education

Lapointe & Rivard/Resistance to IT Implementation

MIS Quarterly Vol. 29 No. 3, pp. 461-491/September 2005 461

RESEARCH ARTICLE

A MULTILEVEL MODEL OF RESISTANCETO INFORMATION TECHNOLOGYIMPLEMENTATION1

By: Liette LapointeFaculty of ManagementMcGill University1001 Sherbrooke Street WestMontreal, Quebec H3A [email protected]

Suzanne RivardHEC Montreal3000 Cte Ste-Catherine RoadMontreal, Quebec H3T [email protected]

Abstract

To better explain resistance to information techno-logy implementation, we used a multilevel, longi-tudinal approach. We first assessed extantmodels of resistance to IT. Using semantic analy-sis, we identified five basic components of resis-tance: behaviors, object, subject, threats, andinitial conditions. We further examined extantmodels to (1) carry out a preliminary specification

1Ron Weber was the accepting senior editor for thispaper. Christina Soh was the associate editor. M. LynneMarkus, Ramiro Montealegre, and Siew Kien Sia actedas reviewers.

of the nature of the relationships between thesecomponents and (2) refine our understanding ofthe multilevel nature of the phenomenon. Usinganalytic induction, we examined data from threecase studies of clinical information systems imple-mentations in hospital settings, focusing on physi-cians resistance behaviors. The resulting mixed-determinants model suggests that group resis-tance behaviors vary during implementation.When a system is introduced, users in a group willfirst assess it in terms of the interplay between itsfeatures and individual and/or organizational-levelinitial conditions. They then make projectionsabout the consequences of its use. If expectedconsequences are threatening, resistance be-haviors will result. During implementation, shouldsome trigger occur to either modify or activate aninitial condition involving the balance of powerbetween the group and other user groups, it willalso modify the object of resistance, from systemto system significance. If the relevant initialconditions pertain to the power of the resistinggroup vis--vis the system advocates, the object ofresistance will also be modified, from systemsignificance to system advocates. Resistancebehaviors will follow if threats are perceived fromthe interaction between the object of resistanceand initial conditions. We also found that thebottom-up process by which group resistancebehaviors emerge from individual behaviors is notthe same in early versus late implementation. In


462 MIS Quarterly Vol. 29 No. 3/September 2005

early implementation, the emergence process isone of compilation, described as a combination ofindependent, individual behaviors. In later stagesof implementation, if group level initial conditionshave become active, the emergence process isone of composition, described as the convergenceof individual behaviors.

Keywords: User resistance, information techno-logy implementation, information system imple-mentation, longitudinal perspective, multilevel ap-proach, resistance behaviors, semantic analysis,case study

Introduction

Early researchers who studied information tech-nology implementation recognized resistance as acritical variable (Keen 1981). While some sawresistance as a barrier to be removed (Kossek etal. 1994), others saw it as a means by which userscommunicate their discomfort with a system thatmight be flawed (Marakas and Hornik 1996). Areview of 20 IT and IT-related journals over thepast 25 years found 43 articles that treatedresistance as a key implementation issue. Whilethey acknowledge the importance of resistance,most treat it as a black box. Only 9 of the 43articles explicitly defined the concept of resistance.Moreover, only four opened the black box andproposed theoretical explanations of how and whyresistance occurs (Joshi 1991; Marakas andHornik 1996; Markus 1983; Martinko et al. 1996).Adopting the premise that better theories ofresistance will lead to better implementationstrategies and, hopefully, to better outcomes(Markus 1983, p. 430), our study follows in thesteps of these earlier efforts. Using case studyresearch, we propose a process model ofresistance to IT implementation that explains howresistance surfaces, progresses, and culminates.

Our model of resistance both resembles anddiffers from earlier models. Like earlier models, itconceptualizes resistance as a result of theinteraction among several antecedents. It alsoadopts a neutral stance toward resistance to IT,

regarding it as neither good nor bad. It departsfrom earlier models by using a case study strategyof theory building (Eisenhardt 1989), suggestingthat time plays an important role in explaining thenature of resistance behaviors, and adopting amultilevel rather than a single-level approach tostudying resistance. Considering the relativelyscarce prior research and theorizing with respectto resistance in the IT literature, the contribution ofour model will be to shed light on the multileveland evolutionary nature of the resistancephenomenon.

We began our theory-building effort by assessingthe four extant models of resistance to IT. Then,using semantic analysis, we identified five basiccomponents of resistance to IT: behaviors, object,subject, threats, and initial conditions. We furtherexamined the extant models to carry out apreliminary specification of the nature of the rela-tionships between these components and to refineour understanding of the multilevel nature of thephenomenon. We pursued our theory buildingeffort by analyzing data from three case studies onthe implementation of electronic medical records.We verified the presence of the basic componentsand their relationships. We also used analyticinduction to uncover new constructs, relationships,and boundaries that could enrich our under-standing of the phenomenon and assist our theory-building process (Patton 2002). Finally, we re-visited prior research on resistance to IT in light ofour findings.

Assessing Extant Modelsof Resistance to IT

Adopting the political variant of interaction theory,Markus (1983) explains resistance in terms ofinteraction between the system being implementedand the context of use. She posits that a group ofactors will be inclined to use a system if theybelieve it will support their position of power. Ifthey think it might cause them to lose power, theywill resist.


MIS Quarterly Vol. 29 No. 3/September 2005 463

Joshi (1991) uses equity theory to propose amodel wherein individuals evaluate a given changeon three levels. They first assess the variation intheir equity status brought about by the system.They then compare it to that of their organization.Finally, they compare it to that of other membersof their reference group. They will resist if theyperceive inequity. Marakas and Hornik (1996)adapted a model of passive resistance misuse toexplain resistance behavior as passive-aggressiveresponses to threats or stresses that an individualwill, rightly or wrongly, associate with a newsystem. Martinko et al. (1996) propose an attri-butional model of individual reactions to IT thatposits that a new technology, internal and externalvariables, and an individuals experience with suc-cess and failure at tasks involving similar tech-nologies evoke causal attributions. In turn, theseattributions influence the individuals expectationsregarding future performance outcomes, whichthen drive his or her affective and behavioral reac-tions toward the technology and its use. The be-havioral reactions result in outcomes, the nature ofwhich influences the nature of future attributions.

All four models share the underlying assumptionabout the desirability, or lack thereof, of resistance.Contrary to commonly held assumptions aboutresistance that regard it as a critical obstaclepreventing organizations from reaping the potentialbenefits of an IT implementation, the modelsconsider resistance to be neither good nor bad.For instance, rather than seeing resistance as anundesirable result to be avoided or overcome,Markus posits that it can have negative or positiveeffects. When it generates conflict and consumestime and attention, resistance is dysfunctional andcan even be destructive. It can nevertheless befunctional for organizations if it prevents the imple-mentation of systems that, by increasing stress orturnover or by eroding performance levels, wouldhave negative impacts. While Joshi does notexplicitly enunciate the assumption about thedesirable or undesirable nature of resistanceunderlying his model, it is clearly implicit in theargument. Indeed, Joshi posits that extremeinequities should be avoided, because highlyinequitable treatment of some users is likely toinfluence the equity perception of others, hence

causing disruptions. Therefore, when resistanceprevents the use of a system that has inequitableconsequences, it plays a useful role. For Marakasand Hornik, resistance is a means by which userscommunicate their discomfort with a system thatmight be flawed. While they suggest interventiontechniques aimed at overcoming resistance,Martinko et al. do not consider resistance intrin-sically bad; rather, their techniques aim at alle-viating those situations where dysfunctionalattributions can have debilitating effects.

A second common assumption pertains to thenature of the relationship between resistance andits antecedents. While only Markus explicitly usesthe term interaction theory, all four models indeedassume that resistance results from the mutualadjustment of several antecedents. Markussmodel portrays resistance as resulting from theinteraction of system features with the intra-organizational distribution of power. When [theinterests and intentions of both users and de-signers] are very similar, resistance rarely occurs(p. 433); conversely, if users perceive that thesystem represents a loss of power, they are likelyto resist. More precisely, the model stipulates thatthe strength of resistance would be related to thesize of the loss and its perceived importance.Similarly, Joshis model posits that in anyexchange relationship, individuals are constantlyconcerned about their inputs, outcomes, and thefairness of the exchange (p. 231). They alsoconstantly compare themselves with others. Whena system is implemented, individuals assess theimportance of variations in inputs and outcomes.If they perceive the variations as detrimental, theywill experience inequity. The stronger the per-ceived inequity, the more likely they are to resist.Focusing on passive resistance misuse, Marakasand Hornik argue that it is situational. It resultsfrom the interaction between the uncertain condi-tions provoked by a new system and individualtraits (e.g., rigidity and resentment). When thisinteraction leads to perceived threats or stresses,the individual may respond with passive resistancemisuse behaviors. Finally, Martinko et al. suggestthat the intensity and nature of resistance to ITdepend on the interaction of a number of factors:internal and external influences as well as theindividuals prior experience with the technology.



Notwithstanding the contributions of these models,we identified a number of areas to be expandedupon. The first is that each model adopts a single-level perspective in explaining resistance. Threeexplain resistance at the individual level (Joshi,Marakas and Hornik, and Martinko et al.). Thefourth focuses on resistance at the group level(Markus). By adopting a single-level approach,each model helps better understand some aspectsof resistance, but dismisses others. It has beenargued that a multilevel perspective, which aims atproviding a more integrated understanding ofphenomena that unfold across levels in an organi-zation, provides a richer portrait of organizationallife and makes explicit the relationships betweenconstructs that were previously unlinked (Klein andKozlowski 2000). Second, while they have strongtheoretical groundings, three of the models arewanting in terms of empirical evidence. Indeed,while Joshi and Marakas and Hornik illustrate theirmodels with examples, either fictitious or fromprevious studies, Martinko et al. keep the dis-cussion at a conceptual level. Only Markus usesdata from a case study. Finally, while the ITimplementation process is not instantaneous andresistance may evolve over time, the models donot explicitly take into account the role of time inthe unfolding of resistance.

Defining Resistance

The next step of our theory building effort was todefine the theoretical phenomenon under studyand its underlying constructs (Eisenhardt 1989).Because the semantic structure of words is said toreflect their meaning (Pitt 1999), we used semanticanalysis in our effort to define the phenomenon ofresistance. In compositional semantics, themeaning of the whole is made up of the meaningof the parts. According to the principle ofdecompositionality, When an expression admitsanalysis as morphologically or syntactically com-plex, assume as an operating hypothesis that thesense of the expression arises from the com-position of the senses of its constituent parts(Wasow et al. 1983, p. 104). Thus, althoughvarious definitions of a concept might exist,

semantic decomposition can be used to revealcommon ground among them by separating aconcept into smaller concepts, down to semanticprimitives. While the term primitive refers to themost-basic concepts, no correct set of primitivesexists for a given domain. A set of primitives isdeemed adequate when knowledge about themcan be presented in a simple, meaningful way(Wilks 1992).

We conducted semantic decomposition using theguidelines set forth by Akmajian et al. (1984).First, we identified syntactically unstructuredexpressions of the language referring to the con-cept of resistance, taking them from severalsources. The primary source was the set of ninedefinitions of resistance previously identified in theIT literature. Because these definitions had beeneither borrowed or adapted from various referencedisciplines, these disciplines (psychology, politicalscience, sociology, and change management)were also searched for additional expressionsreferring to resistance. Second, we searched forcommonalities among the definitions proposed forthese expressions. Third, to establish correspon-dence among definitions, a small number ofrepeatable semantic primitives were identified foreach definition. While not all definitions have allprimitives, the pool of definitions taken from agiven discipline invariably reveals five basic,common primitives: resistance behaviors, objectof resistance, perceived threats, initial conditions,and subject of resistance (Table 1).

Resistance behaviors. Behavior is the primarydimension of resistance, inasmuch as words likereaction (Ang and Pavri 1994), behavior (Markus1983), and conduct (Zaltman and Duncan 1977)are found in almost all definitions. Resistancebehaviors exist across a spectrum, from beingpassively uncooperative to engaging in physicallydestructive behavior (Marakas and Hornik 1996),or from lack of cooperation to sabotage (Carnall1986). The taxonomy proposed by Coetsee(1993, 1999) is useful in this regard, allowing theclassification of the resistance behaviors accordingto four levels of resistance: apathy, passive resis-tance, active resistance, and aggressive resis-tance. Apathy includes behaviors such as inac-

Pr

i

m

i

t

i

v

e

s

P

r

i

m

i

t

i

v

e

s

Table 1. Results of Semantic Analysis of Resistance to ITAuthor Keen (1981, p. 27) Markus (1983, p. 433) DeSanctis & Courtney (1983, p. 737)Expression Resistance to change Resistance Resistance to MISDefinition The tactical approach to implementation sees

resistance as a signal from a system inequilibrium that the costs of change areperceived as greater than the likely benefits

Behaviors intended to prevent theimplementation or use of a system orto prevent system designers fromachieving their objectives

Resistance to the MIS sometimes occurs whenPEOPLE experience changes in the content oftheir jobs and their relative power vis--visothers

Initialconditions

from a system in equilibrium in the content of their jobs and their relativepower vis--vis others

Perceivedthreats

costs of change are perceived as greater than thelikely benefits

changes

Object change implementation or use of a system orto prevent system designers fromachieving their objectives

the MIS

Resistancebehaviors

signal behaviors

SUBJECT PEOPLE

Author Joshi (1991, p. 231) Ang and Pavri (1994, p. 130) Martinko et al. (1996, p. 322)Expression Resistance to inequity Resistance to change Resistance to the implementation of an ITDefinition Equity theory suggests that the greater the

inequity or decline in the net gain, the greater theresulting distress would be, INDIVIDUALS whoexperience the distress of inequity or loss ofequity are likely to resist it by attempting to mini-mize their inputs and others outcomes as wellas attempting to increase others inputs

Resistance to change is a normalpsychological reaction when theperceived consequences (e.g., loss ofpower) are negative

USER resistance [to the implementation of IT]can take on a wide variety of behavioral forms

InitialconditionsPerceivedthreats

suggests the greater the inequity or decline in thenet gain, the greater the resulting distress

perceived consequences (e.g., lossof power) are negative

Object distress of inequity or loss of equity change to the implementation of ITResistancebehaviors

by attempting to minimize their inputs andothers outcomes as well as attempting toincrease others inputs

psychological reaction behavioral forms

SUBJECT INDIVIDUALS USER

Pr

i

m

i

t

i

v

e

s

Table 1. Results of Semantic Analysis of Resistance to IT (Continued)Author Marakas and Hornik (1996, p. 209) Lee and Clark (1996-1997, p. 121) Enns et al. (2003, p. 162)Expression Passive resistance misuse Resistance to reengineering ResistanceDefinition A recalcitrant, covert behavior resulting from

both fear and stress stemming from the intrusionof the technology into the previously stableworld of THE USER

The resistance may be nothing morethan inertia, but it also stems from ahealthy suspicion of new andunproved market systemsFurthermore, PARTIES AFFECTEDADVERSELY by the change areexpected to fight reengineeringefforts

Resistance is displayed when THE TARGETavoids performing the requested action byarguing, delaying, etc.

Initialconditions

previously stable world of the user

Perceivedthreats

fear and stress a healthy suspicion of new andunproved market systems

Object intrusion of the technology Change; reengineering efforts the requested actionResistancebehaviors

recalcitrant, covert behavior inertia, fight avoids performing [the requested action] byarguing, delaying, etc.

SUBJECT THE USER PARTIES AFFECTED ADVERSELY THE TARGET



tion, distance, and lack of interest. Manifestationsof passive resistance are rather mild; they includedelay tactics, excuses, persistence of formerbehavior, and withdrawal. Active manifestationsare typified by strong but not destructive behav-iors, such as voicing opposite points of view,asking others to intervene or forming coalitions.Finally, aggressive resistance behaviors such asinfighting, making threats, strikes, boycotts, orsabotage seek to be disruptive and may even bedestructive.

Object of resistance. The verb resist is transitive,which means that it takes a direct object. Iden-tifying and understanding this object is critical,because resistance is shaped in part by the con-tent of what is being resisted (Jermier et al. 1994).Psychotherapy clients resist the counseling pro-cess (Cowan and Presbury 2000), employeesresist managements efforts to institute change(Dent and Goldberg 1999), and users resist theimplementation of an information technology (Joshi1991).

Perceived threats. All fields of inquiry share theidea that for resistance to occur, some threat hasto be perceived. Dent and Goldberg (1999) arguethat people do not resist change per se. Rather,they react to the threats that they perceive will bebrought by that change. Perceived threats areidentified by expressions such as overwhelmingemotional pain (Freud 1919) or the perception ofa dangerous situation (Marakas and Hornik 1996).Employees resist changes that they believe willcause either loss of status, loss of revenue, or lossof power (Dent and Goldberg 1999). Individualsresist the implementation of a system when theyperceive inequity (Joshi 1991); groups resist itwhen they fear a potential loss of power (Markus1983).

Initial conditions. Some authors emphasize therole of subjectivity in resistance (Jermier et al.1994). They argue that understanding resistancedemands attention to subjectivities (Collinson1994). Some individuals or groups may accept achange, but others may resist it. Apart from theability, or lack thereof, to perceive a threat, someinitial conditions such as distribution of power

(Markus 1983) or established routines (Marakasand Hornik 1996) may influence how threateningan object is perceived to be.

Subject of resistance. The subject of resistance isthe entity that adopts resistance behaviors. Whenresistance is studied from a psychologicalperspective, the subject is the individual (Cowanand Presbury 2000). When it is studied from apolitical perspective, the subject is generally agroup of actors (Jermier et al. 1994). In ITresearch, sometimes the subject is an individual(Marakas and Hornik 1996). At other times, it canbe a group (Markus 1983) or an organization (Angand Pavri 1994).

Toward a Multilevel View ofResistance to IT

In light of this set of primitives, we furtherexamined the four extant models and derived apreliminary model of resistance to IT. Table 2synthesizes the four models in terms of the fiveprimitives. Although their instantiations differ fromone model to another, all five are included in eachmodel. In terms of relationships between theprimitives, all four models posit that perceivedthreats result from the interaction between a givenset of initial conditions and an object. In turn, thepresence of perceived threats is a necessarycondition for resistance behaviors to occur. Whilesuch a preliminary specification allows the analysisof the interplay among initial conditions, an object,perceived threats, and resistance behaviorsadopted by a subject, it fails to capture themultilevel and dynamic nature of resistance.

Table 2 hints at the usefulness of adopting amultilevel perspective to the study of resistance toIT. First, multilevel theory posits that a phenom-enon can be studied at two levels: the individualand the unit (dyad, group, function, or organiza-tion) (Klein and Kozlowski 2000). As shown inTable 2, resistance to IT can be conceptualized ateither level. Joshi (1991), Marakas and Hornik(1996), and Martinko et al. (1996) adopted an

Table 2. The Extant Models of Resistance to IT Implementation

Article Object Initial Conditions Interaction Perceived Threat Resistance Behaviors Subject

Markus1983

Patterns of inter-action prescribed bythe system

Patterns that alreadyexist in the settinginto which the systemis introduced (herethe political setting)

Mismatches [may]create resistance-generatingconditions

Power loss for agroup, power gain foranother

Speaking resentfully ofthe system

Continuing to followformer procedures

Group

Joshi 1991 Inputs and out-comes prescribed bythe system (self,group of reference,employer)

Inputs and outcomesthat already exist(self, group ofreference, employer)

Mismatches [may]create a perceivedthreat

Distress of inequity orloss of equity

Attempting to minimizetheir inputs and othersoutcomes as well asattempting to increaseothers inputs

Individual

Marakasand Hornik1996

New routines andmodes of workbrought about by anew IT

Established routinesand modes of work

Individuals rigidityand resentment

Interaction betweendifference in de-mands of the IT,established modesof work, and theindividuals rigidityand resentment

Stress and fear Passive resistancemisuse

Individual

Martinkoet al. 1996

Characteristics ofthe IT

Attribution schemataof the individualExternal influencesInternal influences

Attributionalprocesses

Efficacy expectationsOutcomeexpectations

Low levels of useLack of useHarmful use

Individual



individual-level conceptualization, and Markus(1983) adopted a unit-level conceptualization.Second, multilevel research categorizes models aseither single-level or cross-level. Single-levelmodels specify the relationships among constructsconceptualized at the same level. Cross-levelmodels link constructs from different levels. FromTable 2, Markuss and Marakas and Horniksmodels are clearly single level. Table 2 suggeststhat the other two models incorporate antecedentsfrom different levels. Joshis model refers to theself, the group of reference, and the employer.Martinko et al.s model posits that externalinfluences help explain individual resistancebehaviors.

Adopting a multilevel perspective, we concep-tualize resistance to IT as a unit-level phenom-enon. We chose to focus on group resistancebecause, in most circumstances, resistance froma single user would not be sufficient to severelyaffect the overall implementation process or leadto system abandonment (Dickson et al. 1974;Markus 1983). By adopting a multilevel perspec-tive, we also acknowledge the possibility that ourresulting model will be cross-level, recognizing thatindividual or unit-level constructs may influencegroup resistance.

In multilevel research, studying a phenomenon atthe unit level calls for examining the process bywhich lower-level properties manifest at a higherlevel (Klein and Kozlowski 2000). In line with thisprecept, we seek to understand the bottom-upprocess by which group resistance behaviorsemerge from individual behaviors. Finally, we takea longitudinal perspective, because the multilevelapproach calls for a consideration of the roleplayed by time because the temporal scopeaffects the apparent origin and direction of manyphenomena (Klein and Kozlowski 2000 p., 23).

Research Method

Studying resistance to IT with a multilevelapproach requires a rich, longitudinal data set.Data from three case studies of implementation of

electronic metical records (EMRs) in hospitalsettings were analyzed. EMRs are systems thatallow access to patients records at all times fromdifferent locations, either to retrieve data, observetreatment regimens, or obtain test results. Weselected hospital settings because of the presenceof clearly identified groups of actors. Our studyfocused on one of these groups, the physicians.Each case describes the implementation processfrom software selection to installation and use ornonuse by the hospitals physicians. This timeframe sets the models temporal boundaries, thusframing the set of antecedents that will lead tophysicians resistance behaviors. Consequently,the model does not seek to explain events thatoccurred before software selection. Rather, theresults of such events are taken as initialconditions.

Research sites were selected to maximizevariation and allow comparison (Guba and Lincoln1989). As shown in Table 3, similarities andvariations pertain to three characteristics of thecases: hospital type, software package, and out-come. In terms of similarities, two sites wereuniversity hospitals, two had selected the sameEMR software, and two implementations hadfailed. In terms of variations, two hospital typesare present, two different EMR packages wereimplemented, and two different outcomes wereobserved.

To reach an appropriate degree of internal validity,three sources of evidence were used: directobservation, documentation, and interviews (Duband Par 2003). One of us spent several days ateach site observing how each EMR was used inunits providing representative or revealing situa-tions. System and project documentation, minutesfrom committee meetings, as well as memoran-dums and letters were analyzed. Data gatheredfrom these sources was used to corroborate,validate, and complement the interview data.

Interviewees represented the major stakeholders:physicians, nurses, and administrators. Initialrespondents were the project manager, thenursing director, and the medical director from



Table 3. Selected Cases

Hospital Type Software Package

Result of theImplementation

ProcessPeople

Interviewed

Case 1 Community hospital Alpha Failure Physicians: 7Nurses: 4Managers: 5

Case 2 University hospital Alpha Success Physicians: 4Nurses: 4Managers: 5

Case 3 University hospital Delta Failure Physicians: 4Nurses: 6Managers: 4

each site. A snowball sampling strategy was usedto identify subsequent respondents. Each initialinterviewee was asked to suggest other respon-dents who were knowledgeable about the project,represented a subset of the hospital population,and/or had exhibited extreme behavior duringimplementation (Crabtree and Miller 1992; Patton2002). The average interview lasted one hour.Respondents were asked to provide a narrative ofthe implementation, from the decision to imple-ment an EMR to project termination. Interviewstypically began with a generic question thatallowed the respondents to express how theyexperienced the implementation process. Morespecific questions were asked as required toensure that the data from each case coveredsimilar material and would allow cross-casecomparisons (Miles and Huberman 1994). Datacollection ended at the point of redundancy, whenefforts to get additional members cannot bejustified in terms of the additional outlay of energyand resources (Lincoln and Guba 1985, p. 233).One of us produced and coded complete tran-scripts using NUD*IST. To ensure validity of thecoding process, three additional researcherscoded 45 excerpts from the cases. The inter-coder reliability was 0.77. The initial codingallowed for the identification of behaviors, pre-cursors, threats, subjects, and objects of resis-tance and was an integral part of the analysis.

Meaning was attributed to the data, and effortswere made to ensure that the coding processpreserved existing relationships in the data (Milesand Huberman 1994). Because qualitative dataanalysis is an open and iterative process, roomwas made for modifications dictated by the dataitself. When elements were not representedadequately by the selected codes, new categorieswere allowed.

Data was analyzed in two stages (Eisenhardt1989). Within-case analysis was performed first toallow the unique patterns of each case to emergeand to provide researchers with a rich under-standing of each case, hence accelerating cross-case comparisons. Second, a cross-case analysisusing analytic induction was conducted in searchof common patterns and unique features. Analyticinduction is an alternative to purer forms ofphenomenological inquiry and grounded theory(Patton 2002). More precisely, in analyticinduction

researchers develop hypotheses, some-times rough and general approximations,prior to entry into the field or, in caseswhere data already are collected, prior todata analysis. These hypotheses can bebased on hunches, assumptions, carefulexamination of research and theory, or



combinations. Hypotheses are revised tofit emerging interpretations of the dataover the course of data collection andanalysis. (Gilgun 1995, pp. 268-269).

For the within-case analysis, data reduction andpresentation techniques were used (Miles andHuberman 1994). First, separate tables werecreated to distinguish between different episodesof resistance. Segments of the transcripts thatreported specific resistance behaviors were thenidentified. The segments were examined toidentify the objects of these behaviors and theirassociated threats and initial conditions. Theywere then organized in conceptually clusteredmatrices so as to build a logical chain of evidence.The resulting chains of evidence, presented in theappendices permitted an explanation-buildinganalytic strategy (Yin 1994).

Cross-case analysis was conducted using twotactics said to enhance the probability of capturingnovel findings among the data (Eisenhardt 1989).First, categories were selected for identifyingpatterns in each dimension of the framework.Second, cases were compared in pairs to identifythe subtle similarities and differences betweenthem. In addition, the chains of evidence devel-oped in the within-case analyses helped build ageneral explanation that could be applied to eachindividual case, while simultaneously taking intoaccount differences between cases.

Within-Case Analysis: Episodesof Resistance

In each case, we identified several episodes ofphysicians resistance. This section presents adetailed narrative of the events so that thedynamics of each episode can be understood. Notonly does it allow the identification of diverseresistance behaviors, but the threats, initial condi-tions, subjects, and objects of resistance can alsobe distinguished. What results is an in-depthunderstanding of the surfacing, progression, andculmination of physicians resistance throughoutthe implementation process.

Case 1

Case 1 is an acute care hospital where physiciansare remunerated on a fee-for-service basis. Thisnew hospital was to serve as a model paperlesshospital. As they were recruited, nurses andphysicians were informed that they would be usingan EMR. However, when the hospital opened, thesystem had not yet been installed. Implementationof Alpha began two years later. Five years aftersystem introduction, only the first module (Phase I,test requisitions/test results) was still in use. Thesecond module (Phase II, nursing care) had beenwithdrawn after major conflicts, first between thenurses and the physicians, then between thephysicians and the administration. Workstationsinstalled at bedside, which had almost never beenused, had been removed. The system wasrunning at 25 percent of capacity, and there wereno plans for expanding its use. As suggested byFigure 1 and in Appendix A, while resistancebehaviors initially consisted mostly of apathy andlack of interest, they later became more aggres-sive. The project was ultimately abandoned andthe Department of Health put the hospital undertrusteeship.

Prologue. A multidisciplinary committee ofphysicians, nurses, and other professionals wasformed to evaluate systems on the market andidentify the one best suited to meeting thehospitals needs. After an exhaustive review, thecommittee selected Alpha. Management inviteddepartment heads (physicians) and some nursesand technicians to travel to Alpha headquarters inCalifornia to try the system. Upon returning, allparticipants approved the decision to purchase thesystem.

Employees and physicians selected thesystem in a positive group process.People went to a center in California thatalready had the system, they saw it inoperation, and they gave their approval.The hospital wasnt quite open.Everyone was in agreement.Severalphysicians took part in the process.(Administrator 7)



Behaviors

Perceivedthreats

Episode

Adoption

Neutrality

Apathy

Passive resistance

Active resistance

Agressive resistance

Case 1

e0 e1 e2 e3 e4

Work

and

econ

omic

Loss

of

status

Loss

of po

wer

(vs.nu

rses)

Loss

of po

wer (v

s.

admi

nistat

ors)

Behaviors

Perceivedthreats

Episode

Adoption

Neutrality

Apathy

Passive resistance

Active resistance


Case 1

e0 e1 e2 e3 e4

Work

and

econ

omic

Loss

of

status

Loss

of po

wer

(vs.nu

rses)

Loss

of po

wer (v

s.

admi

nistat

ors)

Behaviors

Perceivedthreats

Episode

Adoption

Neutrality

Apathy

Passive resistance

Active resistance


Case 2

e0 e1 e2 e3

Reorg

aniza

tion

of wo

rk Abil

ityto

deliv

erca

re

Behaviors

Perceivedthreats

Episode

Adoption

Neutrality

Apathy

Passive resistance

Active resistance


Case 2

e0 e1 e2 e3

Reorg

aniza

tion

of wo

rk Abil

ityto

deliv

erca

re

Figure 1. Episodes of Resistance



Behaviors

Perceivedthreats

Episode

Adoption

Neutrality

Apathy

Passive resistance

Active resistance


Case 3

e0 e1 e2 e3 e4

Work

and

econ

omic

Loss

of po

wer

(vs.nu

rses)

Loss

of po

wer

(vs.ph

armac

ists)

Loss

of po

wer (v

s.

admi

nistat

ors)

Behaviors

Perceivedthreats

Episode

Adoption

Neutrality

Apathy

Passive resistance

Active resistance


Case 3

e0 e1 e2 e3 e4

Work

and

econ

omic

Loss

of po

wer

(vs.nu

rses)

Loss

of po

wer

(vs.ph

armac

ists)

Loss

of po

wer (v

s.

admi

nistat

ors)

Figure 1. Episodes of Resistance (Continued)

Episode 1: Apathy. At the time Phase I wasinstalled, resistance behaviors mostly took theform of inaction and lack of interest, as physiciansdid not participate in the training sessions and putlittle effort into learning to use the system. Withthe systems introduction, the principal tool inmedical practice, the paper file, was replaced witha computer monitor. Hospital paper files areorganized in a standard manner, and most physi-cians had received their medical training with thistype of record. They know how to quickly locatethe information they need, whether it is thepatients vital signs, laboratory tests, or notes fromearlier consultations. As suggested by the quotesin Appendix A, the new system represented asignificant change in the interface between thephysician and the patient file. Rather than beingable to locate different parts of the record throughcolor-coded tabs and documents, they had tonavigate from one window to another according toan unfamiliar logic. The system not only had animpact on work methods, it also required that

physicians spend more time managing records. Itwas estimated that using the system addedbetween 1 and 2 hours of work every day. Giventhat doctors in this hospital were paid by proce-dure, slowing down the process of carrying outmedical procedures threatened their economicwell-being. One doctor describes the impact of heruse of the system as follows:

This means that it cut two office hours outof every day. Plus, since a day only has24 hours, once I have logged 12 hours ofwork, Ive had it. So this gets to be veryfrustrating, when were told that its all thesame, when over time were talking abouta lot of money. (Physician 3)

In this instance, the interaction between thesystems features and the initial conditions of workhabits and compensation system led doctors toperceive a threat to the organization of their workand their economic well-being. Resistance behav-



iors of inaction and lack of interest followed; theirobject was the system itself and its features.

Episode 2: Passive Resistance. Phase II, whichwas introduced 18 months after Phase I, broughtabout new issues. Under traditional medical prac-tice, and even with Phase I, doctors would gener-ally prescribe care and treatment by giving verbalinstructions to nurses. With Phase II, this was nolonger feasible. From this point on, prescriptionshad to be entered in the system, and only doctorswere allowed to do so. Each user had a personalkey that gave access to particular modules. Onlydoctors keys gave access to the prescriptionmodules. The system, therefore, introduced consi-derable changes to methods for prescribingtreatments and care.

Several physicians refused to accept these newresponsibilities, a manifestation of passive resis-tance. Some refused to enter their prescriptions,arguing that they were not secretaries, and thatcaring for the patients was more important thanentering data. A physician reports the following:

[Dr. X told me that] one day he had toprescribe the removal of a catheter.Given the large number of steps requiredto enter the request into the system, hefelt that it was faster to just remove thecatheter himself rather than entering theprescription. (Physician 3).

Physicians felt that the system required them toperform clerical tasks that were inappropriate for adoctor and this represented a threat to theirprofessional status. At this point, in addition tothreatening the physicians organization of workand economic well-being, the system threatenedtheir very status as health professionals. Severaldoctors argued that administratorspeople whodid not understand the impact of the new systemon work methodswere imposing the changes.Once again, the object of resistance was thesystem itself and its features.

Episode 3: Active resistance. While physiciansreacted negatively to Phase II, nurses in generalappreciated it.

Phase II was the computerized plan ofcare. For nurses, it was an excellenttool, it was great, and it made it possibleto produce so many reports, controlquality, there were lots of things we coulddo with it. (Administrator 7)

Eventually physicians started to feel that, inaddition to the other impacts, the system wasundermining their power vis--vis the nurses. Aphysician reported that when doctors refused toenter prescriptions for various treatments in thesystem and verbally asked the nurses to performthe treatments, some would respond: No, I wontremove [an IV or a catheter] unless your prescribeit. No, I wont [take a blood sample or check vitalsigns] unless you prescribe it (Physician 6).Appendix A provides additional evidence of somenurses no longer accepting verbal orders fromdoctors, telling them that they should enterprescriptions into the system.

In terms of initial conditions, the distribution ofpower between physicians and nurses now playeda critical role. While physicians had traditionallyheld more power than nurses, the use of thesystem challenged this distribution of power. Thenurses behavior had diverged from what wastypical for a power structure in this type of organi-zation. Active resistance behaviors from physi-cians followed. They initially reacted to this threatby individually voicing their indignation. Theyquickly joined forces and informed the hospitalsCEO that they could not accept such a situation.The object of resistance no longer was the systemitself but its very significance.

Episode 4: Aggressive resistance. The CEOand the Board responded to the coalitionsdemands by telling physicians to keep on using thesystem while attempts were made to modify somefeatures. At the same time, the Board decreedthat six physicians, identified as champions of theresistance, should be denied the right to admitpatients. In this health care system, physicians actas entrepreneurs, and hospital administrators donot have formal power over them. The physiciansfelt that the Board was trying to usurp their rights.At this point, the power balance between physi-



cians and hospital administrators came into playas an initial condition, as did their mode of asso-ciation with the hospital. The physicians powervis--vis the hospitals administration was threat-ened. The coalition reacted with aggressivebehaviors by rebelling and menacing to terminatetheir association with the hospital. At this point,the system itself and its meaning were irrelevant;the systems advocatesthe hospital adminis-tratorsbecame the object of resistance.

Epilogue. Some doctors resigned. Those whoremained asked their professional association forassistance. As a result of the resignations, theEmergency Room could no longer function, andconcerns arose that the hospital might have toclose. The Department of Health intervened, dis-missed the CEO, and put the hospital undertrusteeship. The system was dramatically down-sized with a mere 25 percent of its functionalityremaining in operation.

Case 2

Case 2 is a university hospital where physiciansare partly salaried and partly remunerated on afee-for-service basis. Implementing an ERM wasa way to prepare for the 21st century. As in Case1, Alpha was the selected software. Initially thesystem was well-received by physicians, who likedthe idea of using IT in their work. However, acrisis occurred after a few months when thepharmacy module was installed and the residentsformed a coalition. In response, the administrationhad the system modified. Four years after imple-mentation, the project was considered a success,with 65 percent of systems functionality opera-tional and plans to have 75 percent of them infunction within a few months. Case 2 had agreedto collaborate with two other hospitals that wereconsidering implementing the same software.

Prologue. The hospital hoped to computerizepatient records and eliminate paper. A multi-disciplinary committee was formed and reviewedthe systems available on the market while givingconsideration to the hospitals specific objectives.After detailed analysis of three systems, the

committee selected Alpha. Sensing that physi-cians cooperation would be essential to theprojects success, the committee consulted withthem, both individually and through departments.To the committees delight, the physicians wereenthusiastic about the project.

The real reason my colleaguesboughtinto the idea was that they said, Itll begreat. In the hospital, we do a lot ofresearch and teaching in addition totertiary care, and we often have to doretrospective and prospective analysesand manage spreadsheets on a dailybasisit is endless.Once we have thesystem, everything will be on it; well justpress a button, and itll be incredible.They saw its advantages in terms ofbetter care, teaching, and research.They bought the system enthusiastically.(Physician 10)

Episode 1: Passive resistance. The firstmodule implemented handled admissions, trans-fers, and discharges. It was followed by a testrequisitions and test results module. Most physi-cians initially adopted the system without anymajor incidents. However, with time, the systemfeatures led them to perceive a threat to how theirwork was organized, mainly because they had toenter prescriptions themselves rather than give theinformation to a nurse as they had in the past.This approach was not only an important changeto how their work was organized, it also slowedthem down, particularly because the system hadan unduly long response time. In this case, therelevant initial condition was their existing workhabits, which they considered efficient andeffective.

As shown in Appendix B, some physicians reactedto this threat with passive resistance expressedthrough humor. For example, a resident, facedwith rounds that took all day to complete instead ofseveral hours, decided to make a practical joke byprescribing a complete blood count every hour onthe hour for five years. This overloaded thesystem and resulted in a shut-down. Anotherexample involved x-ray requests. Traditionally, x-



rays were requested by completing a form thatlisted possible justifications that were ordered soas to present the most frequent reasons first. Inthis system, the window for x-ray requests alsolisted justifications on which to check off theappropriate item. The first item on the list wasbullet wound. The hospital served a small com-munity where such a case might occur once ayear, so some physicians responded to what theysaw as the ridiculous prominence of bullet woundby checking it off as the reason for every x-rayrequest. Here, the system and its features werethe object of physicians resistance.

Episode 2: Active resistance. The implementa-tion of the pharmacy module led to major prob-lems. The module was not only perceived as aninefficient way for prescribing medication, it alsolengthened a response time that was alreadydrawing criticism. The physicians now consideredthe system a threat to patients safety and theirability to deliver quality care. In terms of initialconditions, the system conflicted with the physi-cians social values, under which the quality ofcare played a critical, nonnegotiable role. For thisreason, they actively manifested their resistance,organizing a formal protest. The residents sent aletter to the management in which they demandeda resolution to the response time issue and awithdrawal of the pharmacy module. They alsothreatened to return their keys. This action wouldhave meant a return to paper files. Here again,the object of resistance was the systems features.

Epilogue. In response to the residents demands,the administration took requests for changes intoconsideration and responded wherever possible.As a result, the schedule of system implementationwas relaxed and the pharmacy module withdrawnso that the necessary improvements could bemade. Four years after the implementation pro-cess began, all parties were using the system andthe hospital considered the project a success.

Case 3

Case 3 is a university hospital where physiciansare paid on a fee-for-service basis. When the timecame to change its laboratory, admissions, radio-

logy, and pharmacy systems, the hospital opted foran EMR. The Delta software was chosen and thesurgery units acted as a pilot site. As implemen-tation began, the surgeons had a positive attitudetoward the EMR; indeed, they showed exceptionalenthusiasm. This quickly changed as they devel-oped reservations about how well the system mettheir needs. Soon they demanded, to the greatdisappointment of nurses, that the system bewithdrawn. They agreed to use the system againonly when the hospital appointed a nurse to enterdata for them. Even with this support, thesurgeons felt that the system was inadequate anddemanded a return to the paper file.

Prologue. Once the decision had been made toacquire an EMR, a selection committee includinga representative for physicians and another fornurses was formed. The committee assessedseveral systems and chose Delta, an EMR thathad already been implemented elsewhere in NorthAmerica.

We went to New York [and elsewhere inNorth America] and saw the system inoperation.In an intensive care unit inNew York, I saw the system, I saw resi-dents prescribing with it, I saw them re-ceive information through it. I was thrilled.I saw the interface between laboratory,radiology, and pharmacy, enabling physi-cians to have access to results anywherein the hospital; you could even have themat your office and monitor important meta-bolic data. The nursing care plan wascomputerized, you could have data on vitalsigns, or on a certain number of statisticswith which services could be optimized,and of course we saw the computerizationof the medical record. (Physician-Administrator 3)

To facilitate implementation, management formeda new committee joined with considerableenthusiasm by several physicians. Some physi-cians visited other hospitals where the system wasoperational. Once the system purchase wasfinalized, committee members were responsible foridentifying needs in their sector, a task they tookseriously.



Episode 1: Active resistance. Physicians in thesurgery units had a positive attitude towards thesystem and participated enthusiastically in trainingsessions. They wanted to participate in the pilotproject and invested time and energy preparingfor it.

There were people in surgery who werevery open to the idea of implementing thesystem, there was willingness, even ifsome individuals [grumbled about it],people were still willing to say yes, letstry it out, it will be a pilot project and itllbe fun, itll shake things up a bit, etc.(Nurse 4)

However, a month after the system was intro-duced, several surgeons noted that they werespending considerable time prescribing tests.Because laboratory results did not come back inelectronic form, however, they were not derivingany benefit. The surgeons and the residents com-plained forcefully about the systems complexityand the fact that it had prompted abrupt changesin their work habits. The system was a threatbecause it led to the reorganization of work anddisrupted the initial conditions: the physicianswork habits. Moreover, entering prescriptions wastedious and took time. The system required thatthey spend more time managing records, whichthreatened physicians economic well-being,because they were paid by procedure. Thesystems features became the object of theircomplaints.

Episode 2: Active resistance. Two monthslater, some surgeons mentioned that they wereconsidering not using the system anymore. As aresult, nurses felt cheated. The physicians dis-satisfaction stirred up preexisting conflicts with thenurses and resulted in new confrontations betweenthe two groups. On the one hand, the physiciansfelt like they were doing nurses work. Once theprescription data was entered, it was used tocreate nursing care plans. On the other hand, thenurses refused to enter the data when asked bythe physicians.

Nurses were upset by the physicians refusal touse the system. In the cafeteria, some were evenheard making disparaging comments about thesurgeons, questioning their ability to learn how touse the new system, eliciting rather causticresponses from physicians. Heated discussionsoccasionally led to an exchange of insults. Interms of initial conditions, hospital standards, theadministrative structure, and the traditionaldistribution of power all supported the physicianspower position. The system presented a threat tothis position because it could upset the existingbalance of power between physicians and nurses.It was, therefore, the significance of the system,and not the system itself, that was the object of thesurgeons resistance.

Trying to find a peaceful solution and satisfyphysicians, the administration asked nurses toenter data for the surgeons and even appointed afull-time nurse to do it.

Their damned machine. We hated itwith the problems we had, and the resi-dents were complaining. So at one pointwe said that we wanted nothing more todo with it, that we wouldnt enter anythingmore in it. They said they would give uswhat they called officers to do the dataentry. So there were some people up-stairs entering prescriptions. (Physician7)

Episode 3: Active/aggressive resistance. Aclear split quickly emerged between the twogroups. The nurses maintained that the systemworked, while the surgeons believed that itrepresented a danger to patients. After an incidentwhere a patient did not receive his medication,some even suggested that significant flaws in thesystem had put some patients lives at risk (seeAppendix C, Episode 3). The surgeons met andchose a representative to discuss it with thehospitals CEO. Eight months after the intro-duction of the system, the representative deliveredan ultimatum, demanding that the system bewithdrawn.



The residents still had problemsso wewent into the General Managers officewith the Director of Professional Ser-vicesand gave him an ultimatum: getit out of there, because if you dont,theres gonna be trouble. The problemwas that the machine couldnt keep upwith us. (Physician 7)

Here the threat appears to be associated with thequality of patient care. The relevant initial condi-tion seems to be group values, under which qualityof care is paramount. The object of resistanceappears to be the system. However, as a hospitaladministrator suggested (see Appendix C), it ispossible that this was only a pretext. Indeed, thenurses and physicians other than the surgeonsconsidered the system adequate. According to apediatrician,

Some errors did occur. Perhaps that wasthe point where the system was blamed,but errors can also be made when writing.There can in fact be even more errors ofinterpretation when you read someoneelses writing. (Physician 14)

The real object of resistance was, therefore, thesystems significance rather than the system itself.In turn, the real threat was physicians loss ofcontrol over their jobs because, using the pretextof the EMR, pharmacists were attempting to forcesurgeons to change their prescribing behaviors.Here, the relevant initial condition was the distri-bution of power between physicians and pharma-cists rather than the professional values held byphysicians.

People in the pharmacy said to them-selves: Doctors are supposed to pre-scribe narcotics every three days andyoure not doing it. Were going to makeyou do it. (Physician 7)

Episode 4: Aggressive resistance. The admin-istration responded to the surgeons ultimatum witha threat: if the surgeons refused to use the sys-tem, beds that had been allocated to them wouldbe given to physicians who were more positive

toward the system. The surgeons perceived thisas a threat to their power and status. Theyrebelled, explicitly asking their colleagues in othercare units not to hospitalize patients in any bedsthat became available this way. By threateningphysicians with the closure of hospital beds in thesurgery units, management was challenging thephysicians traditional right to manage admissions.The object of the physicians resistance was nolonger the system itself nor its significance, but itsadvocates.

Epilogue. The other physicians supported thesurgeons by refusing to admit patients into surgeryunits. With beds sitting idle, causing significantfinancial shortfalls, the administration decided, lessthan a year after its implementation, to withdrawthe system from these units.

Cross-Case Analysis: Theoretical Synthesis

The within-case analysis showed how perceivedthreats result from the interaction between relevantinitial conditions and an object of resistance,leading to physicians resistance behaviors. Usinga cross-case analysis, we now examine howresistance behaviors unfold over time and explaintheir bottom-up process of emergence.

How Resistance Unfolds

Our analysis reveals that, in the presence of mixeddeterminants, resistance behaviors vary in natureand intensity as implementation evolves. It alsosuggests the existence of triggers that influenceinitial conditions and the object of resistance.

Mixed determinants of resistance behaviors. Inall three cases, resistance behaviors initiallyincluded indifference, lack of interest, and com-plaints, which are instances of apathy, passiveresistance, and mild active resistance behaviors.In later episodes, we observed an active resis-tance behavior, the formation of coalitions. In the



final episodes of Cases 1 and 3, physicians threat-ened to resign, created trouble, and rebelled,which all correspond to aggressive resistancebehaviors. Along with an increase in the intensityof resistance behaviors, we observed a change inthe level of the perceived threats. Initially, per-ceived threats manifested themselves at theindividual level and were reported independentlyby various physicians (for example, reorganizationof work or economic well-being). In later episodesof Cases 1 and 3, the perceived threats charac-terized entire groups (for example, the balance ofpower between doctors and nurses). Similarly, inthe final episodes of these two cases, the per-ceived threats stemmed from the administrationsattempt to take away the physicians privileges. Ina multilevel perspective, when they characterizean entire group of people, perceived threats canthen be said to be at the unit level.

From a multilevel perspective, this reveals thepresence of mixed determinants of resistancebehaviors. Indeed, antecedents from both indivi-dual and unit levels influenced physicians resis-tance behaviors. Furthermore, it appears that thelevel of perceived threats influenced the degree ofseverity of these behaviors, because perceivedthreats at the individual level were associated withmore benign forms of resistance behaviors thanthose at the group level.

Triggers. Why did perceived threats change fromthe individual to the group level? A potentialexplanation is found in the dynamics of resistanceacross episodes. Because the interaction betweeninitial conditions and the object of resistance is atthe origin of perceived threats, these antecedentshave to be different for the perceived threats tochange levels. We observed that both antece-dents may change during implementation. Whilethe preliminary specification of the relationshipbetween the components of our model is useful forexplaining each episode of resistance, it fails toexplain the changes that occurred from oneepisode to another. Figure 2 proposes a processmodel, taken in a longitudinal perspective, which ishelpful in capturing the essential elements of theprocess under study (Applegate 1994; Langley1999).

In a given episode of resistance, perceived threatsare expected consequences. A longitudinal per-spective makes it possible to consider actual con-sequences of system use and highlights triggersthat can modify either the set of initial conditions orthe object of resistance. Triggers include conse-quences of system use, other actors actions,system advocates reactions to resistance behav-iors, and events related to the implementationprocess.

Effects of triggers on initial conditions. Whileinitial conditions are always present, they can beeither active or inactive. They are active whenthey play a role in the interaction with the object ofresistance; they are inactive when they do not.For instance, the initial condition distribution ofpower between the hospitals administration andphysicians remained inactive through Case 2 andonly became active in the latest episodes of Cases1 and 3. The nature of initial conditions may alsochange during a project. This occurred in Case 3,where the nature of the physicians associationwith the hospital was modified when theadministration offered beds reserved for surgerypatients to other specialties.

Triggers affect the upcoming episode by eithertransforming one or several initial conditions oractivating one that was latent. At the end ofEpisode 1 of Case 1, the perceived threatreorganization of work materialized as an actualconsequence of system use, which modified aninitial condition, work habits. In Case 2, an event,the introduction of the pharmacy module, activatedthe initial condition of group social values. At theend of Episode 3 of Case 1, the administrationthreatened to terminate some physicians asso-ciation with the hospital. The system advocatesresponse modified the rules that govern a physi-cians association with the hospital, thus changingthe initial conditions of Episode 4. FollowingEpisode 2 of Case 1 and Episode 1 of Case 3,nurses were reluctant to enter prescriptions forphysicians. Some even refused to performmedical acts if they had not been entered in theEMR by a physician. These actions by otheractors (nurses) not only activated the initial condi-tion distribution of power between nurses andphysicians, they also modified it.



Initial conditions

Object

Initial conditions

Object

T2T1

TriggersActual consequences

EventsSystems advocates reactions

Other actors actions

Individual or organizational

level

System features

Individual, group, or

organizationallevel

System featuresor

System significance

or System advocates

Interaction InteractionPerceived threatsPerceived

threatsResistance behaviors

Resistance behaviors

Initial conditions

Object

Initial conditions

Object

T2T1

TriggersActual consequences

EventsSystems advocates reactions

Other actors actions

Individual or organizational

level

System features

Individual, group, or

organizationallevel

System featuresor

System significance

or System advocates

Interaction InteractionPerceived threatsPerceived

threatsResistance behaviors

Resistance behaviors

Figure 2. Resistance to IT Implementation: A Longitudinal Perspective

Changes in the object of resistance. Asdepicted in the cases, the object of resistance mayalso change during implementation. We observedthree such object types: the system itself, itssignificance, and its advocates. In the early epi-sodes, numerous physicians in all three hospitalscomplained about specific aspects of the systemsinterface. Later, in Cases 1 and 3, the object ofresistance became the systems significance andthen the system advocates. We observed thatsuch a change occurs simultaneously with activa-tion of the initial condition, distribution of power.Recall that the nurses actions, following Episode 2of Case 1 and Episode 1 of Case 3, modified thedistribution of power between nurses and doctors.In the next episode, the object of resistance wasthe significance of the systems rather than theirfeatures. Also, again in Cases 1 and 3, the objectof resistance changed from system significance tosystem advocates when the hospital administrationresponded to resistance behaviors by trying toundermine the power of physicians. Why was nosuch change in the object of resistance observedin Case 2? Again, the answer lies in the distri-bution of power, because in this case no triggeractivated or modified the balance of power in thishospital.

Dynamics of resistance. We propose a dynamicexplanation of resistance to IT implementation. Attime T1, when a system is introduced, resistancebehaviors will result if a subject (here the group)

perceives threats from the interaction between thesystems features and individual and/or organiza-tional-level initial conditions. Consequences ofsystem use/nonuse, whether or not they had beenforeseen, will occur. These consequences, in turn,may change the nature of, or activate, one orseveral initial conditions. Other triggers (events,other actors actions, or system advocates re-sponses to resistance behaviors) may also modifythe set of initial conditions. The new set willbecome the initial conditions at time T2. If a triggeraffects an initial condition involving the balance ofpower between the group and other user groups,it also changes the object of resistance from thesystem to system significance. If the relevantinitial conditions pertain to the power of theresisting group vis--vis the system advocates, theobject of resistance is also modified from systemsignificance to system advocates. Again, at timeT2,, resistance behaviors will follow if threats areperceived from the interaction between the objectof resistance and initial conditions.

The Emergence Process of GroupResistance Behaviors

Because group resistance behaviors as a unit-level phenomenon are the aggregate of individualbehaviors, adopting a multilevel approach calls foran examination of the bottom-up process by whichindividual resistance behaviors emerge into group



resistance. This process can be one of eithercomposition or compilation. It is a compositionprocess when the unit-level phenomenonemerges from individual members shared per-ceptions, affect, and responses (Klein and Koz-lowski 2000, p. 33). Group norms are an exampleof this type of phenomenon. The process is one ofcompilation when the unit-level phenomenonemerges from different, independent individualcontributions that do not converge. Team mem-bers personality characteristics are an example.As group resistance to IT has never been studiedin a multilevel perspective, no a priori assumptionsabout the nature of the emergence process can bemade.

We proceeded with our cross-case analysis touncover the process from which the resistancebehaviors of physicians, as a group, emerge fromindividual resistance behaviors. The most salientresult is that the nature of the process differs fromearly to late implementation. Early in the imple-mentation, we observed in each case that whileseveral resistance behaviors were similar, theywere relatively independent. In Case 1, forinstance, there was no indication that the decisionof several physicians not to attend training ses-sions was concerted. Similarly, the two residentsin Case 2 who used humor to show their dis-approval of some system features did so inde-pendently of each other. In all three cases, weobserved that resistance behaviors were notidentical early in the implementation. Somephysicians appreciated the system, some wereindifferent, and some adopted passive or activeresistance behaviors. In Episode 1 of Case 1, forinstance, indifference was the most frequentresistance behavior observed, yet all physicianswere not neutral toward the EMR. Some actuallyliked it: Sitting down in front of a keyboard isntsomething that disgusted me, I didnt hate it. Iliked it. I saw it as something positive (Case 1,Physician 6), while others had a negative reaction:The screen froze.The delays were a bigproblem.If I have to deal with a computer systemthat cant respond, it just isnt going to work. Ill getdistracted and then Ill be angry at the system(Case 1, Physician 14).

When individual contributions to the higher-level

phenomenon vary, the dynamics show dispersionand exhibit nonuniform patterns, and individualcontributions are independent of each other, thebottom-up emergence process is described ascompilation (Klein and Kozlowski 2000). This typeof process took place in all three cases early in theimplementation. Physicians appeared to takestock of the systems features and their rela-tionship to their work time, workload, or economicwell-being.

In the later stages, we observed convergence ofbehaviors. In all three cases, physicians formedcoalitions. In Cases 1 and 3, they threatenedeither to resign or cause trouble, and they thenrebelled against the administration. Even physi-cians who initially appreciated the system or wereindifferent to it displayed such behaviors. InCase 1, a physician who initially was indifferenttoward the system later led a coalition thatrequested its withdrawal: On the first day wedidnt say Its worthless, We tried it. When youtry, and it doesnt work and people start to leave,someone has to act, so we acted (Case 1,Physician 4). In Case 3, a physician who initiallyfound the system useful, later joined forces withhis peers out of solidarity: I abandoned the pro-ject largely in solidarity with my colleagues (Case3, Physician 13). The vocabulary used by physi-cians in reporting on these stages further revealsconvergence of individual behaviors, with expres-sions such as collectively, blockade, petition, vote,and got together as a block. In multilevelresearch, when individual contributions to thehigher-level phenomenon are similar and thedynamics exhibit low dispersion, the bottom-upprocess of emergence is one of composition. Onecomment vividly illustrates the change in thenature of the emergence process, from early tolate implementation:

By the end, the physicians were actingas a monolithic group they had forgeda coalition with the CMDP [council ofphysicians, dentists, and pharmacists].They had become a monolithic group, butat the beginning their behaviors varied.During the three or four months thatfollowed the end of the first deployments,you had every sort of attitude. Some



physicians were trying to find a solution,some were trying to get around it, andsome refused to use it at all. Someplayed along. In other words, they dulyfilled out requests for improvements andchanges.Others, who were a bit moreexperienced with computersfoundtricks to get what they wanted.Othersbehavior was that of a typical doctor:when they dont like something, they canalways find a way around it.Theycontinued to prescribe on paper and hadnurses enter the information instead ofentering it themselvesthere wereothers who had an approach that was abit more radical. From the start, theysaid that they wouldnt use the systemuntil the problems had been ironed out.(Case 1, Physician-Administrator 8)

The physicians discourse provides further supportfor this conjecture. Referring to early stages, theyused expressions such as I couldnt care less, Iwanted the results, this has happened to me,and I really lost my temper, reflecting the impor-tance of the individual. Later, they used expres-sions such as we voted, everyone agreed, weall thought, and even referred to themselves as agroup, saying, for example, we physicians or thedoctors.

How can we explain this change in the bottom-upprocess of emergence? As illustrated in Figure 3,in all three cases the turning point appears to bethe activation of one or several group-level initialconditions. In Episode 2 of Case 2, when thepharmacy module was introduced, the groupcultural values of physicians became active asinitial conditions because the module was felt toput patients health at risk. At this time, all hellbroke loose. The residents formed a coalition andasked the administration to withdraw the moduleuntil the system was better structured. In Epi-sode 3 of Case 1, the distribution of powerbetween nurses and physicians became relevantas an initial condition because it had beenmodified by the nurses actions. Doctors initiallyreacted to the threat to their power by individuallyvoicing their indignation. They then came to an

agreement and quickly notified the GeneralManager that [they] couldnt work that way(Case 1, Physician 4). A physician describes thisepisode as pandemonium. In Case 3, the turningpoint was when the distribution of power betweennurses and doctors, and then between physiciansand pharmacists, became the relevant initialconditions. Infighting was quickly followed by theformation of a coalition.

In the terminology of multilevel research, such achange from time T1 to time T2 is termed levelchange up. It represents situations where, withthe passage of time, individuals in a group changefrom being independent of each other to beinginterdependent in a homogeneous group(Dansereau et al. 1999).

Revisiting Prior Modelsof Resistance to ITImplementation

We propose a model of resistance to IT imple-mentation where, as suggested by prior models,resistance behaviors occur following perceivedthreats that result from the interaction betweeninitial conditions and a given object. These modelsrefer to several resistance behaviors such aspassive resistance, active sabotage, oral defama-tions (Martinko et al. 1996), covert procrastination,protesting, criticism (Marakas and Hornik 1996),and not using the system or sabotage (Markus1983). By adopting Coetsees (1993, 1999)taxonomy and populating it with instances ofresistance behaviors to IT implementation, we addto the extant knowledge by organizing thesebehaviors according to their intensity.

In terms of antecedents, each of the prior modelsacknowledges a single instantiation of thesecomponents. For instance, Markus (1983) seesthe object as the patterns of interaction prescribedby the system, the initial conditions as the politicalsetting, and the perceived threat as the power lossfor a group. In Martinko et al.s (1996) model, theobject is the systems features, the initial condi-tions are the attribution schemata of the individual



T1 T2

One or several group level initial conditions are active

Group resistance behaviors emerge from independent individual behaviors

Group resistance behaviors emerge from a convergence of individual behaviors

All active initial conditions are at the individual ororganizational level

COMPILATION PROCESS OF EMERGENCE COMPOSITION PROCESS OF EMERGENCE

T1 T2

One or several group level initial conditions are active

Group resistance behaviors emerge from independent individual behaviors

Group resistance behaviors emerge from a convergence of individual behaviors

All active initial conditions are at the individual ororganizational level

COMPILATION PROCESS OF EMERGENCE COMPOSITION PROCESS OF EMERGENCE

Figure 3. Level Change Up of Group Resistance Behaviors

influences, and the perceived threats are theefficacy and outcome expectations. By contrast,our model recognizes the potential presence ofmultiple instantiations for each antecedent. Theinitial conditions are identified at the individual,group, and organizational levels, the objects canbe either a systems features, a systems signi-ficance or a systems advocates. The perceivedthreats can be at the individual or group level. Byrecognizing the existence of mixed determinants ofresistance behaviors, our model includes a widerarray of antecedents, thus providing a richerportrait of the phenomenon.

In a multilevel perspective, our model posits that,when perceived threats are at the individual ororganizational level, the ensuing resistancebehaviors will be more benign than when they areat the group level. Adopting a single-levelapproach, two of the previous models offer anexplanation for the increased intensity ofresistance behaviors. Joshis (1991) equity modelposits that the stronger the perceived threat(inequity), the more likely an individual is to adoptresistance behaviors. Because we were studyingresistance at the group level, we did not try toverify this contention. Nevertheless, we recognizethe explanatory potential of Joshis argument.Markus posits that if one loses power to a party

who holds a higher position of authority, thestrength of resistance will be milder than if oneloses power to a peer group. At first glance, ourexplanation appears contradictory because physi-cians resistance behaviors were stronger whenthey felt threatened with loss of power vis--vis theadministration than when the nurses representedthe other party. However, recall that the physi-cians were not hospital employees. Instead, theywere independent entrepreneurs. As such, theywere not under the administrations authority.Notwithstanding this explanation, we believe thatthe relationships between perceived threats at agiven level and the intensity of ensuing resistancebehaviors needs to be specified.

Other models acknowledge the importance ofsome of the elements we identified as triggers.Marakas and Hornik (1996) state that the conse-quences of individuals actions will influence theirfuture actions. Martinko et al. imply that outcomesimpact the factors that influence future behaviorsand that other actors, coworkers, and supervisorsinfluence individuals resistance behaviors.Through the notion of political tactics, Markus alsohints at the importance of other actors actions.Our model not only identifies a wider set oftriggers, it also shows how resistance behaviorschange over time under the influence of triggers.



Because the prior models studied resistance at theindividual level, the process by which group resis-tance emerges from individual resistance was notrelevant in three of them. While Markuss modelstudies resistance at the group level, it does notattempt to explain this emergence process. Ouranalysis explains this process and shows how itdiffers from early to late implementation. Byrevealing that early in the implementationindividual behaviors are independent and that theylater converge, the model provides a finer-grainedunderstanding of group resistance to ITimplementation.

Implications for Researchand Practice

Our study is based on cases set in hospitals andhas physicians as its focal group. As a result,caution is required in generalizing our findings.Because of the power physicians hold in hospitals,they are freer to choose whether they use a givensystem than many other types of users. Tovalidate the model, it would be instructive to seehow, in similar settings, the resistance of othergroups, like nurses, evolves. Also, the modelsexternal validity would be improved by studying theimplementation of systems in different settings.

Notwithstanding these limitations, by taking amultilevel, longitudinal perspective, our model notonly explains the dynamics of group-level resis-tance but also shows how group resistance behav-iors emerge from individual behaviors. Whenstudying group resistance to IT in the early stagesof implementation, independent, individual behav-iors need to be analyzed rather than consideringthe group as a unified entity. In later stages, itthen becomes important to understand how andwhy individual resistance behaviors converge.

Some researchers have recently called for a shiftfrom the monolithic mental model of resistance(Dent and Goldberg 1999; Piderit 2000). Theresults from our study show that within the imple-mentation of a given system, resistance has awide variety of antecedents and manifestations

that can evolve and change in nature. By treatingresistance as a black box, researchers andmanagers limit their ability to deal with it. Byopening the black box of resistance, this study notonly helps explain how resistance develops, it alsoconfirms the contention of other investigators thatthere may, at times, be legitimate reasons forresistance.

For managers, our study reveals that inappropriateresponses to resistance behaviors ultimatelyprovoke resistance escalation. More precisely, ourdata suggests that early in the implementation, theobject of resistance is the system itself and itsfeatures. Tyre and Orlikowski (1994) called thisinitial period following the introduction of a systema window of opportunity. They posit that this timeperiod is ideal to adapt or improve the system.Our data supports their argument. It also suggeststhat, in later stages, when the object of resistanceis the significance of the system or the systemadvocates, resistance has become politicized. Asa result, managing resistance becomes a moredifficult undertaking. Our model provides a toolthat managers can use to recognize the first signsand causes of resistance. It should assist them torespond appropriately.

Acknowledgements

This research has been supported by grants fromthe Social Sciences and Humanities ResearchCouncil of Canada (SSHRCC) and by the QuebecFonds concert daide la recherche (FCAR). Weare indebted to the hospital administrators, nurses,and physicians who participated in this study, fortheir willingness to share their experience with us.We are most grateful to the senior editor, RonWeber, for his advice and guidance. We alsothank the associate editor and the three anony-mous reviewers for their valuable critiques andsuggestions.

References

Akmajian, A., Demers, R. A., and Harris, R. M.Linguistics: An Introduction to Language and



Communication (2nd ed.), The MIT Press,Cambridge, MA, 1984.

Ang, J., and Pavri, F. A Survey and Critique ofthe Impacts of Information Technology, Inter-national Journal of Information Management(14:2), 1994, pp. 122-133.

Applegate, L. M. Managing in an InformationAge: Transforming the Organization for the1990s, in Transforming Organizations withInformation Technologies, R. Baskerville, S.Smithson, O. Ngwenyama, and J. I. DeGross(eds.), North Holland, Amsterdam, 1994, pp.15-94.

Carnall, C. A. Toward a Theory for the Evaluationof Organizational Change, Human Relations(39:8), 1986, pp. 745-766.

Coetsee, L. D. From Resistance to Commit-ment, Public Administration Quarterly (23:2),1999, pp. 204-222.

Coetsee, L. D. A Practical Model for the Manage-ment of Resistance to Change: An Analysis ofPolitical Resistance in South Africa, Interna-tional Journal of Public Administration, (16:11),1993, pp. 1815-1858.

Collinson, D. Strategies of Resistance, Power,Knowledge and Subjectivity, in Resistance andPower in Organizations, J. Jermier, D. Knights,and W. Nord (Eds.), Routledge, London, 1994,pp. 24-68.

Cowan, E. W., and Presbury, J. H. Meeting ClientResistance and Reactance with Reverence,Journal of Counseling & Development (78:4),Fall 2000, pp. 411-419.

Crabtree, B. F., and Miller, W. L. Doing Quali-tative Research, Research Methods for PrimaryCare, Volume 3, Sage, Newbury Park, CA,1992.

Dansereau, F., Yammarino, F. J., Kohles, J. C.Multiple Levels of Analysis from a LongitudinalPerspective: Some Implications for TheoryBuilding, Academy of Management Review(24:2), 1999, pp. 346-357.

Dent, E. B., and Goldberg, S. G. ChallengingResistance to Change, The Journal of AppliedBehavioral Science (35:1), 1999, pp. 25-42.

DeSanctis, G., and Courtney, J. F. TowardFriendly User MIS Implementation, Communi-cations of the ACM (26:10), October 1983,pp. 732-738.

Dickson, G. W., Simmons, J. K. and Anderson, J.C. Behavioral Reactions to the Introduction ofa Management Information System at the USPost Office: Some Empirical Observations, inComputers and Management in a ChangingSociety, D. H. Sanders (Ed.), McGraw-Hill BookCompany, New York, 1974, pp. 410-421.

Dub, L., and Par, G., Rigor in InformationSystems Positivist Case Research: CurrentPractices, Trends, and Recommendations, MISQuarterly (27:4), December 2003, pp. 597-636.

Eisenhardt, K. M. Building Theories from CaseStudy Research, Academy of ManagementReview (14:4), 1989, pp. 532-550.

Enns, H. G., Huff, S. L., and Higgins, C. A. CIOLateral Influence Behaviors: Gaining Peers'Commitment to Strategic Information Systems,MIS Quarterly (27:1), March 2003, pp. 155-175.

Freud, S. Introduction to Psycho-Analysis andthe War Neuroses, The Standard Edition of theComplete Works of Sigmund Freud (17), JamesStrachey et al. (Eds.), The Hogart Press andthe Institute of Psychoanalysis, London 1953-1974, pp. 207-210.

Gilgun, J. We Shared Something Special: TheMoral Discourse of Incest Perpetrators, Journalof Marriage and the Family (57), 1995, pp. 265-281; reprinted in Qualitative Research andEvaluation Methods (3rd ed.), M. Q. Patton(Ed.), Sage Pub

Documents

education