thereverse.

CONSTRUAL LEVEL AND EXPLORATORYLEARNING ACTIVITIES

Cognitive representations have proven to be acritical determinant of important organizationalbehavior phenomena (Fiol & Huff, 1992; Huff, 1990;Tversky & Kahneman, 1986; Walsh, 1995). Boundedrationality, in particular, has spawned an influentialstream of research on organizational learning (Levitt& March, 1988). Learning refers to the developmentof routines and processes that create, capture, trans-fer, and mobilize knowledge (Nonaka & Takeuchi,1995). These activities may be directed towardimproving existing products and processes or de-veloping new ones, both at the individual and thecollective level. Thus, learning is organizationallyimportant because it is essential to adaptation. Exist-ing research on organizational learning distinguishesbetween exploitation activities (i.e., learning fromexperience to improve and refine existing processes)and exploration activities (i.e., discovering, creatingand experimenting with new opportunities; March,1991). While exploitation is focused on improvingmean performance, exploration is associated withgenerating variance from which adaptive alter-natives can be selected (March, 1991). Prior theo-rizing (Gavetti & Levinthal, 2000) suggests thatexploratory learning is distinct with respect to pro-cess (i.e., utilizing a cognitive rather than experi-ential process to evaluate alternatives) and thealternatives considered (i.e., exploratory learningconsiders a broader set of alternatives, and ones thatdiverge more from prior experience). Exploratorylearning is more rare than exploitative learning be-cause it is more uncertain and the success rate islower (March, 1991).

There is debate about the relationship betweenexploratory and exploitative learning (Gupta, Smith, &Shalley, 2006), with some scholars conceptualizingthem as competing (e.g., March, 1991) and others

viewing them as orthogonal (e.g., Katila & Ahuja,2002). Exploratory learning is especially relevant toinnovation, which is a critical outcome of knowl-edge work, and is key to organizational survival(Levinthal & March, 1993). Mental representationsare thought to be particularly relevant for explor-atory learning, which is sometimes even termed“cognitive choice” (Gavetti & Levinthal, 2000). Inparticular, whereas exploitation relies upon rein-forcement learning processes and prior experience(i.e., actions are tried, their outcomes are experi-enced, and then actions are revised), explorationuses cognitive representations of the world to gener-ate a broader set of real and hypothetical alternatives(Gavetti & Levinthal, 2000). For these reasons, wefocus on how workers’ mental representation oftheir work roles relates to their exploratory learningactivities.

While the learning literature has considered theextent to which mental representations are involvedin learning (with mental mindsets being more in-volved in exploration than exploitation becausealternatives are cognitively generated in the former;Gavetti & Levinthal, 2000), it has generally failed toconsider or systematically assess what type ofmental representations are involved in learning be-havior. Research methods used to assess explor-atory learning include archival studies of theactions of individuals and collectives, simulationstudies, and a small number of qualitative casestudies and experiments (Gavetti, Greve, Levinthal,& Ocasio, 2012), but these studies generally offercoarse proxies for cognition, or simulate the effect ofdifferent mental representations rather than mea-suring them. Prior research fails to consider whetherthe form of individuals’ mental mindsets (frommore abstract to more concrete) may shape thelearning activities they engage in.

Construal level theory offers a conceptual modeldepicting how the content and consequences ofpeople’s mental mindsets may vary. We suggest thathigher construal level will be associated with higherlevels of exploratory learning activities becausecognitive abstraction is critical to the analogicalreasoning process that enables decision-makers togenerate hypothetical alternatives and to vicariouslylearn from the experiences of others—both keysources of exploratory learning (Gavetti, Levinthal,& Rivkin, 2005; Huber, 1991). Research on con-strual level theory has repeatedly demonstratedthat higher construals allow people to mentallytranscend the here and now, which enables them tothink hypothetically and vicariously (see Trope &

744 JuneAcademy of Management Journal

Liberman, 2010, for a review). That is, in order toimagine and cognitively assess a wide set of alter-natives that diverge from one’s own experience, it isessential to be able to create a simplified mentalmodel that abstracts away superficial features andfocuses on core aspects. Supporting this assertion isexperimental evidence from research on creativecognition (e.g., Ward, 1994). This work suggests thatproperties represented more abstractly are lessconstraining than those anchored in specific in-stances. Therefore, representing a problem more ab-stractly enables more creative problem solving andleads to greater innovation (Ward, 1994). As describedearlier, higher construal broadens individuals’mentalhorizons, focuses attention on central rather than pe-ripheral features of the situation, and draws attentionto strategic goals and away from feasibility concerns(Trope & Liberman, 2003). These capabilities enableexploration. Conversely, more concrete mental mod-els include the distracting details, emphasis on expe-rience, and focus on execution that constrainsexploration. Thus, we suggest:

Hypothesis 2. Construal level will be positivelyrelated to employees’ exploratory learningbehavior.

To this point, we have drawn upon boundedrationality to hypothesize that role integrationbehaviors will lead to higher construal level(i.e., more abstract mental representations) and thathigher construal level will lead to greater explor-atory learning behavior. Synthesizing these argu-ments, we further suggest that role integrationbehaviors will be positively associated with ex-ploratory learning through their effect on construallevel.

As mentioned earlier, exploratory learning isa variance-seeking activity. Specifically, requisitevariety must be generated, with some of the optionsgenerated offering more desirable outcomes, andactors adapt by selecting the options with the mostappealing outcomes (March, 1991). In the organi-zational learning literature, internal variety (definedas access to divergent information, relationships,and alternatives; Schilling, Vidal, Ployhart, &Marangoni, 2003) has been repeatedly demonstratedto be positively associated with exploration acrossstudies including archival research, experiments,and simulations (e.g., Gavetti, 2005; Schilling et al.,2003; Taylor & Greve, 2006). Frequently studied atthe team level but also shown at the individuallevel, internal variety is thought to provide access todifferent knowledge domains, sources of advice,

goals, and expectations, thereby expanding thedecision-maker’s scope and serving as key resourcesfor exploration (Alexiev, Jansen, van den Bosch, &Volberda, 2010; Beckman, 2006; Brass, 1995; Mors,2010; Schilling et al., 2003; Taylor & Greve, 2006).Integration is an essential complement to internalvariety that enables learning (e.g., Jansen, Tempelaar,van den Bosch, & Volberda, 2009; Reagans &Zuckerman, 2001; Taylor & Greve, 2006). Withoutintegration mechanisms, existing boundaries in-hibit transfer of knowledge and perspective. Indeed,research suggests that it is the combination of in-ternal variety and integration that is optimal (Lavie& Rosenkopf, 2006; Taylor & Greve, 2006).

At a collective level, integration may take theform of integration structures or interpersonal andnetwork cohesion (e.g., Jansen et al., 2009; Lavie &Rosenkopf, 2006; Taylor & Greve, 2006). We suggestthat within individuals, it is role integrationbehaviors that complement internal variety withintegration, enabling the transfer of knowledgeacross different domains and cognitive nodes, thatare an essential foundation for exploratory learning(Schilling et al., 2003). Role integration creates op-portunities for the diverse sets of experiences,expectations, mindsets, and knowledge that peopleuse in different contexts to be simultaneously sa-lient, and abstraction eases transfer of knowledgeinto novel domains. Therefore, role integration islikely to enhance exploratory learning activitiesby broadening workers’ focus, schemas andrepertoire.

In contrast, employees with more segmentedroles have more sequential rather than simultaneousawareness of different situations, bodies of knowl-edge, and influences. The spatial, temporal and socialboundaries that role segmentation behaviors erectand maintain increase focus and concrete thinking.These cognitive processes undermine cross-fertilizationand thus constrain exploratory learning.

Consider one’s social network: role integrationenables a broad and diverse network to be cogni-tively accessible simultaneously, while role seg-mentation makes portions of the network that aredistant from a role less accessible by inserting rela-tively impermeable boundaries and narrowingfocus. We suggest that mental abstraction is thecognitive mechanism or underlying psychologicalprocess enabling people to engage in the analogicalreasoning that diminishes attention to superficialfeatures in order to develop a mental representa-tion of the deeper features that generalize acrossdomains (Gavetti, 2005; Schilling et al., 2003). Thus:

2015 745Reyt and Wiesenfeld

Hypothesis 3. Construal level will mediate thepositive relationship between role integrationbehaviors and employees’ tendency to engage inexploratory learning behaviors.

STUDY 1

We explored the relationship between workers’role integration behaviors and their domain-specific construal level using a large corpus ofemails sent by Enron executives between 1997and 2002. In particular, we expected that roleintegration behaviors apparent in email sendingpatterns would be positively associated withemployees’ use of more abstract language in theirprofessional emails—an unobtrusive indicator oftheir construal level at work that also has organi-zational relevance because it reflects informationsharing behavior.

Over half a million emails sent by Enron execu-tives were made public during the investigation thatfollowed the company’s collapse. This dataset isparticularly interesting for organizational researchbecause it contains real-life professional emails ex-changed between employees over the course ofseveral years. For this study, we used a cleanedcorpus prepared by Shetty and Adibi (2005) con-taining 252,759 messages. We further removed46,176 emails that were not sent from the “enron.com” domain as well as 27,530 duplicate emailsthat had the same sender, date/hour and body oftext, arriving at a final set of emails that reflects in-formation exchange and knowledge sharing amongEnron employees over the time period.

The distribution of emails per sender in thedataset was exponential (as is typical in commu-nication data), with many senders sending fewmessages while a smaller number sent many. Wefocused on the users who sent 100 emails or moreduring the five-year period the dataset covers inorder to obtain a reliable assessment of emailsending patterns and level of abstraction for eachuser. We analyzed only sent emails because writingand sending a work email is evidence of work roleenactment. The data do not indicate when emailswere opened and read, making role integrationimpossible to assess from received emails. Weremoved 12 accounts from our dataset either: (1)because they could not be linked to any employee inparticular (e.g., automatic accounts sending calen-dar notices), or (2) because the accounts belongedto people who sent mostly automated messages.Enron employees often used several email addresses

with different structures (e.g., [email protected],[email protected], [email protected], etc.). A list from Shetty andAdibi (2005) enabled us to identify 39 Enronemployees who sent emails from at least two dif-ferent addresses. We manually identified 16 otherEnron employees who used at least two emailaddresses. We therefore merged the email accountsfor 55 users who sent emails from more than oneaddress. Overall, our analyses are based on theemail messages from 236 Enron employees, whosent a total of 119,751 professional emails in ourdataset.

Measures

Role integration behaviors. We measured work/non-work role integration with respect to inte-gration across temporal boundaries between workand life roles. We do not have sufficient informationabout each Enron employee’s different work roles toidentify at-work role transitions in this dataset. Tocapture employees’ role integration behaviors weassessed the proportion of total emails sent by eachemployee outside of traditional working hours(i.e., 9am to 5pm, Monday through Friday). Higherproportions reflect greater integration across tem-poral role boundaries and thus greater role inte-gration for that worker.2

To validate our measure of role integrationbehaviors, we explored how it was associated withBlackberry smartphone use. In particular, in the lastyear of the dataset, a subset of email messages beganto appear that indicated that they were sent viaBlackberry. Messages sent via Blackberry do notalways contain such signifiers and the set ofBlackberry-indicated messages is small (and thusdoes not provide a reliable measure of role in-tegration or construal level). However, becauseBlackberry keyboards are much less convenient touse than computer keyboards, it is reasonable toassume that employees using a Blackberry to sendmessages are likely to be integrating across spatialrole boundaries because they are unlikely to sendBlackberry messages from their offices or tradi-tional work spaces (where more convenient key-boards are available). We compared Blackberry to

2 The Enron dataset does not contain accurate timezones. Therefore, we individually coded 24,043 emailsthat were forwarded, and thus contain accurate emailsending times hard-coded into the body of the message, toidentify the accurate time zones of the emails.

746 JuneAcademy of Management Journal

non-Blackberry users and found that Blackberry usewas significantly correlated with our role integrationmeasure (r 5 .203; p ,. 01).

Construal level. To assess construal level, wemeasured the level of abstraction of each employ-ee’s sent email messages based on the LinguisticCategory Model (LCM; Semin & Fiedler, 1991). TheLCM suggests that words vary in level of abstrac-tion, and thus texts can be assigned an LCMweighted score as a function of the abstractness ofthe type of words they contain. According to theLCM, the most concrete terms are descriptive actionverbs (DAV) that provide an objective description ofan observable action (e.g., to hit, to walk). Slightlymore abstract are interpretative action verbs (IAV),which refer to a multitude of actions rather than toa specific observable behavior (e.g., to hurt, to help).Still more abstract are state action verbs (SAV),which refer to the emotional consequences ofactions rather than to an action (e.g., to anger, toamaze). Even more abstract are state verbs (SV),which refer to enduring cognitive or emotionalstates (e.g., to hate, to love). Most abstract areadjectives (ADJ), which provide a general de-scription that is stable and valid across situation andcontexts (e.g., aggressive, helpful). Prior researchhas validated the use of the LCM as a measure ofconstrual level (e.g., Freitas, Gollwitzer, & Trope,2004; Fujita, Trope, Liberman, & Levin-Sagi, 2006).

In order to obtain an abstraction score for eachuser, we first removed all signatures, replies, andforwarded emails from the email texts. Oncecleaned, some emails were too short to be includedin our analyses because they did not containa complete sentence. Thus, we conducted our lin-guistic analyses on all emails where the number ofcharacters ranged from 50 to 1,500—a total of78,231 emails.3 Using a bag-of-words approach, weprocessed the whole dataset with a part-of-speechtagging algorithm coupled with an 80,000-wordlexicon that enabled us to identify adjectives, verbs,and nouns, and we removed other types of words(e.g., prepositions, pronouns, etc.) from the dataset.We extracted a list of all the verbs in our dataset inorder to manually identify the different types of

verbs. We treated IAV and DAV as one categorybecause context is necessary to reliably differenti-ate them (Coenen, Hedebouw & Semin, 2006). Weweighted the words by their level of abstraction basedon the LCM model (ADJ * 4 1 SV * 3 1 SAV * 2 1IAV1DAV) and divided by the total number ofwords(i.e., adjectives, verbs, nouns), resulting in a ratio inwhich higher numbers represent more abstract com-munication, indicating higher construal level.

Controls. Prior research has found that peoplewho recalled and wrote about a time in which theywere powerful were more likely to report higherconstrual level than those who recalled an instancein which they were powerless (Smith & Trope, 2006;Smith, Wigboldus, & Dijksterhuis, 2008). Therefore,we controlled for employees’ hierarchical level inour analyses. Data from Shetty and Adibi (2005)enabled us to match 95 employees of our datasetwith their job titles within Enron. In order to findjob titles for the remaining employees, we useda spring/summer 1999 Enron Internal Phone Di-rectory (purchased from eBay) containing hierar-chical rank information, as well as other sourcessuch as email signatures and LinkedIn profiles.Overall, we matched 226 employees (96% success-ful matches) with their job titles at the time theemails were sent. We used Diesner, Frantz, andCarley’s (2005) hierarchy of Enron’s job titles to as-sign a score reflecting employees’ hierarchical level,following previous work on the Enron dataset (e.g.,Hossain, 2009). Higher scores indicate higher levelsin the hierarchy; specifically: (1) Associates, (2)Specialists, (3) Traders, (4) Managers, (5) Lawyers,(6) Senior Management, (7) Executive Management.

We also controlled for the number of emails sentby each user to ensure that the distribution of ourdataset did not influence our dispersion indices andabstraction scores.

Results and Discussion

Table 1 reports the means, correlations, andranges for all of the measures in our study. Withrespect to hierarchical level, the 226 Enronemployees were distributed as follows: Associates(73), Specialists (26), Traders (13), Managers (24),Lawyers (22), Senior Management (58), ExecutiveManagement (10).

We tested whether there was a positive relation-ship between employees’ role integration behav-iors and their construal level (Hypothesis 1) usinghierarchical regression. In the first step, we enteredour two control variables: number of emails sent

3 As a robustness check, we also coded the full set ofemails that contained text written by the employee(i.e., deleting forwarded emails) and used those scores inour analyses. Results using this larger set of messages(N5105,041) were nearly identical in pattern and level ofsignificance to those found using the constrained set ofemails.

2015 747Reyt and Wiesenfeld

and hierarchical rank. In the second step, we addedour measure of role integration as a predictor of theLCM abstraction measure.

Table 2 reveals that hierarchical rank is positivelyand significantly associated with construal level,consistent with prior findings linking power toconstrual level (Smith & Trope, 2006; Smith et al.,2008). These findings provide reassuring evidencethat our linguistic analysis of the abstraction ofprofessional emails is associated with a variable(power) that has been previously shown to shapeconstrual level. Of far greater importance, role in-tegration was significantly and positively associatedwith employees’ construal level as measured by theabstractness of professional email communication,supporting Hypothesis 1.

The fact that Study 1 is an archival study of nat-ural behaviors of executives in an organizationalsetting speaks to the external validity of the re-lationship between role integration behaviors andconstrual level. However, a notable limitation of theresearch design is our inability to verify the di-rection of causality (from greater role integration

to higher construal level). Our archival measureof construal level—the level of abstractness ofinformation exchanged via email by companyemployees—is inherently interesting as an indicatorof organizational communication. Also, writtencommunication has been validated as a measure ofconstrual level in prior research (e.g., Freitas et al.,2004). However, the archival measures of construallevel and role integration in this study are in-herently limited and noisy, and they are onlyproxies for the underlying constructs we sought toassess. Furthermore, the role integration measuretaps only one type of micro-role transition (work/non-work) and thus it is not clear whether ourfindings generalize to at-work micro-role tran-sitions. Study 2 was designed to begin to addressthese limitations.

STUDY 2

Study 2 explores whether role integration is as-sociated with construal level by priming partic-ipants’ role integration or segmentation through awriting task and then measuring their construallevel. The purpose of this experiment was to repli-cate our Study 1 findings in a different setting andwith different measures, and to test the causal as-sociation between role integration and how ab-stractly people mentally represent their work.

Participants

One hundred working adults were recruited onan online platform where people participate instudies for payment. Three participants weredropped from the analyses for failing our attentioncheck. Thus, 97 participants (40 women, age: M 533.71, SD 5 10.13) were included in the finalanalyses.

TABLE 1Descriptive Statistics for all Variables (Study 1)

Variables Mean SD Min Max 1 2 3 4

1. Number of emails 507.4 821 100 5,5082. Hierarchical rank 3.49 2.169 1 7 –.0183. Role integration .472 .17 .05 .982 –.135* .1014. Construal level .754 .074 .54 .958 .012 .232** .259**

Note: N 5 226 for hierarchical rank and N 5 236 for all of the other variables.* p , .05** p , .01

TABLE 2Regressions of Construal Level and Role Integration

(Study 1)

Step 1 Step 2

ControlsNumber of emails .007 .039Hierarchical rank .232* .209*

Main effectRole integration — .229*

Adjusted R2 .045 .092Change in R2

— .047

Note: N 5 226 for hierarchical rank and N 5 236 for all of theother variables. The coefficients reported in each column arestandardized b coefficients.

* p , .01

748 JuneAcademy of Management Journal

Procedure

Role integration/segmentation priming task. Atthe beginning of the study, we asked participants tomake a list of at least five activities they typicallyperform in their work. Mobile technology and itseffect on work roles were made salient for allparticipants at the start of the prime. Participantsread, “In the modern workplace, mobile technol-ogies (e.g., smartphones, tablets, laptops) areubiquitous and job roles are complex.” Partic-ipants in the integration condition (n 5 49) werethen told: “In particular, your work activities maybe quite integrated (that is, overlapping andcombined). Please describe the aspects of yourwork activities that are overlapping and com-bined (i.e. integrated).” Participants all wrotea paragraph in response, and they typically de-scribed the technologies, people and skill-setsthat were common across their work activi-ties. One of the examples provided by a tele-communications company employee was, “Forsome issues I might have to call a customer andwhile they are on the phone, I send Instant Mes-sages to other colleagues.” In the segmentationcondition (n 5 48) after making mobile technologysalient participants were told, “In particular, yourwork activities may be quite segmented (thatis, distinct and separate). Please describe theaspects of your work activities that are distinct andseparate.” Participants typically described thedifferences in technologies, places, and peoplebetween their various work activities. As an illus-tration, a research chemist explained, “The workthat I do in the lab is hands on whereas the work Ido with customers is always through email orphone.”

Work-based construal level. As indicated above,behaviors are a central and defining feature of roles(Biddle, 1979) and people represent their actionshierarchically, ranging from more abstract and su-perordinate representations that are more general,central, and incorporate value to more concreteand subordinate representations that are relativelymore detailed, specific, and practical (Vallacher &Wegner, 1987). To assess the level at which peoplementally represent their work role, we drew uponaction identification theory (Vallacher & Wegner,1987), which serves as an important foundation forconstrual level theory (Trope & Liberman, 2010)and addresses how actions are identified, or men-tally represented hierarchically. Construal levelresearch has used Vallacher and Wegner’s (1989)

behavior identification form (BIF) to assess howabstractly actions are represented. The original BIFcharacterizes a set of 25 everyday non-work activ-ities (e.g., brushing teeth) using a low-level de-scription focused on how the action is performedand a high-level description focused on why theaction is performed. For each activity, participantsindicate which description best represents the be-havior for them. In the BIF, an abstraction score iscalculated by counting the number of high-leveldescriptions that are chosen. We created a work-domain specific adaptation of the BIF because theBIF assesses how abstractly people represent theirbehaviors and activities, which are the definingaspects of roles.

To create a measure of how abstractly peoplerepresent their work role, we selected 30 com-mon knowledge work activities (e.g., preparinga report, attending a meeting, proofreading a doc-ument, etc.) from a database of job tasks devel-oped by the US Department of Labor. FollowingVallacher and Wegner’s (1989) methodology, 40pilot subjects were asked to provide as many de-scriptions as possible for each activity. The mostcommonly cited high- and low-level descriptionswere used to build our work-based construal levelmeasure.

We then administered our resulting measure toa separate test sample of 100 knowledge workersand retained the 18 items that had the highest item-total correlation (Cronbach’s a 5 .92 for the testsample; see Appendix 1 for the entire scale). Weused high and low-level activity descriptions asopposite anchors of six-point scales and asked par-ticipants to indicate the point along the scale thatbest described how they see that work activity. Halfwere reverse scored and all of the activities weredisplayed in random order, with responses aver-aged to construct our measure (Cronbach’s a 5 .88in the study sample).

Manipulation check. To ensure that our role in-tegration prime successfully manipulated the in-tegration of various work roles, we drew upon howrole integration and segmentation are defined inprior research (e.g., Ashforth et al., 2000) and askedparticipants to rate the extent to which their workactivities were: (1) “overlapping,” (2) “combined,”(3) “distinct,” and (4) “separate” on Likert scaleswith responses ranging from 1 (“Not at all”) to 7(“To a great extent”). We reverse-scored the lasttwo and averaged the four ratings (Cronbach’s a 5.79) so that higher scores indicate greater roleintegration.

2015 749Reyt and Wiesenfeld

Results and Discussion

Manipulation check analyses suggest that wesuccessfully manipulated role integration. Respon-dents in the role integration condition reported sig-nificantly higher role integration (M 5 4.78) thanrespondents in the segmentation condition (M5 4.00;t(95) 5 5.4; p , .001).

Of far greater importance, respondents in the roleintegration condition reported significantly higherscores on our work-based construal level scale (M54.04) than did those in the segmentation condition(M 5 3.59; t(95) 5 2.83; p , .01), supporting Hy-pothesis 1.

The findings in this study provide evidence sup-porting our prediction that role integration shapesconstrual level in a setting that allows us to drawinferences about causal relationships. While wecannot discount the possibility that construal levelmay also shape role integration, these findingsprovide evidence supporting the causal directionthat we hypothesized. Moreover, the fact that wewere able to replicate the findings of Study 1 ina different setting and with different measures ofrole integration and construal level lends confi-dence to our interpretation of the findings in Study1. However, while construal level provides in-formation about workers’ mental mindsets it is alsoimportant to evaluate whether these effects on cog-nition have organizationally-relevant downstreamconsequences, consistent with our overarchingconceptual framework of bounded rationality. Inparticular, we expect that the same trade-offs andconstraints that drive boundary theory and con-strual level theory should emerge with respect toorganizationally-important choices and informationsearch behavior. Thus, the goal of the study thatfollows is to evaluate the effect of construal level onexploratory learning.

STUDY 3

Study 3 manipulated construal level using acategorization task and measured distant searchbehavior—one of the most commonly-studied formsof exploratory learning (Gavetti et al., 2012) to testHypothesis 2 linking construal level to exploration.

Participants

One hundred and five working adults wererecruited on an online platform where people par-ticipate in studies for payment. Six participants were

dropped from the analyses for failing our attentioncheck. Thus, 99 participants (69 men, age:M 5 32.3,SD 5 10.92) were included in the final analyses.

Procedure

Construal level priming task. We used a primedeveloped by Fujita et al. (2006) to manipulateconstrual level by inducing more abstract or con-crete mindsets. Participants were given a list of 30items (e.g., college, actor, coin). Those in the highconstrual level condition were asked to generate themore abstract overarching category that each itembelongs to, while those in the low construal levelcondition were asked to generate more concretespecific examples of each item. This prime has beenused in the construal level theory literature to ex-plore the effects of abstraction on a wide variety ofattitudinal and behavioral outcomes (for a review,see Burgoon, Henderson, & Markman, 2013).

Exploratory learning behaviors. After primingparticipants for construal level, we assigned theman information search task. Prior organizationallearning research distinguishes less exploratory“local” or “neighborhood” searches from more ex-ploratory “global” or “distant” searches (Levinthal& March, 1993). These forms of search have beenused to measure exploration (e.g., Beckman,Haunschild, & Phillips, 2004; Rosenkopf & Nerkar,2001; Stuart & Podolny, 1996). There are few ex-perimental studies of exploratory learning behavior(Gavetti et al., 2012; but see the “alien game” inBillinger, Stieglitz, & Schumacher, 2014 for a rareexception) and we are not aware of any experi-mental paradigm that simulates traditional mana-gerial tasks and is thus appropriate for a sample ofworking adults. We therefore developed an in-formation search task modeled on a vendor selec-tion process that was actually used in a largefinancial services corporation. Specifically, partic-ipants were told to imagine their employer soughtto outsource its telephone customer service sup-port, and their task was to search for informationabout six potential vendors in order to make arecommendation.

We presented participants with an informationboard (Payne, 1976)—a 6 3 6 matrix with rowsrepresenting alternative vendors and columns in-dicating vendor attributes (call handling time,phone representative experience, phone represen-tative yearly turnover rate, vendor years in business,cost per minute, and customer data security). Sim-ilar information boards have been used in prior

750 JuneAcademy of Management Journal

research (e.g., Biggs, Bedard, Gaber, & Linsmeier,1985; Payne, 1976).

To differentiate local search (based on existingrelationships) from distant search (developing newrelationships; e.g., Beckman et al., 2004), respon-dents were told that two of the vendors providedservice to other divisions of the company while theother four offered opportunities to create new rela-tionships. To simulate the practical reality thatsearch is costly, respondents were able to click ona maximum of 12 cells in the matrix. Exploratorylearning behavior was measured with respect to theproportion of clicks that participants allocated tounfamiliar options, with higher scores reflectingmore distant search.

Exploratory learning intentions. To ensure thatour experimental paradigm captured exploratorylearning behavior, we asked participants to indicatethe extent to which the search decisions were mo-tivated by exploratory learning intentions. Specifi-cally, participants indicated: (1) “During yourdecision-making process, to what extent did youseek to innovate?” (2) “Howwilling were you to takea risk?” and (3) “To what extent did you search fornew vendors?” (Responses ranged from “not at all”(1) to “to a great extent” (7).) Responses to thesethree items were averaged to create a scale mea-suring exploratory learning intentions (Cronbach’sa 5 .81).

Manipulation check. To ensure that the construallevel prime successfully manipulated participants’mental mindsets, we asked participants to describethe search decisions they made using three itemsfrom Burrus and Roese’s (2006) Rating of a LifeEvent measure. This measure has been validated asa measure of construal level and used to assess ex-perimental manipulations of construal level in priorresearch (e.g., Burgoon et al., 2013; Burrus & Roese,2006; Giacomantonio, De Dreu, & Mannetti, 2010;Giacomantonio, DeDreu, Shalvi, Sligte, & Leder,2010). The items we used were: (1) Meaningless–Meaningful, (2) Not important–Important, (3) Influ-ences minor detours in my life–Influences overallpath of my life. We averaged the three ratings to ob-tain a construal level score (Cronbach’s a5 .79) withhigher scores reflecting more abstract mindsets.

Results and Discussion

Our construal level manipulation was successful;respondents in the high construal level conditionreported construing their decisions at a higher levelon our manipulation check scale (M 5 4.94) than

did respondents in the low construal level condition(M 5 4.41; t(97) 5 2.9; p , .01).

We also found evidence in support of our hy-pothesis that higher construal would lead to moreexploratory learning in the form of more distantsearch. Respondents in the high construal levelcondition reported stronger exploratory learningintentions (M 5 4.64) than respondents in the lowconstrual level condition (M5 3.97; t(97)5 2.3; p,.05). Consistent with Hypothesis 2, respondents inthe high construal level condition allocated a higherproportion of their clicks to unfamiliar cells (sug-gesting more distant search; M 5 71%) thanrespondents in the low construal level condition(M 5 58%; t(97) 5 2.73; p , .01). To assess whetherrespondents’ exploratory learning intentions ex-plain the positive relationship between higherconstrual level and distant search, we tested formediation using the bootstrapping method de-veloped by Preacher and Hayes (2008) using 5,000bootstrap resamples. Mediation is said to occurwhen the derived confidence interval does notcontain zero. The indirect effect of construal levelcondition through exploratory learning intentionson search behavior was statistically different fromzero (95% CI 5 .0088 to .1286), suggesting that ex-ploratory learning intentions underlie the positiverelationship between construal level and distantsearch in our study.

Study 3 provides evidence supporting the positiverelationship between construal level and an impor-tant form of exploratory learning (i.e., search behav-ior) in an experimental setting, enabling causalinferences. Our experimental paradigm, while over-simplified, is rare in the existing literature (Billingeret al., 2014), and offers a behavioral measure of in-formation search relevant to real-world managerialdecisions. Combined with the results of Studies 1and 2, the patterns we find suggest that role in-tegration shapes construal level, which in turnshapes exploratory learning. However, a limitationof manipulating the hypothesized mediator is thatit is not possible to explore the entire process fromrole integration to exploration. Also, experimentalmanipulations maximize internal validity but at theexpense of external validity. Study 4 was designed toaddress these concerns.

STUDY 4

Our final study extends the prior findings by ex-ploring how knowledge workers’ role integrationbehaviors and their construal level relate to their

2015 751Reyt and Wiesenfeld

exploratory learning behavior. We examine theserelationships in a field study surveying knowledgeworkers in three waves over a four-month period.

Respondents

We recruited a panel of knowledge workers froman online platform where people can choose toparticipate in surveys for payment. Five hundredadult, US-based full-time workers took a screeningsurvey containing various questions about theirwork activities. In the first part of the survey, wedifferentiated knowledge workers from other kindsof workers by asking them: (1) to describe their workin a few sentences, and (2) to indicate the extent towhich their work corresponded to a common defi-nition of knowledge work (i.e., “knowledge worktasks include planning, analyzing, interpreting, de-veloping, and creating products and services usinginformation, data, or ideas as the raw materials”;Heerwagen, Kampschroer, Powell, & Loftness,2004). In the second part of the screening survey,respondents were asked to indicate how they usedmobile technology to integrate their social roles andalso reported demographic information. Overall,169 of the 500 respondents fit our knowledge workcriteria and were invited to participate in the sur-veys at Time 2 and Time 3. Of those, 138 respondedto the Time 2 survey, which took place four weekslater and assessed work-based construal level (82%response rate) and 80 of those respondents partici-pated in the Time 3 survey, which took place 12weeks after the second survey and included ourmeasure of exploratory learning activities.4 Thus,the response rate for those eligible participating inall three surveys was 47%. To assess how repre-sentative our final sample was, we compared thefinal sample of respondents who completed allthree surveys over the 16-week period to those whoresponded only at Time 1, using the three de-mographic measures included in the screeningsurvey: gender, age, and number of children. The

final sample was not significantly different from theTime 1 sample with respect to gender or number ofchildren. There was a small but significant differ-ence in age, with our final sample (M(age) 5 36.1years) somewhat older than the initial sample (M 531.2 years; t(167) 5 3.051, p 5 .003).

Measures

Role integration behaviors. We measured in-tegration across spatial role boundaries in Time 1with four items. Specifically, we asked respon-dents: “Think about your work activities in a typicalmonth. Please report how often, on average, youperform job-related tasks using mobile devices(smartphone, tablet, netbook, etc.) in the followingsettings”: Respondents indicated the extent towhich they used mobile devices to work away fromtraditional work contexts: (1) “At home, at yourdesk,” (2) “At home, away from your desk (e.g.,living room, bedroom, etc.),” (3) “In your com-pany’s premises, away from your work station (e.g.,meeting room, other office, etc.),” and (4) “Out ofyour company’s premises, away from your home(e.g., public transportation, client’s premises,etc.).” Responses ranged from “Never” (1) to “Allthe time” (7), and scores on the items were aver-aged to form the role integration index, with higherscores indicating greater spatial role integrationbehavior.

Work-based construal level. Work-based con-strual level was assessed at Time 2 (four weeks afterTime 1). To assess work-based construal level, weused the work-based construal level scale describedin Study 2 (Cronbach’s a 5 .91 in this sample).Participants were given the 18 common workactivities and asked to indicate which of twodescriptions—a high-level and a low-level—bestdescribed the behavior for them. Following VallacherandWegner’s (1989) recommendation, participants’abstraction score was the sum of high-level alter-natives chosen.

Exploratory learning activities. We assessedexploratory learning activities in Time 3 (16 weeksafter Time 1) with a five-item scale developedby Mom, van den Bosch, and Volberda (2007;Cronbach’s a 5 .85). Respondents were asked to in-dicate the extent to which they engaged in, for exam-ple, “searching for new possibilities with respect toproducts/services, processes or markets” or “activitiesrequiring you to learn new skills or knowledge” (theresponse scale ranged from “never” (1) to “all thetime” (7)).

4 Only four weeks elapsed between the first two surveyadministrations because we expect that role integrationand construal level (assessed in Surveys 1 and 2 re-spectively) are both domain-specific and somewhat rein-forcing. However, we allowed 12 weeks to elapse betweenSurveys 2 and 3 to increase the likelihood that the ex-ploratory learning behaviors that were most recent andthus salient to respondents in Survey 3 were most likely tobe those that temporally followed the formation of themental mindsets reported in Survey 2.

752 JuneAcademy of Management Journal

Controls. Given that prior research and our find-ings in Study 1 have linked power to levels of mentalrepresentation (Smith & Trope, 2006; Smith et al.,2008), we controlled for supervisory responsibilityby asking respondents to indicate if they superviseddirect reports at work (Non Supervisor 5 0, Super-visor 5 1). We also controlled for gender (Male 5 0,Female 5 1), age and number of children becauserespondent’s personal responsibilities and lifestage may influence their use of technology outsideof work environments. We controlled for tenure(in years) and level of education (high school orbelow 5 0, college 5 1, graduate degree 5 2) whichmay influence exploratory learning and construal level.

Results

Table 3 displaysmeans, correlations, and ranges forall of the measures in our study. The 80 respondentswho took all three waves (56% female, 44% male)were generally college graduates (68% had at least anundergraduate degree). At the beginning of the studytheir average job tenure was 6.12 years and 53%werein supervisory roles. Our respondents’ job functionswere varied, including information technology, hu-man resources, and administrative functions.

We conducted two hierarchicalmultiple regressionanalyses to assess the relationship between partic-ipants’ role integration and each of construal leveland learning activities. In the first step, we controlledfor age, gender, number of children, hierarchical rank,education and tenure (see Table 4, columns 1 and 5).

Hypothesis 1 predicted that the extent to whichemployees engaged in role integration behaviorsin Time 1 would be positively associated withconstrual level in Time 2. Consistent with ourfindings in Studies 1 and 2, the results reveal thatrole integration in Time 1 significantly predictswork-based construal level at Time 2.5 Hypothesis 2predicted that construal level (Time 2) would bepositively associated with exploratory learningactivities (Time 3). Regression results again sup-port our predictions; Table 4 (column 4) reveals

a significant positive relationship between con-strual level and exploratory learning.6

Hypothesis 3 predicted that employees’ work-based construal level would mediate the relationshipbetween role integration and exploratory learning.We tested for mediation using the bootstrappingmethod developed by Preacher and Hayes (2008)using 5,000 bootstrap resamples. The indirect effectof role integration (Time 1) through work-based con-strual level (Time 2) on exploratory learning (Time 3)was statistically different from zero (95% CI 5 .0036to .0523), which supports Hypothesis 3. The findingsreported in Table 4 (column 7) suggest that role in-tegration remained a significant predictor of learningactivity when work-based construal was included inthe model, suggesting significant partial mediation.

Overall, the results in Study 4 replicate and ex-tend the findings of Studies 1–3. Our findings sug-gest that role integration is associated with higherconstrual levels, replicating the results of Studies 1and 2 and lending support to Hypothesis 1. More-over, we demonstrate that work-based construallevel (Time 2) is associated with greater engagementin exploratory learning activities (Time 3), repli-cating the results of Study 3 in a field study and thuslending additional support to Hypothesis 2. More-over, role integration (Time 1) is positively relatedto exploratory learning (Time 3), and work-basedconstrual level (Time 2) partially mediates this re-lationship, supporting Hypothesis 3. Study partic-ipants were a sample of knowledge workers ina variety of job functions, providing support for thegeneralizability of our findings to workers for whomexploratory learning is likely to be important for jobeffectiveness. The multi-wave study design has thebenefit of reducing the likelihood of common meth-ods bias. However, unlike the prior studies whichcontain manipulations or behavioral measures, alimitation of the design of Study 4 is the fact that thedependent variable in this study was self-reported.

GENERAL DISCUSSION

Using diverse methodologies and measures,we obtained convergent findings regarding the

5 We also used the LCMmethod (utilized to analyze theemails in Study 1) to code the construal level of the jobdescriptions respondents provided in the first surveywave. We included this unobtrusive measure of construallevel (Time 1) in our analyses predicting work-basedconstrual level at Time 2. Role integration remaineda significant predictor of subsequent construal level evencontrolling for prior construal level, supporting the causaldirection we hypothesized.

6 We tested for the relationship between construal leveland exploitative learning and found no significant re-lationship. Also, as a robustness check we ran all analyseswithout the controls and our results verified that all of thekey findings regarding role integration behaviors, con-strual level and exploratory learning were robust to thealternative model without controls both with respect todirection and level of significance.

2015 753Reyt and Wiesenfeld

relationships between role integration, construallevel, and exploratory learning. Our analysis sug-gests insights that enrich existing theory and re-search, open up new avenues of inquiry, andintegrate disparate research streams.

Implications for Boundary Theory

Research on boundary theory has explored thechallenges of managing role boundaries and role

transitions and described the form and function ofboundary work (e.g., Kossek et al., 1999; Kreiner,2006; Nippert-Eng, 1996; Rothbard et al., 2005).This literature assumes that role boundaries arepsychologically important but has neglected toevaluate the relationship between boundary man-agement activities (e.g., segmentation and in-tegration) and characteristics of workers’ cognition(i.e., abstractness). We draw upon construal leveltheory to generate novel hypotheses regarding the

TABLE 4Regressions of Construal Level and Learning Behaviors (Study 4)

Column/Hypothesis 1 H1 2 H1 3 H2 4 H2 5 H3 6 H3 7 H3Dependent Variable Work-Based

ConstrualLevel

Work-BasedConstrual

Level

Exploration Exploration Exploration Exploration Exploration

Step 1 Step 2 Step 1 Step 2 Step 1 Step 2 Step 3

ControlsAge .063 .077 .128 .106 .128 .145 .123Gender (male 5 0,female 5 1)

–.230 –.166 –.419** –.337** –.419** –.343** –.297**

Children .010 .036 –.134. –.137 –.134 –.104 –.114Rank (non man. 5 0,manager 5 1)

.091 –.004 .154 .122 .154 .042 .043

Higher education(no 5 0, college 5 1,

.182 .133 .278* .214* .278* .220 .184

Graduate degree 5 2)Tenure –.057 –.039 .111 .091 .112 .123

Main effects .091Role integration — .309* — — — .364** .279**WBCL — — — .364** — — .275**

Adjusted R2 .001 .075 .209 .324 .209 .320 .381Change in R2

— .074 .105 — .111 .061

Note: N 5 80. The coefficients reported in each column are standardized b coefficients.* p , .05** p , .01

TABLE 3Descriptive Statistics for all Variables (Study 3)

Variables Mean SD Min Max 1 2 3 4 5 6 7 8 9

1. Age 36.14 11.13 19 622. Gender .562 .499 0 1 .2053. # of children .512 .914 0 4 .153 .1374. Rank .462 .501 0 1 .097 .060 .0845. Education 1.65 .530 0 2 –.051 .130 .032 –.0536. Tenure 6.11 2.14 1.17 14 .167 .139 –.060 –.183 .395**7. Role integration 3.35 1.50 1 7 –.087 –.197 –.083 .285* .095 –.0818. Work-Based

Construal Level9.56 5.12 1 18 .008 –.195 .005 .085 .122 –.024 .347**

9. Exploration 4.3 1.39 1.2 7 .038 –.353** –.155 .099 .241* .144 .451** .453**

Note: N 5 80.* p , .05** p , .01

754 JuneAcademy of Management Journal

effect of boundary management on workers’ mentalmindsets. Across multiple studies, we demonstratean association between role integration and con-strual level, with implications for how abstractlypeople communicate in their professional emails andtheir propensity to engage in exploratory learning.

The idea that role integration and segmentationbehaviorsmay shapeworkers’mental representationsis important for several reasons. First, it suggestsnovel consequences of boundary management thatextend those considered in prior research. Much ofthe focus in prior research is on the effect of role in-tegration and segmentation for outcomes such asstress, job satisfaction, and well-being (e.g., Edwards& Rothbard, 1999; Kossek et al., 2006; Kreiner, 2006).These consequences are important from the per-spective of the role occupant, and may have down-stream implications for workgroups, families, andwork performance. Complementing this work, ourexploration of the effect of role boundaries onworkers’mental representations suggests that there may alsobe cognitive implications of role integration, such asthe types of alternatives workers consider and thedecisions theymake. These outcomes have importantstrategic and organizational implications, and thusexpand the scope of boundary theory research. Ourfindings suggest new questions; for example, roleintegration behaviors may influence organizationallyimportant outcomes such as innovation, which itselfis related to exploratory learning.

Second, our perspective potentially offers a moreelaborated understanding of the boundary manage-ment process itself. For example, an importantconclusion from prior boundary theory research isthat transitions between roles are facilitated by roleintegration (e.g., Ashforth et al., 2000), but as yet weknow little about the underlying psychologicalprocesses involved in these transitions. Perhapssuch role transitions are enabled by more abstractmental representations. Abstract mental repre-sentations associated with higher construals enablecomparisons of diverse targets and the transfer ofknowledge and experience outside its original con-text, which may be critical to satisfying competingrole demands simultaneously. Likewise, the focusand execution orientation associated with moreconcrete mental representations, which we suggestmay be facilitated by erecting and maintainingthicker role boundaries, might be responsible forsome of the benefits of role segmentation describedin prior work (e.g., Perlow, 1999). These cognitiveprocesses may relate to the social interactions andsocial negotiation of boundaries that have been

explored in prior research on role transitions (e.g.,Kreiner et al., 2009).

Novel costs and benefits of role integration mayalso be suggested by our analysis. For example,while higher construal makes strategic goals moresalient, people operate more efficiently and actmore expeditiously when their construals are lowerand their mental representations are therefore moreconcrete. Thus, the costs may outweigh the benefitsof role integration when the demands of a specificrole require complete and focused attention. Whenattentional requirements are very high (perhaps thecase for Perlow’s (1999) software engineers), roleintegration may be ineffective for such workers atthe same time that it may be helpful to otherknowledge workers whose tasks demand less fo-cused attention (and perhaps more exploratorylearning behavior). Future research may identifywhen workers can operate at a higher level of con-strual, mentally representing their roles more ab-stractly, without sacrificing job performance.

Our exploration of the relationship between roleintegration behaviors and construal level is domain-specific. However, both conceptually and methodo-logically, we focus on the cumulative effects of roleintegration behaviors over time within role. Thisleaves open the possibility that engaging in role in-tegration in themomentmay have a different effect oncognition (perhaps even leading people to thinkmoreconcretely) than the cumulative effects of role in-tegration behaviors that we explored. Future researchmay take a more dynamic perspective than we haveassumed here, evaluating whether momentary effectsof role integration on cognition and other outcomesmay differ from the cumulative effects we study.

Mobile technologies are only one factor that may in-crease role integrationbehaviors, but our investigationsof the cognitive and behavioral consequences of roleintegration often makes mobile technology salient be-causewe believe these technologies to be an importantenabler of role integration, especially for knowledgeworkers. In this respect, ourwork is relevant to researchon telework, which has focused primarily onindividual-level outcomes of technology-enabledworkmodes, including worker satisfaction, commitment,andwork–life conflict (Gajendran & Harrison, 2007). Ifteleworkers engage in mobile technology-enabled roleintegration, then our findings suggest that tele-work may be associated with a host of previously-unexplored organizationally-relevant outcomes suchas exploratory learning behavior. However, giventhe likely salience of mobile technology in ourstudies, an important question for future research

2015 755Reyt and Wiesenfeld

may be whether mobile technology-enabled role in-tegration is a boundary condition for our findings orwhether our model might also apply to role in-tegration enabled via other means, such as multi-cultural experience. For example, it is possible thatso-called “cosmopolitans” who live and work inmultiple countries may engage in higher levels ofrole integration (and have higher construal level)than “locals” whose experience is more confined(Haas, 2006).

Another boundary condition for our model is thatwe focused on micro-role transitions (work/non-work and at-work transitions) but not macro-roletransitions such as those involved in graduation,retirement, or job loss. Future research may explorewhether the nature of the roles being integrated, orthe nature of the transition required between roles(e.g., role integration that is experienced as threat-ening and stressful rather than challenging and en-ergizing), maymoderate the effect of role integrationon construal level and learning behavior. Workers’role integration preferences may play an importantrole, as well; it is possible that feeling forced to in-tegrate when role integration is not one’s preferencewould attenuate the relationship between role in-tegration and higher construal level.

Also, our findings regarding the effects of roleintegration on construal level are similar for in-tegration across temporal (Study 1) and spatial(Study 4) role boundaries. Future research mayfruitfully explore whether role integration has dif-ferent effects on cognition depending upon whethertime, space, or social role boundaries are bridged.

Implications for Construal Level Theory

Our findings extend construal level research,which has generally been applied to targets such asevents and objects that are clearly delineated in timeand space. In traditional conceptualizations ofconstrual level theory, objective distance from atarget yields psychological distance and thus im-pacts mental representations. However, it is notclear how objective distance applies to targets thatare less tangible (and thus less easily bounded intime and space). We extend notions of psychologi-cal distance to apply to the social roles people enact(i.e., employees’ work role) by drawing upon thetemporal, spatial, and social demarcations betweenroles described in boundary theory. People experi-ence micro-role transitions (whether work/non-workor at-work) throughout the day (Ashforth et al., 2000).We suggest that people who behaviorally segment

their work roles may conceptualize their work ata low level of abstraction because each work activityrequires specific contextual information (i.e., locations,timeframes, etc.) to be performed. Conversely, wesuggest that people who behaviorally integrate theirwork roles may conceptualize their work at a highlevel as a functional adaptation to cognitive limitationsand the challenges of role interruptions, overlaps,contamination, and conflicts. Thus, construal levelmay be domain-specific.

By linking construal level theory to work roles, wenot only extend construal level research beyond theobjects and events studied earlier but we also openup new avenues of inquiry regarding the possibleimplications of more abstract or concrete mentalmindsets at work. Mental abstraction may influencea host of decisions relevant to the field of organiza-tional behavior, including motivation, time alloca-tion, and what information is sought or used indecision-making. Our results in Study 1 suggest thatconstrual level may be manifested in employees’communication, and thus there may be novel socialeffects of construal level in organizations that havenot yet been considered. For example, a manager’sconstrual level may shape how they frame a sub-ordinate’s task (more abstractly and perhaps moreempowering but also more ambiguous, or moreconcretely but perhaps in a way that is perceived asmicro-managing). Task framing may, in turn, haveconsequences for how subordinates execute the task.

Implications for Organizational Learning Research

Our findings may also have implications for re-search on exploratory learning. Prior research sug-gests that exploratory learning involves cognitionmore extensively than exploitative learning andexperience-based search does (Gavetti & Levinthal,2000). Our research may be a first step towardopening up the black box of cognition related tolearning by suggesting that the type of cognition mayalso vary, from more abstract to more concrete. Ourfindings suggest that more abstract mental mindsetshave important implications for distant search, orindividuals’ tendency to depart from the familiar,which is one of the most important ways that learn-ing relates to innovation (Gavetti et al., 2012). Wealso find that role integration behaviors are positivelyassociated with exploratory learning (Study 4). Thisparallels prior research on collectives suggesting thatinternal variety combined with integration mecha-nisms enhance exploration (e.g., Alexiev et al., 2010;Jansen et al., 2009; Reagans & Zuckerman, 2001;

756 JuneAcademy of Management Journal

Schilling et al., 2003; Taylor & Greve, 2006) andevidence that having “cosmopolitans” on a teamwho take a broader perspective enhance externalknowledge acquisition (Haas, 2006). Our findingscontribute by extending this to the individual levelof analysis, where role integration may improve theaccessibility and integration of multiple perspectives.Our findings also suggest a cognitive mechanism po-tentially underlying this effect (i.e., construal level ormental abstraction).

Moreover, our studies suggest methodologicalapproaches to studying these constructs that have notbeen used in prior organizational research. For ex-ample, our various studies adapt or develop means ofmeasuring and manipulating individuals’ mentalmindsets. These include coding their communication(Study 1), developing a domain-specific work-basedconstrual level measure (Studies 2 and 4), and usinga manipulation of construal level in which people areprompted to focus on either categories (more abstract)or exemplars (more concrete; Study 3). We also de-velop an exercise (Study 3) enabling the measure-ment of distant search that is relevant to managerialdecision-making and thus more appropriate for usewith working adult participants than the games usedin prior work (e.g., Billinger et al., 2014). The presentstudies may thus facilitate subsequent research byexpanding the methodological options available toscholars interested in these constructs.

Organizational Implications

The present research also has a number of impor-tant implications for management and for organi-zations. As organizations become more complex andglobal, and demands on workers increase, role in-tegration has become more expected and practiced.Our findings suggest that role integration behaviorsare a proximal factor shaping construal level. Some ofwhat predicts role integration behaviors is organiza-tional and perhaps may be influenced by managers.For example, because roles are inherently social,local norms, goals and incentives, the demands ofinteraction partners and the nature of work re-sponsibilities are all likely to define role integration.At aminimum,managersmust understandhow thesefactors influence employees’ mental representationsand behavior, but perhaps this knowledge may helpmanagers proactively shape work contexts to elicitthe most appropriate and beneficial cognitions andbehaviors. For example, role integration may beencouraged—and perhaps facilitated by providingappropriate technology—for employees who would

perform their work better when thinking abstractly(such as those expected to engage in exploratorylearning). On the other hand, managers may need tobuffer employees from role integration pressure,technologically or otherwise, when effectivenessdepends on narrow focus, detail-orientation, andexecution, and when the costs of integration (e.g.,stress) outweigh its benefits. Perhaps ubiquitous“bring your own device” policies (Cisco, 2013) maybe reconsidered for such employees because thesepolicies may enable too much role integration forsome. Future boundary theory research may helpmanagers develop the tools for assessing and moni-toring their employees’ level of role integration andevaluate how managers can buffer their employeesfrom role integration pressure or otherwise influencerole integration behaviors.

Exploratory learning is important to organizationsbecause it is a promisingmeans of enabling adaptation,particularly in dynamic contexts (March, 1991). Thepresent findings suggest that construal level is a prox-imal predictor of employees’ exploratory learning be-havior, suggesting that managers and organizationsmay benefit from efforts to promote particular mentalmindsets among knowledge workers who are re-sponsible for organizational learning activities. Con-strual level research suggests that mobile technologyand role integration pressuremay be part of a larger setof tools through which organizations can influencetheir employees’ mental mindsets. For example,existing construal level research suggests that longertime horizons create psychological distance, which inturn is associated with higher construals (Trope &Liberman, 2003). Thus, managers may be able to frametasks and decisions with longer or shorter time hori-zons to help shape employees’ construal level. Rewardsystems also focus attention and may therefore shapewhether employees represent their actions more ab-stractly or more concretely. For example, incentivesand controls that focus on process might encourageemployees to focus on how they execute their rolewhile those focused on outcomes may draw moreemployee attention to why.

The findings in Study 1 concerning the email com-munications of executives at Enron suggest that man-agers themselves may be impacted by role integrationbehaviors. Specifically, executives engaging in greatertemporal role integration communicated more ab-stractly. This raises the possibility of coordinationproblems, such as when these executives’ subordinatesor other social interaction partners require more con-crete communication. It is not clear whether integra-tors’more abstract communication was intentional or

2015 757Reyt and Wiesenfeld

unintentional—research on multi-tasking suggeststhat people are relatively unaware of, and overlyoptimistic about, the performance effects of multi-tasking. It is possible that people are likewise unawareof the cognitive and behavioral effects of role in-tegration, and can become better managers of them-selves and others through such knowledge.

CONCLUSION

Work roles and organizational life are trans-forming in ways that are both costly and beneficial.This process is facilitated by mobile technologiesthat are pervasive and evolving rapidly. The presentfindings draw attention to how role integrationbehaviors shape mental mindsets—and knowledgeworkers’ construal level in particular—with impor-tant implications for the learning activities theyengage in. This research is thus a first step in un-derstanding the psychological processes underlyingrole integration behaviors, the cognitive basis ofexploration, and the application of construal levelto roles and to the field of organizational behavior.

REFERENCES

Alexiev, A. S., Jansen, J. J., van den Bosch, F. A., & Volberda,H. W. 2010. Top management team advice seekingand exploratory innovation: The moderating role ofTMT heterogeneity. Journal of Management Stud-ies, 47: 1343–1364.

Amodio, D. M., & Frith, C. D. 2006. Meeting of minds: Themedial frontal cortex and social cognition. NatureReviews. Neuroscience, 7: 268–277.

Ashforth, B. E., Kreiner, G. E., & Fugate, M. 2000. All ina day’s work: Boundaries and micro role transitions.Academy of Management Review, 25: 472–491.

Beckman, C. M. 2006. The influence of founding teamcompany affiliations on firm behavior. Academy ofManagement Journal, 49: 741–758.

Beckman, C. M., Haunschild, P. R., & Phillips, D. J. 2004.Friends or strangers? Firm-specific uncertainty,market uncertainty, and network partner selection.Organization Science, 15: 259–275.

Benbunan-Fich, R., & Truman, G. E. 2009. Technicalopinion: Multitasking with laptops during meetings.Communications of the ACM, 52: 139–141.

Biddle, B. J. 1979. Role theory: Expectations, identitiesand behaviors. New York: Academic Press.

Biggs, S. F., Bedard, J. C., Gaber, B. G., & Linsmeier, T. J.1985. The effects of task size and similarity on thedecision behavior of bank loan officers.ManagementScience, 31: 970–987.

Billinger, S., Stieglitz, N., & Schumacher, T. R. 2014.Search on rugged landscapes: An experimental study.Organization Science, 25: 93–108.

Brass, D. J. 1995 Creativity: It’s all in your social network.In C. M. Ford & D. A. Gioia (Eds.), Creative action inorganizations: 94–99. Thousand Oaks, CA: Sage.

Burgoon, E. M., Henderson, M. D., &Markman, A. B. 2013.There are many ways to see the forest for the trees: Atour guide for abstraction. Perspectives on Psycho-logical Science, 8: 501–520.

Burrus, J., &Roese,N. J. 2006. Long ago itwasmeant to be: Theinterplay between time, construal, and fate beliefs. Per-sonality andSocial PsychologyBulletin, 32: 1050–1058.

Cameron, A. F., & Webster, J. 2005. Unintended con-sequences of emerging communication technologies:Instant messaging in the workplace. Computers inHuman Behavior, 21: 85–103.

Chen, G., Goddard, T. G., & Casper,W. J. 2004. Examinationof the Relationships amongGeneral andWork-SpecificSelf-Evaluations, Work-Related Control Beliefs, andJob Attitudes. Applied Psychology, 53: 349–370.

Cisco, 2013. BYOD insights 2013: A Cisco partner net-work study.

Clark, S. C. 2002. Communicating across the work/homeborder. Community Work & Family, 5: 23–48.

Coenen, L. H., Hedebouw, L., & Semin, G. R. 2006. TheLinguistic Category Model (LCM). Manual. FreeUniversity Amsterdam.

Diesner, J., Frantz, T. L., & Carley, K. M. 2005. Commu-nication networks from the Enron email corpus “It’salways about the people. Enron is no different”.Computational & Mathematical Organization The-ory, 11: 201–228.

Edwards, J., & Rothbard, N. 1999. Work and family stressand well-being: An examination of person–environment fit in the work and family domains.Organizational Behavior and Human DecisionProcesses, 77: 85–129.

Fenner, G. H., & Renn, R. W. 2010. Technology-assisted sup-plemental work and work-to-family conflict: The role ofinstrumentality beliefs, organizational expectations andtime management. Human Relations, 63: 63–82.

Fiol, C. M., & Huff, A. S. 1992. Maps for managers: Whereare we? Where do we go from here? Journal ofManagement Studies, 29: 267–285.

Freitas, A. L., Gollwitzer, P., & Trope, Y. 2004. The influenceof abstract and concrete mindsets on anticipating andguiding others’ self-regulatory efforts. Journal of Ex-perimental Social Psychology, 40: 739–752.

Fujita, K., Trope, Y., Liberman, N., & Levin-Sagi, M. 2006.Construal levels and self-control. Journal of Person-ality and Social Psychology, 90: 351–367.

758 JuneAcademy of Management Journal

Gajendran, R. S., & Harrison, D. A. 2007. The good, the bad,and the unknown about telecommuting:Meta-analysis ofpsychological mediators and individual consequences.Journal of Applied Psychology, 92: 15–24.

Gavetti, G. 2005. Cognition and hierarchy: Rethinking themicrofoundations of capabilities’ development. Or-ganization Science, 16: 599–617.

Gavetti, G., Greve, H. R., Levinthal, D. A., & Ocasio, W.2012. The behavioral theory of the firm: Assessmentand prospects. The Academy of Management An-nals, 6: 1–40.

Gavetti, G., & Levinthal, D. 2000. Looking forward andlooking backward: Cognitive and experiential search.Administrative Science Quarterly, 45: 113–137.

Gavetti, G., Levinthal, D. A., & Rivkin, J. W. 2005. Strategymaking in novel and complex worlds: The powerof analogy. Strategic Management Journal, 26:691–712.

Giacomantonio, M., De Dreu, C. K., & Mannetti, L. 2010.Now you see it, now you don’t: Interests, issues, andpsychological distance in integrative negotiation.Journal of Personality and Social Psychology, 98:761–774.

Giacomantonio, M., De Dreu, C. K., Shalvi, S., Sligte, D., &Leder, S. 2010. Psychological distance boosts value-behavior correspondence in ultimatum bargainingand integrative negotiation. Journal of ExperimentalSocial Psychology, 46: 824–829.

Greenhaus, J. H., & Beutell, N. J. 1985. Sources of conflictbetween work and family roles. Academy of Man-agement Review, 10: 76–88.

Gupta, A. K., Smith, K. G., & Shalley, C. E. 2006. Theinterplay between exploration and exploitation.Academy of Management Journal, 49: 693–706.

Haas, M. R. 2006. Acquiring and applying knowledge intransnational teams: The roles of cosmopolitans andlocals. Organization Science, 17: 313–332.

Hartmann, E. 1997. The concept of boundaries in coun-selling and psychotherapy. British Journal of Guid-ance & Counselling, 25: 147–162.

Heerwagen, J. H., Kampschroer, K., Powell, K. M., &Loftness, V. 2004. Collaborative knowledge workenvironments. Building Research and Information,32: 510–528.

Hossain, L. 2009. Effect of organisational position and net-work centrality on project coordination. InternationalJournal of Project Management, 27: 680–689.

Huber, G. P. 1991. Organizational learning: The contrib-uting processes and the literatures. OrganizationScience, 2: 88–115.

Huff, A. S. 1990. Mapping strategic thought, Chichester,UK: John Wiley & Sons.

Jansen, J. J., Tempelaar, M. P., van den Bosch, F. A., &Volberda, H. W. 2009. Structural differentiation andambidexterity: The mediating role of integrationmechanisms. Organization Science, 20: 797–811.

Jarvenpaa, S. L., Lang, K.R., & Tuunainen, V. K. 2005.Friend or Foe? The Ambivalent Relationship betweenMobile Technology and its Users, in DesigningUbiquitous Information Environments: Socio-tech-nical Issues and Challenges, Carsten Sorensen,Younjin Yoo, Kalle Lyytinen, & Janice I. DeGross(eds.), 29–42. Boston, MA: Springer Science andBusiness Media.

Katila, R., & Ahuja, G. 2002. Something old, somethingnew: A longitudinal study of search behavior andnew product introduction. Academy of Manage-ment Journal, 45: 1183–1194.

Kossek, E. E., Lautsch, B. A., & Eaton, S. C. 2006. Tele-commuting, control, and boundary management:Correlates of policy use and practice, job control, andwork–family effectiveness. Journal of VocationalBehavior, 68: 347–367.

Kossek, E. E., Noe, R. A., & DeMarr, B. J. 1999. Work–family role synthesis: Individual and organizationaldeterminants. The International Journal of ConflictManagement, 10: 102–129.

Kreiner, G. E. 2006. Consequences of work–home seg-mentation or integration: A person–environment fitperspective. Journal of Organizational Behavior,27: 485–507.

Kreiner, G. E., Hollensbe, E. C., & Sheep, M. L. 2009.Balancing borders and bridges: Negotiating thework–home interface via boundary work tactics.Academy of Management Journal, 52: 704–730.

Lavie, D., & Rosenkopf, L. 2006. Balancing explorationand exploitation in alliance formation. Academy ofManagement Journal, 49: 797–818.

Leung, L. 2011. Effects of ICT connectedness, permeabil-ity, flexibility, and negative spillovers on burnout andjob and family satisfaction. Human Technology, 7:250–267.

Levinthal, D. A., & March, J. G. 1993. The myopia of learn-ing. Strategic Management Journal, 14(S2): 95–112.

Levitt, B., & March, J. G. 1988. Organizational learning.Annual review of sociology, 319–340.

Liberman, N., & Trope, Y. 1998. The role of feasibility anddesirability considerations in near and distant futuredecisions: A test of temporal construal theory. Jour-nal of Personality and Social Psychology, 75: 5–18.

Liberman, N., & Trope, Y. 2008. The psychology of tran-scending the here and now. Science, 322: 1201–1205.

Louis, M. R., & Sutton, R. I. 1991. Switching cognitivegears: From habits of mind to active thinking. HumanRelations, 44: 55–76.

2015 759Reyt and Wiesenfeld

Mael, F., & Ashforth, B. E. 1992. Alumni and their almamater: A partial test of the reformulated model oforganizational identification. Journal of organiza-tional Behavior, 13: 103–123.

March, J. G. 1991. Exploration and exploitation inorganizational learning. Organization Science, 2:71–87.

Mazmanian, M. A., Orlikowski, W. J., & Yates, J. 2005.Crackberries: The Social Implications of UbiquitousWireless E-Mail Devices, in C. Sørensen, et al. (eds.)Designing Ubiquitous Information Environments:Socio-Technical Issues and Challenges: 337–344.New York, NY: Springer.

Mazmanian, M., Orlikowski, W. J., & Yates, J. 2013. Theautonomy paradox: The implications of mobile emaildevices for knowledge professionals. OrganizationScience, 24: 1337–1357.

Mead, G. H. 1934. Mind, self, and society. Chicago: Uni-versity of Chicago Press.

Michaelsen, S., & Johnson, D. E. 1997. Border theory: Thelimits of cultural politics. Minneapolis: University ofMinnesota Press.

Middleton, C. A. 2008. Do mobile technologies enablework-life balance? Dual perspectives on blackberryusage for supplemental work. In: D. Hislop (ed.)Mobility and Technology in theWorkplace: 209–224.London, UK: Routledge.

Mom, T. J. M., van den Bosch, F. A. J., & Volberda, H. W.2007. Investigating managers’ exploration and ex-ploitation activities: The influence of top-down,bottom-up, and horizontal knowledge inflows.Journal of Management Studies, 44: 910–931.

Mors, M. L. 2010. Innovation in a global consulting firm:When the problem is too much diversity. StrategicManagement Journal, 31: 841–872.

Nippert-Eng, C. E. 1996. Home and work: Negotiatingboundaries through everyday life. Chicago, IL:University of Chicago Press.

Nonaka, I., & Takeuchi, H. 1995. The knowledge-creating company: How Japanese companies cre-ate the dynamics of innovation. New York, NY:Oxford University Press.

Nussbaum, S., Liberman, N., & Trope, Y. 2006. Predictingthe near and distant future. Journal of ExperimentalPsychology. General, 135: 152–161.

Olson-Buchanan, J. B., & Boswell, W. R. 2006. Blurringboundaries: Correlates of integration and segmenta-tion between work and nonwork. Journal of Voca-tional Behavior, 68: 432–445.

Payne, J. W. 1976. Task complexity and contingent pro-cessing in decision making: An information searchand protocol analysis. Organizational Behavior andHuman Performance, 16: 366–387.

Perlow, L. A. 1999. The time famine: Toward a sociologyof work time. Administrative Science Quarterly, 44:57–81.

Preacher, K. J., & Hayes, A. F. 2008. Asymptotic andresampling strategies for assessing and comparingindirect effects in multiple mediator models. Be-havior Research Methods, 40: 879–891.

Rau, B. L., & Hyland, M. A. M. 2006. Role conflict andflexible work arrangements: The effects on applicantattraction. Personnel Psychology, 55: 111–136.

Reagans, R., & Zuckerman, E. W. 2001. Networks, di-versity, and productivity: The social capital ofcorporate R&D teams. Organization Science, 12:502–517.

Rennecker, J., & Godwin, L. 2005. Delays and inter-ruptions: A self-perpetuating paradox of communica-tion technology use. Information and Organization,15: 247–266.

Richardson, K. M., & Thompson, C. A. 2012. High techtethers and work-family conflict: A conservation ofresources approach. Engineering Management Re-search, 1: 29–43.

Rizzo, J. R., House, R. J., & Lirtzman, S. I. 1970. Roleconflict and ambiguity in complex organizations.Administrative Science Quarterly, 15: 150–163.

Rosenkopf, L., & Nerkar, A. 2001. Beyond local search:boundary-spanning, exploration, and impact in theoptical disk industry. Strategic Management Jour-nal, 22: 287–306.

Rothbard, N. P., Phillips, K. W., & Dumas, T. L. 2005.Managing multiple roles: Work–family policies andindividuals’ desires for segmentation. OrganizationScience, 16: 243–258.

Schilling, M. A., Vidal, P., Ployhart, R. E., &Marangoni, A.2003. Learning by doing something else: Variation,relatedness, and the learning curve. ManagementScience, 49: 39–56.

Schlosser, F. 2002. So, how do people really use theirhandheld devices? An interactive study of wirelesstechnology use. Journal of Organizational Behavior,23: 401–423.

Semin, G. R., & Fiedler, K. 1991. The Linguistic CategoryModel, its bases, applications and range. EuropeanReview of Social Psychology, 2: 1–30.

Shetty, J., & Adibi, J. 2005. Discovering important nodesthrough graph entropy the case of Enron emaildatabase. Proceedings of the 3rd internationalworkshop on Link discovery, 74–81.

Simon, H. A. 1955. A behavioral model of rational choice.Quarterly Journal of Economics, 69: 99–118.

Smith, P. K., & Trope, Y. 2006. You focus on the forestwhen you’re in charge of the trees: Power

760 JuneAcademy of Management Journal

priming and abstract information processing.Journal of Personality and Social Psychology, 90:578–596.

Smith, P. K., Wigboldus, D. H. J., & Dijksterhuis, A. 2008.Abstract thinking increases one’s sense of power. Jour-nal of Experimental Social Psychology, 44: 378–385.

Stuart, T. E., & Podolny, J. M. 1996. Local search and theevolution of technological capabilities. StrategicManagement Journal, 17(S1): 21–38.

Tarafdar, M., Tu, Q., Ragu-Nathan, B. S., & Ragu-Nathan,T. S. 2007. The impact of technostress on role stressand productivity. Journal of Management Infor-mation Systems, 24: 301–328.

Taylor,A., &Greve,H. R. 2006. Supermanor the fantastic four?Knowledge combination and experience in innovativeteams. Academy of Management Journal, 49, 723–740.

Towers, I., Duxbury, L., Higgins, C., & Thomas, J. 2006.Time thieves and space invaders: Technology, workand the organization. Journal of OrganizationalChange Management, 19: 593–618.

Trope, Y., & Liberman, N. 2003. Temporal construal.Psychological Review, 110: 403–421.

Trope, Y., & Liberman, N. 2010. Construal-level theory ofpsychological distance. Psychological Review, 117:440–463.

Turner, J. C. 1982. Towards a cognitive redefinition of thesocial group. In H. Tajfel (Ed.), Social identity andintergroup relations: 15–40. Cambridge, UK: Cam-bridge University Press.

Tversky, A., & Kahneman, D. 1986. Rational choice andthe framing of decisions. Journal of Business, 59:S251–S278.

Vallacher, R. R., & Wegner, D. M. 1987. What do peoplethink they’re doing? Action identification and humanbehavior. Psychological Review, 94: 3–15.

Vallacher, R. R., & Wegner, D. M. 1989. Levels of personalagency: Individual variation in action identification.Journal of Personality and Social Psychology, 57:660–671.

Wakslak, C. J., & Trope, Y. 2009. Cognitive consequencesof affirming the self: The relationship between self-affirmation and object construal. Journal of Experi-mental Social Psychology, 45: 927–932.

Walsh, J. 1995. Managerial and organizational cognition:Notes from a trip down memory lane. OrganizationScience, 6: 280–321.

Wanous, J. P., Reichers, A. E., & Hudy, M. J. 1997. Overalljob satisfaction: how good are single-item measures?.Journal of Applied Psychology, 82: 247–252.

Ward, T. B. 1994. Structured imagination: The role ofcategory structure in exemplar generation. CognitivePsychology, 27: 1–40.

Weiss, H. M., & Adler, S. 1984. Personality and organi-zational behavior. Research in Organizational Be-havior, 6: 1–50.

Wrzesniewski, A., McCauley, C., Rozin, P., & Schwartz, B.1997. Jobs, careers, and callings: People’s relations totheir work. Journal of Research in Personality, 31:21–33.

Jean-Nicolas Reyt ([email protected]) is anassistant professor of organizational behavior atMcGill University’s Desautels Faculty of Management.He received his PhD in management jointly fromParis-Dauphine University and the University of Lux-embourg. His research is focused on understanding therelationship between employees’ mental representa-tions of their work and important work outcomesincluding creativity, exploratory learning, and in-terpersonal influence. He also studies mobiletechnology-facilitated work and other organizationalfactors that influence mental representations.

Batia M. Wiesenfeld ([email protected]) is theAndre J.L. Koo Professor of Management at the SternSchool of Business, New York University. She receivedher PhD from Columbia Business School. Her researchinterests include justice, identity, construal level, andsocial regard, particularly in relation to ambiguous andchanging organizational contexts such as virtual work,layoffs, failure events, and emergent online communities.

APPENDIX 1

WORK-BASED CONSTRUAL LEVEL SCALE

High and low-level activity descriptions are oppositeanchors of six-point scales.

Instructions: Imagine yourself performing the followingwork activities, and indicate on the continuum (the verbaldescriptions represent endpoints) the description thatbest describes each activity for you:

DISCRIMINANT VALIDITY

We explored the correlations between our work-basedconstrual level measure and constructs in the nomologicalnetwork related to it. We contacted 1,000 participants inan online subject pool where people participate in sur-veys for payment. Two hundred and ninety three people(29.3% of our sample) were classified as knowledgeworkers following the method described in Studies 2 and

2015 761Reyt and Wiesenfeld

4. We invited the selected knowledge workers to partic-ipate in our study and obtained answers from 223 of them(76% response rate). Five participants were droppedfrom the analyses for failing our attention check. Thus,218 participants were included in the final analyses.

As reported in the table below, all of the correlationswere significant but modest except the “Career”subscale. Thus, we conclude that work-based construalcan be distinguished from other potentially relatedconstructs.

Measures Source Cronbach’s a Correlation with WBCL

WBCL N/A .902 N/AJob self efficacy Chen et al. (2004) .900 .185*Role ambiguity Rizzo et al. (1970). .889 .186*Org. identification Mael & Ashforth (1992) .888 .191*Job satisfaction Wanous et al. (1997) N/A .227*Job/calling Wrzesniewski et al. (1997) . 696 .244*Career Wrzesniewski et al. (1997) .693 .098–

Note: WBCL 5 Work Based Construal Level Measure.* p , .01

Work Activity Low-Level Description High Level Description

Preparing a report * Compiling information Showing progressUsing a computer Typing on a keyboard Processing informationFilling out a business form * Filling in blanks with information Following work protocolObtaining information from someone Asking relevant questions Gaining knowledgeMaking a presentation * Presenting relevant material Communicating knowledgeAssigning work to someone Telling someone what to do Getting things doneCommunicating information to someone Sending an email or talking to someone Keeping someone informedAnalyzing a dataset * Comparing numbers Identifying trendsAttending a meeting Being present and paying attention Staying up to dateDeveloping a procedure * Writing down step-by-step instructions Increasing work efficiencyWriting business correspondence * Composing an email Maintaining a good business relationshipHiring someone Interviewing candidates Maintaining staff levelDeveloping a budget Listing expenses and revenues Managing fundsProofreading a document * Reading carefully for errors Ensuring accuracyTraining someone * Showing someone how to do things Increasing someone’s productivityAnalyzing an operational report Reviewing information Ensuring smooth operationOrienting a new worker * Showing a new worker around Acclimating a new workerEvaluating someone’s performance Reviewing quality of work Providing feedback

Note: “*” designates reverse-scored items.

762 JuneAcademy of Management Journal