Information Technology and Organizational Change: Causal Structure in Theory and Research

Author(s): M. Lynne Markus and Daniel Robey

Source: Management Science, Vol. 34, No. 5 (May, 1988), pp. 583-598
Published by: INFORMS
Stable URL: http://www.jstor.org/stable/2632080
Accessed: 14-09-2015 14:41 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

INFORMS is collaborating with JSTOR to digitize, preserve and extend access to Management Science.

http://www.jstor.org

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

MANAGEMENT SCIENCE
Vol. 34, No. 5, May 1988

Printedin U.S.A.

INFORMATION TECHNOLOGY AND ORGANIZATIONAL

CHANGE: CAUSAL STRUCTURE IN THEORY
AND RESEARCH*
M. LYNNE MARKUS

AND

DANIEL ROBEY

GraduateSchool of Management,Universityof California,Los Angeles, California90024

College of Business Administration,Florida InternationalUniversity,Miami, Florida 33199
This articleconcernstheoriesaboutwhy and how informationtechnologyaffectsorganizational life. Good theoryguides research,which, when applied,increasesthe likelihoodthat
informationtechnologywill be employedwithdesirableconsequencesfor users,organizations,
and otherinterestedparties.
Butwhatis a goodtheory?Theoriesareoftenevaluatedin termsof theircontent-the specific
conceptsused and the human valuesserved.This articleexaminestheoriesin termsof their
structures-theorists'assumptionsabout the natureand directionof causalinfluence.Three
dimensionsof causalstructureare considered-causal agency,logicalstructure,and level of
analysis.Causalagencyrefersto beliefsaboutthe natureof causality:whetherexternalforces
causechange,whetherpeopleact purposefullyto accomplishintendedobjectives,or whether
changesemergeunpredictablyfrom the interactionof people and events. Logicalstructure
refersto the temporalaspectof theory-static versusdynamic-and to the logicalrelationships
betweenthe "causes"and the outcomes.Levelof analysisrefersto the entitiesaboutwhichthe
theoryposesconceptsand relationships-individuals,groups,organizations,and society.
While there are many possiblestructuresfor good theory about the role of information
technologyin organizationalchange,only a few of these structurescan be seen in current
theorizing.Increasedawarenessof the options,open discussionof theiradvantagesand disadvantages,and explicitcharacterization
of futuretheoreticalstatementsin termsof the dimensions and categoriesdiscussedhere should, we believe, promotethe developmentof better
theory.
(INFORMATIONTECHNOLOGY;ORGANIZATIONCHANGE; CAUSAL STRUCTURE)

Introduction
The relationshipbetween informationtechnology and organizationalchange is a
centralconcernin the field of InformationSystems(IS). In the 30 yearssince Leavitt
and Whisler's(1958) seminalarticle,"Managementin the 1980's,"speculationson the
role of informationtechnologyin organizationsand its implicationsfor organizational
designhave flourished.Few researchersin the IS field question the importanceof the
issue.In an empiricalinvestigationof literaturecitationpatterns,Culnan( 1986)traced
the origins of the IS field to Leavitt and Whisler'sarticle and identified"computer
impacts"as a clearsubfieldwithin it.
Unfortunately,the literatureon informationtechnologyand organizationalchange
does not currentlysupport reliablegeneralizationsabout the relationshipsbetween
informationtechnologyand organizationalchange.There are severalreasonsfor this.
The literaturecontains works by researchersfrom several academic disciplinesand
interdisciplinary
specialties,includingorganizationaltheory,managementscience,sociology,and computerscience,eachwithits own preferredconceptsand theoreticaland
methodologicalbiases.It includesconflictingand uncleardefinitionsand measuresof
informationtechnology (Bakopoulos 1985) and organizationalstructure(Fry 1982).
Finally, it mixes and crosses units and levels of analysis from the individual, the
* Acceptedby Arie Y. Lewin,formerDepartmentalEditor;receivedJune 23, 1986.This paperhas been
withthe authors51/2monthsfor 2 revisions.
583
0025- 1909/88/3405/583$0 1.25
Copyright ?) 1988, The Institute of' Management Sciences

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

584

M. LYNNE MARKUS AND DANIEL ROBEY

workgroup,the department,the organization,and society-a practice which leads

some observersto fear improperly specified models and ungeneralizablefindings
(Freeman 1978, Rousseau 1985).

One importantapproachto solving these problemsis to focus on the substanceof

theory, such as concept definition and normative orientation.For example, Kling
(1980) has identifiedand defined six theoreticalperspectivesin social analysesand
empiricalresearchon computing. These perspectivesare distinguishablefrom each
other by their respectivedefinitionsof technologyand social setting,theoreticalconstructs,beliefs about the dynamicsof technicaldiffusion,evaluationsof "good"technology, and ideologiesof the workplace.Kling and Scacchi(1982) have discussedthe
differingassumptionsabouttechnology,infrastructure,and the dynamicsof changein
versus"web"models.
"discrete-entity"
Our approachdiffersfrom but complementsKling'sanalysesof theory substance.
Instead,we focus on the structureof theory, that is, researchers'conceptionsof the
natureand directionof causality.We believethat soundtheoreticalstructure,like good
theoreticalsubstance,is necessaryfor better theory. Our purposein this paper is to
analyzethe causalstructureof the theoreticalmodels found in the literatureon information technologyand organizationalchange.
The causal structure of theoretical models comprises three dimensions: causal
agency,logicalstructure,and level of analysis.Causalagencyrefersto beliefsaboutthe
nature of causality:whetherexternalforces cause change (the technologicalimperative), whetherpeopleact purposefullyto accomplishintendedobjectives(the organizationalimperative)or whetherchangeemergesfromthe interactionof peopleand events
(the emergentperspective).Logicalstructurerefersto the time span of theory (static
versusdynamic)and to the hypothesizedrelationshipsbetweenantecedentsand outcomes:whethercausesare relatedto outcomesin an invariant,necessaryand sufficient
relationship(variancemodels), or in a recipe of sufficientconditions occurringover
time (processmodels). Level of analysisrefersto the entities about which the theory
posesconceptsand relationships-individuals,collectives,or both. Thesethreedimensions of causalstructureare shown in Figure 1.
Causalstructurecannoteasilybe separatedfrom issuesof theorysubstanceand from
variousmethodologicalissues. But a thoroughunderstandingof causal structurerequiresa depthof treatmentnot usuallyfoundin methodologicalcritiques(e.g.,Attewell
and Rule 1984;Rice 1980;Robey 1977).Consequently,our focus on causalstructure
will exclude several importantconcerns which must also figure prominentlyin the
developmentand testingof good theory.
CAUSAL
AGENCY

* Technological
Imperative

LOGICAL
STRUCTURE

* Variance
Theory

* Organizational
Imperative

LEVEL
OF
ANALYSIS

* Macro

* Micro

- micro

* Process
Theory

- mixed

* Emergent
Perspective
FIGURE 1.

D)imensions of Causal Structure.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

585

CausalAgency
Causalagencyrefersto the analyst'sbeliefsaboutthe identityof the causalagent,the
natureof causalaction and the directionof causalinfluenceamong the elementsin a
theory.Pfeffer(1982), for example,has identifiedthreeperspectiveson action in organizationaltheory. In the "situationalcontrol"perspective,externalfactorsor events
constrainor forcepeople and organizationsto behavein certainways. In the "rational
actor"perspective,people and organizationsevaluatealternativecoursesof action and
exercisefree rationalchoice. In the "emergent"perspectiveon action, the behaviorof
people and organizationsemerges from a dynamic interactionof external circumstancesand internalmotives or interests.'
Buildingupon the work of Pfeffer,we have identifiedthree conceptionsof causal
agencyin the literatureon informationtechnologyand organizationalchange.We label
these: the technologicalimperative,the organizationalimperativeand the emergent
perspective.In the technologicalimperative,informationtechnology is viewed as a
cause of organizationalchange. In the organizationalimperative,the motives and
actions of the designersof information technologies are a cause of organizational
change.In the emergentperspective,organizationalchangeemergesfrom an unpredictable interactionbetween informationtechnologyand its human and organizational
users. Each of these perspectivesis discussed more fully below and summarizedin
Figure2.
The Technological Imperative

The essenceof the technologicalimperativeis conveyedby the word "impact."This

perspectiveviewstechnologyas an exogenousforcewhich determinesor stronglyconstrainsthe behaviorof individualsand organizations.The technologicalimperativeis
consistentwith Pfeffer's(1982) situationalcontrol perspectiveon action in organizations. "In this view, action is seen not as the resultof conscious,foresightfulchoice but
as the resultof externalconstraints,demands,or forcesthat the social actor may have
little controlover or even cognizanceof" (Pfeffer1982, p. 8).
For example,Leavittand Whisler(1958) arguedthat informationtechnologywould
alterdramaticallythe shapeof organizationsand the natureof managerialjobs. Organizations would recentralize,levels of middle managementwould disappear,and a top
managementelite would emerge.Leavittand Whislerurgedmanagersto preparefor
theseinevitableimpactsby developingtheirinternaltechnologicalcapabilitiesand their
liaisonsto externaltechnologicalresources.
Simon (1977) was less pessimisticthan Leavittand Whislerin his predictionsabout
the impactof computers,but no less deterministic.Simon contendedthat computers
wouldnot changethe basichierarchicalnatureof organizations,but would recentralize
decisionmaking.Line organizationalstructureswould shrinkin size, and the number
of levelswoulddecrease.Staffdepartmentswouldincreasein numberand size, making
structuresmore complex and requiringmore lateralinteraction.
While the technologicalimperativehas a long historyand makessome compelling
claims, empiricalresearchhas generatedcontradictoryfindingson almost every dimension of hypothesizedcomputer impact (Robey 1977; Kling 1980; Attewell and
Rule 1984). Informationsystems have been found both to enrich and routinizejobs
'Slack (1984) makes similar distinctions among various conceptions of causality in an analysis of communication technologies. "Simple" causality views technology as an independent entity capable of effecting
change in social systems. "Symptomatic" causality also presumes technologies to be discrete phenomena, but
their effects may be mediated by social forces such as the intentions of rational actors. Slack views both simple
and symptomatic conceptions of causality as mechanistic in their basic assumptions. Her descriptions of
"expressive"and "structural"causality correspond more closely to Pfeffer'semergent perspective on action in
that the distinctions between cause and effect are not sharply drawn.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

586

M. LYNNE MARKUS AND DANIEL ROBEY

Technological

Imperative

Organizational
structure

Information
Technology

Imperative

Organizational

Purposes\

Organizational
Structure

Information

Emergent Perspective

USE
INFORMATIONTECHNOLOGY
MEANING
BEHAVIOR

/P/RPOSI7S

SZ'77/NC

Information
Technology
FIGURE2. Causal Agency.

(Kling 1978;Bj0rn-Andersen,Eason,and Robey 1986),both centralizeand decentralize authority(Klatzky1970;Whisler1970;Stewart1971;Blau, Falbe,McKinley,and

Tracy 1976;Carter1984;Fosterand Flynn 1984;Dawsonand McLaughlin1986),and
produce no changeswhere changeswere expected (Robey 1981; Franz, Robey, and
Koeblitz 1986).
Some investigatorshave proposedthat contingenciesaffectthe relationshipbetween
informationtechnology and structuralchange. For example, in a review of studies
conductedduringthe 1960s and early 1970s, Robey (1977) observedthat computing
appearedto supportan existingdecentralizedstructurein organizationswith uncertain
environments.However,in simpleenvironments,computingappearedto strengthena
centralizedauthoritystructure.Robey suggestedthat computingtechnologybe viewed
as a moderatingvariable,affectingthe strengthof a causal relationshipbetweenenvironmentaluncertaintyand organizationalstructure.
Othercontingencieshave also been examined.Leiferand McDonough(1985) controlledfortask routinenessand foundthat departmentsusinga computer-basedsystem
weremorecentralized,less complex,and perceivedless environmentaluncertaintythat
those not usingthe computer.Klatzky(1970) assertedthat size waspartiallyresponsible
for the decentralizationof decision makingthat accompaniedthe use of information
technology. Carter(1984) found that organizationalsize moderatedthe relationship

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

587

betweeninformationtechnologyand the structureof newspaperorganizations.Pfeffer

and Leblebici(1977) controlledfor both size and environmentalcomplexity in their
study of the use of informationtechnology in 38 small manufacturingcompanies.
Majchrzakand Mosher(1987) used threedifferentanalysisstrategiesto examine organizationalstructureas a contingencyin the technology/performance
relationshipat the
individuallevel of analysis.
The Organizational Imperative

Whereasthe technologicalimperativearguesthat informationtechnologyconstrains

or determineshuman and organizationalbehavior,the organizationalimperativeassumes almost unlimitedchoice over technologicaloptions and almost unlimitedcontrol over the consequences.The organizationalimperativecorrespondsto Pfeffer's
"intendedlyrational"perspectiveon action. It assumesthat behaviorsare chosen, that
suchchoicesoccuraccordingto a set of consistentpreferences,that choices occurprior
to the action itself,and that action is goal directed(Pfeffer1982, p. 6). This perspective
holds that humanactorsdesigninformationsystemsto satisfyorganizationalneeds for
information.2Thus, informationtechnologyis the dependentvariablein the organizational imperative,causedby the organization'sinformationprocessingneedsand manager'schoices about how to satisfythem.
In a widely known versionof this perspective,Galbraith(1977) proposeda number
of organizationaldesign alternativesby which organizationscan fill the information
processingneeds generatedby uncertainty.Managersmay reducethe need to process
informationby managingthe environment,by using slack resources,or by creating
self-containedorganizationalunits. Managersmay also increasethe capacityof organizationsto processinformationby developinglateralrelationsand buildinginformation
systems.
A similar causal argumentis evident in the work of Daft and MacIntosh(1978,
1981).They hypothesizedinformationneeds to varywith task varietyand knowledge
aboutthe task and proposeda relationshipbetweeninformationprocessingneeds and
use of informationsystems.Unanalyzable,nonroutinetasks requirerich information,
capableof conveyingcomplexand equivocalmeanings;facialexpressionsand voice are
the media most capableof processingthis rich qualitativeinformation.Concise,computer-basedinformationsystemsaremoreappropriateforsimplerinformationprocessing requirements,accordingto Daft and his colleagues(see Daft and Weick 1984;and
Daft and Lengel 1986).
Not surprisingly,the normative literatureon information system design evinces
considerableoptimism about the degreeof human influenceover the capabilitiesand
characteristicsof informationsystems(Olson 1982;Olson and Lucas 1982;Olson and
Turner1985). Contextualvariables,which an externalor situationalcontrol perspective mightviewas constraintsor determinants,areviewedin the organizationalimperative as contingenciesthat managersshouldtake into account.Among these contextual
variablesareworkunit technology,organizationallevel, environment,decisionmaking
style, and uncertainty(Whisler1975;Gordon and Miller 1976;Waterhouseand Tiessen 1978;Ginzberg1980;Olerup 1982;Gorryand Scott Morton 1971).
The assumptionof designerdiscretionstandsin sharpcontrastto the externaldeterminism of the technologicalimperative.The organizationalimperativeassumes that
systemsdesignerscan managethe impactsof informationsystemsby attendingto both
technicaland social concerns(Bj0rn-Andersenet al. 1986;Mumfordand Weir 1979).
2 Theseneedsare sometimesviewedas externallygenerated-derivedfrom uncertaintyoriginatingin the
environmentor in work technology(Galbraith1977; Daft and MacIntosh1981; Daft and Lengel 1986;
Tushmanand Nadler 1978).

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

588

M. LYNNE MARKUS AND DANIEL ROBEY

This view is sharedby managementand organizationtheoristswho see information

technologyas a tool for solving organizationalproblems(Child 1984; Huber 1984;
Huberand McDaniel 1986).
Empiricalsupportfor the organizationalimperativeis limited.Whilestudiesby Daft
and MacIntosh(1981), Specht (1986), and Olerup (1982) support the notion that
organizationalcharacteristicscorrelatewith information system characteristics,the
models of Galbraith(1977) and Tushman and Nadler (1978) have received mixed
empiricalsupport(Tushman 1979; Morrow 1981; Penley 1982; Triscariand Leifer
1985;Saundersand Robey 1986). Unfortunately,most of these studies fail to assess
designers'intentions and thus cannot be regardedas complete tests of the organizational imperative.
The Emergent Perspective

The emergentperspectiveholdsthat the uses and consequencesof informationtechnologyemergeunpredictablyfromcomplexsocialinteractions.This perspectivecorrespondsto Pfeffer's"emergent"view of action in organizations:"Becauseparticipation
in organizationaldecisionsis both segmentedand discontinuous,becausepreferences
developand changeover time, and becausethe interpretationof the resultsof actions
-the meaningof history-is often problematic;behaviorcannot be predicteda priori
eitherby the intention of individualactorsor by the conditionsof the environment"
(1982, p. 9).
Kling and Scacchi's(1982) distinctionbetween"discrete-entity"models and "web"
models of computingprovidesa useful startingpoint for a discussionof the emergent
perspectiveon information technology and organizationalchange. Discrete-entity
models conceive of informationtechnologyas a tool with identifiablebenefits,costs,
and skill requirements.Like the organizationalimperative,discrete-entitymodels assume that.the goals of designersguide the developmentof computingapplications.By
contrast,web models conceive of informationtechnologyas an ensemble of equipment, applications,and techniquesthat carrysocial meanings.The primaryvirtueof
web modelsis empiricalfidelity,theirabilityto accountforthe detailsand complexities
of actualsituations.Theircharacteristicproblemis analyticalcumbersomeness(1982,
p. 10).

Centralconceptsin the emergentperspectivearethe roleof the computinginfrastructure, the interplay of conflicting objectives and preferences,and the operation of
nonrationalobjectivesand choice processes.For example,Gasser's(1986) studyof the
integrationof computingand routine work examinedthe misalignmentbetween demands of the work setting and the computing resourcesavailable.Gasseridentified
strategiesby which organizationalactorscoped with slippagebetweentechnologyand
work demands and discussedhow the nature of routine work changed as a result.
Ratherthanattributingchangeto actorintentor exogenoustechnology,Gasserfocused
on the dynamicinterplayamong actors,context, and technology.
In anotherrecent example of an emergentmodel, Barley(1986) studiedthe introduction of computerizedtomography(CT scanners)in radiology.He demonstrated
thatthesetechnologiescan alterthe organizationaland occupationalstructureof radiologicalwork.However,"the identicaltechnologiescan occasionsimilardynamicsand
yet leadto differentstructuraloutcomes"in differentsettings(1986, p. 105).In Barley's
analysis,the scannersoccasionedchange not becauseof their inherentcharacteristics
(as the technologicalimperativewould hold), but "becausethey becamesocial objects
whose meaningswere defined by the context of their use" (p. 106). The technology
presentedan "occasion"for structuralchangebut did not determinewhich of a large
varietyof alternativesactuallyemergedfrom the processof structuring.
The emergentperspectiveadmitsgreatercomplexityto the issueof causalagencyand
This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

589

to the goal of predictingorganizationalchangesassociatedwith informationtechnology. By refusingto acknowledgea dominant cause of change,emergentmodels differ
qualitativelyfrom the deterministiccausal argumentsof the two imperatives.Prediction in the emergentperspectiverequiresdetailedunderstandingof dynamicorganizationalprocessesin additionto knowledgeaboutthe intentionsof actorsand the features
of informationtechnology.This addedcomplexitymakesemergentmodels difficultto
construct(Bendifallahand Scacchi 1987;Kling 1987).
Normative Implications

The threeperspectiveson causalagencypresentedheredifferin theirattributionsof

responsibilityfor the outcomes observed.These attributionsimply that particularinterventionswill be more or less efficaciousin producingor increasingthe likelihoodof
desirableoutcomes. Consequently,analystsof differentpersuasionscan often be distinguishedas easily (or more so) by their normativestance as by the specificsof their
models(Mowshowitz1981).
Technologicalimperativeanalysts hold that informationtechnology generallyor
some particularconstellationof technologicalfeaturesis responsiblefor "impacts"such
as change in organizationalstructure,skill enhancementor deskillingof workers,or
change in employment opportunities.Consequently,the policies and remediesproposedby theseanalystscenteron stopping,slowingor acceleratingthe rateof changein
informationtechnologyor selectinginformationtechnologieswith particularpackages
of features.Organizationalimperativeanalystsattributethe consequencesof information technologyto the choicesand behaviorsof managersand systemdesigners.Consequently, these analysts tend to prescribeimproved design and resource allocation
methodsand betterimplementationstrategiesand tactics. Becauseemergentanalysts
attributeoutcomesto an unpredictableinteractionof technologicalfeaturesand actors'
intentions, their normativeposture is less clear than those consideredabove. Some
emergentanalystseschewintervention,arguingthat predictionis impossibleand outcomes areindeterminate;othersadvocate"emancipatory"strategies,such as extensive
user participationin the analysis, design, and implementationof informationtechnology.
LogicalStructure
A second dimensionof theoreticalstructureconcernsthe logical formulationof the
theoreticalargument.On this dimension, Mohr (1982, Chapter2) distinguishesbetween varianceand processtheories.The distinctionin theoreticalstructurebetween
varianceand processtheoriesis somewhatanalogousto the distinctionbetweencrosssectional and longitudinalresearchmethodologies.Variancetheories are concerned
with predictinglevels'of outcome from levels of contemporaneouspredictorvariables;
variancetheoriesare concernedwith explaininghow outcomesdevelop over time.
Variance and Process Theories

Mohr (1982) explainsthe differencebetweenvariancetheoriesand processtheories

in terms of the hypothesizedrelationshipsbetweenlogicalantecedentsand outcomes.
These are summarizedin Figure 3. In variancetheories,the precursor(loosely, that
which might be referredto as the "cause")is posited as a necessaryand sufficient
conditionfor the outcome.3Forexample,in a variancetheorythat hypothesizesuse of
informationtechnology as a cause of organizationalcentralization(cf. Leavitt and
Thereis a superficialresemblancebetweenvariancetheoriesand regressionmodels,but the two are not
identical.Manyregressionmodelscontainvariableswhichpredictthe outcomebut whichare not hypothe(Mohr 1982,Chapter2).
sizedto be "causal"in the necessaryand sufficient'sense

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

M. LYNNE MARKUS AND DANIEL ROBEY

590

THEORY
VARIANCE

PROCESSTHEORY

ROLEOF TIME

Static

Longitudinal

DEFINITION

Tht cause Is
necessary and
sufficient
for the
outcome

Causation consists
of necessary
conditions In
sequence; chance
and random events
play a role

ASSUMPTIONS

Outcome will
Invariablyoccur
when necessary
and sufficient
conditions are
present

Outcomes may not

occur (even when
conditions are
present)

ELEMENTS

Variables

Discrete outcomes

FORM
LOGICAL

IfX, then Y;
if more X,then
more Y

Ifnot X, then not Y;

cannot be extended
to "moreX'or
"moreyU

FIGURE3. Logical Structure.

Whisler 1958), centralizationis expectedalwaysto occur wheneverinformationtechnology is used (where"always"is definedin terms of statisticalconfidencelevels). In
processtheories,the precursoris assumedinsufficientto "cause"the outcome, but is
held to be merelynecessaryfor it to occur.
In general,necessaryconditions alone cannot constitute a satisfactorytheory. For
example, while water may be necessaryfor the growth of plants, it is not sufficient;
therefore,it cannot be consideredthe cause of plant growth. Mohr has observed,
however,that necessaryconditionscan comprisea satisfactorycausalexplanationwhen
they are combined in a "recipethat stringsthem togetherin such a way as to tell the
story of how [the outcome] occurs wheneverit does occur" (1-982,p. 37). In short,
outcomesare (partially)predictablefrom a knowledgeof process,not from the level of
predictorvariables.
A good example of a processtheory is the "garbage-can"theory of organizational
choice (Cohen, Marchand Olson 1972). In this model, decisions result from the (almost) randomcollisionsof participants,choice opportunities,solutions,and decisions.
Many diffusionof innovationtheoriesare processtheories,at least implicitly.Barley's
study (1986) of CT scanning in radiologyalso meets the requirementsof a process
theory.
Markus(1984) proposeda processtheory for explaininguser resistanceto information systems.Includedas a necessaryconditionfor userresistancein this "recipe"is the
introductionof an informationsystem with featuresdifferingfrom the featuresof the
organizationalsetting.This necessarycondition is not believedsufficientto ensurethe
occurrenceof resistance,but it is believedto be necessary.Consequently,this process
theoryrecognizesthat resistancemay not alwaysoccur, even when the necessarycondition of "differingfeatures"is present.In any given case, resistancemay not occur for
severalpossiblereasons:peoplemay like the changesembodiedin the system;they may

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

591

be too apathetic to resist; or they may find ways to circumvent the changes the system
implies. More accurate predictions are only possible when these additional ingredients
of the setting are known and their temporal relationships to each other are understood.
Another example of a process theory of information technology is the work of Zmud
and Apple (1988) on optical scanning technology in supermarkets. Zmud and Apple
distinguish between the routinization of an innovation, defined as the accommodation
of an organization's governance system to the innovation, and its institutionalization,
defined as the organization's achievement of higher levels of use and benefits from the
innovation. They argue persuasively that these two concepts are related in a necessary
but not sufficient fashion. Routinization is necessary for institutionalization, but institutionalization is not certain to occur when an innovation is routinized.
Variance theories, then, differ from process theories in their assumptions about the
relationship between antecedents and outcomes. Variance theories posit an invariant
relationship between causes and effects when the contingent conditions obtain. Process
theories assert that the outcome can happen only under these conditions, but that the
outcome may also fail to happen.4
-Variance and process theories also differ in their conceptualization of outcomes and
precursors. In variance theories, these constructs are usually conceptualized as variables: entities which can take on a range of values. This practice allows the prediction of
the full range of values of the outcome variable. For example, if the use of information
technology is necessary and sufficient for organizational centralization, then increased
use of information technology should lead to greater centralization.
In process theories, however, outcomes are not conceived as variables that can take
on a range of values, but rather as discrete or discontinuous phenomena, that might be
called "changes of state." Process theories cannot be extended, as variance theories can,
to explain or predict what happens when there is "more" of a precursor variable. Thus,
if a process theory specifies that certain conditions are sufficient to cause user resistance, it does not follow that more of these conditions will mean more resistance.
To illustrate, in their study of the adoption of supermarket scanners, Zmud and
Apple (1988) identified three distinct levels of institutionalization, each accompanied
by a unique ideology of the organizational role served by scanners: to enforce retail unit
worker discipline, to possess the capability of better managerial information, and to
provide better managerial information. While a "higher level" of institutionalization
includes the "lower level" states, the levels are better conceived as qualitatively different
outcomes than as varying degrees of a single dimension.
Mohr (1982) believes that variance and process theories can "peacefully coexist," but
that the distinctions between them should not be blurred in an attempt to gain the
advantages of both within a single theoretical approach. He offers three reasons for this
position. First, for any variance theory, it is always possible to specify mechanisms that
intervene between antecedents and outcomes. But, because variance theories posit
sufficiency, the inclusion of intervening variables does not improve prediction of the
outcome variable. In short, intervening variables are redundant, unless one is "puzzled
about the means of bridging the gap between one included phenomenon and another"
(p. 43). Second, process theories can easily bog down under the imposition of conditions thought to increase the likelihood of the outcome (pp. 61-65). Third, while
agreeing that process theories and variance theories may mutually inform each other,
Mohr concludes that "odd bits and pieces of research results cannot be integrated or
interchanged from one theoretical type to the other; the effort produces confusion and
' While contingencytheoriesdeal with necessaryconditions, they are not processtheories.They are
variancetheories,becausethe conditionsthey specify are both necessaryand sufficientfor the outcome
to occur.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

592

M. LYNNE MARKUS AND DANIEL ROBEY

stagnation-the frustration of theory. Sorting the two out and keeping them separate,
however, produces clarity and the basis of progress" (1982, pp. 67-68).

RelationshipbetweenCausalAgencyand LogicalStructure
At first glance, it may appear that all imperative theories are variance theories and all
emergent theories are process theories. Slack (1984) and Mohr (1982) appear to support
this observation. Indeed, we find it hard to imagine how emergent theories could
effectively be cast as variance models. However, both variance and process models are
available to analysts from the perspective of either the technological or the organizational imperative. An example of a technological imperative process model should
suffice to make the point.
An early formulation of the technological imperative can be seen in the work of Ellul
(1964) and Winner (1977). Ellul argued that technology creates social changes which
reach far beyond its original applications. Once developed, technology follows a selfsustaining evolutionary path with the dynamic that whatever can be developed must be
developed. Thus, techniques carry in themselves the seeds of new applications. Winner
extended Ellul's theory by noting the role of supporting infrastructures, such as the
huge power plant required to supply electricity for small appliances. These supporting
infrastructures institutionalize and perpetuate the technologies they were originally
created to support.
Initially, the inevitability of technological impact-the hallmark of the technological
imperative-appears to be the salient feature of Ellul's and Winner's arguments. Yet a
long historical perspective reveals many technologies that have been abandoned with.out trace or consequence. What seems to make the difference for a technology to have
an extensive and enduring impact is the formation of an infrastructure, which, once
established, perpetuates itself and institutionalizes the use of the technology. In process
theory terms, the development of the infrastructure is a necessary condition (but not
necessary and sufficient) for social changes to occur.

Advantagesof ProcessTheories
To say that process models are possible does not provide a convincing rationale for
their use. While we do not argue that process models are superior to variance models,
process models have been unjustly neglected in favor of the more common variance
theories.5 Analysts should consider the following advantages of process theories when
formulating the logical structure of their theories.
By their very structure, variance theories posit an invariant relationship between
antecedents and outcomes. This assumption may simply be too stringent for social
phenomena. Put differently, if a behavioral outcome occurs only some of the times
when its antecedents are present, then it may not be possible to establish an invariant
relationship;between the antecedents and outcome, even with generous statistical confidence levels. As Sutherland has put it, "not all real-world phenomena will ultimately
become deterministic if we spend enough time analyzing them." (1973, p. 145) In
circumstances like these, process theories may enable researchers to find patterns in
empirical data that variance theorists might miiss.
Process theories have another advantage. While they retain the empirical fidelity of
the emergent perspective, they also preserve the belief in the regularity and predictability of social phenomena that characterizes the technological and organizational imperatives. Prediction of patterned regularities over time is one of the goals of process theory
I
Some of this neglect may stem from the disrepute of information systems "stage" models. Mohr (1982)
describes stage models as incomplete process models, because they generally lack specification of the mechanism by which subsequent stages come about.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

593

research.By contrast,some criticsof variancetheory researchtake a strongantipositivist stance, arguingthat predictionof human action (such as the use of information
systems)is not a legitimatepursuit.6
Thus,while empiricalprocessresearchtypicallyrevealsthat thingsare more complicatedthanvariancetheoryrepresentsthem to be, such researchshouldnot be dismissed
as isolatedstories or illustrativecases only. With care, findingscan be generalizedto
other-settings,and predictionscan be testedin laterresearch.The advantageof process
theoriesis that these predictionsmay correspondmore faithfullyto actual events in
organizationsthan do the typicalpredictionsof varianceformulations.
In summary, we believe that process theories are useful preciselybecause, while
recognizingand acceptingthe complexityof causalrelationships,they do not abandon
the goals of generalizabilityand prediction.By acceptinga more limited definitionof
prediction,one in which the analystis able to say only that the outcome is likely (but
not certain)undersome conditionsand unlikelyunderothers,processtheoristsmay be
able to accumulateand consolidatefindingsabout the relationshipbetweeninformation technologyand organizationalchange.
Level of Analysis
The specific theories and researchstudies discussed in this paper concern three
differenttypes of entities,or levels of analysis:individuals,organizations,and society.
Questionsaboutthe appropriatelevel of analysishavebeen widelydebatedin the social
sciencesgenerally,but have rarelybeen explicitlydiscussedwithin those researchcommunities concernedwith the causes and consequencesof informationtechnologyin
organizations.The debatecenterson two issues:problemsof inferenceand ideological
biases(Pfeffer1982).
Problemsof inferencearisewhen conceptsaredefinedand dataarecollectedat levels
of analysisinappropriatefor the theoreticalpropositionsbeingexamined.For example,
researchersinterestedin organizationalgoals often collect data on the goals of key
individuals.When inferencesdrawnfromthese data referonly to organizationalgoals,
levels of analysishave been confused.Avoidingsuch inferenceproblemsrequiresthe
researcher"to bound the organizationin such a way that observedunits are unambiguouslyseparablefromeach otherand fromtheirenvironmentsin both spaceand time"
(Freeman1978, p. 336).
Ideologicalbiasesoriginatein the orientationsof differentdisciplinarygroups(Rousseau 1985). The customary division of levels of analysis into "macro-level" and
"'micro-level"theories reflectdisciplinaryboundaries,each with its favoredrescarch
questions,acceptablemethodologies,and conventionsfor reportingresults.The concepts in macro-leveltheories are propertiesof large-scalecollectives (organizations,
populations,societies);this level of analysis is favoredby macrosociologists,macroeconomists,and evolutionarytheorists.The conceptsin micro-leveltheoriesare propertiesof individualsand smallgroups;this level of analysisis favoredby socialpsychologistsand microeconomists.
Macro-leveland micro-levelproponentstend to preferdifferentcausal structures.
Macro sociologicaltheory typicallyexplains phenomenaby referrng only to macro6 For example, Boland (1985) argues for a phenomenology of information science in which researchers
"read the interaction during system design and use it in order to interpret the significance and potential
meanings they hold" (p. 196). Hirschheim and Klein (1985; Klein and Hirschheim 1983; Hirschheim 1985)
argue for a "consequentialist" perspective, based on hermeneutic analysis. Phenomenology and hermeneutics
strive for subjective "understanding," and view prediction as an illegitimate goal. Lyytinen and Klein (1985)
take a critical theory perspective toward information system research. In critical theory, the aim is intervention to alleviate current conditions. Prediction is irrelevant to critical theorists, since outcomes are assumed to
be under the control of actors.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

594

M. LYNNE MARKUS AND DANIEL ROBEY

level concepts.Stinchcombe(1968), for example,discussesthreetypes of causalstructures-demographic, functional, and historicist-which explain social phenomena
without introducingsuch concepts as individual attitudes, intentions, motives and
choices.
Proponentsof macro-levelanalysisarguethat it is a productivestrategyfor generating "falsifiable,parsimonious,and readilycomprehensibleexplanationsfor behavior"
(Pfeffer1982, p. 256). Criticsarguethat macro-leveltheoriessufferfrom data insufficiency, failureto explainhow macrorelationshipscome about,and the needto assume
the existenceof a social systemas a startingpoint (Coleman 1986).
An exampleof a propositionfrommacro-leveltheoryis Leavittand Whisler's(1958)
claim that the introductionof informationtechnologycauseschangein organizational
structure.This propositioncontains no concepts about the individualswho populate
the organization,design the informationtechnology,use the informationtechnology,
and so on. As partof a variancetheory,this statementdoes not requireany articulation
of the individualbehaviorsand processesby whichthe outcome occurs.By contrast,a
correspondingmicro-levelconstructionof Leavitt and Whisler'sproposition might
includeindividuals'choices to use informationtechnology,changedcoordinationpatternsin the workunit, and managers'needsto maintaincontrol.
Proponentsof micro-levelanalysisarguethat only people can act; collectivebodies
are incapableof action. Further,social collectivesconsist of individuals,and macro
conceptslike organizationalstructurearepermissibleonly when it is possibleto ground
them in the individualbehaviorsand the micro-levelevents and processesthat comprise them (Pfeffer 1982). Critics of the micro level of analysis accuse it of logical
fallacy,confusingthe question of causalitywith the assertionof an answer,reducing
social phenomenato biologicalphenomena,and relianceon concepts "that logically
residein the headsof people"(Pfeffer1982, p. 22).
In contrastto our cautionagainstmixingprocessand variancetheory,we believethat
mixinglevels of analysismay be usefulin researchand theoryon informationtechnology and organizationalchange. In defense of mixed-leveltheory, Rousseau (1985,
1986)assertsthat technologiessuch as officeautomationare neitherstrictlymicro nor
macro in character.She believesthat mixed-levelresearchshouldaboundin an interdisciplinaryfield where mixed-level phenomena are the inevitable subject of study
(1985; pp. 2-3). That it does not is a disturbingcommentaryon the power of discipline-basedresearchgroups.
In one of the few systematicattemptsto addressthe level of analysisissue for studies
of computing,Kling(1987) proposescriteriafor establishingthe analyticboundariesof
computerapplications.He arguesthat populations,equipment,spatialand temporal
elementsshouldbe includedwithinthe boundariesof an analysiswhen theseelements
eitherconstrainactorsinvolvedin the specificapplicationor are takeninto accountby
the actors in constructingtheir actions. These criteriaexpand the focal situationto
includelargersocial arenas,thus creatinga need for mixed-levelresearchstrategies.
Coleman(1986) proposesone such mixed-levelstrategy:"not to remainat the macrosociallevelbut to move down to the level of individualactionsand backup again''(p.
1322).An exampleof a mixed-leveltheory of informationtechnologyand structure.is
seen in Barley's(1986) work. The introductionof a new computer-basedtechnology
into a worksetting(macro-level)affectsthe skillsand competenciesof the peoplein the
work unit (micro-level).Interactionsamong people at differentlevels of skill create
patternsof seekingand givingadvice (micro-level).Ultimately,these patternsbecome
institutionalizedas formalorganizationalstructure(macro-level).Barley'sinsightson
the relationshipbetweentechnologyand structuredepend on moving carefullyacross
levels of analysis.
The role that mixed-leveltheorygives to humanpurposeand intentionis consistent
with Stinchcombe'sconception of technologyas "a descriptionof the causalconnecThis content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

595

tion between the ends people have and what they have to do to achieve those ends"
(1983; p. 122). Both macro-level theory and much micro-level research tend to ignore
human intentions.7 While the mixed-level strategy preserves macro-level concepts, it
grounds these concepts in individual purposes and behavior and so remains "methodologically individualist" (Coleman 1986). Consequently, a mixed-level strategy remains
vulnerable to the criticisms of macro-level proponents mentioned above.
In summary, theorists and researchers who study the relationship between information technology and organizational change have given little explicit attention in their
writing to the choice of an appropriate level of analysis -macro, micro or mixed.
Choice of any level is subject to criticism by proponents of the others, but researchers
will be better able to respond to these criticisms after deliberate and thoughtful choice of
the appropriate level of analysis for their own work.

Summaryand Conclusions
Social theoriesembody researchers'conceptionsof causality.In this paperwe have
discussedthe causal structuresfound in theoriesabout the relationshipbetweeninformationtechnologyand organizationalchange.We have evaluatedtheoryand research
in the field by focusingour discussionon three dimensionsof causal structure:causal
agency,logicalstructure,and level of analysis.
Causalagencyrefersto analysts'assumptionsabout the identityof the causalagent
and the directionof causalinfluence.Muchof our thinkingaboutinformationtechnology's consequences in organizations has been guided by theories with fairly simple

notions of causal agency. The technologicalimperativeviews technology as causal

agent,and the organizationalimperativeviewshumanbeingsas agentsof socialchange.
A third conception of causalagency,the emergentperspective,attributescausalityto
complex indeterminantinteractionsbetweentechnologyand human actorsin organizations.Centralto the emergentperspectiveis the social meaningascribedto information technology.This perspectiveaccountsfor conflictingresearchfindingsabout impacts by demonstratingthe differentmeaningsthat the same technologyacquiresin
differentsocial settings.
Logicalstructurein theoryrefersto the natureof the relationshipbetweenelements
identifiedas antecedentsand those identifiedas outcomes. In variancetheories,antecedentsare conceivedas necessaryand sufficientconditionsfor the outcomesto occur.
In processtheories,antecedentsare necessarybut not necessaryand sufficient.Process
theoriesdepend heavily on the specificationof temporalrelationsamong theoretical
elements, much like a recipe stringstogetheringredientsfor a finishedmeal. Process
theorieshaveloweraspirationsabout"explainedvariance,"but providericherexplanations of how and why the outcomesoccurwhen they do occur.
Finally,level of analysisdistinguishesmore or less inclusiveentities as the focus of
analysis.The macrolevel of analysisfocuseson societiesand formalorganizations;the
micro level addressesindividualsand small groups.Causalstructuresfor macro levels
of analysisexplainsocialphenomenawithoutusingthe constructsof individuals'mental processes,and many micro analysisignorehuman purposesand intentions.Since
theoriesabout informationtechnologyin organizationsare difficultto confine naturally to one level of analysis, mixed levels of analysishave become attractiveto re7Coleman (1986) traces the popularity of micro-level theories of social phenomena to the development of
statistical survey research methods in the 1940s. Inferences drawn from statistical associations within survey
data formed the basis for causal explanations of behavior, but the methods had little natural affinity for the
intentions or purposes of individuals. The causes of human behavior were theorized as either characteristics of
the individual or characteristics of the individual's environment, without recourse to individual purpose or
intention (1986, p. 1314). Coleman rejects macro-level theory, particularly because it attributes purpose or
intention to collectives, but he regretsthe tendency of much micro-level research to ignore individual purpose
or intention.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

M. LYNNE MARKUS AND DANIEL ROBEY

596

searchers in this field. By consciously mixing levels of analysis, researcherscan

explore the dynamic interplay among individuals, technology, and larger social
structures.
It is no secretthat researchon informationtechnologyand organizationalchangehas
producedconflictingresultsand few reliablegeneralizations.By carefullyconsidering
each of the dimensionsof causalstructurediscussedin this paper,researchersshouldbe
able to constructsoundertheoriesto guide more fruitfulresearch.While attendingto
researchdesign, sampling,measurement,statisticalanalysis,and other researchtechniques will also improve the state of the art, we believe that the more fundamental
issuesof theoryconstructionmust be addressedfirst.When assumptionsabout causal
agency, logical structure,and levels of analysis are addressedexplicitly, subsequent
decisionsabout researchstrategyand techniquewill be betterinformed.For example,
ethnographyappearsbetter suited to emergent process researchat mixed levels of
analysisthan does surveyresearch.Too often, we fear,researchmethodsareelectedfor
reasonsotherthan their utility in servinga particulartheoreticalapproach.
We cannot exclusivelyendorse any single combinationof the dimensionswe have
discussedas the "correct"causalstructurefor researchin informationsystems.All can
servethe goals of interestingdiscoveryand rigoroustestingto which researchersnormallyaspire.However,some causalstructuresfarebetterwhen evaluatedagainstcriteria of simplicityand parsimony;othersfarebetterwhen evaluatedfor "empiricalfidelity," the abilityto mirrorfaithfullythe phenomenaunderstudy.We find little balance
within our fieldbetweenthese two sets of criteriafor evaluatingtheories.Much of the
researchon resistanceto systems,growthof the informationsystemsfunctionin organizations;systemsimplementation,and the like, has adoptedvariancetheory formulations of logical structureand an imperativeconception of causal agency. We suspect
that greateruse of theoreticalstructureswhichemphasizeempiricalfidelitywill stimulate more and betterresearchon these phenomena.
In conclusion,carefulexaminationof causalstructuresis a productiveexercisein any
field.Researchersinterestedin the consequencesof informationtechnologyfor organizationsshouldmakeclearand consciouschoicesregardingthe causalstructuresof their
theories. These choices are at least as important as more technical researchissues,
perhapsmore so. The discussion of causal structurein this paper should facilitate
choice and critical thinking both for researchersand for those who apply research
findings.8
8 We extend our thanksto the many people,too many to list here,who have given us helpfuladviceon
earlierdraftsof this paper.We especiallythankJohn King, Rob Kling,Ron Rice, Jon Turner,and several
anonymousreviewerswho havecommentedextensivelyon severalof our drafts.

References
ATTEWELL, P. AND J. RULE, "Computing and Organizations: What We Know and What We Don't Know,"

Comm. ACM, 27 (1984), 1184-1192.

BAKOPOULOS,J. A. Y., "Toward a More Precise Concept of Information Technology,"

Proc.SixthInternat.

Conf: Information Systems, Indianapolis, IN, 1985, 17-24.

BARLEY, S. R., "Technology as an Occasion for Structuring: Evidence from Observations of CT scanners and
the Social Order of Radiology Departments," Admin. Sci. Quart., 31 (1986), 78-108.
S. AND W. SCACCHI,
BENDIFALLAH,
"Understanding Software Maintenance Work," IEEE Trans.Software
Engrg., 13, 1 (1987), in press.
BJ0RN-ANDERSEN,
N., K. EASONAND D. ROBEY,ManagingComputerImpact:An InternationalStudyof
Management and Organization, Ablex, Norwood, NJ, 1986.
BLAU,P. M., C. M. FALBE,W. MCKINLEYAND P. K. TRACY,"Technology and Organization in Manufacturing," Admin. Sci. Quart., 21 (1976), 20-40.
BOLAND,RICHARDJ., JR., "Phenomenology: A PreferredApproach to Research on Information Systems," in
E. MUmford, R. Hirschheim, G. Fitzgerald, and T. Wood-Harper (Eds.), Research Methods in Information Systems, North-Holland, New York, 1985.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

INFORMATION TECHNOLOGY AND ORGANIZATIONAL CHANGE

597

CARTER, N. M., "Computerization as a Predominate Technology: Its Influence on the Structure of Newspa-

per Organizations," Acad.ManagementJ., 27 (1984), 247-270.

A Guideto Problemsand Practice,2nd ed., Harper and Row, London, 1984.
CHILD,J., Organization:
COHEN, M. D., J. G. MARCH AND J. P. OLSON,"A Garbage Can Model of Organizational Choice," Admin.
Sci. Quart., 17 (1972), 1-25.
COLEMAN, J. S., "SocialTheory,Social Research,and a Theoryof Action,"Amer.J. Sociology,91 (1986),
1309-1335.
CULNAN, M. J., "The Intellectual Structure of Management Information Systems, 1972-1982: A Co-citation
Analysis," ManagementSci., 32 (1986), 156-172.
DAFT,R. L. AND R. H. LENGEL, "Organizational Information Requirements, Media Richness and Structural
Design," Management Sci., 32 (1986), 554-571.
AND N. B. MACINTOSH, "A New Approach to the Design and Use of Managemient Information,"
CaliforniaManagementRev., 21 (1978), 82-92.
AND , "A Tentative Exploration into the Amount and Equivocality of Information Processing
in Organizational Work Units," Admin.Sci. Quart.,26 (1981), 207-224,
AND K. E. WEICK, "Toward a Model of Organizations as Interpretative Systems," Acad. Management
Rev.,9 (1984), 284-295.
DAWSON,P. AND I. MCLAUGHLIN,
"Computer Technology and the Redefinition of Supervision:A Studyof
the Effects of Computerization on Railway Freight Supervisors," J. ManagementStudies,23 (1986),
116-132.
ELLUL,J., The Technological
Society,Knopf, New York, 1964.
FOSTER,L. W. AND D. M. FLYNN,"Management Information Technology: Its Effects on Organizational
Form and Function," MIS Quart.,8 (1984), 229-236.
FRANZ,C. R., D. ROBEYAND R. R. KOEBLITZ,
"User Response to an Online Information System: A Field
Experiment," MIS Quart.,10 (1986), 29-42.
FREEMAN,J. H., "The Unit of Analysis in Organizational Research," in M. W. Meyer and Associates,
Environments
and Organizations,Jossey-Bass, Inc. San Francisco, 1978.
FRY, L. W., "Technology-Structure Research: Three Critical Issues," Acad. ManagementJ., 25, 3 (1982),
532-552.
GALBRAITH,
J. R., Organization Design, Addison-Wesley, Reading, MA, 1977.
GASSER,L., "The Integration of Computing and Routine Work," ACMTrans.OfficeInformation
Systems,4,
3 (1986), 205-225.
GINZBERG,M., "An Organizational Contingencies View of Accounting and Information Systems Implementation," Accounting,Organizations,
and Society,5 (1980), 369-382.
GORDON,L. A. AND D. MILLER,"A Contingency Framework for the Design of Accounting Information
and Society, 1 (1976), 59-69.
Systems," Accounting,Organizations,
GORRY,G. A. AND M. S. SCOTTMORTON,"A Framework for Management Information Systems," Sloan
ManagementRev., 13 (1971), 55-70.
HIRSCHHEIM,
R. A., OfficeAutomation:A Social and Organizational
Perspective,Wiley, New York, 1985.
AND H. K. KLEIN,"Fundamental Issues of Decision Support Systems: A Consequentialist Perspective," DecisionSupportSystems, 1 (1985), 5-24.
HUBER,G. P., "The Nature and Design of Post-industrial Organizations," ManagementSci., 30 (1984),
928-95 1.
ANDR. R. MCDANIEL,JR., "Exploiting Information Technologies to Design More Effective Organizations," In M. Jarke (Ed.), Managers,Microsand Mainframes,Wiley, New York, 1986.
KLATZKY,S. R., "Automation, Size, and the Locus of Decision Making: The Cascade Effect," J. Business,43
(1970), 141-151.
KLEIN,H. K. AND R. A. HIRSCHHEIM,
"Issues and Approaches to Appraising Technological Change in the
Office: A Consequentialist Perspective," Office:Technologyand People,2 (1983), 15-42.
KLING,R., "The Impacts of Computing on the Work of Managers, Data Analysts, and Clerks," Public Policy
Research Organization, University of California, Irvine, 1978.
"Social Analyses of Computing: Theoretical Perspectives in Recent Empirical Research," Computing
Surveys,12 (1980), 61-110.
, "Defining the Boundaries of Computing Across Complex Organizations," in R. Boland and R.
Hirschheim (Eds.), CriticalIssues in InformationSystemsResearch,John Wiley, New York, 1987.
AND W. SCACCHI,"The Web of Computing: Computer Technology as Social Organization," Ad,
vancesin Computers,21, Academic Press, New York, 1982.
LEAVITT,H. J., ANDT. L. WHISLER,"Management in the 1980s," HarvardBusinessRev.,36 (1958), 41-48.
LEIFER,R., AND E. F. McDONOUGHIII, "Computerization as a Predominant Technology Effecting Work
Unit Structure," Proc.Sixth AnnualConf InformationSystems, 1985, 238-248.
LYYTINEN,K. AND H. K. KLEIN,"The Critical Theory of Jurgen Habermas as a Basis for a Theory of
Information Systems," in E. Mumford, R. Hirschheim, G. Fitzgerald, and T. Wood-Harper (Eds.),
ResearchMethodsin InformationSystems,North-Holland, New York, 1985.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions

598

M. LYNNE MARKUS AND DANIEL ROBEY

MAJCHRZAK,A. AND P. MOSHER, "Technological Change and StructuralContingency Theory," unpublished

paper, 1987.
MARKUS,M. L., Systemsin Organizations:
Bugsand Features,Pitman, Marshfield, MA, 1984.
MOHR,L. B., ExplainingOrganizational
Behavior,Jossey-Bass, San Francisco, 1982.
MORROW, P. C., "Work Related Communication, Environmental Uncertainty, and Subunit Effectiveness:A
Second Look at the Information Processing Approach to Subunit Communication," Acad.ManagementJ., 24 (1981), 851-858.
A., "On Approaches to the Study of Social Issues in Computing," Comm.ACM,24 (1981),
MOWSHOWITZ,
146-155.
MUMFORD,E. AND M. WEIR, ComputerSystemsin WorkDesign-The ETHICSMethod,Wiley, New York,
1979.
OLERUP, A., A ContextualFramework
for ComputerizedInformationSystems,Nyt Nordisk Forlag Arnold
Busck, Copenhagen, 1982.
OLSON,M. H., "New Information Technology and Organizational Culture," MIS Quart.,6 (1982), 71-92.
AND H. C. LUCAS,JR., "The Impact of Office Automation on the Organization: Some Implications for
Research and Practice," Comm.ACM,25 (1982), 838-847.
AND J. A. TURNER,"Rethinking Office Automation," Proc.SixthAnnualConf.Information
Systems,
(1985), 259-269.
PENLEY, L. E., "An Investigation of the Information Processing Framework of Organizational Communication," HumanCommunication
Res., 8 (1982), 348-365.
PFEFFER, J., Organizations
and OrganizationTheory,Pitman,Marshfield,MA, 1982.
AND H. LEBLEBICI,"InformationTechnologyand Organizational Structure," PacificSociological
Rev.,20 (1977), 241-261.
RICE,R. E., "The Impacts of Computer-Mediated Organizational and Interpersonal Communication," in M.
Williams (Ed.), Annual Review of InformationScience and Technology,15, Knowledge Industry
Publications, White Plains, NY, 1980.
ROBEY,D., "Computers and Management Structure: Some Empirical Findings Re-examined," Human
Relations,30 (1977), 963-976.
, "Computer Information Systems and Organization Structure," Comm. ACM, 24 (1981), 679-687.
ROUSSEAU, D. M., "Issues of Level in Organizational Research: Multi-level and Cross-level Perspectives," in
Research in Organizational Behavior, 7, JAI Press, Inc. Greenwich, CT, 1985, 1-37.
"Choosing an Appropriate Level of Analysis for Studying Computerized Work Systems," Paper
Presented at the Annual Meeting of the Academy of Management, Chicago, IL, August, 1986.
SAUNDERS, C. S. AND D. ROBEY, "Task Characteristics, Information Systems, and Interdepartmental Communication," unpublished paper, 1986.
SIMON,H. A., TheNew Scienceof ManagementDecision,Prentice-Hall, EnglewoodCliffs,NJ, 1977.
SLACK,J. D., Communication
TechnologiesandSociety:Conceptionsof CausalityandthePoliticsof TechnologicalIntervention,
Ablex, Norwood, NJ, 1984.
SPECHT, P. H., "Job Characteristics as Indicants of Information System Requirements, MIS Quart., 10
(1986), 271-287.
STEWART, R., How Computers
AffectManagement,MIT Press, Cambridge, 1971.
STINCHCOMBE,A. L., Constructing
Social Theories,Harcourt,Braceand World,New York, 1968.
, EconomicSociology,Academic Press, New York, 1983.
SUTHERLAND, J. W., A GeneralSystemsPhilosophyfortheSocialand BehavioralSciences,George Braziller,
New York, 1973.
TRISCARI, T. AND R. LEIFER, "Information Processing in Organizations: An Empirical Test of the Tushman-Nadler Model," Paper presented to the national meetings of the Academy of Management, San
Diego, August, 1985.
TUSHMAN, M. L., "Impacts of Perceived Environmental Variability on Patterns of Work-related Communication," Acad.ManagementJ., 22 (1979), 482-500.
AND D. A. NADLER, "Information Processing as an Integrating Concept in Organizational Design,"
Acad.ManagementRev., 3 (1978), 613-624.
WATERHOUSE, J. H. AND P. TIESSEN, "A Contingency Framework for Management Accounting Systems
Research,"Accounting,Organizations,and Society,3 (1978).
WHISLER, T. L., TheImpactof Computerson Organizations,
Praeger Publishers, New York, 1970.
"Man, Organization and Computer-A Contingency Analysis," in E. Grochia and N. Szyperski
(Eds.), InformationSystemsand Organizational
Structure,de Gruyter, Berlin, 1975.
as a Themein PoliticalThouight,
WINNER, L, AutonomousTechnology:Technics-out-of-control
MIT Press,
Cambridge, 1977.
ZMUD, R. W. AND L. E. APPLE, "Measuring the Infusion of a Multi-Business Unit Innovation," Unpublished
working paper, Florida State University, 1988.

This content downloaded from 140.119.81.207 on Mon, 14 Sep 2015 14:41:23 UTC
All use subject to JSTOR Terms and Conditions