Executive Overview - National Center for Border Security ...borders.arizona.edu/classes/mis696a/resources/readings/Lee-1989... · Case Study Methodology science is.' The model used

Executive Overview

This article provides practitioners with some valuable insights into the nature of MIS research. Oneof the more persistent research issues in our field is the value of detailed case studies. The classicalresearch paradigm, drawn from the physical sciences, requires the proposal of a theory from whichconclusions can be deduced and verified. In addition to testability, a successful scientific theory re-quires logical consistency and predictive power. The scientific method calls for control over experimen-tal variables in order to measure their effects on outcomes. Mathematical rigor and replicability areesteemed in such research as a means of better assuring objectivity.

The classical research requirements cannot be met in a case study. Such a study requires the obser-vation of uncontrolled, and often unmeasurable, events. Confounding variables typically make it ex-ceedingly difficult to sort our causal relationships. The imposition of classical experimental controlsand rigor, aimed at overcoming these problems, may require such an artificial environment that thevalidity of the results is called into question. And yet, detailed case studies can potentially add agreat deal to our understanding of important MIS issues.

The author argues persuasively that a case study can meet the requirements of rigorous research.Theories can be proposed that can then be tested against observed results. Although control cannotbe exercised over importance variables, advantage can be taken of the natural variation that occursin the real world. The study can often be replicated in other organizations in order to subject a theoryto further verification and add to its generality.

This is a very Interesting and scholarly article, well worth the effort to read for the practitioner, whodesires a better understanding of how our experience can increase our accumulated general knowl-edge so that each problem encountered does not have to be treated as a unique case.

32 MIS Quarterly/March 1989

Case Study Methodology

A ScientificMethodoiogy for iVIiSCase Studies

By: Allen S. Lee502 Hayden HallNortheastern UniversityBoston, MA 02115

AbstractA methodology for conducting the case studyof a management information system (MIS) ispresented. Suitable for the study of a singlecase, the methodology also satisfies the stan-dards of the natural science model of scientificresearch.

This article provides an overview of themethodological problems involved in the studyof a single case, describes scientific method,presents an elucidation of how a previously pub-lished MIS case study captures the major fea-tures of scientific method, responds to the prob-lems involved in the study of a single case, andsummarizes what a scientific methodology forMIS case studies does, and does not, involve.

The article also has ramifications that go beyondmatters of MIS case studies alone. For MIS re-searchers, the article might prove interesting foraddressing such fundamental issues as whetherMIS research must be mathematical, statistical,or quantitative in order to be called "scientific."For MIS practitioners, the article's view of sci-entific method might prove interesting for em-powering them to identify, for themselves, thepoint at which scientific rigor is achieved in anMIS research effort, and beyond which furtherrigor can be called into question, especially ifpursued at the expense of professionalrelevance.

Keywords: Information systems, case studies,research methods, research design,organizational impacts

AOM Categories: K.4.3, K.6.0, K.6.1

IntroductionThere is a strong case-study tradition in the aca-demic field of management information systems(Benbasat, et al., 1987; Fulk and Dutton, 1984;Kling, 1978; Kling and lacono, 1984; Kling andScacchi, 1982; Kraemer, et al., 1987; Laudon,1974; Leonard-Barton, 1987; Markus, 1983;1986). At the same time, case researchers ingeneral are still attempting to clarify themethodological basis upon which to conductcase studies (Benbasat, et al., 1987; Datta,1982; Oukes, 1965; George and McKeown,1985; Herriot, 1982; Hersen and Barlow, 1976;Huberman and Crandall, 1982; Louis, 1982; Lu-thans and Davis, 1982; Miles, 1979; 1982; Yin,1981a; 1981b; 1982a; 1982b; 1984). The ob-jective of this article is to present a scientific meth-odology with which to conduct case studies ofmanagement information systems (MIS). Indoing so, the article applies and builds upon con-cepts that pertain to case-study methodologyand that the author developed in his previousresearch (Lee, 1985; 1986; 1987b; forthcoming).In order to provide a practical demonstration ofhow these methodological concepts are usablein case studies of management information sys-tems, the article illustrates them with extensivematerial taken from an actual, published MIScase study — Markus' (1983) "Power, Politics,and MIS Implementation."

In particular, this article (1) provides an over-view of the methodological problems involvedin the study of a single case, (2) offers a de-scription of scientific method, (3) elucidates howthe MIS case study by Markus fits this descrip-tion, (4) responds to the problems involved inthe study of a single case, and (5) summarizeswhat the article's scientific methodology for MIScase studies does, and does not, involve.

What is science?The formulation of a particular scientific meth-odology for conducting MIS case studies andidentifying the methodological problems associ-ated with this type of research depends on whatis meant by "science" in the first place. In deter-mining this meaning, there are numerous modelsof science from which to select. Indeed, philoso-phers of science — the scholars who make ittheir job to observe scientists and to explain whatit is that scientists do — have not yet settled,among themselves, on a single model of what

MIS Quarterly!March 1989 33


science is.' The model used in this article is whatsocial scientists call the "natural science model"of social science (Behling, 1980, p. 483; Burrelland Morgan, 1979, p. 4; Daft, 1983, p. 539;Schon, et al., 1984, p. 9; Susman and Evered,1978, pp. 582-583). According to this model, natu-ral science is the ideal on which social scienceshould model itself. There are three reasons forselecting this particular model of science.

First, among management researchers, the natu-ral science model is a well-known and widelyaccepted model for conducting studies in socialscience.^ As such, the natural science model sug-gests itself as a useful device for introducing schol-ars, unfamiliar with case studies, to this type ofresearch.

Second, many of the criticisms directed againstcase studies are voiced from the perspective ofthe natural science model. It is the critics of casestudies, not scholars already working in the case-study tradition, who need to be convinced of thelegitimacy of case studies. Recognizing this, thearticle demonstrates the legitimacy of case stud-ies by using the standards of the criticsthemselves.'

Third, the article recognizes that a scientific meth-odology, which applies the natural sciencemodel, actually complements and supports themethods traditionally associated with case stud-ies. The natural science model is primarily amodel for testing theories, not formulating theo-ries in the first place. Methods traditionally used

' "Is scientific advance registered by increasing prob-ability (Carnap), by discrete verifications (the logicalpositivists) or falsifications (Popper), by revolutions(Kuhn), by growing consensus (Polanyi), by progres-sive versus degenerating research programs, con-ducted over long periods (Lakatos), or what? Thedebate goes on" (Vice and Slavens, 1983, p. 418).

^ In his article, "The Case for the Natural ScienceModel for Research in Organizational Behavior andOrganization Theory," Behling (1980) advocates thenatural science model: "Research methods similarto those used in the natural sciences have long beenthe norm in organizational behavior and organiza-tion theory" (p. 483). He adds: "Clearly, the authorsof mainstream texts in organizational behavior andorganization theory accept the natural science modelof good research. Those who Include research meth-ods chapters . . . clearly follow this approach andgenerally appear to owe an intellectual debt to Ker-linger, a strong proponent of the natural sciencemodel" (p. 483).

by case researchers to formulate theories (Ben-basat, et al., 1987; Filstead, 1970; Garfinkel,1967; Geertz, 1973; Kirk and Miller, 1986; Lee,forthcoming; Louis, 1982; Sanday, 1979; Taylor,1979; Van Maanen, 1979; Yin, 1981a; 1981b;1982b; 1984) may therefore still be applied inaddition to the methods specified by the naturalscience model. Thus, this article respects andpreserves the traditional function of case stud-ies in suggesting hunches and generating theo-ries for later testing — a function recognized byscholars of all persuasions. Specifically, the meth-odology formulated in this article (1) allows MIScase researchers to continue using the tools theyhave traditionally used, and (2) enables MIScase researchers to conduct case studies thattest theories by using the natural science model.

Methodological ProbiemsRaised by the Study of aSingle CaseIn this article an MIS case study refers to theexamination of a real-world MIS as it actuallyexists in its natural, real-world setting. In hold-ing MIS research to the standard of the naturalscience model, four problems can be identifiedin MIS research that is conducted in the form

^ On the one hand, the natural science model of socialscience represents a view of science taken by many(if not most) social scientists; see the quotation inthe previous footnote, as weli as Burrell and Morgan(1979, p. 4),.Daft (1983, p. 539), Schon, et al. (1984,p. 9), and Susman and Evered (1978, pp. 582-583).On the other hand, the philosophy of science wouldrecognize the natural science model to be a descen-dant of logical positivism — a view of science thatphilosophers of science themselves had originated,but subsequently abandoned (Bernstein, 1978; 1983;Schon, 1983, pp. 37-49). Whereas it is the article'sresponsibility to point out the disparity in viewpointsof these two communities of scholars (philosophersof science on the one hand, and many (if not most)social scientists on the other), it is beyond the arti-cle's scope to investigate the ramifications of the dis-parity, much less to resolve it. For a detailed exami-nation of this matter, see Lee (1987b; forthcoming).

Practically speaking, the audience for this article —management scholars critical of, or unfamiliar with,case studies — largely subscribe to the natural sci-ence model. For this reason, an effective strategyfor approaching this audience would be to proceedwith a framework they already find familiar and ac-ceptable — namely, the framework of the natural sci-ence model.



of case studies. These problems are discussedin the next four sections.

wrong?"(Miles, 1979, p. 590; emphasis in theoriginal).

Problem 1: Making controlledobservationsThe first problem concerns how to make con-trolled observations. In testing for relationshipstheorized to exist among different factors,natural scientists routinely observe the influenceof one factor on another factor, where the po-tentially confounding influences of all other lac-tors are somehow removed or "controlled for."Laboratory experiments in the natural sciencesaccomplish this through the use of controlgroups and treatment groups. In statistical ex-periments, it is accomplished with the help ofstatistical controls, such as those availablethrough a multivariate regression analysis. Un-fortunately for the MIS case researcher, (1) thestudy of a real-world MIS in its real-world set-ting precludes, by its very nature, the laboratorycontrols of laboratory experiments, and (2) thestudy of a single case commonly yields morevariables than data points — a situation that ren-ders inapplicable the statistical controls of sta-tistical experiments (Yin, 1981b).

Problem 2: Making controlleddeductionsThe second problem concerns how to make con-trolled deductions. Making controlled or logicaldeductions with mathematical propositions — asis commonly done in the natural sciences — isa standard, non-controversial practice. However,since it is rare (though certainly not undesirable)for a case study to be quantitative, the MIS caseresearcher is typically denied the methodologicalconvenience of working with numerical data andmathematically stated propositions. Instead, thecase researcher must somehow manage withqualitative data and verbally stated propositions.Making controlled deductions with verbal propo-sitions (I.e., qualitative analysis), while certainlypossible, is more problematic: "For quantitativedata, there are clear conventions the researchercan use," such as the widely accepted and well-known rules of algebra through which the valid-ity of mathematical deductions is known, "butthe analyst faced with a bank of qualitative datahas very few guidelines for protection againstself-delusion. . . How can we be sure that [aqualitatively deduced] finding is not, in fact.

Problem 3: Allowing for replicabilityThe third problem concems how to allow for repli-cability. Research in the natural sciences is rou-tinely replicated as a means of assuring the ob-jectivity of the research. However, the MIS caseresearcher is unlikely to observe the same setof events — namely, the same configuration ofindividuals, groups, social structure, hardware,and software — unfold again in the same way.The non-replicability of the same observationswould clearly hinder subsequent attempts by in-dependent investigators wishing to verify the find-ings of a particular case study.

Problem 4: Allowing forgeneralizabilityThe fourth and last problem concerns how toallow for generalizability. An often admired qual-ity of theories in the natural sciences is their ap-plicability to a range of settings. (In this sense,theories in the natural sciences are said to be"nomothetic," as opposed to "idiographic") How-ever, the fact that the study of a single caseis marked by unique and non-replicable eventswould make the study vulnerable to charges thatits findings cannot be extended to other settings.

The following section describes a scientific meth-odology for MIS case studies — a methodologythat follows from the natural science model. Useof this methodology allows the final section ofthis article to address the problems (Identifiedabove) associated with case studies.

A Description of ScientificMethodIn modeling MIS case studies on natural science,we must ask: How does inquiry in natural sci-ence proceed? Specifically, what Is meant by"the natural science model"?

In his classic text. Introduction to Logic, Copi(1986) provides a lucid description of the logicof reasoning used in the natural sciences. "Fewpropositions of science," he explains, "are di-rectly verifiable as true. In fact, none of the im-portant ones are. For the most part they con-

MIS Quarterly/March 1989 35


cem unobservable entities, such as moleculesand atoms, electrons and protons, chromosomesand genes" (p. 483). As a result, the mannerof verification in tbe natural sciences is indirectratber tban direct. "Tbe pattern of indirect test-ing or indirect verification consists of two parts.First one deduces from the proposition to betested [tbe proposition being the tbeory] one ormore other propositions capable of being testeddirectly [these latter propositions being tbe pre-dictions]" (p. 486). In the tenninology of logic,a theory's predictions are its conclusions. "Tbenthese conclusions are tested and are found tobe either true or false." The researcher then com-pares what the theory predicts and wbat is actu-ally observed. "If the conclusions are false, anyproposition that implies them [namely the tbeory]must be false also. On the other hand, if tbeconclusions are true, tbat provides evidence fortbe truth of tbe proposition being tested, wbichis tbus confirmed indirectly" (p. 486).

Karl Popper (1968) describes tbe same proce-dure in The Logic of Scientific Discovery, wberebe calls it the deductive testing of theories (pp.32-33, p. 60, pp. 109-111). It is deductive in tbefollowing sense. The natural scientist applies atheory (for example, "All men are mortal") to aset of facts or initial conditions ("Socrates is aman"), from wbich a conclusion or prediction isdeduced ("Socrates is mortal"). It is tbe predic-tion — as a deduced statement — tbat is tbentested against an observation statement (for ex-ample, "Socrates dies").

In tbis procedure, an observation tbat contra-dicts a prediction would be sufficient to castdoubt on (perhaps to the point of falsifying) thetheory from whicb the prediction follows. On theother hand, an observation that confirms a pre-diction is never regarded as conclusively estab-lishing the tbeory's trutb. The reason is tbat adifferent set of empirical circumstances, or initialconditions, to wbich the same tbeory may beapplied would result in yet anotber prediction(e.g., "Plato is mortal" or "Superman is mortal"),which in turn would open up the same theoryto yet anotber opportunity for its falsification.Tbus, the ever-present possibility for contradic-tory evidence to surtace in a subsequent testrequires tbat a tbeory always be regarded asfalsifiable. Indeed, falsifiability is tbe demarca-tion criterion that Popper uses to distinguish sci-ence from non-science (pp. 40-42).

Scientific method, in tbe form of the deductivetesting of tbeories, is widely known. Kuhn (1970),

whose school of thought is a rival to Popper's,expresses tbat "no field is potentially a science"unless its theories are cast according to "SirKarl's demarcation criterion" (p. 245). The nowcommon characterization of tbeories, in botb natu-ral science and social science, as falsifiable, re-futable, testable, or disconfirmable, is an indica-tion of tbe widespread extent to wbicb thedeductive testing of theories is practiced.

Falsifiability is just one requirement tbat a theorymust satisfy in order to be scientific. Tbere aretbree additional requirements, wbich are all as-sociated with tbe concept of the deductive test-ing of theories (Popper, 1968, pp. 32-33). Oneof tbese requirements is logical consistency: aslong as the different predictions that may be de-duced from tbe tbeory are not mutually contra-dictory, the tbeory can be said to be logicallyconsistent. Another requirement is that tbetbeory must be at least as explanatory, or pre-dictive, as any competing tbeory. Tbe last re-quirement is tbat the tbeory, while falsifiable,must survive tbe actual attempts made at itsfalsification.

In the way that scientific method appears in tbenatural science model, the notion of controlledobservation comes into play in tbe last step(namely, the step where tbe researcher makesa comparison between what is predicted andwbat is observed). In this step, tbe researcbermust be able to show tbat the observed effectcan be attributed to tbe factor being tested andtbat tbe potentially confounding effects of otherfactors bave been removed or "controlled for"(Campbell and Stanley, 1963). Tbis article hasalready mentioned laboratory controls and sta-tistical controls as examples of how observationscan be made in a controlled way.

An Exemplar for ScientificM\S Case StudiesMarkus' (1983) "Power, Politics, and MIS Im-plementation" captures the major features of sci-entific method tbat Copi and Popper describe.As such, the MIS case study by Markus maybe regarded as an exemplar for scientific MIScase studies in general, wbere tbe meaning of"scientific" is tbe one embodied in tbe naturalscience model.

Markus' research is tbe intensive study of asingle case, involving tbe entire configuration of


Case Study f\Aethodology

individuals, groups, social structure, hardware,and software in the setting of an organizationthat she calls Golden Triangle Corporation(GTC). Through interviews and documents,Markus observes events in the way they un-folded in their natural setting at GTC.

By conducting her research in this way, Markusinvokes the severe methodological problems men-tioned earlier. They are the problems of: (1) howto make controlled observations, (2) how tomake controlled deductions, (3) how to allow forreplicability, and (4) how to allow forgeneralizability.

Despite these problems, the MIS case study byMarkus still succeeds in crafting a theory aboutMIS implementation that conforms to the require-ments of fatsifiability, logical consistency, pre-dictive power exceeding that of competing theo-ries, and survival of the empirical tests aimedat falsifying it.

Markus presents three alternative theories on anequal footing and then compares the deductions(the predictions) of each against observationsmade in the setting at GTC. All are theoriesabout resistance to MIS implementation efforts.The case itself involved people's resistance toGTC's newly computerized Financial InformationSystem (FIS).

The people-determined theory involves "factorsinternal to the person" (p. 431). When peoplefactors such as human nature, cognitive styles,or personality traits are incompatible with the re-quirements of a computerized informationsystem, the system's intended users, accordingto the people-determined theory, will resist itsutilization.

The system-determined theory involves "factorsinherent in the application or system being im-plemented" (p. 431). Markus cites the followingas examples of system factors that incur resis-tance: lack of user-friendliness, technically defi-cient systems, and poor ergonomic design (p.431). According to the system-determinedtheory, when such factors are present, thesystem's intended users will resist its utilization.

The interaction theory, the most sophisticatedof the three theories, involves people factors aswell as system factors. However, Markus says,"This explanation identifies neither the systemnor the organizational setting as the cause ofresistance, but their interaction" (p. 431). She

describes the interaction theory in the followingways:

. . . resistance is explained as a product ofthe interaction of system design featureswith the intraorganizational distribution ofpower, defined either objectively, in termsof horizontal or vertical power dimensions,or subjectively, in terms of symbolism (p.432).

Resistance-generating conditions are mis-matches between the patterns of interac-tion prescribed by a system and the pat-terns that already exist in the setting intowhich the system is introduced (p. 438).

The primary assumption . . . is that informa-tion systems frequently embody a distribu-tion of intraorganizational power among thekey actors affected by its design. Intraor-ganizational power̂ is an attribute of indi-viduals and subgroups . . . .

. . . When the introduction of a computerizedinformation system specifies a distributionof power which represents a loss to certainparticipants, these participants are likely toresist the system. Conversely, when the dis-tribution of power implied in the design ofan information system represents a gain inpower to participants, these participants arelikely to engage in behaviors that might sig-nify acceptance of i t . . . .

. . . (A necessary condition for resistance tothe implementation of a system is that]people perceive the system to represent apower loss . . . (p. 442).

All three theories refer extensively to the exis-tence of phenomena that are neither directly ob-servable nor easily discernible: human nature,cognitive styles, personality traits, factors inter-nal to the person, user friendliness, technical de-ficiency in a system, ergonomics, horizontal andvertical dimensions of intraorganizational power,power in terms of symbolism, and (perhaps mostimportant and most directly unobservable of all)the distribution of power implied in the designof an information system. None of the three theo-ries is therefore directly verifiable as true.

In this context, Copi's remarks on scientificmethod, already quoted above, are worth repeat-ing: "Few propositions of science are directlyverifiable as true. In fact, none of the important



ones are. For the most part they concern unob-servable entities, such as molecules and atoms,electrons and protons, chromosomes andgenes." Hence, because direct verification is notfeasible, theoretical propositions are tested indi-rectly instead. This manner of testing involveswhat Popper calls the deductive testing of theo-ries, the details of which were described earlierin the article.

As it turns out, Markus provides an exemplarydemonstration of how an MIS case study is ca-pable of carrying out the deductive testing of theo-ries. She tests for the truth of the three theories(and for the presence of the unobservable phe-nomena) in the following way. Whereas the threetheories refer to phenomena that are not directlyobservable, they nonetheless yield predictionsof events that (if the given theory is true) wouldbe observable. Thus Markus' strategy is to usethe contrary theories to make contrary predic-tions about what would happen in the same set-ting. The theory that emerges unfalsified in thiscompetition would be judged scientific.

"The people-determined theory leads to the pre-diction that replacing individual resistors or co-opting them by allowing them to suggest improve-ments to the system might reduce or eliminateresistance" (p. 437). However, resistance to thenew information system persisted despite GTC'spractice of job rotation and mobility. Markusgives the example of an accountant, one of thedesigners and advocates of the system, who hadoriginally been working in corporate accountingand then became the controller in one of thedivisions that had resisted the system all along.He subsequently came to resist the system him-self. This observation falsifies the people-determined theory.

"The system-determined theory predicts thatfixing technical problems eliminates resistance"(p. 437). However, the resistance continued de-spite corrective actions taken to address anumber of major technical problems. Signifi-cantly, the problems were identified by a taskforce whose members were (as characterizedby Markus) "resistors." The corrective actionswere the installation of a larger computer witha more powerful operating system, a change inprocessing mode from batch to online, and asimplified method in the software by which man-agers could create new accounts. Markus (1983)writes, "when data were collected for this study,about one year after the last of these changes

was installed, informants in the [resisting] divi-sions still spoke of FIS [Financial InformationSystem]. Many felt strongly that the systemshould be replaced . . . " (p. 438). Markus reportsno mitigation in the resistance — a mitigationthat the system-determined theory predicts.These observations falsify the system-deter-mined theory.

The interaction theory predicts that nei-ther changing the people (by removingthem, by educating them, or by attempt-ing to coerce them), nor changing techni-cal features of the system will reduce re-sistance as long as the conditions whichgave rise to it persist. [The prediction isthat there will be resistance as long asthere are] mismatches between the pat-terns of interactions prescribed by asystem and the patterns that already existin the setting into which the system is in-troduced (Markus, 1983, p. 438).

Markus describes the interaction pattern thatwas already in place at GTC as consisting ofautonomy experienced by the divisional account-ants and dependence experienced by the cor-porate accountants. The divisional accountantscontrolled their own data (often in thick, manu-ally maintained ledger books) and could there-fore reconcile unusual situations before releas-ing reports. The corporate accountants had togo through the divisional accountants to obtainfinancial data, which was "a valued' resource"(p. 438). This conflicted with the interaction pat-tem prescribed by FIS, under which "all finan-cial transactions were collected into a single [com-puterized] database under the control ofcorporate accountants . . . . At any time, corpo-rate accountants had the ability to 'look into'the database and analyze divisional perform-ance" (p. 438). Markus' observation showed re-sistance to FIS, just as predicted by the interac-tion theory.

The interaction theory satisfies the four require-ments that Popper observes to be satisfied byall scientific theories. First, it is falsifiable (e.g.,the interaction theory would have been falsifiedif Markus had observed acceptance of FIS de-spite the difference between the interaction pat-tern that was in place and the one that FIS pre-scribed). Second, its logical consistency isknown through the mutual compatibility of thedifferent predictions that Markus considers (pp.437-438). Third, it is confirmed, not falsified, by



the observations in the GTC case study. Fourth,and most important, its predictions succeed,whereas the predictions of its rival theories —the people-determined theory and the system-determined theory — fail. In crafting the interac-tion theory so that it satisfies the four require-ments, Markus not only attains her specific re-search goal of explaining resistance to MISimplementation efforts, but also demonstrateshow an MIS case study is able to capture themajor features of scientific method in the waythat scientific method is embodied in the naturalscience model.

It must be emphasized that the conclusionsdrawn by Markus in her case study are only ten-tative at best. After all, (1) there may exist (orcome into existence) some corporate account-ants at GTO who, when transferred into one ofthe divisions, will continue to accept, and nevercome to resist, FIS; (2) the improvements in thetechnical features of the system may not yethave reached the threshold at which the resis-tance would diminish observably; and (3) the re-sistance may persist even when the interactionpattern, required by FIS, comes to be more likethe interaction pattern already in place. Thesepossibilities, however, do not weaken Markus'case study, but actually strengthen it by empha-sizing the extent of the interaction theory's falsi-fiabiiity and allowing the case researcher to im-prove the theory by pointing out where latersurprises may occur. No scientific explanation— whether Markus' interaction theory or a theoryof physics — may ever be conclusively proventrue. According to the logic of the deductive test-ing of theories, a theory can only be shown tobe false, or not (yet) false. In scientific research,further tests are always in order.

Response to the ProblemsInvolved in the Study of aSingle CaseThis section discusses the four severemethodological problems associated with caseresearch in the way they are manifested in thecase study by Markus. The discussion estab-lishes the context in which the article in its finalsection, will articulate what a scientific method-ology for MiS case studies does, and does not,involve.

How to make controlledobservationsA critic of caSe research may point out, correctly,that Markus fails to utilize either laboratory con-trols or statistical controls when making obser-vations to test the three theories. However,Markus solves the problem of how to make con-trolled observations by utilizing natural controls.

A simple but clear example of this (referred toearlier) is Markus' test of the people-determinedtheory, in which a particular accountant, uponmoving from his position in corporate account-ing to controiler in one of the divisions, changesfrom being an advocate of FIS to one of its re-sistors. This particular test "controls for" or"holds constant" the people factors by focusingon just one person (the accountant), and "varies"or "treats" the situation external to the personby observing his move from corporate account-ing to a division. Thus, Markus is able to cleanlyattribute the accountant's new behavior (resis-tance to FIS) to the "treatment" (the change inthe situation external to the accountant) ratherthan to the "control" (the people factors — thefactors internal to the accountant). Indeed, bymaking this controlled observation, Markus fal-sifies the people-determined theory, which pre-dicts no change in behavior where there is nochange in people factors.

In utilizing natural controls and treatments to testpredictions, the case researcher must do morethan wait passively for desired controls and treat-ments to materialize. Rather, the case re-searcher must actively apply his or her ingenu-ity in order to derive predictions that takeadvantage of natural controls and treatmentseither already in place or likely to occur. For ex-ample, in Markus' prediction concerning thepeople-determined theory, the control (the hold-ing constant of people factors) is already in placeby virtue of focusing on just one person, andthe treatment (the variation in the environment)takes advantage of the person's move from onepart of the organization to another. In general,it is incumbent upon the case researcher to scanthe empirical material for the presence of natu-ral controls and treatments that may be incorpo-rated into the formulation of a prediction. (Thisis no different from the activity of the statisticianwho, in utilizing a multiple regression analysisto analyze 1980 census data, scans the datato identify what factors might serve as the ap-

MIS Quarterly!March 1989 39

Case Study Methodoiogy

propriate independent variables and, hence, asthe statistical controls).

MIS case researchers who wish to utilize natu-ral controls wiii find themselves in good com-pany, investigators in some of the natural sci-ences, such as astronomy, geology, and humanbiology, are also unable to conduct laboratoryexperiments for obvious reasons and are there-fore also prevented from utilizing iaboratory con-trols in order to make controlled observations.Instead, these investigators routinely conduct natu-rai experiments in which they utilize natural con-trols, through which they have been able toachieve impressive resuits (Nagel, 1979, p. 452).MIS case researchers who invoke natural con-trols would therefore be employing a researchstrategy no different from, and no less scientificthan, what is employed by these natural scien-tists. In using naturai controls, MIS case research-ers wouid therefore be keeping within the natu-ral science model. To pursue this line of think-ing would take us beyond the scope of this arti-cle, but wouid lead to the conclusion that casestudies can be conducted as a form of naturaiexperiment, which is already a conventional formof research practiced in the "hard" sciences(Lee, forthcoming).

How to make controlled deductionsIn qualitative anaiysis, as performed by Markusin her case study, how can deductions be madein a controlled (i.e., logical) way? In mathemati-cai analysis, the validity of deductions invoivingmathematicai propositions can be readilychecked by turning to the rules of algebra. Inqualitative analysis, there is no correspondingbody of ruies as succinct or easily applied asthe rules of algebra for verifying the validity ofdeductions invoiving verbal propositions.

To respond to this problem, it must first be em-phasized that mathematics is a subset of formallogic, not vice versa. Logicai deductions in thegeneral case do not require mathematics. AnMIS case study that performs its deductionswith verbal propositions (i.e., qualitative analy-sis) therefore only deprives itself of the conven-ience of the rules of algebra; it does not de-prive Itself of the rules of formal logic, to whichit may therefore still tum when carrying out thetask of making controlled deductions.

Indeed, Markus herself provides examples of con-troiled deductions involving verbal propositions.

(Namely, she deduces several different, verballyexpressed predictions from the three different,verbally expressed theories as applied to the ver-bally expressed facts of the situation at GTC.)With regard to logical form, Markus' deductionsinvoiving verbal propositions are identical to andno less valid than, the deduction of the verbalproposition, "Socrates is mortal" (the prediction)from the two other verbal propositions, "Ail menare mortal" (the theory) and "Socrates is a man"(the facts or initial conditions).

Like the situation pertaining to the utilization ofnatural controls, MIS case researchers will findthemselves in good company with regard to anaiy-sis that utilizes the medium of verbal proposi-tions, as opposed to mathematical propositions.Consider biology and the theory of evolution. ForDarwin, the medium of logical deduction waswords and sentences, not numbers and mathe-matics (Kaplan, 1964, pp. 245-246).

How to allow for replicabilityHow might an independent investigator go aboutreplicating the findings of the MiS case studyby Markus?

One way — perhaps the most conceptuallystraightfonward way — would be to attempt toreplicate the case study in exactiy the way thatMarkus performs it. For the independent investi-gator, this would involve the attempt to appiythe same three theories to the same set of in-itial conditions in order to deduce the same pre-dictions as Markus, and then test these predic-tions against the same observations made byMarkus. The obvious difficulty with this proce-dure is that, in an MIS case study, any observedconfiguration of individuals, groups, sociai struc-ture, hardware, and software in a reai-world set-ting is highly unlikely to recur and be observedagain. Thus, an independent investigator couidnot verify the findings of the MIS case study byMarkus, at least not through this conceptuallystraightforward .procedure.

Fortunately, there is at least one alternative pro-cedure. The independent investigator couldapply the same theories as tested in the originalcase study to a different set of initial conditions(for exampie, the facts of the situation at AAACorporation or XXX Corporation), thereby result-ing in different predictions (for exampie, if thepeople-determined theory is true, then individu-als who share the same people factors at XXX



Corporation wiil display no difference in theirlevel of resistance to, or acceptance of, the com-puterized information system at XXX, regardiessof the rank and location of their position in theorganization). In other words, the investigatorwouid be working with a new prediction, "Platois mortal," as opposed to the original prediction,"Socrates is mortal"; even though the predic-tion would be different; it would still be the sametheory being tested. Such new predictions wouidcaii for observations different from the onesmade by Markus and would therefore relieve theindependent investigator of the impossible taskof attempting to replicate the observations madein the original case study. Consequently, eventhough the observations in a particular MIS casestudy are non-replicable, the case study's find-ings (that a particuiar theory is confirmed or dis-confirmed) would be replicable.

How to allow for generalizabilityThe fact that Markus' case study of GTC ismarked by unique and non-replicable events ren-ders it vuinerable to the charge that its findingscannot be extended to other settings. However,such a criticism, applied to the study of a singlecase, wouid be misplaced. A comparison to ex-penments conducted in the natural sciences clari-fies the issue.

Consider a naturai science theory that has sofar been confirmed in just a single experiment(whether a laboratory, statistical, or natural ex-periment). Of course, the theory wouid not begeneralizabie on the basis of the single experi-ment, since the experiment would have testedthe theory against just a single set of empiricalcircumstances. Instead, the theory wouid be gen-eralizabie to other sets of empirical circum-stances oniy on the basis of actually being con-firmed by additional experiments that test itagainst those other sets of empirical circum-stances. The same point holds true for case stud-ies. No theory concerning MIS wouid be gener-aiizabie on the basis of a single case study,since the singie case study would have testedthe theory against the empirical circumstancesof just a single setting. Instead, the theory con-cerning MIS would be generalizabie to other set-tings only on the basis of actually being con-firmed by additional case studies that test itagainst the empirical circumstances of thoseother settings.

In other words, generalizabiiity is a quaiity de-scribing a theory that has been tested and con-firmed in a variety of situations, whether suchtesting is conducted through case research, labo-ratory experiments, statistical experiments, ornatural experiments. As such, generaiizabilityposes no more, and no less, of a problem forMIS case research than it does for the studiesconducted in the naturai sciences. In taking thisposition, the MIS case researcher would, again,be in step with the natural science model.

What a ScientificMethodology for MIS CaseStudies Does, and Does Not,involveIn suggesting how MIS case studies might becarried out, this articie has offered a scientificmethodology that involves (1) the deductive test-ing of theories, where (2) the theory must be(a) falsifiabie, (b) logically consistent, (c) morepredictive than other theories, and (d) not faisi-fied by the tests it experiences. As such, thisscientific methodoiogy is no different from, andtherefore no iess rigorous than, scientific meth-odology as it is practiced as the natural sciences.

At the same time, this article takes the positionthat the scientific methodology of the natural sci-ence modei does not invoive, as objectives, theutiiization of any of the foliowing, even thoughthey may often be regarded (in this article's view,improperly) as necessary elements in scientificresearch: iaboratory controls, statistical controls,mathematicai propositions, and repiicable obser-vations. Instead, each one of these happens tobe a means to an objective in scientific researchrather than the objective itself. MIS case stud-ies are capable of achieving the same scien-tific objectives through different means.

Laboratory controis and statistical controls, forexample, constitute a means to controlled ob-servation — an objective MIS case studies areable to achieve through naturai controls. Like-wise, mathematical propositions constitute ameans to controiled or iogical deduction — anobjective MIS case studies are able to achievethrough verbal propositions that apply the ruiesof formal logic, of which the ruies of mathemat-ics are but a subset. Finally, replicabie observa-tions constitute a means to the replication of atheory's confirmation or disconfirmations — an



objective MIS case studies are able to achieveby testing the same theory through new predic-tions, thereby calling for new observations ratherthan replications of old ones.

The article has also taken the position that sci-entific methodology does not involve generalizabil-ity based on the result of a single test, whetherit is a single test taking place in an MIS casestudy or a single test taking place in a labora-tory experiment of the natural sciences. Instead,generalizability is a product of successive test-ing across a range of settings, not a single testin a single setting.

It is worth emphasizing how the particular sci-entific framework described in this article allowsus to identify some case studies as having moreanalytical rigor than others. There are two waysin which analytical rigor may be assessed.

First, there is simply the matter of whether agiven case study explicitly addresses each ofthe four requirements. As a check for falsifiabil-ity, does the case study consider any predic-tions through which the theory of interest couldbe proven wrong? As a check for logical consis-tency, are all the predictions considered consis-tent with one another? As a check for empiricalvalidity, does the case study confirm the theorythrough empirical testing? Finally, as a checkfor relative predictive power, does the case studyrule out rival theories? These questions pre-sume, of course, that the theory of interest isstated explicitly in the first place and that predic-tions following from the theory are also explicitlystated. MIS case studies that satisfy all four re-quirements explicitly and successfully are morerigorous than MIS case studies that satisfy anyof the four requirements implicitly orunsuccessfully.

Table 1 compares a number of MIS case stud-ies with respect to the four requirements. Ofcourse, none of the cited case studies was con-ducted with the avowed purpose of fitting thisarticle's scientific framework, so any value judg-ments pertaining to the quality of these studieswould be inappropriate. Instead, the table is in-tended simply as a guide for the reader whowishes to pursue additional examples of MIScase studies that illustrate, to varying extents,the scientific methodology this article hasdescribed.

The second matter pertaining to analytical rigoris one of degree. Some case studies may sat-

isfy the four requirements better than other casestudies. For instance, first, consider the require-ment that the theory of interest be confirmedthrough empirical testing. Confirming the theoryby successfully testing it through just one of itspredictions would not be as rigorous as success-fully testing it through several of its predictions.Likewise, confirming the theory in just one or-ganizational setting would not be as rigorous asconfirming it in two or more organizational set-tings. As the number of explicitly derived predic-tions or the number of organizational settingsis increased, the theory's degree of confirma-tion may also be increased.

Second, the requirement that the theory of in-terest be more predictive than any rival theorymay, of course, be satisfied more rigorously byincreasing the number of rival theories againstwhich its predictive performance is compared.As the number of rival theories considered isincreased, the theory's degree of relative pre-dictive power may also be increased."

Third, the requirement that the theory of interestbe logically consistent may be more rigorouslysatisfied by increasing the number of predictionsderived from it, then making sure that none ofthe predicted events contradicts or precludesone another. In contrast, deriving just two pre-dictions from the theory, and checking that theydo not contradict each other, would provide mini-mal rigor. Thus, as the number of explicitly de-rived predictions is increased, the theory'sdegree of logical consistency may also beincreased.

Finally, the requirement that the theory of inter-est be falsifiable may be more rigorously satis-fied by increasing the number of predictions de-rived from it and through which the theory couldbe proven wrong. In other words, as the numberof explicitly derived predictions is increased, thetheory's degree of falsifiability may also beincreased.

Campbell (1975), too, has referred to the situ-ation of the case researcher who pursues ana-lytical rigor in the ways just suggested. "In somesense, he has tested the theory with degreesof freedom coming from the multiple implicationsof any one theory" (p. 182). In other words,

' With respect to the research strategy of using asingle case study to distinguish among competingtheoretical explanations, Allison (1971) establishedthe primary scholarly modei used by Markus (1983),Kling and lacono (1984), and others.



Table 1. Checking MIS Case Studies Against the Four Requirements

Case studyauthor(s)

Markus (1983)

Kraemer.Dickhoven.Tierney andKing (1987)

Kling andlacono (1984)

Laudon (1974)

Kling (1978)

Kling andScacchi (1982)

Leonard-Barton (1987)

Fulk andDutton (1984)

Main theoryof interest

"Interaction"theory

Theory otsuccessful modelimplementationin federalagencies

"Organizationalpoiiticsmetaphor"

Theory olresistance tocentraiizedcomputing instate and localgovernments

Theory of interplaybetween technicalfeatures andsocial setting

"Web models"

Theory of factorsInfluencing useracceptance

Theory oforganizationaluses of video-teleconferencing

Does the casestudy considerany predictionsthrough whichthe theory couldbe proven wrong?

Yes'

. Yes"

Yes"

Yes"

Yes"

Yes"

Yes'

Notapplicable'

The Four Requirements

Are aii thepredictionsconsistent withone another?

Yes"

Yes

Possibly

Yes

Possibl/

Yes

Possibly

Notapplicable"

Does the casestudy confirnnthe theorythroughemplricaitesting?

Yes'

No'

Yesi

No"

Yes"

Yes'

Partially*

No-

Does the casestudy ruie outrivai theories?

Yes"

No9

Yes"

No'

Yes'

No"

No'

No"

See explanation for responses in "Notes for Table 1 and Table 2," following Table 2.

Campbell is extending the concept of "degreesof freedom" beyond its traditional statistical mean-ing. There are three ways Campbell's extendednotion of "degrees of freedom" can be appliedto describe the analytical rigor of a case study.

First, there are the degrees of freedom in thenumber of predictions the case study considers.As the degrees of freedom are increased in thiscategory, the theory's degree of falsifiability,degree of logical consistency, and degree of con-firmation can all be correspondingly increased.Increasing the degrees of freedom in this cate-gory therefore allows the case study tostrengthen the extent to which it satisfies threeof the four requirements.^

Second, there are the degrees of freedom in thenumber of cases or organizational settings in

which a given theory is tested. As the degreesof freedom are increased in this category, thetheory's degree of confirmation can be corre-spondingly increased. Increasing the degrees offreedom in this second category therefore allowsthe case study to strengthen the extent to whichit satisfies one of the four requirements.^

° Since increasing the nunnber of predictions also In-creases the number of different ways in which thecase study's finding (e.g., that the theory of interestis confirmed) could be subsequently replicated, thedegree of replicability can also be increased.

° Increasing the number of cases or organizational set-tings also allows the degree of generalizability to becorrespondingly increased. It should be noted thatincreasing the number of cases or organizational set-tings is not necessarily the same thing as increasing


Case Sfudy Mettiodology

Case studyauthor(s)

Markus (1983)

Kraemer.Dickhoven,Tierney andKing (1987)

Kling andlacono (1984)

Uudon (1974)

Kling (1978)

Kiing andScacchi (1982)

Leonard-Barton (1987)

Fulk andDutton (1984)

Table 2. Checking theMIS Case Studies

Main theoryolinterest

"Interaction" theory

Theory ofsuccessful modelimplementationin federaiagencies

"Organizationalpoliticsmetaphor"

Theory ofresistance tocentralizedcomputing instate and iocalgovernments

Theory of interplaybetweentechnicalfeatures andsocial setting

"Web models"

Theory of factorsinfluencing useracceptance

Theory oforganizationaluses of video-teleconferencing

Degrees of Freedomin Three Categories

Of

Three Categories of Degrees of Freedom

Number ofpredictionsconsidered

Several'

30»

Several"

4"

Several"

5

Several"

Number of casesor organizationaisettingsconsidered

1

2»

1

4'

1

2aa

1

1

Number ofrivaitheoriesconsidered

2"

0

3"

0

v

r

0

0

See explanation for responses in "Notes for Table 1 and Table 2." following Table 2.

Third and last, there are the degrees of freedomin the number of rival theories against which thetheory of interest is compared. As the degreesof freedom are increased in this category, thetheory's degree of relative predictive power canbe correspondingly increased. Increasing the de-grees of freedom in this third category thereforeallows the case study to strengthen the extentto which it satisfies one of the four requirements.

the number of data points in a statistical study. Thelatter is often done simply to increase the "level ofconfidence" associated with a single, statistically in-ferred observation (e.g., the observation that "the truemean is different from zero"), v\/hereas the formerinvolves increasing the total number of observations.In other words, increasing the number of data pointswill increase the "degrees of freedom" only in theconventional statistical sense of this term, not in theadditional senses this article is explaining.

Table 2 compares the same MIS case studiesconsidered in Table 1 with respect to the threecategories of degrees of freedom. As the tableshows, the greater a case study's degrees offreedom in each category, the greater the casestudy's analytical rigor. A particular case study's"analytical strategy" might therefore be de-scribed in terms of the number of degrees offreedom it pursues in each category.

The central concern of this article has been toaddress certain methodological issues pertain-ing to MIS case studies. However, the article'sanalysis and conclusions may have ramificationsthat go beyond matters of MIS case studiesalone. These ramifications might prove interest-ing to scholars and practitioners alike.

For MIS scholars, the article's discussion of sci-entific method might prove interesting for its rele-



Notes for Table 1 and Table 2

• One prediction is explicitly stated (in Table II, p. 437): "Changing individuals and/or fixing technicalfeatures will have little effect on resistance." Other predictions pertaining to the interaction theoryare considered implicitly in the discussion (pp. 437-438).

" This refers to predictions that the case study treats explicity as well as implicity.

' See p. 438.

" The case study rules out the "people-determined" theory and the "system-determined" theory (p.AOO/

• The case study states 30 predictions (these are the "propositions' on pp. 256-287). Each predictionmakes it possible to refute the theorized impact of a specific variable. (See Figure A. I , p. 257, for alist of the 30 variables.)

' The case study's objective is theory formulation, not theory testing. Still, the theory and predictionsare consistent with the facts of the two cases considered (the "TRIM/MATH" computer model andthe "DRI" computer model).

' Only one theory is formulated and considered.

" All predictions treated are implicit. For example, the case study states (p. 1225): "The CBIS did notsimply evolve along a natural path nor did it drift, rather it was pushed in a specific direction whichwould increase the power and control of key actors within the organization." Thus the reader mayinfer the prediction that, if the organizational politics metaphor is true, then we should observeneither evolution along a natural path, nor drift, but observe development in a direction that wouldincrease (not decrease or keep constant) the power and control of key actors within theorganization.

' The case study provides sufficient material for the reader to infer predictions (as the prediction innote " was inferred) that may then be compared.

' For example, the prediction mentioned in note " is confirmed.

" The case study rules out the technological evolution metaphor, the economic rationality metaphor,and the organizational drift metaphor (pp. 1222-1223). The reader may infer additional predictions!pertaining to these three theories, that are implicit in this portion of the case study.

'" The theory contains the variables of "homogeneity," "interdependence," and "internal integration"(pp. 67-75). The case study states four predictions explicity: (1) "We hypothesize that ceterisparibus, the more organizations are homogeneous with respect to tasks — the production of similarproducts or services — the more likely they share similar environmental and internal problems, themore likely it is that they will interact with each other in dealing with shared problems, and the morelikely they are to pursue collective solutions to those problems [such as sharing and using acentralized, computerized database systeml" (p. 69); (2) "Here we hypothesize that high andincreasing levels of interdependence among social units are conducive to higher levels of socialintegration among those units, and supportive of efforts attempting to increase integration [such assharing and using a centralized, computerized database system!" (p. 71); (3) "Therefore, wehypothesize a tradeoff between homogeneity and interdependence in relation to integration of asocial system. If both qualitites are high in a system, increases in integration would be supported. Ifboth are low, further integration would be most difficult. If of opposite sign, one low and the otherhigh, the effects should tend to cancel out" (p. 72); and (4) "For these reasons we hypothesize thatunder conditions of high internal unit integration, resistance to system integrating efforts will be veryhigh, and/or the terms under which such units are included into larger systems will be very favorab-le. .. Furthermore, we suggest that if resistance remains high, and if the demands of highly inte-grated units are very high, the integrating effort will cease or force will be resorted to" (p. 73).



Notes for Table 1 and Table 2 — continued

" The study qualifies itself by taking the position that the theory "is not itself proved by the [four] casestudies [but] is intended to serve as a guide to the cases" (p. 91). However, this qualification may beread as a sign of modesty, since the study (in Table 3, p. 75) offers what appears to be the favorableresults of empirically testing the predictions mentioned in note "•. Moreover, the study even de-scribes specific sets of empirical conditions pertaining to the bureaucratic reform process andstates the level of resistance predicted for each. These sets of empirical conditions are presentedunder the headings of the "pluralist model"; the "collegial model"; the "notables model"; and the"reputational elite model" (pp. 80-90).

" All predictions treated are implicit, but the case study provides sufficient discussion for the reader toinfer them. The theory of interest is that a computerized information system's impacts are a jointproduct of its technical features and its social setting. One prediction the reader may infer is that, ifthe theory is true, then deficient technical features alone will not bring about a lack of impact or anegative impact. Whereas this prediction is clearly refutable, it is confirmed by the facts the casestudy reports.

' See p. 492.

' The case study rules out the theory that either the technical features alone, or the social settingalone, can determine the impacts of a computerized information system.

» The case study states five predictions (these are the propositions on p. 26).

' See pp. 55-63.

" The case study compares "web" models to "discrete-entity" models, but states (p. 70): "We havenot organized this article to test the relative explanatory power of the discrete-entity and webmodels."

' One prediction is explicit (p. 10): "It seems reasonable to hypothesize that the first adopters of SSAmight be younger, more highly educated in the computer field, and more skilled in computerlanguages than their colleagues who are not yet using, and may never use, SSA." Other predictionstreated are implicit.

* The prediction (mentioned in note') was refuted (p. 14): "Age, type of education, and skill in Fortranand PL1 showed no relationship to SSA use." Other predictions were confirmed.

• This case study avows (p. 106): "Our purpose was not to provide a controlled experimental com-parison . . . , but rather to gather exploratory and descriptive data." This case study is included as areminder that case studies may also be legitimately used for the purposes of exploratory analysisand theory generation, not just theory testing. With respect to theory generation, this particular casestudy may be regarded as a useful contribution toward the development of a theory of the organiza-tional uses of video-teleconferencing.

» See pp. 21-22.

" See pp. 12-14.

•" See pp. 40-53.



vance to MIS research in general, not just MIScase studies. Research methodology in thestudy of management information systems hasbeen gaining attention as a problem, in itself,that deserves investigation.' In this regard, thearticle's view of scientific method could helpsecure the emerging position of qualitative re-search in MIS^ and perhaps, at the same time,reconcile the perceived differences between quan-titative and qualitative approaches in MIS re-search. In this larger methodological context, itis interesting to observe that this article's schemafor assessing analytical rigor (illustrated in Tables1 and 2) recogriizes no differences between quan-titative and qualitative approaches. (Indeed, thecase study by Leonard-Barton (1987) involvesstatistical inference.) The degrees of freedom ineach category (which follow from the four re-quirements of the natural science modei) canbe greater or smaller, whether the theory of in-terest is stated in the form of mathematical propo-sitions or verbal propositions. In other words, aqualitative case study can possess more ana-lytical rigor than a statistical study using LISREL,just as the reverse may be true. In this sense,any distinctions between quantitative and quali-tative approaches are artificial and inconsequen-tial. Neither type of research is inherently morerigorous than the other. In other fields of aca-demic research, the perceived differences be-

' In fhe spirit typical of methodological inquiry, MIS aca-demics are making their own research methodologytheir object of study. (For example, see Ein-Dor(1986); Jenkins (1986); Kauber (1986); Klein (1986);Naumann (1986); Wand and Weber (1986). Thesepapers were presented at the Management Informa-tion Systems Researcher's Workshop, held at the No-vember 1986 annual meeting of the Decision Sci-ences Institute.) This methodological inquiryaddresses not only what it means for our researchto be "scientific," but also such matters as the roleof frameworks, epistemology, and paradigms in MISresearch.

° For a provocative view of the qualitative methods thatare emerging in MIS research see Bjorn-Andersen(1986); Goldstein (1986); Markus (1986); Rosen(1986). These papers were presented at the panelon the "Use of Qualitative Methods in MIS Re-search," held at the December 1986 annual meetingof the International Conference on InformationSystems.

For an introduction to the use of qualitative methodsin general, see Filstead (1970); Kirk and Miller (1986);Yin (1984), as well as the December 1979 Issue ofAdministrative Science Ouarterly.

tween quantitative and qualitative approacheshave, unfortunately, become institutionalized intoopposing camps.* Some of the methodologicalconcepts in this article may prove helpful in avoid-ing a similar fate in the academic field of MIS.

For MIS practitioners, the article's discussion ofscientific method might prove interesting for de-mystifying the aura of MIS research that claimsto pursue scientific rigor, whether it involves thequalitative study of a single case or the utiliza-tion of a sophisticated statistical tool such asLISREL. The discussion of scientific method —especially the four requirements that a scientifictheory must satisfy — may empower MIS practi-tioners themselves to identify the point at whichscientific rigor is achieved in an MIS researcheffort, and beyond which further rigor, especiallyif pursued at the expense of professional rele-vance, can be called into question.

Finally, it is important to point out that, in theactual formulation of scientific knowledge, nei-ther natural scientists nor social scientists nec-essarily think in terms of the formalized proce-dures of any model of science, including thenatural science model. Lee (1987a) states:"These procedures do not address the private,mental process by which a scientist formulatesscientific knowledge, but rather the public proc-ess by which the scientist will on occasion retro-spectively test the truth of the already formu-lated knowledge for acceptance by his or her

' In the field of organizational studies, the existenceof opposing camps is clearly evident. Morey and Lu-thans (1984), in their review of the organizational lit-erature, describe the opposition between the twocamps as objective versus subjective (Burrell andMorgan, 1979), nomothetic versus idiographic (Lu-thans and Davis, 1982), quantitative versus qualita-tive (Van Maanen, 1975), outsider versus insider(Evered and Louis, 1981), and etic versus emic. Thishas generated concem over what Morey and Luthanscall the "widening gap between the two major orien-tations to organizational research" (p. 29) — a gapso wide that some authors have called for a rap-prochement between the two approaches (Everedand Louis, 1981; Luthans and Davis, 1982; Moreyand Luthans, 1984).

In the field of operations research, the interdiscipli-nary quantitative/qualitative approach, which charac-terized the field shortly after its founding during WorldWar II, eventually gave way to the dominance of quan-titative approaches. Only recently has the non-mathematical camp re-emerged to challenge, or com-plement, the mathematical camp. See Ackoff (1979)for a historical review of the events in the develop-ment of the field of operations research.



peers" (p. 577). Kaplan (1964) refers to theformer process as the actual "logic in use" bya scientist, and the latter as a "reconstmctedlogic" (p. 8). in recognizing the natural sciencemodel as one among many possible recon-structed logics of science, the article also recog-nizes the need for future research to investigatethe ramifications that alternative models of sci-ence could have for MIS case-studymethodology.

ReferencesAckoff, R. "The Future of Operational Research

Is Past," Journal of the Qperational ResearchSociety (30:2), February 1979, pp. 93-104.

Allison, G. Essence of Decision: Explaining theCuban Missile Crisis, Little, Brown, & Co.,Boston, MA, 1971.

Behling, O. "The Case for the Natural ScienceModel for Research in Organizational Behav-ior and Organization Theory," Academy of Man-agement Review (5:4), October 1980, pp. 483-490.

Benbasat, I., Goldstein, D. and Mead, M. "TheCase Research Strategy in Studies of Infor-mation Systems," MIS Quarterly (11:3), Sep-tember 1987, pp. 369-386.

Bernstein, R. The Restructuring of Social andPolitical Theory, University of PennsylvaniaPress, Philadelphia, PA, 1978.

Bernstein, R. Beyond Objectivism and Rela-tivism: Science, Hermeneutics, and Praxis, Uni-versity of Pennsylvania Press, Philadelphia,PA, 1983.

Bjorn-Andersen, N. "Action Research on theImpact of IS on Clerical Workers," abstractedin Proceedings of the Seventh InternationalConference on Information Systems, SanDiego, CA, December 15-17, 1986, p. 338.

Burrell, G. and Morgan, G. Sociological Para-digms and Organisational Analysis, Heine-mann, London, 1979.

Campbell, D. " 'Degrees of Freedom' and theCase Study," Comparative Political Studies(8:2), July 1975, pp. 178-193.

Campbell, D. and Stanley, J. Experimental andQuasi-Experimental Designs for Research,Houghton Mifflin, Boston, MA, 1963.

Copi, I. Introduction to Logic, Macmillan, NewYork, NY, 1986.

Daft, R. "Leaming the Craft of OrganizationalResearch," Academy of Management Review(8:4), October 1983, pp. 539-546.

Datta, L. "The Politics of Qualitative Methods,"American Behavioral Scientist (26:1), Septem-ber-October 1982, pp. 133-144.

Dukes, W. "N = 1," Psychological Bulletin(64:1), 1965, pp. 74-79.

Ein-Dor, P. "An Epistemological Approach to theTheory of Information Systems," Proceedingsof the 1986 Annual Meeting of the DecisionSciences Institute (1), Honolulu, HI, Novem-ber 23-25, 1986, pp. 563-565.

Evered, R. and Louis, M. "Alternative Perspec-tives in the Organizational Sciences: 'Inquiryfrom the Inside' and 'Inquiry from the Out-side'," Academy of Management Review(6:3), July 1981, pp. 385-395.

Filstead, W. Qualitative Methodology, Markham,Chicago, iL, 1970.

Fulk, J. and Dutton, W. "Videoconferencing asan Organizational Information System: Assess-ing the Role of Electronic Meetings," Systems,Objectives, Solutions (4:2), April 1984, pp. 105-118.

Garfinkel, H. Studies in Ethnomethodology, Pren-tice-Hall, Englewood Cliffs, NJ, 1967.

Geertz, C. The Interpretation of Cultures, BasicBooks, New York, NY, 1973.

George, A. and McKeown, T. "Case Studies andTheories of Organizationai Decision Making,"in Advances in Information Processing in Or-ganizations (2), L. Sproull and P. Larkey(eds.), JAI Press, Greenwich, CT, 1985, pp.21-58.

Goldstein, D. "The Use of Qualitative Methodsin MIS Research," abstracted in Proceedingsof the Seventh International Conference onInformation Systems, San Diego, CA, Decem-ber 15-17, 1986, p. 338.

Herriot, R. "Tension in Research Design and Im-plementation: The Rural Experimental SchoolsStudy," American Behavioral Scientist (26:1),September-Qctober 1982, pp. 23-44.

Hersen, M. and Bariow, D. Single Case Experi-mental Designs, Pergamon Press, New York,NY, 1976.

Huberman, A. and Crandall, D. "Fitting Wordsto Numbers: Multisite/Multimethod Researchin Educational Dissemination," American Be-havioral Scientist (26:1), September-Qctober1982, pp. 62-83.

Jenkins, A.M. "Management Information Sys-tems Researcher's Workshop," Proceedingsof the 1986 Annual Meeting of the DecisionSciences Institute (1), Honolulu, HI, Novem-ber 23-25, 1986, pp. 559-562.

Kaplan, A. The Conduct of Inquiry, Chandler,New York, NY, 1964.



Kauber, P. "What's Wrong with a Science ofMiS?" Proceedings of the 1986 Annual Meet-ing of the Decision Sciences Institute (1),Honoiuiu. HI, November 23-25,1986, pp. 572-574.

Kiri<, J. and Miiier, M. Reliability and Validity inQualitative Research, Sage University PaperSeries on Qualitative Research Methods (1),Sage Publications, Beveriy Hiiis, OA, 1986.

Klein, H.K. "The Oritical Social Theory Perspec-tive on Information Systems Development," Pro-ceedings of the 1986 Annual Meeting of theDecision Sciences Institute (1), Honolulu, HI,November 23-25, 1986, pp. 575-577.

Kling, R. "Automated Welfare Olient-Trackingand Service Integration: The Political Econ-omy of Computing," Communications of theACM (21:6), June 1978, pp. 484-493.

Kling, R. and lacono, S. "The Oontrol of Infor-mation Systems Developments After Implemen-tation," Communications of the ACM (27:12),December 1984, pp. 1218-1226.

Kling, R. and Scacchi, W. "The Web of Comput-ing: Computer Technology as Social Organi-zation," in Advances In Computers (21), M.Yovits (ed.). Academic Press, New York, NY,1982, pp. 1-90.

Kraemer, K., Dickhoven, S., Tierney, S. andKing, J. Datawars: The Politics of Modelingin Federal Policymaking, Columbia UniversityPress, New York, NY, 1987.

Kuhn, T. "Reflections on My Critics," in Criticismand the Growth of Knowledge, I. Lakatos andA. Musgrave (eds.), Cambridge UniversityPress, New York, NY, 1970, pp. 231-278.

Laudon, K. Computers and BureaucraticReform, Wiley, New York, NY, 1974.

Lee, A. "The Scientific Basis for ConductingCase Studies of Organizations," Academy ofManagement Proceedings, R. Robinson andJ. Pearce (eds.), San Diego, CA, August 11-14, 1985, pp. 320-323.

Lee, A. "The Case Study of an Organization asa Scientific Research Strategy," WorkingPaper 86-25, Coiiege of Business Administra-tion, Northeastern University, Boston, MA,1986.

Lee, A. "Ouixotic Communication: The Case ofExpert Witness Testimony," Knowledge (8:4),June 1987a, pp. 549-585.

Lee, A. "Integrating Positivist and Interpretive Ap-proaches to Organizational Research," pres-entation at the Annual Meeting of the South-ern Management Association, New Orleans,LA, November 4-7, 1987b.

Lee, A. "Case Studies as Natural Experiments,"Human Relations, forthcoming.

Leonard-Barton, D. "Implementing Structured Soft-ware Methodologies: A Case of Innovation inProcess Technology," Interfaces (17:3), May-June 1987, pp. 6-17.

Louis, K. "Multisite/Multimethod Studies: An In-troduction," American Behavioral Scientist(26:1), September-October 1982, pp. 6-22.

Luthans, F. and Davis, T. "An Idiographic Ap-proach to Organizational Behavior Research:The Use of Single Case Experimental Designsand Direct Measures," Academy of Manage-ment Review (7:3), July 1982, pp. 380-391.

Markus, M.L "Power, Politics, and MIS Implemen-tation," Communications of the ACM (26:6),June 1983, pp. 430-444.

Markus, M.L. "Case Study Research and theUse of Communication Systems in Organiza-tions," abstracted in Proceedings of the Sev-enth International Conference on InformationSystems, San Diego, CA, December 15-17,1986, p. 339.

Miles, M. "Qualitative Analysis as an AttractiveNuisance: The Probiem of Analysis," Admin-istrative Science Quarterly (24:4), December1979, pp. 590-601.

Miles, M. "A Mini-Cross-Site Analysis," Ameri-can Behavioral Scientist (26:1), September-October 1982, pp. 121-131.

Morey, N. and Luthans, F. "An Emic Perspec-tive and Ethnoscience Methods for Organiza-tional Research," Academy of ManagementReview (9:1), January 1984, pp. 27-36.

Nagel, E. The Structure of Science, Hacket, In-dianapolis, IN, 1979.

Naumann, J.D. "The Role of Frameworks in MISResearch," Proceedings of the 1986 AnnualMeeting ofthe Decision Sciences Institute (1),Honolulu, HI, November 23-25,1986, pp. 569-571.

Popper, K. The Logic of Scientific Discovery,Harper Torchbooks, New York, NY, 1968.

Rosen, M. "Ethnographic Study of the Role ofInformation Technology in Organizational Con-trol Systems," abstracted in Proceedings ofthe Seventh International Conference on In-formation Systems, San Diego, CA, Decem-ber 15-17, 1986, p. 339.

Sanday, P. "The Ethnographic Paradigms," Ad-ministrative Science Quarterly (24:4), Decem-ber 1979, pp. 527-538.

Schon, D. The Reflective Practitioner: How Pro-fessionals Think in Action, Basic Books, NewYork, NY, 1983.

MIS QuarterlylMarch 1989 49


Schon, D., Drake W. and Miiier, R. "Social Ex-perimentation as Reflection-ln-Action," Knowl-edge (6:1), September 1984, pp. 5-36.

Susman, G. and Evered, R. "As Assessment ofthe Scientific Merits of Action Research," Ad-ministrative Science Quarterly (23:4), Decem-ber 1978, pp. 582-602.

Taylor, C. "Interpretation and the Sciences ofMan," in Interpretive Social Science, P. Rabi-now and W. Sullivan (eds.). University of Cali-fomia Press, Berkeley, CA, 1979, pp. 25-71.

Tice, T. and Slavens, T. Research Guide to Phi-losophy, American Library Association, Chi-cago, IL, 1983.

Van Maanen, J. "The Fact of Fiction in Organ-izational Ethnography," Administrative Sci-ence Quarterly (24:4), December 1979, pp.539-550.

Wand, Y. and Weber, R. "On Paradigms in theIS Discipline: The Problem of the Problem,"Proceedings of the 1986 Annual Meeting ofthe Decision Sciences Institute (1), Honolulu,HI, November 23-25, 1986, pp. 566-568.

Yin, R. "The Case Study Crisis: Some Answers,"Administrative Science Quarterly (26:1),March 1981a, pp. 58-65.

Yin, R. "The Case Study as a Serious ResearchStrategy," Knowledge (3:1), September1981b, pp. 97-114.

Yin, R. The Case Study Strategy: An AnnotatedBibliography, the Case Study Institute, Wash-ington, D.C, 1982a.

Yin, R. "Studying Phenomenon and ContextAcross Sites," American Behavioral Scientist(26:1). September-October 1982b, pp. 84-100.

Yin, R. Case Study Research: Design and Meth-ods, Sage Publications, Beverly Hills, CA,1984.

About the AuthorAllen S. Lee holds a Ph.D. in social scienceand policy analysis from M.I.T. In addition tocase-study methodology, he is conducting re-search on the implementation of managementinformation systems, particularly human resourceinfonnation systems. He is currently assistant pro-fessor of management science at NortheasternUniversity.

50 MIS Quarterly!March 1989

Documents

Executive Overview - National Center for Border Security ...borders.arizona.edu/classes/mis696a/resources/readings/Lee-1989... · Case Study Methodology science is.' The model used