Professional Documents
Culture Documents
ConsequencesforDispersedCollaboration
CatherineDurnellCramton
Mail Stop SF5, Fairfax,Virginia22030-4444
School of Management,George Mason University,
@som.gmu.edu
ccramton
Knowledge
withan idea fora businessthatwould use theInternet
categoncal)
in
Contextual or
Problem: some way,(2) writinga businessplan,and (3) creatinga
local differences
Member differences
presentationfor investorsor an online storefront.
Less shared reality
Biased discussion
P
rquality -
The
Performance
Geographic
dispersion project spanned a seven-weekperiod. Communication
deciso
decision
Dispersed
toolsused bytheteamsincludedelectronicmail,Internet-
making
task-related
information
based "chat"tools,an Internet-based votingtool (Dennis
Task
et al. 1996), telephone,and fax. The projecthome page
Informationload Interdependence
was a commonpointof referenceforthe teams.Home
characteristics ... ... ..
-- . --...- ------
-- - Time pressure . ........ndenc
--
messages thatan e-mail was missing,and I earmarked puttingthe13 cases intoan accessibleformwithoutcom-
thesepoints.The countof e-mailrecordsby teamranged promisingtheirrichness,(2) understanding each case on
from61 messages(Team 30) to 217 messages(Team 6). itsowntermsbeforeattempting to generalizeacrosscases
Each team's communication log listedtheteam's on- (Eisenhardt1989, Miles and Huberman1994), (3) con-
line chatsand use of otherWeb-basedtools such as the ductingan embeddedinformation-processing analysisof
votingtool.The numberofchatsheldbyteamsor subsets episodesofconflict, or confusion,(4) creating
frustration,
ofteamsrangedfromnone(Teams 11, 18,and 26) to five and refiningconstructsthatcut across cases, (5) identi-
(Teams 5, 21, and 39). The teamsturnedin copies of the fyingothervariablesof interest, (6) reviewingall cases
textsof mostof theirchats.However,on some occasions of constructsand build evidence
to refinethe definition
theyfailedto makea printout so thetextwas notavailable to measureor refutethem,(7) integrating theconstructs
foranalysis.Descriptiveinformation about each of the intoa tentativemodel,and (8) reviewingall cases to re-
cases appearson theleftside of Table 1. futeor refinethemodel.
GMU studentswroteindividualsix-pageanalysispa-
Data Management.I followed the process used by
pers aftertheprojectended.They were instructed to an-
Gersick(1988) in her studyof eightcollocatedproject
alyze one or two eventsin thelifeof theirteamthatthey
teamsto graduallycondensethevoluminouscase histo-
consideredto be significant,
usinge-mailandchatrecords
ries so theycould be reviewedtelescopicallyas well as
as a resource.They were requiredto do theirbest to ex-
microscopically.I wantedto make it possible to follow
amine these events fromthe perspectivesof the other
the flowof each case while preservingtightlinksto the
teammembersas well as fromtheirpersonalperspective.
originalpieces of data. Each of the 13 teams'e-mailwas
To understand theexperiencesoftheteams,I also drew
readintoAskSam,a text-management softwareprogram.
on my experiencesas a memberof the geographically
Missing e-mail identifiedthroughthe crosscheckwith
dispersedfacultyteam and my work withthe students
TCU was added to the files.Each piece of e-mail was
engaged in the project.In addition,severalmembersof
assignedan identification number.
the facultyteam wroteabout theirexperiencesand cir-
AskSam was used to createfieldswithinwhichto an-
culated these narratives.Finally,my colleague at TCU
notateeach piece ofe-mail.In one suchfield,myresearch
sentme copies of thegradesshe awardedto each of the
assistantsummarizedthe literalcontentof the e-mail.
13 teams thatincludedGMU studentsso thatI would
This paralleledGersick's literalsummariesof the team
have two perspectives,hers and my own, on team per-
meetingsshe studied.In anotherfield,myresearchassis-
formance.
tantrecordedherinterpretations oftheactivityintheteam
Cursoryreviewof the cases suggestedthattherehad
and thequestionsthatcame to hermind.She had been a
been a greatdeal of conflictin theteams.In sevenof the
memberof one of the teamsand recognizednuancesof
13 teams,conflictescalatedto thepointthathostilecoa-
situationsthata newcomerto thecomplexprojectprob-
litionsformed.In fiveof these teams,membersat two
ably would have failed to grasp. Her attentionwas di-
sites began to complainabout partnersat the thirdsite,
rectedprimarily to themicroscopiclevel-recordingthe
refusingin some cases to sendthempieces of theteam's
literalcontentof each piece of e-mail-with secondary
workor put theirnames on finishedwork.Two teams
attention to theflowof events.
evidenced shiftingcoalitions among subgroupsat the
threesites. Close examinationof episodes of conflict, Case Analysis.I studiedeach case and recordedmy
or confusionin theteamsseemedto be mer-
frustration, observationsin a fieldcreatedforthispurpose.Creation
ited. ofthesummariesmadeitpossibleforme to reviewentire
cases quicklywhenI wishedto,tracking theoverallflow
Data Analysis of events.In addition,countsof the numberof e-mails
Data managementand analysisproceduresare summa- each teamexchangedduringeach day oftheprojectwere
rizedin Table 2. My objectiveswereto analyzeepisodes generatedthroughthe softwareand transformed into
of conflict, orconfusionintheteams,examine graphsof the team's communicationactivity-another
frustration,
the significanceand consequencesof these episodes in perspectiveon theflowof thewhole.I also examinedthe
thecontextof each case as a whole,and look forpatterns e-mailmicroscopically, comparingmyimpressionswith
across cases. I followed Eisenhardt's(1989) specifica- thesummariesand commentsof myresearchassistant.I
tionsforanalysisof multiplecase studies,withthe ad- wantedto be surethatthe summariesshe preparedwere
ditionofan embeddedinformation-processing analysisof sufficiently
descriptive,so thatI could relyon themwhen
episodesofconflict, orconfusionwithincases
frustration, I wished to move quicklythroughthe material.Using
(Yin 1994, Coulam and Smith1985). Steps included(1) AskSam,one can double-clickon thesummary ofa piece
C, U, R, S, 0, and T = serious problemsofthistype in the team. c, u, r,s, q, and t = some problemsof thistype in the team.
E = frequentuse of information fromexternalsources in the team. e = some use of information fromexternalsources in the team.
representsthe numberof chats involvingat least two locationsduringwhichprojectworkwas conducted
numberadjusted to reflectdifferencesin gradingcriteria
Stages of Work
and Key Observations Activities Purpose
Data Manipulation
Check completenessofe-mailrecordsbycomparinge-mail
turnedinbyteammembersat ECU and SWU.
Add to themasterfileany newlydiscoverede-mails.
all e-mailintotextmanagementsoftwareprogram.
Import
Adjusttimestampsto EasternStandardTimeso e-mailcan
be sequenced correctly
and resequence.
Assemblerecordsand reportsofteam "chat"sessions.
Researchassistantwritesa literalsummary ofeach ofthe
1,649 pieces ofe-mailand ofeach chatsession. Focus
on contentofindividualpieces ratherthanoverallflowof Gersick1988: Beginprocess ofgraduallycondensing
events.Observationsand questionsnotedinseparate voluminous ofteamactivity
transcripts into
field. summariesofeventsequences.
Data analysis
butPaul did notchangehis understanding of theteamor team to be low and the pace to be slow. Accordingly,
see how differently thisexchangemusthave appearedto theymayfurther reducetheirpace, orberatemembersfor
Don. Even thoughPaul figuredout thathe had Don's because theydo notknowoftheirefforts.
inactivity Anal-
name and addresswrongand offeredto sendDon all the ysisoftheteamhistoriessuggeststhatthesekindsofper-
early messages he had written,Paul still presentedthe ceptionscan be excruciatingly to identifyand
difficult
exchangeas a powerstrugglein his analysispaper.Paul changewhena teamis dispersed.Private"conversations"
stillthoughtthatDon deliberatelyignoredthe workhe maycreatemuchmoreconfusionfordispersedteamsthan
had done. face-to-face
teams.
In relationshipsconductedface-to-face,it is a chal-
Differences in theSalience ofInformation. Teams also
lengingcognitiveexerciseto interpret a set of factsfrom
encounteredproblemsthathingedon differences in the
theperspectiveof anotherperson.It is farmoredifficult
salience of information among team members.Writers
to determinehow theinformation beforetheotherparty
tendedto assumethatwhatwas salientto themwouldbe
differsfromone's own, and then see thingsfromthe
salientto theirreaders.Scholarshave observedthatme-
other'sperspective.Geographicdispersionmakes these
diatedcommunication oftenlacks cues to meaningsuch
two activitiesmoredifficult because of undetected "leaks
as facial expressions,body language,and tone of voice
in thebucket,"because partnersseem to have difficulty
(Kiesler and Sproull 1992, Sproull and Kiesler 1986).
retaininginformation about remotelocations,and be-
However,thesetypesof cues also signalthesalience to
cause feedbackprocessesare laborious.In addition,the
thecommunicator of one piece of informationrelativeto
data suggestthatteammemberswithcompleteor correct
another.Dispersedteammemberswerenotsuccessfulin
information may not speak up when erroneousconclu-
communicating to theirpartnerswhatpartsof theirmes-
sions are voiced in theteam.
sages, or whichmessages,theyconsideredmostimpor-
Problemsstemmingfromunevenlydistributed infor-
tant.When an e-mail message addressedseveraltopics,
mationwere notlimitedto cases involvingerrorsin ad-
partnerssometimesdifferedon whichtopicstheyfound
dresses and undeliveredmail. Sometimespeople knew
salient.For example, as describedpreviously,a GMU
theywere exchangingmail withonly partof the team,
partnerwho wanteda change made in the team home
butfailedto understand howthisaffected theperspectives
page raisedtheissue on thetelephonewithTCU Partner
of teammemberswho did notreceivethemail,or how it
1 and in a postscriptto a four-paragraph e-mailto TCU
affectedthe dynamicsof the team as a whole. In Team
Partner2. The proposedchangewas thelast of threeis-
11,Lisa in Portugalsuggesteda focusfortheteam'sproj-
sues addressedin thee-mail.TCU Partner2 laterinsisted
ect. Team membersat GMU and TCU exchangede-mail
indignantlythat he had never heard of the request.
aboutLisa's idea withoutcopyingher.They agreedthat
Clearly,thepostscript had greatersalience to the sender
heridea was creativeand interesting buttoo complexfor
thanto thereceiver.
the team's timeframe.When the team voted electroni-
Differencesin information salience were exacerbated
cally,Lisa's idea was notselected.FromLisa's perspec-
by unwieldyfeedbackprocessesin the dispersedteams
tive,her idea was metby silence.Therewas no discus-
and the makingof indirectrequests.Analyzinga tense
sion, praise, or criticism-just a vote. When team
exchangein Team 6, a GMU memberobserved:
membersbegan volunteering via e-mail to develop par-
ticularpartsof thebusinessplan,thePortuguesepartof Withso muchinformation goingback and forth, it was difficult
the team was silent.I suspectthatthe GMU and TCU formyteammatesto absorbeverydetail ... Because I couldn't
"see" if thereceiverwas payingattention, I didn'tknow if my
memberswere moreconscious of the silence fromPor-
message had to be repeated.Yet it is time-consuming to let the
tugalthantheirown silencein responseto Lisa' s idea. E-
senderknowmyperceptionof theirmessage.
mailconcerning Lisa's idea was exchanged.Itjustwasn't
sentto Lisa. In othercases, writersdid not use in theire-mailkey
It is not news thatprivateconversationscan create wordsthattheythought theyhad used. In Team 18,Anna
problemsin a team.However,the dynamicsand conse- wantedto clear up irritationthathad resultedwhen she
quences of this behaviorin dispersedteams are worth and George disappeared for threedays duringGMU's
noting.Privateexchangesof e-maildistortperceptions of springbreak. In her analysis paper, she wrote,"When
thevolumeof activityin a team.This can confusemem- everything done, I thoughtit was timeto clear our
was
bers' sense of pace and timing.Memberswho are receiv- misunderstandings.I didn'twantto brushthingsaway. I
ing all thee-mailwill perceivesome membersas active sentan e-mailsayingthatwe need to have a chatas there
and othersas relativelyinactive.Those who are not re- were some misunderstandings to be cleared." The chat
ceivingall themail will perceivetheenergylevel of the nevertook place. One reasonmay be thatAnna did not
actuallyuse the words"clear our misunderstandings" in and McGrath1994, Waltherand Burgoon 1992). How-
here-mail.Instead,she wrote,"Thereseemsto be a com- ever, this studysurfaceda second typeof problemin-
municationslapse betweenus. Georgeand I thoughtthat volvingspeed: relativedifferences amongteammembers
we probablyneed to discuss certainaspects fromthe in speed of access to information. One manifestation of
home page. Is it possible to chat today?. .. This is im- this problem stemmed from differences among team
portant,so please let us know soon." It is unlikelythat membersin access to communications technology. Some
Anna's wish to resolvetheissue of whyshe and George membershad 24-houre-mailaccess whiletheirpartners
disappearedwas clear to her dispersedteammatesbe- had access onlywhenat theiruniversity. If some mem-
cause she said thatshe wantedto "discusscertainaspects berssee e-mailonlyonce a day or once everyfewdays,
fromthehomepage." Confusiondue to indirectwording thislimitsthe interaction thatis possible and slows the
is notconfinedto computer-mediated communication, as pace of theteam.Observeda memberof Team 3, "Some
Tannen(1994) has shown.But thecharacteristics ofsome problemsdraggedon for days while the suspicionsof
of the communicationtechnologiesused by dispersed groupmembersintensified. In reality,theproblemcould
groupsprobablymake it difficult formembersto recog- have been as simpleas someonenotbeingable to getto
nize themeaningand importance of indirectrequestslike thecomputerlab to checktheirmessages."
thisone. A second manifestation seemed to stemfromdiffer-
In addition,therewas a tendencyto requestfeedback ences in the speed of electronictransmissionsamong
fromthe team indirectly, yet to expect quick responses partsof a team.This was exemplified by therelationship
fromeveryone.(Kiesler et al. 1984 predictedthelatter.) amongmembersofTeam 5, whichincludedtwomembers
"EverytimeI sent an e-mail requesting"any thoughts" in Australiaand fourmembersin theUnitedStates,two
fromeveryone,I expectedto receiveone fromeveryone. of themin Virginiaand two in Texas. The teamheldfive
And when I didn't,I feltthatthose who didn'trespond onlinechatsduringwhichtensionsbetweenthemembers
were notholdingup theirend of thebargain,"reporteda in theUnitedStatesand Australiawereevident.Near the
memberof Team 5. A memberof Team 17 observed, end of the team's fifth chat,an Americanteammember
"People alwayssaid, 'Hope to hearfromyou soon.' Who observedthattheAustralianmembersalways seemedto
thenhas responsibility forinitiating communication?" A be "25 minutesbehindthediscussion"and suggestedthat
memberof Team 11 recalled sendinge-mail "into the thiscould be an artifact of thespeed of transmission be-
abyss," and GMU-based membersof Team 6 discussed tween the continents.The team membersat the two
thefeedbackproblemamongthemselves,butneverwith UnitedStateslocationscould carryon a relativelyrapid
othermembersof theirteam. In an analysispaper,one exchangeuntilbeing "interrupted" by teammembersin
wrote: Australiawho referred to subjectsfromwhichtheothers
We wantedacknowledgment of thetimewe spenton thedeliv-
had movedon.
erable as well as a feelingthatwe were on target.No one re- I was not able to verifythishypothesisthroughchat
sponded. We sent anothere-mail sayingthatwe hadn't heard room transcripts or othermeans.However,theexplana-
fromanyone. Finally,we heard fromone group member,but tionproposedbytheteammemberis credible.Telephone
even thatmessage containedminimalinformation. lines carriedmost of the Internettrafficbetween the
UnitedStatesand Australiaand theyfrequently became
Clearly,the salience of the requestforfeedbackwas
overloaded,resultingin breakdownsand timelags. This
higherforsendersthanreceivers.Thus, whenelectronic
would mean thatpartsof the teamwerecommunicating
communication is voluminous,sendersand receiversun-
at different rates-one ratebetweenthe two sitesin the
wittingly may differin whattheyfindmostsalientand
UnitedStatesand anotherratebetweentheUnitedStates
fail to fulfilltheirdistantpartners'expectations.This
andAustralia.This is a recipeforfrustration andirritation
problemmay be complicatedby a tendencyto statere-
forall. IfthemembersinAustraliarespondedtomessages
quests indirectly,yet expect quick responsesfromall
theinstanttheyreceivedthem,theirresponsesstillwould
members.In general,thelevel of feedbackamongmem-
appearin thechatroomtraffic well aftertheconversation
bers of a dispersedgroupis not likelyto be as high as
betweentheUnitedStatespartners had movedon because
memberswould wish,and maynotbe sufficient to ensure
of timelags comingand going.Moreover,fromtheper-
sharedunderstanding.
spectiveof theAustralianpartners, a streamof unrelated
RelativeDifferences in Speed ofAccess toInformation. commentsby the United States partnerswould always
Researchhas shownthatteamsusingcomputer-mediated follow theirmessages. It would appear thattheircom-
communication operateat a slowerratethanteamsmeet- mentswereignored.
ing face-to-face(Lebie et al. 1996, Straus 1997, Straus Both types of problems concerningrelative speed
Unevenlydistributedinformation
Propositions:Dispersed collaboratorsfailmoreoftenthantheyrealize to distributethe same information
to all members.Causes include
humanand technologicalerror,and selective distribution
withoutawareness of all itsconsequences.
Uneven distribution of information
resultsin team members havingdifferent perspectives because of the different
information
theyhave.
Partnerswithcomplete or correctinformation may notspeak up when erroneousconclusions are voiced in the team. Impressionscreated
on the basis of unevenlydistributedinformation oftenpersistin the face of correctinginformation.
Uneven distributionof information
distorts
perceptionsofthe volume of activityin a team, and confuses the team's pacing and timing.
Type of problem
Serious problem 54 69 23 23 46 69
Some problem 23 23 46 38 54 23
Not a problem 23 8 31 38 0 8
performance (i.e., gradesreceived),theincidenceof hos- less than half the numberthat many otherteams ex-
tile coalitions,and the numberof e-mails and chats changed.The averagegradeitreceivedon itsprojectwas
logged. I also looked to see whetherteams could be second highestamong the thirteenteams studied.Indi-
groupedby theirtaskand relationship management strat- vidual introductionswere briefand businesslike,butthe
egies. The data suggestfourconstellations: (1) good per- tone of communication was graciousoverall.One team
formance,task focus, moderaterelationshipdemands, memberrepeatedlymodeledthisbehavior,whilestilltak-
relativelylow volumeof communication, and low coali- ing a strongstandon a few issues. The team's commu-
tion activity;(2) good performance, hightask and rela- nication is peppered with commentssuch as "Good
tionshipdemands,relatively highvolumeofcommunication, ideas!," "Thanksforthevoteof confidence," "Greatpic,"
and highcoalitionactivity;(3) weakerperformance, rela- and "What a talentedgroup!"The groupencountereda
tivelyhighvolumeof communication, manyand diverse seriouschallengeearlyon because of errorsin e-mailad-
information problems,and highcoalitionactivity;and (4) dressesinvolvingboththeGMU and TCU partners. The
weakerperformance, relationshipfocus,tasksecondary, upshotwas thatfourof the six team membersvoted to
relativelyhighvolume of communication, and low coa- accept a projectidea thatthe two GMU membershad
litionactivity.Therewereno obviousdirectrelationships nevereven received,whilea projectproposalfromGMU
in thissmall samplebetweentheincidenceand severity bouncedback undeliveredfromTCU. Both GMU mem-
of particularinformation problemsand performance. In- bersnotedin theiranalysispapersthattheywereirritated
formation problemsseemed to be moredamagingto re- by thisevent;however,one of themadded,"I do under-
lationshipsthanto task performance. It also seemed to standwhythisoccurred.An initialphase of confusionis
matter how membersdealtwiththeinformation problems to be expectedin virtualteamsthatrelyprimarily on e-
thatoccurred:theextentto whichtheyextendedtheben- mail forcommunication." In otherwords,he made a sit-
efitof thedoubtwhenremotepartnersdid notbehaveas uationalratherthanpersonalor categoricalattribution.
When new challengesinvolvingdifferences in context,
expected,and theextentto whichtheywereable to iden-
tifyaspectsof thesituationthathelpedexplainbehavior constraints,and ideas arose fortheteam,theywerehan-
to expectations. dled in sophisticatedways. Confronting disagreements
thatwas contrary
over elementsof the businessplan, a team memberin
Thus,teamswithan averagegradeof 90% and above
Canada wrote, "Greetings, friends. We seem to
seem to constitute two types:thosethatexchangedrela-
be at odds overwhatour companyshouldbe doing. . ."
tivelylittlee-mail and did notdevelop coalitionactivity,
In the otherthreeteamsthatearneda gradeof 90 or
and thosethatexchangedmoree-mailand developedco-
above (Teams 15, 11, and 17) some memberswere de-
alitionactivity.Teams 26, 30, and 21 fitthefirstcategory
terminedto turnin a qualityproductand managedto do
and Teams 15, 11, and 17 fitthesecond.The threeteams
so. However,lifeinsidetheteamwas turbulent. Attempts
thatearneda grade of 90 or above and avoided the de-
were made to extendthe benefitof the doubtto remote
velopmentof hostilecoalitionsemphasizedminimalism partners;however,attributions grewharsh.For example,
and closure in theirworkprocessesand tendedto give a memberof Team 11 eventuallywrote,"We gota cheap
remotepartnersthebenefitof thedoubt.In otherwords, excuse (fromthePortuguesepartners) thatI did notbuy."
theykepttask and social demandsrelativelylow. They Ultimately, partnersat two of the threelocationsbegan
exchangedsome personalinformation at the outset,but planningto do theprojectas a groupoffouranddiscussed
otherwisetendedto focuson thetask at hand. The tone whetherto withholdthenamesoftheout-group members
was businesslikebutgracious.Throughskillorluck,they fromthefinishedproduct.It maybe thattheapproachto
agreedquicklyon a viable idea fora businessand then thetasktheseteamstookwas relativelydemanding,and
focusedon implementing the idea. Givingremotepart- it overwhelmedtheteam's social capacityand available
ners the benefitof the doubtwas important, as each of communication media.For example,theseteamsdebated
these teams did encountersignificantinformation ex- longeron the kind of businesstheywould design than
changeproblemsearlyin theirlivesthateasilycouldhave did Teams 26, 30, and 17. When information problems
escalated. However, members'interpretations of these occurred,theywere difficult to resolve and did consid-
events,attributions made aboutremotepartners, and con- erabledamage.
sequentbehaviorsdid notacceleratethedevelopmentof The same problemsand processesappearedin a more
hostilerelationships in theteam.Theystruggled to main- virulentformin Teams 5, 3, and 39, whose gradesaver-
taina gracioustonedespite unanswered questions,seem- aged below 90 percentand who had a relativelyhigh
ing slights,and frustration. volume of communication, frequentand diverseinfor-
For example,Team 26 exchangedonly 65 messages, mationproblems,and fiercecoalitionactivity.Almostall
Figure 2 The Mutual Knowledge Problem and Its Consequences for Dispersed Collaboration
factors:
Structural Processes:
Group including
structure, Themutual
dispersion
geographic knowledge
Useof [2] problem
Available
communication communication (points1-5)
media ia
med
Taskcharacte/stics [3] shared?
Context
Taskcharacteristics / exchange
Information Uneven
distribution?
(interdependence, [1]
load,
information Degreeof Feedback
time
pressure) *
integration
required Lags? [4]
required?
Effort Meaningofsilence?
task?
Interdependent Relative speeds?
cycle Interpretation Salience
shared?
information?
task-related
Dispersed [5] ofmessage
orlocaldifferences?
Contextual [
Member differences?
[7] Causal[6]
Cohesion attnbution
Outcomes: Modifynorms?
Coalition
activity? Personal?
Viability Disengagement? SPtuational?
SCtategoncal?
Categorical?
Performance
andits
consequences
(points6-7)
theeffortrequiredto seek and give feedbackwhendis- by all parties(Point [4]) and led to negativeattributions
persed,andrelativedifferences thatmayexistinfeedback to individualsand subgroups(Point [6]) and thedisinte-
speeds amongpartsof thegroup(Point [5]). Slow feed- grationof relationships (Point[7]). Sometimes,problems
back cycles (Point [5]) reduce correctivefeedbackand weremitigatedwhenerrorswerecaughtin thecourseof
increasethelikelihoodoferroneousinterpretations (Point a feedbackcycle (Point [5]). However, therewere in-
[4]) and exaggeratedattributions(Point[6]). stancesin whichthiscorrectionwas so slow in coming
Finally,theremay be a tendencyto generalizesuch thatthoseinvolvedwere unable to traceand modifyall
social perceptions,particularlynegativeones, to the lo- thefaultyconclusionstheyhad drawn.The cycledepicted
cationalsubgroupof whicha personis a member,which in Figure2 can be self-reinforcing: The problemsof in-
sets in motionin-group/out-group dynamics(Point [7]) formationexchange,interpretation, de-
and attribution
thatare destructive to groupcohesion.In some cases that scribedhere,and theirdisintegrative effecton teamre-
I studied,collocatedpartnersreliedincreasingly on each lationships,add to the already substantialintegration
other,criticizingtheirremotepartners amongthemselves challengeconfronted bya geographically dispersedgroup
and sometimesdisengagingfromthe group's work.In (Point[1]).
othercases, subgroupsin twolocationsexchangedcritical
e-mailaboutthethirdsubgroupand refusedto sendthem
work products.Typically,an important piece of infor- Discussion
mationwas not sent,sentto the wrongplace, or lost in This paperproposesthata centralproblemof geograph-
transit(Point [3]), whichaffectedtheconclusionsdrawn ically dispersed collaborationis maintainingmutual
as demonstrated work ethic,fairdealings,and consis- McGrath 1994, Walther 1992, Waltherand Burgoon
tency"(Jarvenpaaet al. 1998,p. 3 1). The authorspropose 1992). My findingcalls attention to theconsequencesof
thattrusting actionand demonstrated reliability
increase partsof a groupcommunicating at differentrates.Future
trustin dispersedteams.However,myworksuggeststhat researchshouldexplorewhetherunevenfeedbackcycles
human and technicalerrorsin information distribution withina grouphave a different impactthanuniform feed-
may be commonin dispersedcollaboration,particularly back cycles. Uneven feedbackcycles across partsof a
duringtheearlyphasesof activity.If theseareinterpreted groupcould be moredestructive thana uniformly slow
as failuresofpersonalreliability,
theyare likelyto inhibit pace because subgroupsgrowout of syncwith,and iso-
thedevelopmentand maintenanceof trust. latedfrom,thegroup.This could resultin theirbecoming
scapegoats. Ironically,feedbackcycles may be slower
Failures of Interpretation and moreunevenamongpartsof a groupunderjust those
This studyalso identifiedthreeproblemsthatdisrupted conditionsforwhichrapidcyclesare mostneeded:when
shared interpretation of informationin the dispersed thecontextsof sendersand receiversdiffersubstantially.
teams: difficulty communicating and understanding the Forexample,feedbackcyclesmaybe unpredictable when
salienceof information, differences in speed of access to partof a teamis travelingconstantly or locatedin an area
information, and difficulty interpreting the meaningof witha weak communications infrastructure.
silence. When problemsof salience occurred,partners Finally,I observedthatteammembersoftenmisinter-
had the same information but attendedto different parts pretedthe meaningof theirremotepartners'silence.
of it and misunderstood each otheras a result.Whilethis Physicaldispersionand dependenceon communications
problemis not unique to dispersedteams,it probablyis technologyadd sourcesof uncertainty aboutthemeaning
exacerbatedby theuse of computer-mediated communi- of silencebeyondthoseexperiencedby groupsthatmeet
cation.The mediumdoes notprovidetheparaverbaland Partnersmayfall silentbecause theyfindit
face-to-face.
verbalcues thatpeople use in conversationto signalthe
difficult and time-consuming to convey sensitiveissues
importanceof one piece of information relativeto an-
in text,or because of technicalfailures.They may be
other.Furthermore, failureto communicatesaliencemay
silentbecause theyagree,because theydisagree,or be-
be morecostlyin dispersedthancollocatedcollaborations
cause theyarephysicallyabsent.In themutualknowledge
because of slow feedbackchannelsand restricted back-
literature, Brennan(1998) describeshowlack offeedback
channelfeedback.Slow and effortful feedbacklimitsde-
(i.e., silence) leads to failuresof groundingin conversa-
tectionand correctionof misunderstandings. Kieslerand
tionswithand through computers. Withoutfeedback,one
Sproull(1992, Sproulland Kiesler 1986) describedhow
does notknowwhethera computeris working,has com-
computermediationrestricts cues to themeaningofcom-
pletedthetask,has malfunctioned, or is waitingforad-
munication.I focuson theproblemof signalingtheim-
portanceof one piece of information relativeto another
I
ditional inputs. broaden her point by showing how
and how thisproblemis manifested in dispersedcollabo- physical dispersion presents additional sourcesof uncer-
rationand exacerbatedby the otherproblemsidentified of
taintyas to themeaning partners' silence.
in thisstudy.
The membersof thedispersedteamsI studiedalso had ConsequencesforAttribution
difficultyworkingtogetherwhenthe speed of feedback The failuresof information exchangeand interpretation
cyclesdiffered amongpartsof thegroup.Some members identifiedin thisstudyhave consequencesforattribution
were in frequentcontact with rapid feedback cycles, processes.They illuminatetwo reasonswhypeople are
while contactwithotherswas limitedand slowerpaced. likelyto makepersonalratherthansituational attributions
These differences were caused by differences in access concerningremotepartners. First,failureto shareand re-
to communicationtechnologyand the distributionof memberinformation about remotesituationsand con-
membersacross distanceand time.The teamswere not texts,and unevendistribution of information, mean that
synchronized in theiraccess to information and abilityto remotepartnersoftenlack information to make situa-
detect and correct misunderstandings, so had a dif-
and tionalattributions.Accordingto theattribution literature,
ficulttimemaintaining mutualknowledge. when people do not have situationalinformation, they
This observation revealsa new side of theissue ofrate tendto makepersonalattributions, i.e., theirexplanations
in computer-mediated communication. Researchershave focuson thedispositionsof individuals(Jonesand Nisbett
shown thatthe rate at which computer-mediated com- 1972,Nisbettet al. 1973).
municationproceedsaffectsgroupproductivity the and The studyalso demonstrates the complexityof com-
developmentof relationships(Straus 1997, Straus and municatingand collaboratingacross distance and via
Amplifying
and ModeratingForces practicesthatshouldmoderateproblemsincludemethod-
Feedback lags seem to amplifythe problemsof infor- ically seekingout situationaland uniquelyheld infor-
mation exchange and interpretation identifiedby this mation,givingpromptfeedbackwheneverpossible,fo-
study.Krauss and Bricker(1966) demonstrate thatfeed- cusing on the overall structureand processes of the
back lags disruptthe abilityof sendersand receiversto systemof relationships ratherthanon individuals,reex-
establishcommonreferents, a buildingblock of mutual amininggroup operatingpracticesand norms,and ex-
knowledge.In addition,thisstudysuggeststhatfeedback tendingthebenefitof the doubtratherthanengagingin
lags contributeto the exaggerationof negativeattribu- the creationof out-groups.The overall effectof these
tionsconcerningremotepartnersand make it moredif- practicesis to directattentionto group-leveldiagnosis
ficultfordispersedcollaboratorsto diagnosetheirsitua- and learning.
tion.
Withoutfeedback,deprivedcollaboratorsare leftto Limitations
speculatewhytheirexpectationshave notbeen fulfilled The mode of generalization to case studyre-
appropriate
and when feedbackwill come. In the absence of situa- searchis analyticgeneralization-generalization to the-
tionalinformation, theyare likelyto makenegativeattri- ory ratherthan statisticalgeneralization(Yin 1994).
butionsconcerningthedispositionsof theirremotepart- Therefore,itis important howtheteamsstud-
to articulate
ners. These attributionscan grow more negative as ied here may be typicaland atypicalof geographically
waitingcontinues.Deprivedcollaborators also sometimes dispersedwork groups.Geographicallydispersedwork
amplifytheirdemands, triggeringan exaggeratedre- groupstake manydifferent formsin practice(Goodman
sponse fromtheirremotepartners.The situationis like and Wilson 1998, Leonard et al. 1998, Maznevski and
thatof thepersonwho ultimately turnsthehotwaterup Chudoba2000, Mazchrzaket al. 2000, Snow et al. 1996).
too highin theshowerbecause oftimelags betweenturns The teamsI studiedprobablyare atypicalin the limita-
of thefaucetand response.One participant tionstheyfacedaroundmeansofcommunication. Travel,
said as much:
videoconferencing, and telephoneconferencing werenot
"Some problemsdraggedon fordayswhilethesuspicions
an optionforthem,and theywere limitedby personal
of group membersintensified. In reality,the problem
expensein theiruse of thetelephone.Occasional face-to-
could have been as simpleas someonenotbeingable to
face meetingsand moretelephonecontactmightmoder-
get to the computerlab to check theirmessages." The
ate the processesobserved;however,thereis reason to
speed of feedbackcycles may constitutea criticalcon-
thinkthatbasic tendenciesmightbe thesame.Additional
straintforgeographically dispersedgroups.
modesof contactcould contribute to unevenexchangeof
In addition,feedbacklags and dispersedinformation
information amongpartsof a teamifused extensively by
make it extremely difficult
forpeople to getan overview
dyadsor subgroups.
of thestructure and functioning of a dispersedsystemof
Severalteamdesignfactorsshouldalso be noted:group
relationships. Actionsand reactionsare difficult to inter- identityand timeframe(discussedin Walther1997), in-
pret when disruptedby lags in feedback.Blaming is a terdependence, and composition.The local university-
common response when individualsdo not grasp the based subgroupsoftheteamsstudiedprobablyhad a con-
structureand dynamicsofcomplexsystemsofwhichthey siderablystrongerbasis for identitythan the teams as
are a part(Bowen 1985, Senge 1990). In futureresearch, wholes.Whiledispersedteamsin practiceare oftencom-
systemsdynamicstheory(Sterman1989) mightcontrib- posed of people frommultipleorganizationsand sub-
ute to our understanding of theimpactsof distributed in- groupswith strongidentities,the weak basis for team
formationand feedbacklags on dispersedcollaboration identityshould be taken into account. Likewise, the
and computer-mediated communication. teams' seven-weektimeframeis notunusualin business
In both dispersedand collocatedcollaboration,prob- practice;however,its relativeshortnessand the teams'
lems establishingand maintaining mutualknowledgeare low expectation offutureinteractionshouldalso be noted.
mostlikelyto occurwhenthereis a greatdeal ofuniquely As discussedpreviously,theteam's level of interdepen-
held task-relatedand contextualinformation and limited dence could be characterizedas moderate.They carried
communication channels.Exacerbatingfactorscan be ex- anda range
outa projectthatrequiredresearch,creativity,
pectedto includeheavycognitiveload, a complexinter- of skillswithoutcomesof significance forthemembers.
dependenttask, tighttime limits,and a complex team The teamswere composedof adultprofessionalsrepre-
design-particularlyone involvingstrongsubgroupiden- sentinga rangeof ages withmoderateinternational and
tities,whichmayreinforce local perspectives. technicalexperience.We mustcontinueto examinethe
For situationsin which these factorsare operating, formsthatdispersedcollaborationtakes in practiceand
Fishbein,eds. CurrentStudiesin Social Psychology.Holt,Rine- Lipnack,J.,J. Stamps. 1997. VirtualTeams: Reachingacross Space,
hartand Winston,New York. Timeand OrganizationswithTechnology.Wiley,New York.
Jarvenpaa,S., K. Knoll, D. Leidner. 1998. Is anybodyout there?An- Majchrzak,A., R. Rice, N. King,A. Malhotra,S. Ba. 2000. Computer-
tecedentsof trustin global virtualteams.J.ManagementInform. mediatedinter-organizational knowledge-sharing:Insightsfroma
Systems14(4) 29-64. virtualteam innovatingusing a collaborativetool. Inform.Re-
, D. Leidner. 1999. Communicationand trustin global virtual sourcesManagementJ. 13(Jan-Feb)44-53.
teams.Organ. Sci. 10(6) 791-815. Maznevski,M., K. Chudoba. 2000. Bridgingspace over time:Global
Jones,E., R. Nisbett.1972. The actorand theobserver:Divergentper- virtualteamdynamicsand effectiveness. Organ. Sci. 11(5) 473-
ceptionsof thecauses of behavior.E. Jones,D. Kanouse, H. Kel- 492.
ley,R. Nisbett,S. Valins,B. Weiner,eds. Attribution:Perceiving McDonald, D. 1995. Fixing blame in n-personattributions: A social
theCauses ofBehavior.GeneralLearningPress,Morristown, NJ identitymodel forattributional processesin newlyformedcross-
79-94. functionalgroups.M. Martinko,ed. Attribution Theory:An Or-
Karakowsky,L., J. Siegel. 1995. The effectof demographicdiversity ganizationalPerspective.St. Lucie Press,DelrayBeach, FL 273-
on causal attributionsof workgroupsuccess and failure:A frame- 288.
workforresearch.M. Martinko,ed. Attribution Theory:An Or- McGrath,J., A. Hollingshead. 1994. Groups InteractingwithTech-
ganizationalPerspective.St. Lucie Press, Delray Beach, FL 77 nology.Sage, ThousandOaks, CA.
289-313. McGuire,T., S. Kiesler,J.Siegel. 1987. Groupand computer-mediated
Kelly, H., J. Michela. 1980. Attribution theoryand research.Annual discussioneffectsin riskdecisionmaking.J. Personalityand So-
Rev. Psych.31 457-501. cial Psych.52 917-930.
Kemske,F. 1998. HR 2008. Workforce (January)47-60. McLeod, P. 1996. New communicationtechnologies for group
Kiesler,S., J. Siegel, T. McGuire. 1984. Social psychologicalaspects decision-making: Towardan integrative framework. R. Hirokawa,
of computer-mediatedcommunication.Amer. Psych. 39(10) M. Poole, eds. Communication and GroupDecision Making,2nd
1123-1134. ed. Sage, ThousandOaks, CA 426-461.
L. Sproull. 1992. Group decision makingand communication Miles, M., A. Huberman.1994. QualitativeData Analysis,2nd ed.
technology.Organ. Behaviorand HumanDecision Processes 52 Sage, ThousandOaks, CA.
96-123. Nisbett,R., C. Caputo,P. Legant,J.Marecek. 1973. Behavioras seen
, D. Zubrow, A. Moses, V. Geller. 1985. Affectin computer- by theactorand as seen by theobserver.J.Personalityand Social
mediated communication: An experiment in synchronous Psych.27(2) 154-164.
terminal-to-terminal discussion. Human Comput.Interaction1 O'Hara-Devereaux,M., R. Johansen.1994. Globalwork:BridgingDis-
77-104. tance,Cultureand Time.JosseyBass, San Francisco,CA.
Krauss,R., P. Bricker.1966. Effectsof transmission delay and access Redding,W. 1972. Communication WithintheOrganization.Industrial
delay on the efficiencyof verbal communication.J. Acoustical CommunicationCouncil,New York.
Rommetveit,R. 1968. Words,Meanings and Messages. Academic
Society41 286-292.
Press,New York.
, S. Fussell. 1990. Mutual knowledgeand communicativeeffec-
Senge, P. 1990. The FifthDiscipline: The Art and Practice of the
tiveness.J.Galegher,R. Kraut,C. Egido, Eds. IntellectualTeam-
LearningOrganization.Doubleday/Currency, New York.
work: Social and Technological Foundations of Cooperative
Siegel, J.,V. Dubrovsky,S. Kiesler,T. McGuire. 1986. Groupprocess
Work.LawrenceErlbaum,Hillsdale,NJ 111-146.
in computer-mediated communication. Organ.Behaviorand Hu-
S. Lewis, L. Swezey. 1982. Listenerresponsivenessand theco-
man Decision Processes 37 157-187.
ordinationof conversation.J. Personalityand Soc. Psych.43(4)
Snow, C., S. Snell, S. Davison, D. Hambrick.1996. Use transnational
718-731.
teamsto globalize yourcompany.Organ. Dynam.24(4) 50-76.
Lea, M., R. Spears. 1991. Computer-mediated communication,de-
Sproull,L., S. Kiesler. 1986. Reducingsocial contextcues: Electronic
individuationand group decision-making.Internat.J. Man-
mail in organizationalcommunication. ManagementSci. 32(1 1)
Machine Stud.34 283-301.
1492-1512.
1992. Paralanguageand social perceptionin computer-
Stasser,G., D. Stewart.1992. The discoveryof hiddenprofilesby
mediatedcommunication. J. Organ. Comput.2 321-341. decision-makinggroups: Solving a problem versus making a
1993. Love at firstbyte?Buildingpersonalrelationships judgment.J. Personalityand Social Psych.63 426-434.
over computernetworks.J. Wood, S. Duck, eds. Under-studied G. Wittenbaum.1995. Expertroles and information ex-
Relationships.Sage, ThousandOaks, CA 197-233. change during discussion: The importanceof knowing who
Lebie, L., J. Rhoades, J. McGrath. 1996. Interactionprocess in knowswhat.J. Experiment. Social Psych.31 244-265.
computer-mediated and face-to-face
groups.ComputerSupported , W. Titus. 1985. Pooling of unsharedinformation in groupde-
CooperativeWork4 127-152. cisionmaking:Biased information samplingduringdiscussion.J.
Leonard, D., P. Brands, A. Edmondson,J. Fenwick. 1998. Virtual Personalityand Social Psych.48 1467-1478.
teams: Using communicationstechnologyto manage geographi- 1987. Effectsof informationload and percentagesof
cally disperseddevelopmentgroups.S. Bradley,R. Nolan, eds. sharedinformation on thedissemination of unsharedinformation
Sense and Respond: CapturingValue in the NetworkEra. Har- duringgroupdiscussion.J. Personalityand Social Psych.53 81-
vardBusiness School Press,Cambridge,MA 285-298. 93.
371
ORGANIZATIONSCIENCE/Vol. 12, No. 3, May-June2001