_Ha Vu Kien - 2012 - A Meta-analysis of the consequences of virtualness on team functioning

<span class='text_page_counter'>(1)</span><div class='page_container' data-page=1>

A

meta-analysis

of

the

consequences

of

virtualness

on

team

functioning

Ana

Ortiz

de

Guinea

a,1

,

Jane

Webster

b,

*

,

D.

Sandy

Staples

b,2

a_HEC_Montre´al,_3000,_chemin_de_la_{Coˆte-Sainte-Catherine,}_Montre´al,_Que´bec_H3T_2A7,_Canada
b_Queen’s_School_of_Business,₁₄₃_Union_Street,_Kingston,_Ontario_K7L_3N6,_Canada

1. Introduction

Thepracticeofusingvirtualteamsinorganizationshasbecome

popular. Thus, it is important to organizations and team

researcherstounderstandhow‘‘virtualness’’affectsteam

perfor-mance.Wedecidedtoreviewtheempiricalresearchonthistopic

throughameta-analysis.

Virtualteamshavebeendefinedasgroupsofindividualswho

work together in different locations on interdependent tasks,

sharing the responsibility for outcomes, while relying on

technologytoprovidemostoftheircommunication.Whileearly

virtual team research usually examined ‘‘virtualness’’ as a

dichotomy, either face-to-face or computer-mediated (without

physicalinteraction),virtualnesshasevolvedtoincludedegreeof

separation of members(distance), proportionof memberswho

workvirtually (configuration),and theproportion of time that

teammembersworkapart.WeadoptedSchweitzerandDuxbury’s

[19]suggestionthat‘‘tobeconsideredvirtual,ateammusthave

somememberswhodonotworkineitherthesameplaceand/orat

thesametime,andtherefore,cannotcollaborateface-to-faceallof

the time’’. Thus, we focused on the degree of virtualness, a

continuum,rangingfromnotatallvirtualtohighlyvirtual.We

onlyincludedarticlesinourstudythatmeasuredorvariedoneor

moredimensionsofvirtualness.3 _In_doing_so, _this_article_differs

fromothersthatincludedarticlesthatdonotassessorvarythe

level ofvirtualness[16]orthatdo notanalyzetherelationship

betweenvirtualnessandothervariables[10].

Narrativereviewsoftheliteraturehavehighlighted

inconsis-tent findings, such as positive and negative [e.g., 7] relations

between virtualnessandperformance.Understandingwhysuch

inconsistencies exist could help practitioners. The narrative

reviewshave alsopointedout a predominanceof certaintypes

ofstudies,suchasshort-termexperimentalresearch,andquestion

whether theresults are generalizable. In contrast, quantitative

reviews, suchasmeta-analyses,helptoaddressthese

inconsis-tencyissuesandquestions.

Meta-analyses quantitatively integrate results reported in

existing studies. The analysis increases the power of the

conclusions made on the relations between variables. Unlike

narrativereviews, meta-analysesadjust forsamplesizeandthe

reliability of measures, thus providing better estimates of the

ARTICLE INFO

Articlehistory:
Received8July2010

Receivedinrevisedform6June2012
Accepted21August2012
Availableonline25September2012

Keywords:
Meta-analysis
Distributed
Virtualness
Group
Performance
Knowledgesharing
Satisfaction
Conflict
Time

ABSTRACT

Virtualteamsinorganizations havenowbecomeareality,but therehave beenonlyahandfulof

quantitativereviewson‘‘virtualness’’(i.e.,teamsthataremoreorlessvirtual).Wedecidedtoconducta

meta-analytic review of the effects of virtualness on team functioning (conflict, communication

frequency,knowledgesharing,performance,andsatisfaction).Toexplaininconsistenciesintheresults

of publishedmaterial on thetopic, we also examinedthe moderatingeffects of levelofanalysis

(individual/group),method (experiment/survey), andtime frame (short/long). Eightystudies were

foundthatcoveredsomepartofthisdomain.Resultsseemtodifferintherelativeimportanceofthe

factors.Thusthoughaggregatedfindingssuggestednegativeeffectsofvirtualnessonteamfunctioning,

resultsvaried in strength and direction of themoderators, indicating thatit was not possibleto

generalize.Forexample,thenegativeeffectsheldonlyforshort-termteams,whileinlonger-termteams

theeffectsweakenedordisappeared.

ß2012ElsevierB.V.Allrightsreserved.

*Correspondingauthor.Tel.:+16135333163;fax:+16135332325.
E-mailaddresses:(A.OrtizdeGuinea),

(J.Webster),

(D.S.Staples).

1

Tel:+15143407817;fax:+15143406132.

2

Tel:+16135332314;fax:+16135332325.

3_We_excluded_articles_that_did_not_have_either_a_measure_of_virtualness_or_a

comparisonofvirtualwithface-to-faceteams.Neitherdidwequantitativelyreview

articlesatotherlevelsofanalysis,suchasvirtualcommunitiesororganizations,as
thiswouldconfoundtheresults.

ContentslistsavailableatSciVerseScienceDirect

Information

&

Management

j our na l ho me pa ge : w ww . e l se v i e r . com / l oca t e / i m

0378-7206/$–seefrontmatterß2012ElsevierB.V.Allrightsreserved.

</div>
<span class='text_page_counter'>(2)</span><div class='page_container' data-page=2>

relationsbetweenvariables.Theyalsoallowforthediscoveryof

moderatorsthrough thecoding ofdifferences between studies.

Thus,progress intheorybuildingand cumulative knowledgeis

achievablewithmeta-analysis.

Theobjectivesofourmeta-analysiswereto:(1)examinethe

extent to which virtual teams research has built cumulative

knowledge,(2)quantifythestrengthoftherelationshipsbetween

virtualnessand team processes and outcomes, and (3) explore

reasonsforcontradictoryfindings.

2. Theconsequencesofvirtualness

Toexaminetheeffectsofvirtualnessinteamwork,weusedthe

well-knowninput–process–output(IPO)modelthatincludesinput

factorssuchasteammembercharacteristics,team-and

organiza-tional-levelarchitecture,contextualinformation,andvirtualness;

processeswhicharetheinteractionsbetweenteam4membersor

communicationsfromindividualsinvolvingteamwork,thatmay

beeitherexpressive(interpersonal)orinstrumental(task-related);

andoutputsthatmeasureteameffectiveness,suchasperformance

onthetaskorthesatisfactionofgroupmembers.

The IPOmodel hasbeenmodifiedas morehasbeenlearned

aboutteameffectiveness; forexample,somemediating

mecha-nisms, initially considered processes, have been identified as

emergentstates(e.g.,cognitive,motivational, oraffectivestates

such as potency and cohesion). Therefore, the general term

mediatorisusedtodescribeprocessesandemergentstates.Time

hasalsobeenrecognizedtohaveanimportantrole,whichwasnot

capturedwellintheIPOmodel.ThesechangeshaveledtoanIMOI

(input–mediator–output–input)framework[14].Withinthisnew

framework,itis possibletoexplaintheeffectsofvirtualnesson

mediatorsand/oroutcomeswiththetheoreticalperspectives of

mediarichness/socialpresence,attributiontheory,and

categori-zation.

Today,groupsaremorelikelytobedispersed,withelectronic

communicationdominatingtheirinteraction.Mediarichnessand

socialpresencetheoriessuggestthatsuchcommunicationisless

personal, with fewer nonverbal cues. Also, attribution theory

suggeststhatpeopletrytoexplaintheirownorothers’behaviors

bymakingtwotypesofcasualattributions:internal(dispositional)

orexternal(situational).Aperson’sinitialattributionofanother’s

behaviorisusuallydispositional(i.e.,thefundamentalattribution

error)andthencorrected,dependingonknowledgeoftheperson

and/orsituation.Virtualteamsarelikelytomakeattributionerrors

becausemembers haveless knowledgeoftheir teammatesand

theirenvironments.Thusthereispotentialforattributionerrorsto

go uncorrected [6]. Categorization provides a third theoretical

perspective:socialidentitytheory,self-categorizationtheoryand

thesimilarity/attractionparadigmallsuggestthatpeople

catego-rize themselves into subgroups according to salient cues.

Individualsidentify more closely with people they perceive as

beingsimilartothemselves[5].Invirtualteamsettings,subgroups

maydevelop.Asin-andout-groupcharacteristicsbecomesalient

insubgroups,individualsbecomemorebiasedtowardstheirown

subgroups[20].

3. Developmentofresearchquestionsandhypotheses

Virtualnessistheonlyinputvariableweexamined.Considering

thenumberofmediatorandoutcomeconstructs,thenumberof

possiblerelationshipsisquitelarge.Wesearchedtheliteraturefor

empiricalstudiesrelatedtovirtualteams.Whenevervirtualness

wasincluded in a study and its relationship to a mediator or

outcomevariablehadbeenexaminedinatleastoneotherstudy,

themediator/outputvariablewasincludedinourmeta-analysis.

Therefore,ourfirstgeneralresearchquestionwas:

Question 1: What are the strengths and directions of the

relationships between virtualness and team mediators and

outputs?

We also examined the generalizability of the findings.

Specifically, by introducingmoderatingvariables, meta-analysis

techniques can determine whether between-study differences

werepartiallydue tothedifferentconditionsin themoderator

variables.Researchdesignsandsamplecharacteristicsaretypical

moderatorsexaminedinmeta-analysiswork. Accordingly,three

possible moderators that varied most frequently across the

primarystudieswereinvestigated:thelevelofanalysis(individual

versusgroup),thestudymethod(experimentversussurvey),and

the time theteams worked together (short versus long-term).

Therefore,oursecondresearchquestionreferredtomoderators:

Question2:Towhatextentdolevelofanalysis,studymethod,

and/or time duration moderate the relationships between

virtualnessandteamprocessesandoutputs?

Toinvestigatethesequestions,wedevelopedspecific

hypothe-sesabouttheeffectsofvirtualnessontheselectedmediatorsand

outcomes.Thechoiceofprocessesandoutcomestoincludeinour

meta-analysiswasdrivenbythreefactors.First,weneededtypical

processes and outcomes to be represented. Second, we were

restrictedtosituationsinwhichmultipleempiricalstudiesonthe

relationshipbetweenvirtualnessanda specificvariableexisted.

Third,wewishedtofocusonvariableswheretheempiricalresults

wereambiguous,sothatourmeta-analysiswouldhelptoclarify

them. Three mediating variables (conflict, communication

fre-quency, knowledge sharing) and two output variables

(perfor-mance,andsatisfaction)metourcriteria.

3.1. Maineffecthypotheses

Conflict represents perceived discrepancies, incompatible

desires, and wishes of the parties involved in a team. Conflict

maybedividedintothreetypes:relationshipconflict(theaffective

componentofconflictthatconcernstheawarenessof

incompati-bilities intheinterpersonal realm);taskconflict (differencesin

opinionsaboutagrouptask);andprocessconflict(differencesof

opinionaboutthewayataskshouldbeperformed).Researchers

have usedconflict measureswithlabels suchas: proportion of

disagreement, personal attacks, tension, task conflict, process

conflict,relationshipconflict,affectiveconflictandgeneralconflict,

oracombinationofthese.Allthreetypesrelatetovirtualteam

functioning[9].Forexample,itcanbearguedthatmoderatetask

conflictcanbebeneficialbecauseitresultsindiscussionthatcan

provideabettersolutiontosolvethetask.Indeed,Masseyetal.

[13]foundthatmoreproductivevirtualteamsexperiencedmore

conflict.However,manystudiesfindthatthethreetypesofconflict

inter-relatestronglybutnegativelytoteamfunctioning[e.g.,8].

Wethereforeexpectedthatthoseinmorevirtualteamswill

experiencemoreconflict,whilethoseinlessvirtualteamsenjoy

more face-to-face communication, resulting in more informal

interaction and socialization. This is likely to result in greater

rapport and collaboration, stronger team identity, and lower

conflict.Becauseofdistancebetweenteammembers,conflictis

difficult tomanage in morevirtual teams. Differences in work

locationcanalsoleadtotheformationofin-groupsandout-groups

withinteams.Thus,conflicttypicallyarisesbetweenthesubgroups

duetofavoritismoflocalmembers.

Althoughmoststudiesfoundpositiverelations(higherconflict

for more virtual teams), a few have demonstrated a negative

4

</div>
<span class='text_page_counter'>(3)</span><div class='page_container' data-page=3>

relationship [e.g.,21]. Nevertheless, given that the majority of

empiricalresearchshowsapositiverelationbetweenvirtualness

andconflict,weexpectedthatmorevirtualteamswillexperience

moreconflict:

Hypothesis1:Virtualnessrelatespositivelytoconflict.

Our secondhypothesis concerns theeffectof virtualness on

communication frequency (the amount of communication

be-tweenteammembers). Withoutcommunication,lackofmutual

knowledge,misunderstandings(attributionerrors),andalackof

contextualknowledgemayoccur.Unfortunately,communication

representsa challengeforvirtualteams.Althoughthosein less

virtualteamsmayusethesamecommunicationtechnologiesand

communicate at the same frequency, patterns of face-to-face

communicationcan differ. Co-locationpromotesinformal

com-municationandthereforemoreoverallcommunication.

Mostempiricalresearchsupportsthis,forexample,O’Learyand

Cummings [17] found that communication frequency related

negatively to virtualness. Thus, we expected that more virtual

teamswouldcommunicatelessandhypothesized:

Hypothesis2:Virtualnessrelatesnegativelytocommunication

frequency.

Knowledgesharingrepresentsanotherimportantteamprocess.

Researchershaveusedmeasuressuchasamountofinformation

shared, transferred, or exchanged within teams. In traditional

teams,thesharingofexpertiseisanessentialgroupprocessfor

teameffectiveness.Sharingofinformationandknowledgeiseven

more critical for cross-functional teamwork and good decision

making[18].

Few studies have examined the effect of virtualness on

knowledge sharing, and within those, the results have been

somewhat mixed,though most reported negative relationships

[e.g.,1].Itis,ofcourse,moredifficulttosharerichinformationand

knowledge through electronic media than face-to-face. Also,

geographic dispersion may lower employees’ attention to the

virtualteamtask.Geographicdiversitycanalsoleadtothecreation

ofsub-groupswithinateam,wherein-groupsfavorlocalmembers

andsharelessknowledgeoverall.Teamidentitycanalsobehigher

inlessvirtualgroupsduetomoreinteractionandsocialpresence.

Finally,becausevirtualteamsareoftenculturallydiverse,language

skillsmightlimitknowledgesharing.Thus:

Hypothesis 3: Virtualness relates negatively to knowledge

sharing.

Outcomes represent the results and by-products that have

valuetotheorganization; theymaybesplitinthreeclasses(or

dimensions):performance,which relates totheteam’staskand

includesteam efficiency, productivity,and innovation; affective

reactions, which include satisfaction and commitment; and

behaviors,whichincludeindicatorsofthegroup’sabilitytoexist

overtimesuchasturnoverandabsenteeism.

We focused on two outcomes: team performance and

satisfaction. Performance represents the effectiveness of the

outcome with respect to the specific task or project at hand.

Researchershaveincludedmeasuressuchasproject outcomes

(e.g., met requirements, within budget, within schedule) and

confidenceinthedecisionreachedbythegroup.Becausemuchof

thecommunicationsinvirtualsettingstakeplaceelectronically,

causal factors, such as lower social presence and slower

communication, might affect performance. Asynchronous and

slower-pacedcommunicationmayleadtolessfocusandattention

totheteamactivity,resultinginlower performance.Invirtual

teams,trusttakeslongertodevelopandcanbemorefragilethan

inface-to-faceteams.Thus,lowertrustcannegativelyinfluence

teamperformance.Developingmutualknowledgeabout

team-mates’environmentsisalsochallengingwhenthereisgeographic

separation,andthelackofthisknowledgecanleadtoattribution

errorsandundermineteameffectiveness.

Nonetheless,virtualnesscanalsoprovidebenefits.Forexample,

creativity of the team may be enhanced by bringing diverse

perspectives to bear and encouraging people to share their

opinions. Lessproduction blockingtakesplace inasynchronous

communicationsincepeoplecanworkatthesametime,aswellas

rehearsetheircommunications.However,mostempiricalarticles

suggest anegativerelationshipbetweenvirtualnessand

perfor-mance.Therefore:

Hypothesis 4: Virtualnessrelates negatively toteam

perfor-mance.

Thereareseveralwaysinwhichvirtualnesscouldaffectteam

satisfaction. In virtual settings, relationships take longer to

develop.Strongerinterpersonalrelationshipsandteamtieshave

been linkedtomotivation and less process loss,which in turn

affects team satisfaction. More developed teams communicate

more constructively, have more mutual understanding and

knowledge (i.e., fewer attribution errors), are able to manage

conflictmoreeffectively,andaremorecohesive.Thereforevirtual

teams may have lower satisfaction, at least in theshort-term.

Cyber-ostracism, where one ormore individualsis excluded in

groupinteractions,canalsonegativelyimpactteamsatisfactionin

virtualteams.Therefore:

Hypothesis5:Virtualnessrelatesnegativelytoteam

satisfac-tion.

3.2. Moderatinghypotheses

Someofthereportedresearchfindingsaremixed;forexample,

Martinsetal.[12]notethemixed(positiveandnegative)relations

between virtualness and team functioning. The examination of

moderatorsmighthelptoshedlightontheseconflictingresults.

Therefore,weconsideredhowourmoderators(levelofanalysis,

studymethod,andteamduration)couldalterourfivehypotheses.

Resultsatonelevelofanalysismaynotgeneralizetoanother

andthusmixinglevelsofanalysismayintroduceambiguitiesinto

theresults[2].Teamshavecharacteristicsthataredifferentfrom

thatofindividualswhomaynotagreewiththeperceptionoftheir

team-mates.Further, resultsmaybestronger attheteam level

becauseconstructsusuallycapturesharedbeliefs.Therefore:

Hypothesis6:Thelevelofanalysis(groupversusindividual)

moderatesthevirtualnessrelationships.Morespecifically,for

groups, rather than for individuals, virtualness relates more

positivelyto(i)conflictandmorenegativelyto(ii)

communi-cation frequency, (iii) knowledge sharing, (iv) team

perfor-mance,and(v)satisfaction.

Most experiments are made on ad hoc, temporary groups,

whosemembersareparticipatingforcoursecreditandarelocated

inoneuniversity;incontrast,mostsurveysaremadeon

longer-term groups of employees from multiple organizations and

countries. Becauseparticipantsinexperimentsgenerallyhavea

specifictasktocompletewithinashorttimeframe,allparticipants

(generally students) communicate to the extent necessary,

(throughwhatever mediaareavailable) tocompletetheir task.

Thus,experimentalparticipantstendtobededicatedtothetask

andtheeffectsofdistanceareminimized.Thiswouldsuggestthat

weakerrelationshipsforvirtualnessoccurinexperimentsthanin

surveys. However, one could also take the opposite view and

</div>
<span class='text_page_counter'>(4)</span><div class='page_container' data-page=4>

experimentstobeshorter-term(andthelogicof H8,presented

next),wemightexpectteamprocessesandoutcomestobeless

positiveforexperiments.Aswecannotpredictthedirectionofthe

moderation,weposit:

Hypothesis 7: The study methodmoderates thevirtualness

relationships. Morespecifically,virtualnesswillrelate

differ-ently to (i) conflict, (ii) communication frequency, (iii)

knowledgesharing,(iv)teamperformance,and(v)satisfaction

forsurveysandexperiments.

Thethirdmoderatorwastheamountoftimethatteamsworked

together.Thisisimportantbecauselonger-termteamsappearto

bemorelikelytoexistinactualorganizationalsettings,andteam

processesandcommitmentchangeovertime.Thenegativeeffects

ofdiversityarelikelytobeneutralizedasmembersspendmore

timetogether[15].Similarly,communicationpatternscanchange,

starting with unidirectional communication but ending with

mutualcommunication.Therefore,asteamsdevelopmorefully,

differencesbetween face-to-face and distributed teamstend to

disappear,and teammembersbecomemorewillingtoworkto

understandtheproblemsofotherteammembersovertime[3].

Thus,team duration canchangetheamountof timethat team

memberswanttointeractandtheperceivedbenefitsofinvesting

insocialandworkingrelationships.Therefore:

Hypothesis 8: Team duration moderates the virtualness

relationships. More specifically,forshorter termrather than

longer termteams,virtualness relates morepositively to(i)

conflictandmorenegativelyto(ii)communicationfrequency,

(iii) knowledge sharing, (iv) team performance, and (v)

satisfaction.

4. Method

By using meta-analysis, the strengths of the relationships

betweentwoormoreconstructscanbeassessedbyquantitatively

combining resultsfrom existing (primary) empirical studies to

determine the overall relationships. Inconsistent findings are

commonamongprimarystudiesduetomeasurementerror,low

statisticalpower, anddifferentresearchcontexts.Meta-analytic

techniquescanhelpovercomesomeoftheseproblems.

Weusedseveralwaystofindempiricalstudiesandupdatedthe

datawithmorerecentsearches.Articleswerenotrestrictedtoany

discipline and were found by searching databases (PsycINFO,

ProQuest,WebofScience)andconferenceproceedings(Academyof

Management (AOM), Americas Conference on Information

Sys-tems, Computer Supported Cooperative Work, International

Conferenceon Information Systems), obtaining workingpapers

throughpersonalcontacts,searchingtwo Websitesfocused on

virtual team research (VoNet and virtualteamresearch.org),

posting calls for unpublished and ‘in press’ papers (AISWorld,

AOM’sOrganizational CommunicationandInformation Systems

division),reviewingcollections of articleson virtual teamwork,

conducting a general Google Web search, and examining

referencesinreviewedarticlestoidentifynewleads.Thisresulted

inasubstantialvolumeofarticlesbeingretrieved;forexample,

ProQuestcoversapproximately1800businessjournalsfrom1971

tothepresent time.Our searchterms were:virtual, virtualness,

distributed, dispersed, global, or remote combined with team or

group.Weexcludedstudiesthatcontainedonlyqualitativedata,

examinedarelationshipthatwasnotusedbyanyotherauthor

and/ordidnotcontainallofthedataweneeded5_._Although_the

searches provided us with almost 400 articles that might be

includedinourmeta-analysis,manyhadtobeexcludedbecause

they did not match our criteria. For example, for one of our

searches(aftertheremovalofqualitativeandtheoreticalarticles),

62%didnotcapturevirtualness(eitherasameasureofvirtualness

orasacomparisonbetweenvirtualandface-to-faceteams),30%

didnotcontainthequantitativedataneeded,6%didnotinclude

enoughinformation for theirinclusion, and 2%werearticlesin

which the data had already been published in another article

alreadyinourmeta-analysis.Fromalloursearches,wefound80

uniqueandusefuldatasetsfromthe79articleslistedintheonline

Appendix(see />

4.1. Codingthedata

Meta-analysiscanbeusedtosummarizerelationshipsbetween

variables (effect sizes) using correlations, mean differences, or

proportions.Manymeta-analystssummarizecorrelationsbecause

oftheirprevalenceinresearcharticles,whichiswhatwedid.For

our analysis, one author coded all of the papers based on

discussions with the other two authors. Once the coding was

complete,wehiredagraduatestudentwhowasblindtoourcoding

toensurethatthedirectionoftheeffectswasappropriatelycoded

foreachstudy6_._Any_{discrepancies}_were_resolved_by_the_authors.

Moststudiescapturedvirtualnessasadichotomousmeasure

(co-located [0] or distributed teams [1]). Only sixteen studies

examined virtualness as a continuous variable (e.g., combined

spatialdistanceandcommunicationmediaasameasure).When

studies captured virtualness as a dichotomous measure, we

transformed thedata to point-biserial correlations which were

used with the correlations for the continuous measures; this

allowedustoexaminetheoveralleffectsofvirtualness.

As with any meta-analysis, we needed to make a series of

coding decisions.Thefirstconcerned studiesinwhich labels of

variables appeared inconsistent with the content of their

measures.Thiswas doneby comparing thelabels in thestudy

withtheitemsusedtooperationalizethem.Atleasttwoauthors

discussed thecontent of a measure and reconciled any

discre-pancieswithitslabel.Thesecondconcernedmoderatorlevels.For

example,regardingteamduration,wehadtodeterminethecut-off

between short- and long-term teams. Moderator levels are

determinedbythedataavailableintheempiricalstudies.Based

onthis,teamswereconsideredshort-termwhentheirmembers

workedtogetherforadayorlessandclassifiedaslong-termwhen

theyworked togetherfor more thana day. Althoughthis time

periodmayseemarbitraryandsmall,wewereconstrainedbythe

availabledata.

Anotherissueconcernedthenecessityofaggregatingnarrow

measures into larger measures. For example, some studies

reported process satisfaction while others reported outcome

satisfaction;wecodedthesedifferentaspectsofsatisfactioninto

thesameoutcomevariablecalled‘‘Satisfaction’’.Wedidthisfor

threevariables:conflict, satisfactionand performance.Tocheck

the appropriateness of aggregation, we ran separate analyses

onthedisaggregatedmeasuresandthencomparedtheconsistency

oftheresults.Forconflict,weconductedseparateanalysesfortask,

relationship,process,andmiscellaneousconflict(suchasoverall

conflict or conflict behaviors); for performance, we conducted

analysesformeasureslabeledasperformanceandmeasureswith

otherperformancelabels(suchas‘creativityofsolution’);andfor

5

Manyofthearticlesdidnotreportallofthedataweneeded.Wecontacted
authorswhoprovidedpartialdata;somewereabletosendusthedata.

6_Some_of_the_studies_reported_mean_differences_(or_some_other_statistic)_on_a

</div>
<span class='text_page_counter'>(5)</span><div class='page_container' data-page=5>

satisfaction,weconductedanalysesforsatisfactionwithsolution,

satisfaction with process, and other satisfaction (e.g., general

satisfaction).Theresultsshowedthatthedisaggregatedmeasures

hadthesamerelationshipwithvirtualnessastheoverallmeasure,

except for conflict: task conflict was positively related to

virtualness,whiletheotherconflict measureswere

non-signifi-cant.Therefore,taskconflictispresentedseparatelyfromtheother

typesofconflict.

We also made coding decisions for studies that presented

multiplecorrelations forthesame constructs.Because multiple

correlationsinasinglestudycreatethepotentialforoverweighting

studiesthatcontainnon-independentdata,theresultsofa

meta-analysismaybedistorted.Toavoidthis,weusedaconservative

approachofchoosingonecorrelationperstudytominimizethe

problems associated with non-independent coefficients. When

morethanonecorrelationwasreportedforthesameconstructat

thesamepointoftimeinthesamestudy(multipledimensionsof

the same construct), we averaged the correlations into one

correlationinordertocapturethedimensionalityoftheconstruct.

Whenmakingthesedecisions,weusedthedatathatmostclosely

representedteamsinorganizations,makingtheresultspotentially

closer to practice. For example, when data were reported for

multiple points in time, we used the data that most closely

represent actual employee teams (e.g., if correlations were

presentedafter 1 monthand 3 monthsfor thesame variables,

wechosethecorrelationfor3months).Also,iftherewasmore

thanoneface-to-faceconditionintheexperiment(e.g.,bothwith

and without computer support), we chose the condition with

computersupport,asthismostcloselyresemblesactualteamsin

organizations.

5. Analyses

Meta-analysisprovides an estimateofthe‘‘true’’ population

effect size for a given relationship via a series of steps. We

cumulatedcorrelationcoefficientsintoaverageeffectsizes[4,11].

5.1. Computationofeffectsizes

Toexaminethefirstfivehypotheses,weusedsimpleeffectsizes

ofcorrelations7_._A_number_of_studies_did_not_report_correlations

but we obtained some either by contacting the authors or by

converting otherreported data (suchas t-tests, F-tests,etc.) to

correlations.

We calculated overall mean effect sizes (r) by weighting

correlationsbysamplesize,correctingboththecorrelationsand

weightings for unreliability, performing Fisher’s

Z-transforma-tions, calculating the overall mean correlations, and then

un-transformingtheoverallmeancorrelations.Theseweretestedfor

significanceusingaz-testandconfidenceintervalsarereported.

(In addition, we report the mean effect sizes (rxy) without

correctingforunreliability).

5.2. Moderatordetectionandestimation

HypothesesH6–H8 wereaddressedbydeterminingwhether

thestudiesincludedin themeta-analysis came from

heteroge-neouspopulations(i.e.,whethermoderatorvariableswerepresent

ornot),and,ifso,whetherthemoderatorvariablesaccountedfora

significantamountoftheresidual variance.The Qstatistic was

used as a test of effect size homogeneity, with a statistically

significantQindicatingheterogeneityofeffectsizes.For

statisti-callysignificantQ’s,weseparatedtheresultsbythemoderator

Table
1
Meta-analyses
of
virtualness
with
processes
and
outcomes.
Team
Inputs
Conflict
(H1)
Communication
frequency
(H2)
Knowledge
sharing
(H3)
Performance
(H4)

Satisfaction
(H5)
k
a
nr
xy
b
r
c(CI)
d
k
a
nr
xy
b
r
c(CI)
d
k
a
nr
xy
b
r
c(CI)
d
k
a
nr
xy

b
r
c(CI)
d
k
a
nr
xy
b
r
c(CI)
d
Degree
of
virtualness
e
(overall)
Task
6
630
0.13
0.14
*
13
2544
0.10
0.11
*
(
0.15, _0.06)

7
542
0.21
0.22
*
(
0.31, _0.13)
30
3813
0.09
0.09
*
(
0.12, _0.06)
23
2717
0.07
0.08
*
(
0.12, _0.04)
Other
14
908
0.06
0.07
Degree
of

virtualness (categorical measure)

Task
5
471
0.28
0.31
*
9
840
0.06
0.06
3
173
0.60
0.61
*
23
2483
0.15
0.15
*
18
1880
0.15
0.16
*
Other
12
571
0.02

0.02
Degree
of

virtualness (continuous measure)

Task
1
159
0.26
0.33
*
4
1704
0.12
0.14
*
4
369
0.01
0.01
7
1330
0.00
0.00
5
837
0.09
0.11
*

Other
2
337
0.13
0.16
*
a
k
=
number
of
studies.
b
rxy
=
m
ean
weighted
coefficient.
c
r
=
coefficient
corrected
for
the
unreliability
of
predictor
and

criterion.
d
CI:
95%
confidence
interval.
e
A
higher
number
represents
higher
virtualness.
*
z-test:
correlatio
n
is
different
from
0
(p
<
0.05).
7

</div>
<span class='text_page_counter'>(6)</span><div class='page_container' data-page=6>

variables (individuals/groups, experiments/surveys, and short/

long-term)todetermineifthecorrelationsdifferedacrossthese

moderators.

6. Results

Allofourmainandhomogeneityeffectsweresignificant,and

thereforeposthocpoweranalyseswerenotneeded.Nevertheless,

we calculated power and found that it was >0.90 for all

relationshipsexceptforknowledgesharing(power=0.59).Table

1 presents the meta-analyticresults for therelations between

virtualnessandteamprocessesandoutcomes(H1–H5).

We first examined the results for the overall (combined)

measure of virtualness. Hypothesis 1 was supported for task

conflict(r=0.14*),buttheresultswerenotsignificantforother

typesofconflict.Thefindingswereconsistentwiththerestofthe

hypotheses: Hypotheses 2 (r= 0.11*), 3 (r= 0.22*), 4

(r= 0.09*),and5(r= 0.08*)weresupported.

TurningtothetwomeasuresofvirtualnesspresentedinTable1,

fourofthefivehypothesizedrelationshipsdiffersignificantly.For

each of these, the relationships were more likely tobe in the

hypothesizeddirectionforthecategoricalthanforthecontinuous

virtualnessmeasure(seeTable1).

6.1. Moderators

Whenobservingtherelationshipswithintheindividualstudies,

wesawdifferencesbetweenstudies,suggestingthepresenceof

moderators. For team performance, we found 30 studies that

empiricallyexaminedtherelationshipwithvirtualness:sixwith

mediumtorelativelystrongpositiveeffects (definedasarhoof

0.20orlarger),sevenwithmediumtorelativelystrongnegative

effects,and17withamixofweak(positiveornegative)effects.

Studiesalsoshowedconflictingresultsfortheeffectsofvirtualness

on communication frequency, conflict, knowledge sharing, and

satisfaction.

Thepossibilityofmoderators(H6–H8)wasexaminedviatheQ

statistic.AsallhadasignificantQ(p<0.05),wefurtheranalyzed

theserelationshipsbysplittingthedataonthethreemoderatorsof

individual/group, experiment/survey, and short-term/long-term

(seeTable2).Allcomparisonsforlevelofanalysis(groupversus

individual)asa moderator (H6)weresignificant.However, the

relationshipfor satisfaction wasin the opposite directionthan

proposed (with individual level of analysis beingstronger). All

comparisonsfor studymethod(surveyversusexperiment)asa

moderator(H7)weresupported,exceptforotherconflict(withno

significantdifferencefound).Allcomparisonsforteamduration

(short-versuslong-term)weresupported,exceptthat

communi-cationfrequencywasintheoppositedirectionthanthatproposed

(withlong-termbeingmorenegative).

7. Discussionandimplications

Ourfirstgoalforthismeta-analysiswastoexaminetheextent

to which research presents a cumulative body of knowledge

regardinghowvirtualnessaffectsteamfunctioning.Wefoundthis

tobelacking.DrawingonthetraditionalIPOmodel,thetheoretical

perspectivesofmediarichness/socialpresence,attributiontheory,

andcategorizationhelpedtoexplainourhypotheses.Theoverall

results, summarized using80 datasets (representingresponses

from over several thousand participants), supported our main

hypotheses:morevirtualteamsexhibithighertaskconflictand

lower communication frequency, knowledge sharing,

perfor-mance,and satisfaction. Althoughthesefindings are consistent

withmorerecentresearch,ourresultssuggestthattheresultsdo

not generalize to all types of teams and methodological

approaches.

Our choice of moderators was dependent on the empirical

articles available, and we were able to examine only one

theoretically-based moderator, time. We found that higher

virtualness in short-termteams (a dayor less in ouranalysis)

didnegativelyaffecttheteam(e.g.,moreconflict,less

communi-cation and knowledge sharing, weaker performance and lower

satisfaction).However,inlonger-termteams,virtualnessdidnot

havenearlythesamedetrimentaleffects.Therewasnonegative

effectonteamperformanceandsatisfaction,andteamconflictwas

reduced as virtualness increased. While still havinga negative

effectoncommunicationfrequency andknowledgesharing,the

effect wasconsiderably weakerthan in short-term teams.Itis

importanttonote,however,thatmostofthestudiesinvestigating

short-term teams (a day or less) involved students, whereas

studiesfocusingonlonger-termteams(morethanaday)involved

studentsaswellasemployees.

Virtualnessclearlyhasadifferenteffectonteamsdependingon

thelengthoftimetheteamworkstogether.Forshort-termteams,

leaner media, misattributions, and subgroups all potentially

contributeto lesseffective teams.For example,mediarichness

theorywouldsuggestthattheslowerabilitytocommunicaterich

Table2

Hypothesizedmoderators.

Relationships Q Levelofanalysis
(H6)

ra

(kb

,n) Method(H7) ra

(kb

,n) Teamduration
(H8)

ra

(kb

,n)

Virtualness–Taskconflict 77* _Individualc _0.05_(2,433) _Experimentc _0.33*_(3,404) _Shortc _0.37*_(2,355)

Groupc

0.33*

(4,197) Surveyc

0.20*

(2,226) Longc

0.18*

(4,275)
Virtualness–Otherconflict 59*

Individualc

0.21*

(3,420) Experiment 0.05(10,504) Shortc

0.03(8,372)
Groupc

0.05(11,488) Survey 0.09*

(4,404) Longc

0.16*

(5,358)
Virtualness–Communication

frequency

252*

Individualc

0.04(9,2239) Experimentc

0.01(7,600) Shortb

0.05(6,517)
Groupc

0.30*

(4,305) Surveyc

0.11*

(6,1944) Longc

0.18*

(6,758)
Virtualness–Knowledgesharing 67*

Individualc

0.15*

(3,318) Experimentc

0.61*

(3,173) Shortc

0.67*

(1,59)
Groupc _0.34*_(4,224) _Surveyc _0.01_(3,353) _Longc _0.10*_(5,417)

Virtualness–Performance 276* _Individualc _0.04_(9,2220) _Experimentc _0.17*_(12,889) _Shortc _0.27*_(11,₇₅₀₎

Groupc _0.19*_(21,1593) _Surveyc _0.08*_(16,2924) _Longc _0.07_(16,1919)

Virtualness–Satisfaction 262*

Individualc

0.11*

(14,1990) Experimentc

0.03*

(16,1525) Shortc

0.22*

(12,1178)
Groupc

0.00(9,727) Surveyc

0.12*

(7,1192) Longc

0.05(10,1361)

a

r=coefficientcorrectedfortheunreliabilityofpredictorandcriterion.

b

k=numberofstudies.

</div>
<span class='text_page_counter'>(7)</span><div class='page_container' data-page=7>

informationandinteractbackandforthviaelectronicmediawould

havea muchlargerimpactonteameffectivenessforshort-term

teamsthanforteamswithseveraldaysorweekstoaccomplish

their tasks. (However students in large courses mayalso have

manypeopleintheirgroupswhoknowandhaveworkedwithone

another).Forlongertermteams,memberswouldalsobelesslikely

tomakemisattributionstotheperson:asteammembersinteract,

theybuildknowledgeabouteachotheranddeveloprelationships.

Also, as knowledge of the underlying context and the parties

involvedinthecommunicationsgrows,theabilityfor

understand-ingandtherichnessofthemessagesalsogrow.Similarly,aspeople

start to identify with a team and build a team identity and

psychologicalandaffectivetiestotheteam,thishelpstoovercome

othersourcesofdifferences.

Intermsofmethodsvariables,wefounddifferingresultsbased

on several comparisons.First, resultsaremore encouragingfor

studies using continuous rather than categorical measures of

virtualness. That is, for studies with continuous measures of

virtualness,therelationshipwithtaskconflict ismore negative

(lower conflict for more virtual teams), the relationships with

knowledgesharingandsatisfactionaremorepositive,andthereis

noimpactonperformance.Measuringvirtualnessasacategorical

variablemightbeasimplisticviewthatdoesnotcapturethereality

of virtual work in a natural setting. It might be portraying a

‘‘spurious’’ view of the effects of virtualness on processes and

outcomesintherealworld.Studiesusingacategoricalmeasureare

morelikelytobeexperimentswithstudentsandexperimentsare

designedusingstrongermanipulations,resultinginbiggereffect

sizes.Becausestudiesusingcontinuousmeasuresaremorelikely

to investigate longer-term employee teams using a survey

methodology,wemightbeabletobemoreoptimisticconcerning

virtual team functioning in organizations. Furthermore, past

researchon virtualteamshasoftenconfoundedtechnology use

withmeasuresofvirtualness.Thatis,theycompareddistributed

groupsusingonetool(suchase-mail)withco-locatedgroupswith

no tools. Further, for the articles that provided bundles of

technologies to team members, there was no single set of

technologiesprovidedacrossthearticlesandthereforewecould

not determine the effects of particular technologies for virtual

teams.

Wefoundthatanothermethodsvariable,thelevelofanalysis

(groupversusindividual)alsomoderatedthevirtualness

relation-ships.Asexpected,therewerestrongereffectsofvirtualnesson

taskconflict,communicationfrequency,knowledgesharing,and

performance at the group level than at the individual level.

However, the effects of virtualness on other conflict and

satisfaction were stronger at the individual level than at the

grouplevel.

7.1. Implicationsforresearchandpractice

Researchers shouldbevery specific aboutthe natureof the

teamstheyarestudying(e.g.,definehowlongtheyhaveexisted)

and use theoretical arguments and empirical findings that

appropriately match the types of teams. Our results provide

valuable advice for designers and leaders of virtualteams. For

practitioners, ourresultssuggested that virtual teamprocesses

wouldbemorenegativeintheshort-term.Organizationswillneed

toexpectthesedifficultiesand putintoplacestructurestohelp

minimizetheirimpacts.Recognizingthatvirtualteamswillhave

communication and conflict challenges (which influence team

functioning),weneedtohelpidentifywaystoimprove

effective-ness. For example,team membersshould betrained tohandle

conflicttominimizeitsnegativeimpactonproductivity.For

long-termteams,practitionersshouldexpectinitialchallenges,butthey

couldmovemorerapidlytolessnegativestages.

7.2. Limitations

Aswithanyresearch,this meta-analysishaslimitations that

mustbeconsideredwheninterpretingtheresults.Meta-analytical

workisonlyasmeaningfulastheprimarystudiesfromwhichitis

derived.Wefoundalackofconsistencyconcerningmeasuresand

reportingstandards.Becauseofthelargevarietyofmeasuresused,

we sometimes had to collapse narrow measures into larger

categories, reducing the variance in some measures. A further

weakness in the literature is the small number of studies

examiningsomerelationships.Althoughwesearchedfor

unpub-lished papers, this study is still subject to the so-called ‘‘file

drawer’’problem,orthedifficultyoffindingnon-significantresults

sincetheytendtostayunpublished.Wealsofoundarticlesthatdid

notreportneededinformationsuchasthelevelofanalysis.More

disturbingwasthelackofreporteddataneededtoconductthe

analyses(effectsizes,samplesizes,etc.):wehadtoexcludemany

papersbecauseofalackofnecessarydata(whoseauthorscould

notprovideit).

Itisalsoimportanttonotethatmoderatorsoftencanrelateto

each other(e.g., in our reviewedstudies, we found that study

method[experiment/survey]correlated0.68withtime duration

[short/long]).Fewstudiestrytoseparatetheirindependenteffects.

Althoughweexaminedtheseeffectswithourmoderatoranalyses,

itisnotclearwhetherresultsforstudiesusingcontinuousvirtual

measuresaremoreencouraging,becausethemethodtendstobe

non-experimentalandtheseteamshaveaccesstoawiderrangeof

communication tools, and because teams are made up of

employees,orarelonger-term.

8. Conclusion

Ourmeta-analyticreviewdemonstratedthatnotallvirtualness

measuresarethesameandthatsomeconflictingresultsaredueto

thisandothermoderatorvariables(suchasstudymethod).Asa

result,thisstudyhighlightsandhelpsexplainconflictingresults,

andpointstoempiricalgapsinourknowledgeandthenecessityto

informpracticebybuildingacumulativebodyofresearchinwhich

variablesaresystematicallyinvestigated.

Theresultsofourmeta-analysissuggestthatteamprocesses

andoutputsarenegativelyinfluencedbyvirtualness.However,

theseresultsdonotgeneralizetoalltypesofteamsandanalytical

approaches.Themoderationanalysisindicatesthatvirtualness

hasadifferentinfluenceonteamfunctioningdependingonhow

itismeasuredaswellasonthelengthoftimetheteamworks

together.Thenegativeeffectsfoundinshort-termteamsweaken

or disappear in teams that havelonger lives. Ourresults also

pointtoimportant differencesontheeffectsof virtualnessfor

differentlevelsof analysis(individual vs.group)andmethods

(experiments vs.surveys). Weneedto know more about how

differentlevelsofanalysisinteractwitheachother,aswell as

utilizefinerandmultidimensionalmeasuresofvirtualness.Allof

thiscanhelpusunderstandhowtoimproveteamprocessesand

outcomestomorefullytakeadvantageofthepotentialbenefits

ofvirtualwork.

Acknowledgements

We would like to thank Derek Chapman, Peter Dacin, and

SandyHershcovisfortheiradviceonthemeta-analyticanalyses.

An earlier version was presented at the Symposium on High

PerformanceProfessionalTeams,Queen’sUniversity,Canada.This

paperwasfinanciallyassistedbytheIndustrialRelationsCentre,

Queen’s University, and by Social Sciences and Humanities

Research Councilof Canada grants to Sandy Staplesand Jane

</div>
<span class='text_page_counter'>(8)</span><div class='page_container' data-page=8>

References

[1]B.Alge,C.Wiethoff,H.J.Klein,Whendoesthemediummatter?
Knowledge-buildingexperiencesandopportunitiesindecision-makingteamsOrganizational
BehaviorandHumanDecisionProcesses9,2003,pp.26–37.

[2]D.J.Beal,R.R.Cohen,M.J.Burke,C.L.McLendon,Cohesionandperformancein
groups:ameta-analyticclarificationofconstructrelations,JournalofApplied
Psychology88,2003,pp.989–1004.

[3]S.T.Bell,Deep-levelcompositionvariablesaspredictorsofteamperformance:a
meta-analysis,JournalofAppliedPsychology92,2007,pp.595–615.
[4]M.Borenstein,L.V.Hedges,J.P.T.Higgins,H.R.Rothstein,Introductionto

Meta-Analysis, Wiley,Chichester,England,2009.

[5]T.Carte,L.Chidambaram,Acapabilities-basedtheoryoftechnologydeployment
indiverseteams:leapfroggingthepitfallsofdiversityandleveragingitspotential
withcollaborativetechnology,JournaloftheAssociationforInformationSystems
5,2004,pp.448–471.

[6]C.D.Cramton,K.L.Orvis,J.M.Wilson,Situationinvisibilityandattributionin
distributedcollaborations,JournalofManagement33,2007,pp.525–546.
[7]C.D.Cramton, S.S.Webber,Relationshipsamonggeographicdispersion,team

processes,andeffectivenessinsoftwaredevelopmentworkteams,Journalof
BusinessResearch58,2005,pp.758–765.

[8]C.K.W.DeDreu,L.R.Weingart,Taskversusrelationshipconflict:team
perfor-mance, and team membersatisfaction: ameta-analysis, JournalofApplied
Psychology88,2003,pp.741–749.

[9]P.J.Hinds,R.Bailey,Outofsight,outofsync:understandingconflictindistributed
teams,OrganizationScience14,2003,pp.615–632.

[10]C.Lin,C.Standing,Y.C.Liu,Amodeltodevelopeffectivevirtualteams,Decision
SupportSystems45,2008,pp.1031–1045.

[11]M.W.Lipsey,D.B.Wilson,PracticalMeta-Analysis, SagePublications,Thousand
Oaks,CA,2001.

[12]L.L.Martins,L.L.Gilson,M.T.Maynard,Virtualteams:whatdoweknowand
wheredowegofromhere?JournalofManagement30,2004,pp.805–835.
[13]A.P. Massey,M. Montoya-Weiss,M. Hung,Because time matters:temporal

coordinationinglobalvirtualprojectteams,JournalofManagementInformation
Systems19,2003,pp.129–156.

[14]J.Mathieu,M.T.Maynard,T.Rapp,L.Gilson,Teameffectiveness1997–2007:a
reviewofrecentadvancesandaglimpseintothefuture,JournalofManagement
34,2008,pp.410–476.

[15]M.T.Maynard,J.E.Mathieu,T.L.Rapp,L.L.Gilson,Something(s)oldand
some-thing(s)new:modelingdriversofglobalvirtualteameffectiveness,Journalof
OrganizationalBehavior33,2012,pp.342–365.

[16]J.R. Mesmer-Magnus,L.A. DeChurch,M. Jimenez-Rodriguez,J. Wildman, M.
Shuffler,Ameta-analyticinvestigationofvirtualityandinformationsharingin
tea,OrganizationalBehaviorandHumanDecisionProcesses115,2011,pp.214–
225.

[17]M.O’Leary,J.N.Cummings,Thespatial,temporal,andconfigurational
character-isticsofgeographicdispersioninteams,MISQuarterly31,2007,pp.433–452.
[18]N.R.Quigley,P.E.Tesluk,E.A.Locke,K.M.Bartol,Amultilevelinvestigationofthe

motivationalmechanismsunderlyingknowledgesharingandperformance,
Or-ganizationalScience18,2007,pp.71–88.

[19]L.Schweitzer,L.Duxbury,Conceptualizingandmeasuringthevirtualityofteams,
InformationSystemsJournal20,2010,pp.267–295.

[20]S.M.B.Thatcher,K.A.Jehn,E.Zanutto,Cracksindiversityresearch:theeffectsof
diversityfaultlinesonconflictandperformance,GroupDecisionandNegotiation
12,2003,pp.217–241.

[21]R.L.Wakefield,D.E.Leidner,G.Garrison,Amodelofconflict,leadership,and
performanceinvirtualteams,InformationSystemsResearch19,2008,pp.434–
455.

Ana Ortiz de Guinea is an assistant professor of
InformationSystemsatHECMontre´al.SheholdsaPhD
inInformationSystemsfromQueen’sUniversity,aMSc
fromthe University ofLethbridge, and adegree in
ComputerScienceandEngineeringfromthe
Universi-daddeDeusto. Prior toreturning to academia,she

workedas an Information Systemsconsultant. Her
researchhasbeenpublishedinComputersinHuman
Behavior,theInternationalJournalofHumanResource
Management,theInternationalJournalofe-Collaboration,
theJournalofGlobalInformationManagement,andMIS
Quarterly.

Jane Webster received her PhD from New York
University and is the E. Marie Shantz Professor of
MISintheSchoolofBusinessatQueen’sUniversityin
Canada.She has served asa Senior Editorfor MIS
Quarterly,VPofPublicationsforAIS,andadvisorfor
AIS-SIGCHI.She haspublished in a variety of journals
includingtheAcademyofManagementJournal,
Commu-nicationResearch,InformationandOrganization,
Infor-mationSystemsJournal,InformationSystemsResearch,
JournalofOrganizationalBehavior,MISQuarterly,and
OrganizationScience. Her current research concerns
information systems and technologies to support
environmentalsustainability

</div>

<!--links-->
<a href=' />