The Nature of Theory in Information Systems Author(s): Shirley Gregor Reviewed work(s): Source: MIS Quarterly, Vol. 30, No. 3 (Sep., 2006), pp. 611-642 Published by: Management Information Systems Research Center, University of Minnesota Stable URL: http://www.jstor.org/stable/25148742 . Accessed: 09/05/2012 21:46
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Management Information Systems Research Center, University of Minnesota is collaborating with JSTOR to
digitize, preserve and extend access to MIS Quarterly.
http://www.jstor.org
Gregor/The Nature of Theory in IS
RESEARCH
ESSAY
lite!,! V VXZllV
The Nature of Theory in Information Systems1
By: Shirley Gregor isgivenfor the legitimacy and value of each theorytype. The School of Accounting and Business Information building of integrated bodies of theory that encompass all Systems theory types is advocated.
College of Business and Economics
The Australian National University Keywords: Theory, theory taxonomy, theory structure, infor
Canberra ACT 0200 mation systems discipline, philosophy of science, philosophy
AUSTRALIA of social sciences, interpretivisttheory,design theory,design
Shirley.Gregor@anu.edu.au science, explanation, prediction, causality, generalization
Abstract Introduction^ ^^ ^ ^ The aim of this research essay is to examine the structural nature of theory in Information Systems. Despite the impor The aim of this essay is to examine the structuralnature of tance of theory,questions relating to itsform and structure theory in the discipline of Information Systems. There are a are in comparison with to episte number of grounds for believing that thismeta-theoretical neglected questions relating
mology. The essay addresses issues of causality, explanation, exploration is both necessary and timely. Calls continue for prediction, and generalization thatunderlie an understanding "good theory" in IS (Watson 2001) and the development of our "own" theory (Weber 2003). Despite the recognition of of theory.A taxonomy isproposed thatclassifies information the need for theory development, however, there is limited systems theories with respect to themanner inwhich four central are addressed: discussion in IS forums ofwhat theorymeans in IS and what goals analysis, explanation, predic tion,and prescription. Five interrelated types of theoryare form contributions to knowledge can take.
distinguished: (I) theoryfor analyzing, (2) theoryfor ex To place thisdiscussion incontext, consider thequestions that plaining, (3) theoryfor predicting, (4) theoryfor explaining and predicting, and (5) theoryfor design and action. arise about thebodies of knowledge or theories encompassed in a discipline. These questions fall into a number of inter Examples illustrate thenature of each theory type. The appli cability of the taxonomy is demonstrated by classifying a related classes2:
sample ofjournal articles. The paper contributes by showing 1. Domain What are of interest in that multiple views of theory exist and by exposing the questions. phenomena the discipline? What are the core problems or topics of assumptions underlying different viewpoints. In addition, it is suggested that the type of theory under development can interest?What are theboundaries of the discipline?
influence thechoice of an epistemological approach. Support 9 "The last three of these four classes have parallels in the three sets of issues distinguished by Godfrey-Smith (2003) for thinking about the philosophy of science: (1) the logical structure of science, (2) epistemological and methodological issues, and (3) scientific thinking, or the social organization Allen Lee was the accepting senior editor for this paper. M. Lynne Markus, of science. When thinking about one discipline in particular, we need to add Michael D. Myers, and Robert W. Zmud served as reviewers. the first class to define the range of phenomena of interest in that discipline.
MIS Quarterly Vol. 30 No. 3, pp. 611-642/September 2006 611
Nature of Theory in IS
Gregor/The
- Structural or ontological questions. What is theory? positivist and interpretivist paradigms (for example, see How is this termunderstood in the discipline? Of what Orlikowski and Baroudi 1991) or between qualitative and is theory composed? What forms do contributions to quantitative methods. Some have argued for pluralism in knowledge take? How is theoryexpressed? What types methods (Mingers 2001) or for integratingapproaches (Lee of claims or statements can be made? What types of 1991). There has been littleor no recognition to date in IS of questions are addressed? theview that the research approach adopted could vary with differenttypes of theory in IS,which is a view underlying this
- Epistemological questions. How is theory constructed? essay. How can scientific knowledge be acquired? How is theory tested? What research methods can be used? Socio-political questions, in the fourth category, address What criteria are applied to judge the soundness and diverse issues. Into this category fall questions concerning rigor of research methods? the historical development of scientific thought in a disci plinary community (as inKuhn 1996). An example in Infor
- Socio-political questions. How is thedisciplinary knowl mation Systems is the analysis of how the interpretivist edge understood by stakeholders against thebackdrop of paradigm has emerged historically in contrast to positivism human affairs? Where and by whom has theory been (Walsham 1995). There is also discussion of political, power, developed? What are thehistory and sociology of theory and prestige issues for thediscipline. The benefits and costs evolution? Are scholars in the discipline in general of diversity in IS research to the discipline have been agreement about current theories or do profound differ considered by Benbasat andWeber (1996) and Robey (1996). ences of opinion exist? How is knowledge applied? Is the knowledge expected to be relevant and useful in a Questions of relevance topractice of IS research also fall into this category. Further,what is termed critical theory expli practical sense? Are there social, ethical, or political issues associated with the use of the disciplinary citly addresses ethical and moral questions, by seeking to be emancipatory and bring about improvements in the human knowledge? condition (see Ngwenyama and Lee 1997). Each of these classes of questions has received attention. Examination of each category, however, shows thatquestions Returning to questions in the second category, discussion of the structural nature or ontological character of theory in falling into the second category have received limited treat ment in the extant literature. Each of these classes of Information Systems is scattered and there is scanty recog nition that these questions are even of interest.Here theword questions is considered in turn, leaving the second category until last. ontology is used in the sense that it refers to a language for talking about the nature and components of theory (for With respect to the firstcategory, questions about thedomain example, the different types of statements that are incor of interest of IS research have remained a topic of interest porated). Many IS researchers who use theword theory since the inception of the discipline. Argument about the repeatedly in theirwork fail to give any explicit definition of definition ofmanagement informationsystems dates back to their own view of theory. A number of papers thatdiscuss the 1970s (Benbasat 2001) and many writers have debated different research paradigms (for example, Klein andMyers aspects of our domain identitysince that time. A selection of 1999; Mingers 2001) offer little in theway of definitions or articles illustrates the range and history of this stream of discussion of thenature of theoryor types of knowledge that debate. Weber (1987) was concerned with identifying the can be expected to result fromdifferentresearch approaches. unique nature of IS that distinguished it from other disci Recognition thatdifferent types of theory exist can be found plines. Orlikowski and Iacono (2001) argued for attention to in some proponents of constructive or design theory (Iivari the information technology artifact as the core subjectmatter 1983; Markus et al. 2002; Walls et al. 1992). A wider view of the IS discipline. Benbasat and Zmud (2003) proposed a on theoryand knowledge types is found in only a handful of core set of phenomena to define the IS field, generating papers in IS (Cushing 1990; Gregor 2002a 2002b; Iivari 1983; furtherdebate ina series of articles inCommunications of the Markus and Robey 1988). Association of Information Systems (2003, Volume 12). Table 1presents examples of theories in IS of differentonto Epistemological questions, in the third category, have also logical types to demonstrate thatmultiple views of theory received considerable attention. Numerous articles argue the exist. These initial examples are presented briefly. Further merits of differentparadigms for conducting research in IS. delineation of these views and more is the raison d'etre of Frequently, debate is framed in termsof distinctions between this essay.
612 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
I Table 1 Some Differing Views of Theory in Information Systems
Theory as statements thatsay how something should be done inpractice:
An early textbook by Davis and Olson (1985) articulates theway inwhich MIS should be designed, implementedand
managed. This theory provides prescriptions to be followed in practice, with the implicitexpectation that the
prescribed methods will insome sense be "better"than alternatives (Cushing 1990).
Theory as statements providing a lens forviewing or explaining theworld:
Orlikowski and Robey (1991) drew on structuration theory and empirical work to construct a theory inwhich the
organizational consequences of ITare viewed as the products of bothmaterial and social dimensions. Such theory
isseen as a desirable end product; formal testingof such a theory is not envisaged (Walsham 1995).
Theory as statements of relationships among constructs thatcan be tested:
The technology acceptance model (Davis 1986) posits that two particular beliefs on the part of users, perceived
usefulness and perceived ease-of-use, are of primary relevance for computer acceptance behaviors. This theory leads
to testable propositions thatcan be investigated empirically (see Davis et al. 1989).
Examination of what is meant by theory occurs in other social system, or even the two side by side; in disciplines. An issue of theAcademy ofManagement Review addition, it investigates thephenomena thatemerge (1989, Volume 14,Number 4), focused on theoryand theory when the two interact. development. Similarly, an issue of Administrative Science Quarterly (1995, Volume 40, Number 3) contained articles Thus, we have a discipline that is at the intersectionof knowl about what theory is,what theory is not, and how theorizing edge of the properties of physical objects (machines) and occurs. Descriptions of theory in the social sciences can also knowledge of human behavior. IS can be seen to have com be found in Dubin (1978), Freese (1980), Kaplan (1964), monalities with other design disciplines such as architecture Merton (1967), and Weick (1989). More established disci or engineering,which also concern both people and artifacts, plines have considerable histories of enquiry into the nature orwith other applied disciplines such as medicine, where the of theory. In thephilosophy of science therehas been discus products of scientificknowledge (such as drugs or treatments) sion of scientific knowledge and the formulation of theory are used with people. To understand IS, theory is required over a very long period (for example, Hume 1748; Locke that links thenaturalworld, the social world, and the artificial 1689; Nagel 1979; Popper 1980). Fundamental ideas from world of human constructions. Thus, thebody of knowledge this prior work are drawn upon in this essay, but they are that is needed draws on natural science, social science and adapted for the IS context. what has been termeddesign science (Cross 2001; Hevner et al. 2004; March and Smith 1995; Simon 1996). The attributes It is importantto examine thenature of theory in IS separately of such a body of knowledge areworthy of exploration,which from other disciplines as the four classes of questions is the aim of this essay. depicted earlier are interrelated. The domain of interestfor a discipline can be expected to influence the nature of its Thinking clearly about the nature of theory in Information theory. Theory inmathematics and music, for example, Systems has significance for research and practice. Our means differentthingsand knowledge isdeveloped, specified, leading journals expect thatpublished research articles will and used in differentways. Thus, the nature of theory in IS have a stronggrounding in theory (MISQ 2004). Developing could differ from that found in other disciplinary areas. A theory is what we are meant to do as academic researchers characteristic thatdistinguishes IS from other fields is that it and it sets us apart from practitioners and consultants. In concerns the use of artifacts in human-machine systems. Lee addition, there is the view that "nothing is so practical as a (2001, p iii) uses thesewords: good theory" (Lewin 1945). Theories are practical because theyallow knowledge tobe accumulated ina systematicman research in the information systems field examines ner and this accumulated knowledge enlightens professional more than just the technological system, or just the practice.
MIS QuarterlyVol. 30 No. 3/September2006 613
Nature of Theoryin IS
Gregor/The
Personal experience with doctoral students in particular Issues discussed include the nature of theory in general, the suggests that theyoftenhave limitedunderstanding ofwhat is need for generalization, the nature of causality and the core meant by theory. Exposure to conflicting or simplistic de goals of explanation and prediction. It is impossible in a scriptions of different research paradigms (for example, single paper to condense the extensive discussion of these interpretivismversus positivism) sometimes leads to con topics over many hundreds of years into a meaningful repre fusion. The distinction made earlier among the different sentation of all thathas been said. The approach adopted is classes of questions about research suggests thatpieces of the to give an outline of the perspectives considered and to puzzle these novice researchers are facing are missing. The highlight those differences in thought that are intimately discussion of differences among paradigms is frequently connected with different approaches to theory, as well as framed around epistemology and the practice of doing some importantcommonalities. research in a community and thepossibility that theremay be differenttypesof theoryappropriate indifferentcircumstance A wide rather than a narrow view of theory is taken so that is not explored. An initial premise for the paper is that the subject matter of the essay is not restricted. Dictionary differenttypesof theoryexist in Information Systems and that definitions show that the word theory can take on many all can be valuable. The exploration of theory that follows meanings, including "a mental view" or "contemplation," a has been framed to be inclusive and does not depend on the "conception ormental scheme of something tobe done, or the adoption of a specific epistemological commitment (that is, method of doing it; a systematic statement of rules or prin how knowledge is acquired and justified). The paper is ciples to be followed," a "system of ideas or statements held intended to be of interestto a range of scholars with different as an explanation or account of a group of facts or phenom
personal preferences for research approaches. ena; a hypothesis thathas been confirmed or established by observation or experiment, and is propounded or accepted as The remainder of this paper proceeds as follows. First, it accounting for the known facts; statements of what are held considers general notions of theory inmore detail, including to be the general laws, principles, or causes of something differentconceptions of causality, explanation, and generali known or observed," a "mere hypothesis, speculation, conjec zation, which are central to differentways of developing and ture" (OED 2004). Thus, theword theorywill be used here expressing knowledge. An argument ismade that an appro ratherbroadly to encompass what might be termed elsewhere priate taxonomy for IS depends on classifying theorieswith conjectures, models, frameworks, or body of knowledge. respect to the degree and manner inwhich they address four central goals of theory: analysis, explanation, prediction and prescription. The five different types of IS theory distin Different Perspectives on Theory guished are labeled: (1) theory for analyzing, (2) theory for explaining, (3) theoryforpredicting, (4) theoryfor explaining Differences in views of theory depend to some degree on and predicting (EP theory), and (5) theory for design and philosophical and disciplinary orientations, yet thereare also action. The differenttypesof theoryare interrelatedand some commonalities. This essay draws upon writings from the bodies of theory could comprehensive, well-developed philosophy of the natural sciences, the social sciences, from include components from all the types of theory discussed. the interpretivisttradition, and from the sciences of the arti Illustrations of relevant work in IS are provided under each ficial, all ofwhich are relevant to Information Systems. heading, as are related research methods (briefly), and the form a contribution to knowledge could take. The applica In general, philosophers of science writing in the traditionof bility of the taxonomy is demonstrated by classifying a the physical or natural sciences are likely to see theory as sample of articles from recent journal issues. The paper con providing explanations and predictions and as being test cludes with a discussion of questions that arise from con able. For example, Popper3 (1980) held that theorizing, in sideration of these different views of Information Systems part, involves the specification of universal statements in a theoryand suggestions for further work. form that enables them to be tested against observations of
Popper was an effective critic of Marxism and Freud's psychoanalytic
theories and was the first insightful critic of logical positivism. Popper's
About TheoryH^H^HHHHHHIH contributions to the philosophy of science continue to be significant. Godfrey-Smith (2003) saw that he had an appeal tomany working scientists This section presents underlying ideas relevant to theory to and was regarded as a hero by many. Magee (1998, p. 256) places Popper the leading philosophers of the 20th century, along with Russell, preface the subsequent discussion of theory in Information among Wittgenstein, and Heidegger, and believes that there will be continued Systems. It is necessary to express these ideas to show the discovery and development of his positive views in comparison with his underlying philosophical positions on which the essay relies. critiques.
614 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
what occurs in the real world. Popper described theory as Many of the ideas in this approach stem from theGerman follows (p. 59): intellectual traditionof hermeneutics and theVerstehen tradi tion in sociology, fromphenomenology and from critiques of Scientific theories are universal statements. Like all positivism in the social sciences. Knowledge in thisparadigm linguistic representations they are systems of signs takes on a differentperspective. or symbols. Theories are nets cast to catch what we call "the world"; to rationalize, to explain and to Knowledge consists of those constructions about master it.We endeavor tomake themesh even finer which there is a relative consensus (or at least some and finer. movement towards consensus) among those com
petent (and in the case of more arcane material,
Similar views of theory can be found in the social sciences. trusted) to interpretthe substance of the construc Doty and Glick (1994, p. 233), writing about typologies as a tion. Multiple "knowledges" can coexist when form of theory in organizational studies, thought that the equally competent (or trusted) interpretersdisagree minimal definition of a theorywas that itmust meet threepri (Guba and Lincoln 1994, p. 113). mary criteria: (1) constructsmust be identified; (2) relation ships among these constructsmust be specified; and (3) these A furtherapproach to theory is evident when the concern is relationships must be falsifiable (that is, able to be tested). with the construction of technological artifacts. The classic work that treats technology or artifact design as a special The philosophy of science offers a rich variety of views, prescriptive typeof theory isHerbert Simon's The Sciences of which continue to develop (see Godfrey-Smith 2003). One theArtificial (1996) firstpublished in 1969. Simon (p. xii) specific and rathernarrow position, thatof logical positivism, notes that in an earlier edition of his work he described a has had a lingering influence on themanner inwhich theory central problem thathad occupied him formany years: is regarded by some philosophers of science. Logical positi vism was developed inEurope after World War I bywhat was How could one construct an empirical theory? known as theVienna Circle. At thebase of logical positivism is the famous Verification Principle: only assertions thatare I thought I began to see in theproblem of artificiality in principle verifiable by observation or experience can an explanation of thedifficulty thathas been experi convey factual information.Experience was thought tobe the enced in filling engineering and other professions only source of meaning and the only source of knowledge with empirical and theoretical substance distinct (Magee 1998). Many philosophers of science regard logical from the substance of their supporting sciences. positivism as defunct (Passmore 1967) and Popper takes Engineering, medicine, business, architecture and
credit for "killing" it as early as 1934 (Popper 1986). painting are concerned not with the necessary but However, traces of logical positivism can still be detected in with the contingent?not with how things are but thephilosophy of science: for example, in the reluctance to with how theymight be?in short,with design. refer to causality in theoretical statements (see Dubin 1978), causality not being directly observable. An ontological position is adopted in this essay that is consistent with a number of these different of conceptions The intrepretivisttradition steers researchers toward a dif theory. Theory is seen as having an existence separate from ferent outlook, where the primary goal is not to develop the subjective understanding of individual researchers. This theory that is testable in a narrow sense (although itsvalidity position corresponds to ideas expressed by both Habermas or credibilitymay still be assessed), but in and Popper. Habermas (1984) recognizes three different worlds: the objective world of actual and possible states of understanding the complex world of lived experi affairs, the subjective world of personal experiences and ence from the point of view of those who live it. beliefs, and the social world of normatively regulated social This goal is variously spoken of as an abiding con relations. These threeworlds are related to Popper's Worlds cern for the lifeworld, for the emic point of view, 1, 2, and 3 (Popper 1986). World 1 is the objective world of forunderstanding meaning, for grasping the actor's material things;World 2 is the subjective world of mental definition of a situation, forVerstehen. The world of states; and World 3 is an objectively existing but abstract lived reality and situation-specific meanings that world ofman-made entities: language, mathematics, knowl constitute the general object of investigation is edge, science, art, ethics, and institutions. Thus, theory as an
thoughttobe constructed by social actors (Schwandt abstract entitybelongs toWorld 3. An individual can have a
1994, p. 118). subjective view of what a theorymeans, at which point an
MIS QuarterlyVol. 30 No. 3/September2006 615
Gregor/The Nature of Theory in IS
understanding of the theory resides in a personal World 2. Theories can be classified by their level of generalization. This essay, however, is concerned with theory as World 3 Meta-theory is at a very high level of abstraction and provides entities, existing outside an individualmind (as, for example, a way of thinking about other theories, possibly across in journal articles). disciplines. Giddens (1984) describes his structurationtheory as being ameta-theory. Examples of theories of thisnature in To sum up, thisdiscussion of differentperspectives on theory Information Systems are not readily apparent, although this at a general level shows theories as abstract entities thataim essay is one example, being a "theory of theories," and to describe, explain, and enhance understanding of theworld Markus and Robey's (1988) work is another. Theories with and, in some cases, toprovide predictions ofwhat will happen sweeping generalizations that are relatively unbounded in in the futureand to give a basis for interventionand action. space and time are referred to as grand theories (Bacharach The following sections explore the fundamental goals of 1989). Weber argued that IS is in need of such powerful, explanation and prediction inmore depth, as well as the general theories that recognizably belong to the discipline, a related issues of causality and generality. motivation for his work with Yair Wand on theories of representation (Weber 1997).
The generality of a theory is also indicated by itsbreadth of
Generalization focus (Neuman 2000). Substantive theory is developed for a specific area of inquiry, such as delinquent gangs, strikes, A number of different views of theory are encompassed in or race based on analysis of observations divorce, relations, this essay with a view tobeing inclusive. There is still a limit, and is contrastedwith formal theory,which is developed for however, to what is classed as theory. Abstraction and a broad conceptual area such as deviance, socialization, or generalization about phenomenon, interactions, and causation power. Another termused ismid-range theory, referring to are thought to be at the core of a theory.We do not regard a theory that ismoderately abstract, has limited scope, and can collection of facts, or knowledge of an individual fact or easily lead to testable hypotheses. Merton (1968) saw mid event, as theory. "Data are not theory" (Sutton and Staw range theoryas particularly importantforpractice disciplines. 1995, p. 374), although data may form the foundation for theoreticaldevelopment. For thisreason, theword knowledge Definition of the level of generality or scope of a theory when used in this essay does not refer to knowledge of includes specifying the boundaries within which it is ex specific events or objects, but means body of knowledge, or pected tohold and providing thequalifying words, themodal theoretical knowledge. qualifiers, thatare used in theoretical statements (words like some, every, all, and always). For we could example, specify Views differ on the degree to which generalization or a theory about information system failure and say that ithad universality is required in theory. Popper's (1980) view is boundaries in that it applied only to large and complex that the natural sciences should aim at strictly universal systems. A very general theorymight have statements that statements and theories of natural laws (covering laws), applied to all systems,where "all" is a modal qualifier. although these laws can never be held with certainty.4 The notion of prediction entails some conception of generality. In The level of generality has not been used in this essay as a order to predict what will happen in the future,we need a primary characteristic for distinguishing theory types. The generalization that includes futureevents. processes bywhich generalizations are developed may differ with the tradition inwhich work is carried out (see Lee and The possibility of true"laws" similar to the laws of nature in Baskerville 2003), yet there appears to be reasonably wide researchers of many social affairs is thoughtunlikely (Audi 1999, p. 705; Cook acknowledgement among persuasions
and Campbell 1979, p. 15; Hospers 1967, p. 232) primarily that varying degrees of generality are possible in theories. of the very large number of conditions (Xn) that Theories in each of the five theory types distinguished in this because essay could be subjected to secondary classification on the might impact on any outcome (Y). Nevertheless, we expect basis of the level of generality to give a two-dimensional in the social sciences (and IS) that theory should still include classification scheme?a potential area for further work. generalizations to some degree.
Examples of natural laws are Boyle's Law or the law of gravity or e= mc2. Popper showed these laws are not empirically verifiable, acknowledging that Causality David Hume had made this observation two and a half centuries before. The problem is the problem of induction: from no finite number of observations, The idea of causality, or the relation between cause and event, however large, can any unrestrictedly general conclusion be drawn that would be defensible in logic. is central tomany conceptions of theory. When theory is
616 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
taken to involve explanation, it is intimately linked to ideas of sis difficult to undertake. "To say thatC is the cause of causation. Often, to ask for an explanation of an event is to E is to assert that the occurrence of C, in the context of ask for its cause. Similarly, the ability tomake predictions social processes and mechanisms F, brought about E, or from theorycan depend on knowledge of causal connections. increased the likelihood of E" (Little 1999, p. 705). The concept of causality is extremely problematic but is of 4. Manipulation or teleological causal analysis. In this fundamental philosophical importance. The 18thcentury Scottish philosopher, David Hume, for example, pointed out view, a cause is an event or state that we can at produce thatwe are unable to see or prove that causal connections will, or otherwise manipulate to bring about a certain other event as an effect. on an every exist in theworld, thoughwe continue to think and act as if This analysis relies
we have knowledge of them (Norton 1999, p. 400). Kant's day understanding of a cause as an act by an intentional (1781) position was thatunderstanding in termsof cause and agent, for example, flicking a switch causes a light to turn on. effect was an a priori characteristic of the human mind underlying all human knowledge. Cook and Campbell (1979) give a more detailed coverage of There are many ways of reasoning about causality and to causality, although they believe that "The epistemology of some extent differenttypes of theoryreflectdifferentways of causation, and of the scientificmethod more generally, is at present in a productive state of near chaos" (p. 10). They ascribing causality in the phenomena we observe around us presentMill's criteria for causality as being of practical use: and thedifferenttypes of explanation thatarise. An important (1) the cause has to precede the effect in time, (2) the cause point, however, is thatthevarious arguments for causality are and effectmust be related, and (3) other explanations of the notmutually exclusive and at different times and indifferent cause-effect relationship have to be eliminated. Pearl (2000) circumstances we will rely on differentreasons for ascribing points out that statisticians (including Karl Pearson) have causality. Four prominent approaches to the analysis of event avoided or argued against the concept of causality altogether, causation can be distinguished (see Kim 1999): because it is a mental construct that is notwell-defined, pre ferring to deal only with correlations and contingency tables. 1. Regularity (or nomological) analysis. Universal regu laritygives rise touniversal or covering laws. "There are Different approaches to theorizing concern themselves with some causes, which are entirelyuniform and constant in causality to varying degrees and rely on differentways of producing a particular effect; and no instance has ever reasoning about causality for explanations, as seen in the been found of any failure or irregularityin their in their following section. operation" (Hume 1748, p. 206). This type of regularity is sought in the natural sciences, with examples of covering laws being Boyle's Law and Ohm's Law in Explanation and Prediction physics. Many believe this type of regularity should not be expected or sought in the social sciences (for example, Central tomany understandings of theory are the twin goals Little 1999). of explanation and prediction. These goals can be recognized in Popper's view of above, and also in the views of theory 2. Counterfactual analysis. Under this approach, what Nagel (1979), who sees the distinctive aim of the scientific makes an event a cause of another is the fact that if the enterprise as being theories thatoffer systematic and respon cause had not occurred, the event would not have (the sibly supported explanations. cause is a necessary If counterfactuals are condition). understood in termsof covering laws, thencounterfactual Explanation is closely linked to human understanding, as an analysis may turn into a form of regularity analysis. explanation can be provided with the intentof inducing a subjective state of understanding inan individual. The nature 3. Probabilistic causal analysis. This typeof causality was of explanations has been studied in thephilosophy of science recognized by Hume (1748, p. 206) with an entertaining (see, for example, Achinstein 1983; Craik 1943; Hempel and example. Compared to universal laws, "there are other Oppenheim 1948; Nagel 1979), in relation to everyday causes, which have been foundmore irregularand uncer reasoning and thenature of argumentation (see Toulmin 1958; tain; nor has rhubarb always proved a purge, or opium a Toulmin et al. 1979), as well as in the subbranches of IS soporific to everyone, who has taken thesemedicines." relating to knowledge-based systems (see Gregor and This view of causal analysis is thought tobe suited to the Benbasat 1999). The relationship between scientific explana social sciences, where the lack of a closed system and the tion and human understanding is also a subject of interest effects of many extraneous influences make other analy (see, for example, Friedman 1974).
MIS QuarterlyVol. 30 No. 3/September2006 617
Gregor/The Nature of Theory in IS
Approaches to explanation in the philosophy of science can Dubin assigns reasoning about causality in theory to be distinguished in termsof theunderlyingmode of reasoning "rhetoric" (p. 96) and rests his arguments in part on views about causality, and views on this matter have changed expressed by an early 20thcenturyphysicist and philosopher considerably over the last 50 years (Mayes 2004). Two broad of science, Pierre Duhem, who also held approaches to offering explanations can be contrasted. First, a particular fact or event is explained by pointing out the A physical theory is not an explanation. It is a scientific law thatgoverns itsoccurrence. This "covering law system ofmathematical deduced from propositions, theoryof explanation" was developed byHempel and Oppen a small number of principles, which aim to represent heim (1948) and is a tenet of logical empiricist philosophy. as simply, as completely, and exactly as possible a To explain something ismerely to show how to derive it in a set of experimental laws (Duhem 1962, p. 7). logical argument from premises that include a covering law. A number of problems have been detected with thisapproach, Other philosophers of science offer a contrasting view of the one ofwhich is theasymmetryproblem, illustratedby thecase role of explanations in theory. Nagel (1979) distinguishes of a flagpole and its shadow. The covering law approach theories from experimental laws, believing that theories are gives an explanation of the lengthof the shadow fromknowl more comprehensive, whereas an experimental law can be a edge of theheight of the flagpole and theposition of the sun, single statement. A theory is a system of interrelated state but italso supports an argument that the lengthof the shadow ments, possibly containing abstract theoretical terms that explains the height of the flagpole (and the position of the cannot be translated into empirical measures. The theory sun), which is rather an unsatisfying "explanation" in the common sense of theword (see Godfrey-Smith 2003). might also include statements about causality, with varying concepts of causality, including teleological causation, so that the theoryprovides causal explanations. Later attempts at providing a betterbase forexplanations have moved away from this logical conceptualization to include the notion that explanation is a communicative process (see, for Apart from explanations, theories can also aim at predictions, which allow the theory both to be tested and to be used to example, Achinstein 1983). This second communicative school of thoughtargues that it is importantfor explanations guide action. Prediction goes hand inhand with testing. For to include notions of causality thatdo not depend on law-like example, we can make a prediction "If a systems test is not carried out, then a systemwill fail." This proposition can generalizations or statistical association alone, but refer to both be tested against what happens inpractice with projects, other causal mechanisms including teleological-type causes. In the remainder of this essay, the terms explanation or causal and also used to guide action, if it is believed to be accurate.
explanation refer to this second wider view of explanation, Some theories can focus on one either or rather than the narrower goal, explanation covering-law approach. prediction, at the expense of the other. That is, it is possible The differentviews on causality and explanation are evident to achieve precise predictions without necessarily having of the reasons outcomes occur. an in different approaches to theory. On the one hand, Dubin understanding why Using
(1978) specifically excludes causal relationships,5 stating that everyday example, it is possible to predict from the appear ance of clouds that itabout to rain,without having an under
Empirically relevant theory in the behavioral and standing of how precipitation occurs. Moreover, it ispossible
social sciences is built upon an acceptance of the to have models that are powerful in contributing to under
notion of relationship rather than the notion of standing of processes without providing, at the same time,
causality (p. 96). precision inprediction. Case studies of information systems
implementationmight give us a good understanding of how
lack of involvement of users can lead to user dissatisfaction
with a completed system. Itwould still be difficult topredict
with any degree of accuracy thedegree of user dissatisfaction
It is somewhat surprising to find these views in the latter part of Dubin's text. In an early chapter, Dubin describes clearly how a theory can contain arising from lack of involvement over a wide range of propositions of two types (Dubin 1978, pp. 30-31): one for prediction of out systems and settings. Dubin refers to these situations as the comes and one for describing processes. He sees causal process-type state precision paradox and the power paradox respectively. ments as useful in a chain of statements, to link and justify the outcome-type statements. However, he does not follow through with this view when specifying the components of theory,where he excludes process-type causal The distinction between the goals of explanation and predic explanation. It is possible that a number of researchers who use Dubin as a tion is central to the differentiation among types of theory in reference do not agree with his view on the omission of causality from the taxonomy proposed in thispaper. theory, or are not aware that it is a part of his arguments.
618 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/TheNature of Theory in IS
with theprimary goals6 of the theory. Research begins with
Classifying Theory in_ a problem that is to be solved or some question of interest. InformationSystems The theory that is developed should depend on thenature of this problem and the questions that are addressed. Whether A central question for this essay is how to construct a classi the questions themselves are worth asking should be con ficatory scheme for theories in Information Systems. sidered against the state of knowledge in the area at the time. McKelvey (1982) reviews a number of methods that have The four primary goals of theorydiscerned are been proposed for the construction of taxonomies and illus trateshis arguments in the context of organizational classi Analysis and description. The theoryprovides a descrip fication. This branch of enquiry is referred to as systematics tion of the phenomena of interest,analysis of relation (Simpson 1961) and dates back to the logical enquiries of ships among those constructs, the degree of generaliza Aristotle and Plato and their study of the hidden nature or bility in constructs and relationships and theboundaries form or essence of things. A difference between the illus within which relationships, and observations hold. trationsof classificatory schema given byMcKelvey and the objects of enquiry in this essay is that the formerconcerns the Explanation. The theoryprovides an explanation of how, objects of natural and social science, such as organizations, why, and when things happened, relying on varying views of causality and methods for argumentation. This plants, and animals, while the concern in this essay iswith the structuralnature of theories,which are abstract entities. The explanation will usually be intended to promote greater classification system proposed resulted from an iterative understanding or insightsby others into the phenomena of interest. process involving the study of the nature of theories evident in IS, analysis of prior work, and refinement of an analytic Prediction. The theory stateswhat will happen in the classification method thatdistinguished among the classes of future if certain preconditions hold. The degree of cer theoryon the basis of their importantattributes. Alternative tainty in the prediction is expected to be only approxi methods would include forms of empirical classification, mate or probabilistic in IS. which could, for instance, involve the study ofwhat forms of theoryhave occurred in IS publications over a period of time Prescription. A special case of prediction exists where and allow groupings to emerge from this study on thebasis of the theory provides a description of the method or the characteristics of the theories observed, possibly using a structureor both for the construction of an artifact (akin statistical technique such as cluster analysis. An analytic to a recipe). The provision of the recipe implies that the method is adopted here as it is believed to be more suitable recipe, if acted upon, will cause an artifact of a certain when the defining attributes of theory can be extracted from type to come intobeing. a considerable literature. Combinations of these goals lead to the five types of theory The advantage gained by classifying objects of interest in a shown in the left-handcolumn of Table 2. The distinguishing features of each theory type are shown in the right-hand taxonomy is that like properties of a class of phenomena can column. It should be noted that the decision to allocate a be identified and a means is provided for comparing and theory to one class might not be straightforward. A theory contrasting classes. Identification ofwhich class a proposed that is primarily analytic, describing a classification system, piece of work falls intoprovides some guidelines as to how can have implications of causality. For example, a framework the theorydeveloped should look and how itcan be evaluated. that classifies the important factors in information systems The criteria for evaluating classification schema and taxon development can imply that these factors are causally con omies should be considered (see Doty and Glick 1994). nected with successful systems development. Some judge These theories (ofType I in the taxonomy) are expected tobe ment may be needed to determine what theprimary goals of complete and exhaustive; that is, they should include classes a theory are and towhich theory type itbelongs. that encompass all phenomena of interest. There should be decision rules,which hopefully are simple and parsimonious, to assign instances to classes and the classes should be A theory is an artifact in that it is something thatwould not exist in the real world without human intervention. The word goal here means the goal of the mutually exclusive. In addition, as taxonomies are proposed artifact in the sense that it is the causa finalis, the final cause or end of the to aid human understanding, we would like the classes to be artifact (following Aristotle's writing on the four explanations of any "thing" easily understood and to appear natural. in The Four Causes, from a translation by Hooker 1993). The goal of a theory is "what the theory is for": analyzing, explaining, predicting, or prescribing. For a more commonplace artifact such as a table, the causa The method for classifying theoryfor IS proposed here begins finalis iswhat the table is for (e.g., eating from).
MIS QuarterlyVol. 30 No. 3/September2006 619
Gregor/The Nature of Theory in IS
Table 2. A Taxonomy of Theory Types in Information Systems Research
Theory Type Distinguishing Attributes
I. Analysis Says what is. The theorydoes not extend beyond analysis and description. No causal relationships among phenomena are specified and no predictions are made.
II. Explanation Says what is, how, why, when, and where. The theoryprovides explanations but does not aim to predictwith any precision. There are no testable propositions. III. Prediction Says what isand what will be. The theoryprovides predictions and has testable propositions but does not have well-developed justificatorycausal explanations. IV. Explanation and Says what is, how, why, when, where, and what will be.
prediction (EP) Provides predictions and has both testable propositions and causal explanations.
V. Design and action Says how to do something. The theorygives explicit prescriptions (e.g., methods, techniques, principles of formand function) for constructing an artifact
Table 3. Structural Components of Theory_I Theory Component (Components Common to All Theory) Definition Means of representation The theorymust be represented physically insome way: inwords, mathematical terms, symbolic logic, diagrams, tables or graphically. Additional aids for repre sentation could include pictures, models, or prototype systems.
Constructs These referto the phenomena of interest inthe theory (Dubin's "units"). All of the
primaryconstructs in the theoryshould be well defined. Many differenttypes of
constructs are possible: for example, observational (real) terms, theoretical
(nominal) termsand collective terms.*
Statements of relationship These show relationships among the constructs. Again, these may be ofmany types: associative, compositional, unidirectional, bidirectional, conditional, or
causal. The nature of the relationship specified depends on the purpose of the
theory.Very simple relationships can be specified: forexample, "x isa member of
class A."
Scope The scope isspecified by the degree of generalityof the statements of relationships
(signified by modal qualifiers such as "some," "many," "all," and "never") and
statements of boundaries showing the limitsof generalizations.
Theory Component (Components
Contingent on Theory Purpose) Definition
Causal explanations The theorygives statements of relationships among phenomena thatshow causal reasoning (not covering lawor probabilistic reasoning alone). Testable propositions (hypotheses) Statements of relationships between constructs are stated insuch a formthat they can be tested empirically.
Prescriptive statements Statements inthe theoryspecify how people can accomplish something inpractice construct an artifact or develop a (e.g., strategy).
*Dubin (1978) defines a real unit as one forwhich an empirical indicator can be found, and a nominal unit as one forwhich an empirical indicator cannot be found. Collective units are a class or set of units while member units are the members of the class or set. Further distinctions are made between enumerative, associative, relational, statistical, and complex units.
620 MIS Quarterly Vol. 30 No. 3/September 2006
Gregor/TheNature of Theoryin IS
Table 3 shows the components of theories across the tax (2) a descriptive level, atwhich the explanatory conjectures onomy. This specification allows IS researchers (1) to and hypotheses are generated and tested; and (3) a prescrip identifywhat theory is composed of in general and (2) to tive level, atwhich methods for constructing systems are put analyze the components of theirown theoryand the theoryof forward, with recommendations for their practical use. others. This framework is used in the following section for Iivari's views are congruent with what is proposed here, the analysis of examples of theories. although presentedwith less detail and thedistinctions among the levels are less fine-grained. Some components of theory are necessary for other compo nents. Each theory must have some means of representation.7 Cushing (1990) distinguished frameworks, descriptions of The focus in thispaper is on the structuralanalysis of theory facts, empirical generalizations, and theoryas separate steps that is accessible tomore than one person, that is, it can be in a program of scientific research. His breakdown of steps communicated. Thus, themeans of representation for theory has some similaritieswith the classification schema proposed include words, either spoken or written,mathematical sym here, except his steps are means toward a single form of bols, operators from symbolic logic, diagrams, graphs, and "scientific theory," rather than each step being recognized as other pictorial devices. It is possible thatworking models or a legitimate form of theorizing in its own right. could also be used to represent constructs or rela prototypes
tionships. A single concept can have more than one physical Markus and Robey (1988) also distinguished theorypartly in "=" structural terms, considering (1) the nature of the causal representation: for example, themathematical symbol represents the same concept as the words "is equal to." Each agency (technological, organizational or emergent); (2) the theorymust also have constructs,which refer to the entities logical structure (whether variance or process theory); and that the theory concerns. These entities could be physical (3) the level of analysis. The first dimension defines the phenomena or abstract theoretical terms. All the other com adoption of a particular theoretical stance, rather than ameta ponents of theorydepend on these basic components. theoretical dimension. The thirddimension of level of analy sis is related to thedegree of generality of a theory.As such, Theoretical statementsare composed ofwords or symbols that it is a possible candidate for classifying theory,potentially e = represent constructs (for example, mc2). Statements of giving a two-dimensional classificatory schema. relationship, scope, explanation, prediction, and prescription are all different types of statements. The active words, or Markus and Robey's second dimension deserves further verbs, in a statementwill distinguish the type of statement. consideration, as there are divergent views on the nature of Membership of a class or category is indicated bywords such process-type theory and variance-type theory. In one view, as to" or "is a." Words such as "led "influ variance-type theory is seen as possessing laws of interactions "belongs to,"
ences," or "constrains" imply causality. The terms "asso (relationships) such that ciated with" and "linked" are frequently used but are less informative. could mean "correlated "came 1. Given variation in the values of a unit (A), They with," 2. what other units (B, C, ...K) can be linked to before," "composed of," "located next to," or something else. the first (by laws of interactions) Prescriptive statements can take an imperative form: "A 2. so that the variance in values of the original unit system of type x should include functions a, b and c." (A) may be accounted for by the linked varia The taxonomic method proposed can be compared with other tions of values in the other units (B, C, ...K)? taxonomies proposed for theory types on the basis of their (Dubinl989,p. 92). structural character, although few systematic attempts at Process theory is seen as offering classifying theories across paradigms can be found. Fawcett and Downs (1986), working in the field of nursing, classified An explanation of the temporal order inwhich a theories as descriptive, explanatory, or predictive. Their tax discrete set of events occurred, based on a story or onomy, while being an influence on the currentwork, ex historical narrative (Huber and Van de Ven 1995, p. cludes prescriptive theoryof the type needed for design and action and includes theory that is purely descriptive. Iivari vii).
(1983) distinguished three levels of theorizing for IS: (1) a Some authors that process-theory and argue variance-theory conceptual level, atwhich the objects of enquiry are defined; should be kept separate. Mohr (1982) believed that the
This paragraph gives a very simple account of notions of representation, attempt tomix them constituted "a significant impediment, one source of the frustration of theory" (p. 37) and other signs, and meaning, which are dealt with at great length elsewhere (e.g., in semiotics). writers have adopted this view (for example, Seddon 1997).
MIS QuarterlyVol. 30 No. 3/September2006 621
Nature of Theoryin IS
Gregor/The
It is important to realize thatMohr presents some contro analysis of data (Glaser and Strauss 1967). Similarly, theory versial views, equating variance theorywith the laws of can be referred to as deductive, meaning that it has been nature?yet, as seen earlier, a number of philosophers, going deduced from an existing body of theory, or inductive, back toHume, have argued that thedegree of regularityfound meaning that it is derived fromdata. in the laws of nature should not be adopted or sought in the social sciences (see Little 1999). Further,Mohr argues that Critical theoryseeks tobring about improvements in the lives variance theoryhas explanations of causality reliant on the of human actors. Theory labeled normative has an ethical or identification of sufficient and necessary antecedent condi moral dimension in addressing what should be done. tions for an outcome. This view of causation presents a Normative theory differsfrom theory for design and action number of difficulties, including that of the asymmetry of (Type V) in that the latterdoes not imply thatone course of causation as discussed under the covering-law model of action is better than another in a moral or ethical sense. explanation, and also the problem that the identification of Again, all of the theory types depicted here could have social necessary and sufficient conditions is considered unlikely in or political implications. The mere act of classifying people "open" systems such as social systems, compared with the or things intogroups and giving themnames ("black" versus relatively "closed" systems found in the experimental "white") can give rise to stereotypical thinking and have sciences. An alternative view sees process and variance political and social consequences (see Foucault 1971). theorizing (as defined by Dubin rather thanMohr) as inter However, while attributes of theory that relate to epistemol related, with both necessary for developing satisfying and ogical and socio-political questions are very real and sound theory and arguing for causality in differentways. important, they are beyond the scope of the currentwork, Research might carry out process-type case studies of the which focuses on the structuralaspects of theory. context, content, and settings inwhich information systems are introduced and isolate some of themore importantcondi The following sections describe the five different types of tions and events that lead to various outcomes. Variance-type theory that are identified as germane to IS. studies could further investigate the degree of the relation ships among the identified events, conditions and outcomes using statistical techniques and larger samples (for a fuller explanation of this view, see Huber and Van de Ven 1995).8 Five Types of Theory in InformationSystems WtK^BKtKKKKK^tk Lee, Barua and Whinston (1997) discussed theory in IS in terms of underlying causal relationships, but primarily from A detailed description of each type of theory follows, with a statistical viewpoint, which gives a narrower focus than that of this essay. examples as illustrations. There is some variation within each theory type,with different types of work depending on the focus of work undertaken and the scope of the theory. The The current taxonomy builds on thepriorwork on the struc for each type are for evi examples given theory analyzed tural nature of theory in IS. It is regarded as an advance, dence of all seven theory components identified earlier: however, as it offers a fuller and more systematic basis for means of representation, constructs, relationships, scope, classifying theory. causal explanations, falsifiable statements, and prescriptive statements. This analysis of existing work is not straight Note that some furtherdistinctions among theories in the forward, as theories are rarelypresented explicitly in termsof literaturedo not depend on the structuralnature of the theory. these seven components and some interpretive licence has Rather, differenttheory types are distinguished depending on been employed in presenting the examples. In addition, as theirassociation with particular epistemological positions or noted earlier, the classification is dependent on themain or with particular socio-political aims. The distinctions made primary goals of the theory,rather than goals thatare present concern the thirdand fourthclasses of questions distinguished only to a minor degree. For example, a theory that focused in the introduction rather than the second. Thus, we find primarily on prediction yet had some explanations of a very theories identified by their origin or research method; for rudimentary typewould be classified as Type III. example, grounded theoryrefers to theory thatemerges from thegrounded theorymethod, which involves close and careful
g Type I: Theory forAnalyzing The view thatprocess and variance explanations can coexist in one theory does not mean that it is a simple matter to draw a box-diagram model Analytic theories analyze "what is" as opposed to explaining representing the theory as a whole, or that both process and variance com causality or attempting predictive generalizations. These ponents can easily be tested in a single study (see also the views of DeLone theories are and McLean (2003) in their re-specification of their success model).
622 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
Table 4 An Example of a Taxonomic Theory
Theory Overview
Iivari,Hirschheim and Klein (2000-2001) propose a dynamic frameworkforclassifying IS development approaches and methodologies. The framework is intended to serve as an organizing and simplifyingdevice thatcontributes tomethodology comparisons by pointingout similarities and differences between them.
Theory Component Instantiation
Means of representation Words, diagrams, tables.
Primary constructs Paradigms, approaches, methodologies, and techniques. Statements of relationship The four tiers of the framework are comprised of paradigms, approaches, methodologies, and techniques. Entities at one levelare represented as inheriting the features of the class towhich theybelong at the next levelof abstraction inthe framework,allowing the recognitionand modeling of genealogical dependencies of j methodologies.
Scope Methodologies thathave been proposed inthe scholarly literature.A procedure for
includingnew methodologies as theyarise is proposed.
Causal Explanations Not present. Testable propositions Not present.
Prescriptive statements_ Not present._
themost basic type of theory. They describe or The term typology is used more or less synonymously for
classify specific dimensions or characteristics of taxonomy and classifications, although Doty and Glick
(1994) argue that itsuse should be restricted to the special
individuals, groups, situations, or events sum
by
marizing the commonalities found in discrete case where there is a conceptually derived interrelated sets
observations. They state "what is." Descriptive of ideal types. They argue that these typologies are in
theories are needed when nothing or very little is tended to predict variance in dependent variables, because
known about thephenomenon inquestion (Fawcett
for organizations, the types identified are developed with
and Downs 1986, p. 4).
respect to a specified organizational outcome. In this case,
as the ideal types are developed with the intention of
Fawcett and Downs referredto these theories as descriptive, but this term is not entirely appropriate as this class of explaining or predicting outcomes throughfalsifiable rela theories goes beyond basic description inanalyzing or sum tionships, theDoty and Glick typology is an example of marizing salient attributes of phenomena and relationships theoryType III or IV in the taxonomy. among phenomena. The relationships specified are classi or associative, not explicitly causal. Frameworks, classification schema, and taxonomies are ficatory, compositional, numerous in IS. A classic early case is Gorry and Scott Variants of this theory type are referred to as classification Morton's (1971) framework formanagement information schema, frameworks, or taxonomies. McKelvey (1982) systems. Table 4 gives a more recent example. Other a of the of tax gives comprehensive coverage subject examples of theorizing in this category include research on onomies and classification for organizations under the thedelineation of constructs and theirassociated measures. heading of systematics, drawing upon work in biology, For Davis' work on defining and measuring ease example, zoology, and botany, where the challenges for systematics are immense. He sees this kind of science as "the science of-use and usefulness analyzed the properties thatdefined these constructs and allowed them to be measured (Davis of diversity" (p. 12). McKelvey points out the importance of systematics as a prerequisite to good scientificmethod, 1989). Some examples of grounded theory can also be inproviding clear delineation of theuniformities of classes examples of Type I theory,where the grounded theory of phenomena to be studied. method gives rise to a description of categories of interest.
MIS QuarterlyVol. 30 No. 3/September2006 623
Gregor/The Nature of Theory in IS
What constitutes a contribution toknowledge with theoryof A device of sudden enlightenment. From this this type? Theory thatdescribes and analyses is valuable, perspective theory is complex, defamilarizing, rich as stated above, when little is known about some phenom inparadox. Theorists enlighten not through con ena. Any evidence gathered would be expected to have ceptual clarity... but by startling the reader into credibility. Descriptions presented should correspond as far satori. The point of theory, in thisview, is not to as possible to "what is" (Miles and Huberman 1994). because are generalize, many generalizations
widely known and ratherdull. Instead, theory is a
Further evaluation depends on the subtype of the theory. machine"...a set of and "surprise categories The evaluative criteria for classification schema have been domain assumptions aimed at clearing away con mentioned before. Ifany classification system isdeveloped, ventional notions to make room for artful and implicit claims are that the classification system isuseful in exciting insights. aiding analysis in some way, that the category labels and groupings aremeaningful and natural,9 and thathierarchies High-level Type II theories include structurationtheory,an of classification are appropriate (most importantdivisions understanding of the world as reciprocal relationships are shown at thehighest level). The logic for theplacement between action and social structure (Giddens 1984), actor of phenomena into categories should be clear, as should the network theory,an understanding of inanimate objects and characteristics that define each category. In addition, material systems as actors or co-agents of human intentional importantcategories or elements should not be omitted from actors (Latour 1991) and the situated-action perspective, a the classification system, that is, it should be complete and model thatcontrasts routine activity situated in an environ exhaustive. A previous classification system could be re ment with theories of deliberative action (Agre 1995). vised as new entities come to light,or some preferable way of grouping or naming categories is identified.A judgement In a second subtype of theory for explaining at a lower as to the degree towhich the theory satisfies these criteria allows one to assess the contribution to knowledge. level, explanations are given for how and why things happened in some particular real-world situation. Many case studies fall into this category. A nice example from history could be a case study of Napoleon's march on Type II: Theory forExplaining Moscow. Such a case study could analyze the causal factors that contributed to a military defeat (such as cam This type of theory explains primarily how and why some paigning inwinter without good supply lines). Similarly, phenomena occur. These theories are formulated in such a case studies of failure in IS can give a good understanding way, however, thatmaking testable predictions about the ofwhat not to do when building systems. For example, an future is not of primary concern. Explanations of how, when, where, and why events occurred be analysis of three case studies of fairly catastrophic IS may presented, failures showed that, in all cases, therehad been a lack of giving rise to process-type theory. This class could well be labeled theoryfor understanding, as these theories often managerial attention to recognized IT governance and have an emphasis on showing others how theworld may be project management principles (Avison et al. 2006). viewed in a certainway, with the aim of bringing about an It can be seen that forms of this type of theory correspond altered understanding of how things are orwhy they are as they are. reasonably closely to some views of theory in the inter pretivist paradigm (Klein andMyers 1999), although other At least two subtypes of work may be distinguished here. interpretivetheory is framed in such a way that it includes In the first,theory isused as a "sensitizing device" at a high testable propositions, making itpossibly Type IV. Table 5 level to view theworld in a certainway (Klein andMyers shows an example of theoryfor explaining in IS that fits the
1999, p. 75). DiMaggio (1995, p. 391) describes theoryas interpretivistparadigm, where the theory itself is an end enlightenment, where theory serves as product and is not expected to lead to predictive, deter ministic theory. Q Foucault (1971, p. xv, citing Borges) gives a striking example of a taxonomy that violates our innate sense of order. A "certain Chinese Research approaches that can be used to develop this type encyclopedia" is reputed to have written that "animals are divided into: of theory include case studies (Yin 1994), surveys, ethno (a) belonging to the Emperor, (b) embalmed, (c) tame, (d) sucking pigs, (e) sirens, (f) fabulous, (g) stray dogs, (h) included in the present classi graphic, phenomenological, and hermeneutic approaches fication, (i) frenzied, (j) innumerable, (k) drawn with a very fine camel-hair (Denzin and Lincoln 1994), and interpretive field studies brush, (1) et cetera, (m) having just broken thewater pitcher, (n) that from a long way off look like flies." (Klein andMyers 1999).
624 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
Table 5 Example of Theory forExplaining
| Theory Overview Orlikowski (1992) developed a new theoreticalmodel, the structurationalmodel of technology,which made the claim that technology is both constituted by human agency and constitutes human practice.
Theory Component Instantiation
Means of representation Words, diagrams, tables.
Primary constructs Technology, with various conceptions intermsof both itsscope and role. Structural features
of organization, including rules and resources.
Statements of relationship An example:
Technology is an outcome of such human action as design, development,
appropriation, and modification (p. 410).
Scope The statements of relationships include no modal qualifiers. A very high level of generality
is suggested. No boundaries to the theoryare stated.
Causal Explanations The statements of relationship include causal explanations, forexample:
Technology facilitates and constrains human action through the provision of
interpretativeschemes, facilitiesand norms (p. 410).
Testable propositions Not present. It isstated (p. 423) that themodel should not be applied deterministically.
Prescriptive statements Not present.
What constitutes a contribution to knowledge with theoryof Judgment regarding the contribution to knowledge for this this type? The theorydeveloped, or the conjectures, need to typeof theory ismade primarily on thebasis ofwhether new be new and interesting,or to explain something thatwas or interesting insights are provided, and also on the basis of poorly or imperfectly understood beforehand. With case plausibility, credibility, consistency, and transferabilityof the studies,more than just a "story" is expected, as to qualify as arguments made.
theorizing the exercise must lead to conclusions with some generality. Klein and Myers (1999, p. 75) argue thatwith interpretivefield studies there is a philosophical basis for Type III: Theory forPredicting abstraction and generalization: Theories aiming at prediction say what will be but not why; Unique instances can be related to ideas and parts of the system remain a "black box." These theories are concepts that apply tomultiple situations. Never able to predict outcomes from a set of explanatory factors, theless, it is important that theoretical abstractions without explaining theunderlying causal connections between and generalizations should be carefully related to the thedependent and independent variables inany detail. There field study details as theywere experienced and/or are several reasons for leaving part of the system a black box. collected by the researcher. First, the focus of the theoreticalmodel could be on predic tion, because that is the theorist's primary interest and Again, we expect plausibility and credibility of any accounts detailed explanation of lower-level supportingmechanisms given of events in the real world and justification for are thoughtunnecessary. Some economists admit that they generalization. An aim of this type of theory is to explain are not so concerned if the assumptions underlying their how and why events happened as they did; therefore,we theory are implausible, so long as they get high predictive expect any ascriptions of causality tobe made very carefully. power (Friedman 1953). Second, reasons to justify theascrip The identification of a cause is subject to the same set of tion of causality in regularityrelationshipsmight not yet have difficulties as with other research approaches. Possible alter been uncovered. Captain Cook theorized to good practical native explanations as to what caused a particular outcome effect that regular intakes of citrus fruits helped prevent should be examined and assessed (internal validity). scurvy,without knowing exactly why thiswas so. Others
MIS QuarterlyVol. 30 No. 3/September2006 625
Gregor/The Nature of Theory in IS
would use the labels empirical generalizations (Cushing 1990; can be of interest if thesewere unknown before, especially if Kaplan 1964) or experimental laws (Nagel 1979) for this the theory's predictive power is of considerable practical category rather than theory. Third, a leaning toward this type importance, as in theprediction of share prices in finance and of theory can be detected in the logical positivist view that inpredicting theweather. The methods used to develop and theory should not include statements of causality (explana test this theoryare primarily quantitative, so rigor is expected tions) (Dubin 1978; Duhem 1962). in statistical design and methods.
Examples of this type of theory in IS do not come readily to The limitations of this type of theory should be recognized. hand, suggesting thattheyare not common. One example that The existence of regularities or correlations between two vari is related to IS isMoore's Law. In 1965, Gordon Moore of ables does not necessarily implya causal relationship. Height Intel suggested that the number of transistors,and thus the and weight are related but one does not cause the other. The power, of an integratedcircuitwould double every two years, number of ice creams sold at beaches has a strong positive while the cost remained the same. Moore later revised this relationship with thefrequency of shark attacks. We would estimate to a doubling every 18months. Table 6 presents this not conclude that ice cream eating caused shark attacks. In example. Further examples are thealgorithmic approaches to both cases, a thirdvariable, which is a determinant of both, is software cost estimation, including the COCOMO model, of more interest. This variable is temperature,with higher where the cost model is developed from empirical observa temperatures leading tomore people at beaches and more tion. Many of themathematical formulae included have an people in thewater where they can be attacked by sharks. exponential component, as experience has shown thatcosts do not normally increase linearlywith size (Sommerville 2001). In addition, our practice can improve ifwe understand why are more common in finance and econometric two variables are related. Use of a proxy such as Examples organi studies,where the researcher appears to choose independent zational size, although itmay have high predictive power in variables because they increase theR2 ina regression analysis, many circumstances, can lead to inconsistent results (Goode rather thanforany other reason. Hitt and Brynjolfsson (1996, 2001). From a pragmatic viewpoint, we are interested in p. 138) in developing models of value creation from IT note which variables can be manipulated to bring about an out that they improved thefitof themodels by "adding commonly so we need to know where causal exist. come, relationships used control variables." A furtherexample provides elabora Organizational size could be less easy tomanipulate than tion. Organizational size is used to predict organizational organizational resources, which may be the "real" pre innovativeness, thisvariable having consistently been found condition for innovation. to be positively related to innovativeness (Rogers 1995). However, many studies offer littleor no supportive analysis or justification of theiruse of organizational size as a predic tor variable; that is, they give no causal explanation for its Type IV: Theory forExplaining and be a surrogate inclusion. Organizational size could for several Predicting (EP Theory) dimensions that lead to innovation, including organizational resources, organization levels and economies of scale (Goode This type of theory says what is, how, why, when, and what 2002). will be, and corresponds to commonly held views of theory in both thenatural and social sciences, (although Type III theory Associated research approaches include statistical techniques is thought to be thenatural-science typemodel by others). It such as correlational or regression analysis and data mining. is difficult to find an appropriate short label for this theory Correlational work can be longitudinal; that is,we can show class without resorting to nomenclature such as "scientific how Y varies with a number of independent variables (X1? type" theory,which is not appropriate because of the con X2 ...) over a time period. Correlation studies can also be flicting views within the philosophy of science. Thus, this multidirectional; that iswe can say larger values of X are class will be referred to as EP theory. EP theory implies both related to largervalues of Y, and also largervalues of Y are understanding of underlying causes and prediction, as well as related to larger values of X (as in height and weight of the description of theoretical constructs and the relationships population). Neural net techniques allow models to be among them. constructed thatgive an accurate prediction of outcomes from input variables, although the reasons for the weightings Type IV theories include "grand theories" such as general applied to inputvalues are not transparent. system theory (Ashby 1956; von Bertanlanffy 1973) and the related information theory of Shannon (1948). General What constitutes a contribution to knowledge with theoryof system theory provides a very high-level way of thinking this type? The discovery of regularities thatallow prediction about many of the open systems of interest in IS. Open sys
626 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
Table 6. Example of Theory for Predicting
Theory Overview Moore's law (1965) proposed thatas technology evolved, largerand largercircuit functionscould be crammed onto a single semiconductor substrate, meaning costs would continue to fall.
Theory Component Instantiation Means of representation Words, graph.
Primary constructs silicone base materials, cost per component ina |Semiconductor integratedcircuits, |circuit. Statements of relationships The complexity (number of components per integratedcircuit) forminimum component cost will increase at a rate of roughlya factorof two per year.
Scope Stated as a general law using themodal qualifier "roughly"for the rate of increase.
Assumed that silicone was likelyto remain the base material for the semiconductor
circuits. Expected that the relationshipwould hold forabout 10 years.
Causal Explanations The general nature of the relationshipbetween improved technology and lowercosts is explained, but there is no causal explanation as towhy the power doubles. This factor was determined empiricallyby plottingthe graph of the logof the number of components I per integratedfunctionagainst the year from 1959 to 1965. j Testable propositions Yes. The predicted relationship could be tested.
Prescriptive Not statements_ present._
Table 7 Example of a Theory for Explaining and Predicting
Theory Overview Bhattacherjee and Premkumar (2004) proposed a theory thatshows the causative drivers and emergent mechanisms driving temporal changes inuser beliefs and attitude toward IT usage. This theorybuilds on expectation-disconfirmation theoryand the technology acceptance model.
Theory Component Instantiation Means of representation Words, diagrams.
Primary constructs Antecedent conditions, beliefs, attitude, disconfirmation,satisfaction, intentions Statements of relationships An example: Perceived usefulness and attitude ina pre-usage stage are linkedwith those in the usage stage.
Scope The theoreticalmodel isgiven ina very general form;boundaries are not stated and the
hypotheses have no modal qualifiers. Intesting,only one usage-related belief (perceived
Iusefulness) was examined, student subjects were used and the technologies examined
j
were computer-based trainingand rapid application development software.
Causal Explanations Yes. See p. 234: We hypothesize usage-stage belief as thejoint outcome ofpre-usage belief and disconfirmation,and usage-stage attitude as being determined jointlyby pre- j usage attitude and satisfaction.
Testable propositions Yes. The theorywas tested throughstatisticalmethods and with qualitative data.
Prescriptive statements Not themain thrust,although recommendations forpractice are given. I
MIS Quarterly Vol. 30 No. 3/September 2006 627
Gregor/The Nature of Theory in IS
ternsare seen as being in a continuous state of exchange with Many researchmethods can be used to investigate aspects of theirenvironment and interactingwith other systems outside theEP theory type, including case studies, surveys, archival of themselves. They aremodeled in termsof the familiar con studies, experiments, the grounded theory approach, quasi cepts of input, throughput,output, feedback, boundary, and experiments, statistical analysis, and field studies. Although environment. General system theoryprovides testable propo theproblems with terminologywere noted earlier, this typeof sitions, such as the law of requisite variety: only variety in a theory can have contributions from both process studies, system's responses can keep down variety inoutcomes when which look at theunfolding of events over time,and variance the system is subjected to a set of disturbances (Ashby 1956). studies,which look at the degree towhich one variable can Ashby gives very detailed explanations as towhy this law predict changes in another variable (Huber and Van de Ven applies tomany systems. General system theory has com 1995). Note also thatwith this type of theory it is legitimate monalities with other high-level approaches to theory include to have an overall dynamic theory (with feedback loops as in
cybernetics, the soft systems approach, and complex systems. general system theory),yet testhypotheses deduced from the theory in cross-sectional (variance) studies. Further examples of type IV theorycan be distinguished. The What constitutes a contribution to knowledge with theoryof Technology Acceptance Model (TAM) (Davis et al. 1989) and DeLone and McLean's this type? Studies can usefully contribute to either theory dynamic model of information success (1992 2003) both aim to explain and predict. Weber building or theory testing.Many authors provide discussion of how "scientific" knowledge should be generated and tested (1997) gives a theoryof representation,which aims tomodel thedesirable properties of informationsystems at a deep level (for example, Cook and Campbell 1979; Popper 1980) and a and be a theorynative to IS. long listof potential criteria for "good theory" can be formed, including clarity, parsimony, elegance, internal consistency, agreement with evidence, absence of disconfirmation, sound Doty and Glick (1994) show how typologies can be another ness of argument, internal and external validity, and consis form of Type IV theory, citing the example of Miles and Snow (1978), who describe the prospector, analyzer and tencywith other theory. defender types as ideal types of organization that are maxi mally effective. Organizations that resemble more closely any one of the ideal types are predicted to be more effective. Type V: Theory forDesign and Action Table 7 gives a furtherexample of EP theory. This type of theory says how todo something. It is about the principles of form and function,methods, and justificatory Investigation of how authors specify EP theory in practice theoretical knowledge thatare used in the development of IS proves interesting. Inmany cases, itappears authors have not (Gregor 2002a; Gregor and Jones 2004; Walls et al. 1992). quite made up theirmind as towhether causality is allowable in a theoryor not, orwhere itcan properly be mentioned. In There are diverging views on the status of design theory and specifying the theoryor conceptual background in a paper, it itsrelationship toother typesof theory.Relevant work can be is common togive a web of interrelateddefinitions and causal found, although it is scattered and appears under different statements that comprise the theory. Bhattacherjee and labels. Associated research has been referred to as software Premkumar's (2004) description of expectation-disconfirma engineering research (Morrison and George 1995), as a tion theory is an example, relying explicitly on causal constructive type of research (Iivari 1991; Iivari et al. 1998), reasoning and using words such as change and determine and as prototyping (Baskerville and Wood-Harper 1998), as a a process have systems development approach (Burstein and Gregor 1999; mentioning model. These authors research
hypotheses that are also stated in causal terms. In other Lau 1997; Nunamaker et al. 1990-91), and as design science research articles, the language changes in specifying a (Hevner et al. 2004 March and Smith 1995; Simon 1996). researchmodel, propositions, and hypotheses so thatany hint of causality is eliminated. Instead, vague words such as It is clear thatwork of this type occupies an importantplace associated with or linked to, rather than the strongerwords in IS. A review by Morrison and George (1995) of three influences, leads to, or determines, which are explicitly leading management IS journals showed that software causal. In the interests of developing stronger and clearer engineering related research represented about 45 percent of theory of this type, the author believes researchers should the IS articles found in the 6-year period from 1986 to 1991. make theircommitment clear and couch theirpropositions in Of thefive bodies of knowledge identifiedby Davis (2000) as terms thatshowwhat theyreallymean: that is,use statements being unique or somewhat unique to IS, two relate towhat such as "Failure to perform a systems test can be a cause of could be termed design science: IS development processes systems failure." and IS development concepts.
628 MIS QuarterlyVol. 30 No. 3/September2006
Nature of Theory in IS
Gregor/The
Design science has been given validity in IS through thework Interrelationships among Theory Types ofMarch and Smith (1995), who developed a framework to demonstrate the relationship, activities and outputs of design Figure 1 depicts graphically the interrelationshipsamong the and natural science research. March and Smith (1995) saw classes of theory. The most basic type of theory, analytic four products of design science: constructs, models, methods, theory, is necessary for the development of all of the other and implementations. These authors, however, saw the term types of theory. Clear definition of constructs isneeded inall theory as the preserve of natural science. These ideas are theoryformulation. Both theoryfor explaining and theoryfor furtherdeveloped inHevner et al. (2004), where the "artifact predicting can sow seeds for the development of EP theory itself is emphasized as the prime or only contribution of that encompasses both explanation and prediction. Design design science, although it is recognized thatthe contributions theory can be informedby all the other classes of theory. A of design science can include foundational constructs,models design methodology can build on particular idiographic and methods for the design science knowledge base, and studies ofwhat has worked inpractice, on predictive relation evaluation methods and metrics. Still, the focus of thiswork ships that are known but not fullyunderstood (such as the is on design as an activity, rather than theproblem ofwhat a relationship between organizational size and innovativeness), special design-type theorymight look like. and on fullydeveloped EP theories such as those relating to data or human behavior. representation In contrast, the notion of design theory has been adopted by Markus et al. (2002) in specifiying a design theoryforknowl Design theory and EP theory are strongly interrelated. edge management systems (see Table 8). Other examples of Knowledge of people and informationtechnology capabilities design theory include methodologies?for example, struc informs the design and development of new information tured systems analysis methods (Gane and Sarson 1979) and system artifacts. These artifacts can thenbe studied in terms Multiview (Avison andWood-Harper 1990)?and prescrip of EP theory:what impacts do the artifactshave in thework tions forbuilding specific applications?for example, decision place and in society. Note that there are many examples in support systems as inTurban and Aronson (2001). Codd's the history of science of the interaction between scientific (1970) theory of relational database design is a further knowledge and technologies. The invention of the telescope example. allowed Galileo to make astronomical observations and confirm predictions made from theory about the phases of Discussion of research approaches for this theory type can be Venus (Gribbin 2002). Yet, a design theory for a telescope found inBurstein and Gregor (1999), Hevner et al. (2004), relies on knowledge of optics for itsdesign principles. March and Smith (1995), and Nunamaker et al. (1990-91). Action research is seen as particularly appropriate Some broad bodies of theory include all classes of theoryand (Baskerville and Wood-Harper 1998). In support of this are perhaps most likely to be found inmonographs. An view, Lau's (1997) review of action research in IS studies example is thework on electronicmarkets byAjit Kambil and over a 25 year period showed that 11 of the 30 articles Eric van Heck and theircolleagues at Erasmus University in reviewed were categorized as "systems development," Rotterdam, which is described in a series of research articles covering the areas of analysis, design, development and and themonograph Making Markets (Kambil and van Heck implementation of IS and decision support systems. The case 2002). This body of theory includes Type I theory.A frame studywas thepreferredmethod of investigation. Iivari et al. work was developed showing 11 key market processes that (1998) argue for the particular importance of thesemethods are related to the execution of tradingof any kind (pp. 26-28). for applied disciplines, such as IS and computer science. Key insights that "resonate time and again throughout the book" are Type II theory, for example, "Electronic markets What constitutes a contribution to knowledge with theoryof are not technological interactions supported by humans. They this type?March and Smith and Hevner et al. point out some are human interactions supported by technology" (p. 3). conditions under which theybelieve a contribution toknowl Further Type II theory is represented by the insightsfrom the edge in design science has occurred. Their criteria include numerous case studies included. A study by Jochem Paarle utility to a community of users, thenovelty of the artifact,and burg is an example of Type III theory (p. 69). This study thepersuasiveness of claims that it is effective. Models and examined 194 consumer Web auctions in 8 countries and methods can be evaluated for completeness, simplicity, developed an index for thematurity of each in termsof trade consistency, ease of use, and the quality of results obtained processes. Statistical analysis showed that auctions with a throughuse of themethod. Simon sees "interestingness" as higher level ofmaturity had a higher transactionvolume. The a valid claim. authors note this correlation, but do not attribute causality.
MIS QuarterlyVol. 30 No. 3/September2006 629
Gregor/The Nature of Theory in IS
Table 8 Example of a Theory for Design and Action
Theory Overview Markus et al. (2002) proposed a design theoryforsystems thatsupport emergent knowledge processes.
Theory Component Instantiation Means of representation Words, tables, screen shots, diagrams.
Primary constructs Users, work context, information requirements, system architecture, processes system (and more).
Statements of relationships An example: Design forcustomer engagement by seeking out naive users (p. 188), that is, involvement of naive users leads to a greater degree of customer engagement.
Scope Systems thatsupport emergent knowledge processes.
Causal Explanations Underlyingmicro or kernel theories thatexplain the design includean understanding of the nature of emergent knowledge processes. One requirement of these processes is that general expert knowledge must be contextualized when making decisions, something that familiar expert systems are unlikely to supply, so that supplementation by personal communication systems should be considered.
Testable propositions The claim ismade that the design theorywill assist designers of other systems with similar requirements.
Prescriptive statements The design theoryspecifies themeta-requirements addressed by the theoryand gives a set of system features and a set of system principles forsystem development tomeet these requirements. The user requirements include a need to translate expert knowledge into actionable knowledge for non-experts. One development principle is that knowledge translation requires radical iteration with functionalprototypes.
(IV)Theoryfor
explaining and _ (V) Theory for
predicting design and action
(EP theory)
for
\
(II)Theory (III) for
/ Theory
explaining \ /predicting
(I)Theoryfor
analyzing
Figure 1. Interrelationships among Theory Types _I
630 MIS Quarterly Vol. 30 No. 3/September 2006
Gregor/The Nature of Theory in IS
Type IV theorywas tested in several ways. Experiments by schema is (1) exhaustive, (2) understandable, and (3) does not Otto Koppius to test the effects of information feedback in have unnecessary categories. The activity showed that the multi-attribute reverse auctions are reported (p. 82). Type V schema appears to be exhaustive: no papers were found that theory is evident in the guidelines that are given for the could not be classified under the schema. The decision rules construction of electronic markets: "You must also create a to allocate theory to a category appear to be understandable, social context for trading, one that is similar to traditional as the junior researcherwas able to apply them independently markets" (p. 3) and the strategies for introducing dynamic and arrive at similar conclusions to the author. The question markets (Chapter 8). The integrationof different sources of arises as towhether theType III category is really necessary. evidence and different types of theorizing in thismonograph Only one articlewas foundwith theory thatfell into category is impressive and the book has appeal to practitioners. III: prediction without causal explanation. It is believed that Personal experience has shown that itwas read and appre this category should be retained, even if instances in this class ciated by themanager of a farmers' cooperative thatwas in are few, both for analytic completeness and as it is a type of the early stages of developing an electronic market. More theoryrecognized by a number of authors. Table 9 shows the over, the underlying academic work satisfies aca relative frequency of each theory type. rigorous demic standards. Koppius' (2002) thesis studywas awarded the International Conference on Information Systems (ICIS) Best Dissertation Award in 2002. Together, these endeavors have led to a comprehensive, impressive body of theory. Questions forDiscussion ^^ B
It has been argued in other disciplines that different This essay has presented a number of ideas about theory in IS approaches to theory should be recognized and combined. In and proposed a taxonomy for classifying the theories we the field of management, DiMaggio (1995) suggests that develop. Many of the ideas about theorypresented are not are new and have been taken from a considerable literature. many of the best theories hybrids, combining the best qualities of covering-law, enlightenment, and process ap Nevertheless, the selection and combination of these ideas proaches. DiMaggio sees each of these types of theory as leads to a taxonomy that differs considerably in emphasis valid, but limited. In a similar vein,Weick (1995) sees many fromwhat is commonly found in discussion of IS research. An aim of thepaper is to challenge and promote discussion of products that are labeled as theory as valuable, but as approximations to theory,with "strong" or "full-blown" thenature of theoryas a topicworthy of discussion in itsown theorya result ofmany interim struggles in theorizing. right. To this end, discussion follows under the headings of questions thatencourage dialectic and some personal opinions and value judgements are included.
Application of the Taxonomy ^^^^M Do some theory types belong to particular research
paradigms?
The taxonomywas used to classify 50 research articles in two leading journals to demonstrate its applicability. Issues of An unequivocal "no" is the answer to this question. In MIS Quarterly and Information Systems Research from principle, one individual could theorize using any of the forms March 2003 to June 2004 were used as the source of the of theory. Theory Types II toV require some form of realist articles. All articles except issues and opinions, review as constructs in theoretical statements can refer to ontology, articles, and research essays were included in the classifi entities in the realworld. Type I theorydoes not necessitate cation process. Classification was performed by the author reference to real-world entities, but could be purely analytic, and a junior researcher independently, using the distin as inmathematics and logic. Apart from this qualification, guishing attributes in Table 2 as the primary basis for deci none of the theory types necessitate a specific ontological or sions to assign an article to one of Types I toV. Of the 50 epistemological position (for example, an interpretative articles studied, 45 were classified identically by each stance, or naive realism, or value-free enquiry, or quantitative researcher separately in the initial round. Five articles methods). While it is argued thatno theory class is exclu required discussion before a decision was made on classifi sively the province of any paradigm, proponents of specific cation. The appendix shows the results of this classification. paradigms favor some forms of theorymore than others. Proponents of the interpretivistparadigm have argued for the The purpose of this classification activitywas not to deter recognition of theoryType II: theory for explaining but not mine the relative frequency of publication of different theory precise prediction. Logical positivism, characterized as an types, but to test on a small scale whether the classification extreme form of empiricism, can be linked to some propo
MIS QuarterlyVol. 30 No. 3/September2006 631
Gregor/The Nature of Theory in IS
Table 9 Classification of articles inMIS Quarterly and Information Research - June Systems (March 2003 2004)
Theory Type Frequency of Occurrence
I.Analysis 3
II. Explanation 4 III. Prediction 1 IV. Explanation and prediction (EP) 33 V. Design and action 9 Total 50
nents of theoryType III. Itwould be wrong to conclude, Types III and IV theorymatch common conceptions of however, that all people who develop Type III theory are scientific-type theory, according to the natural sciences logical positivists or thatall people who develop theoryType model, as they provide theory that yields testable propo II are interpretivists,or that interpretivistsbuild theoryType sitions. As such, they have found ready acceptance in IS. II only. What is commonly held to be natural-science type The words scientific and science, however, are used in dif
theory is Type III or Type IV theory,where there is theory ferentways indifferentquarters, so it isunwise to attach them thatcan be tested. It appears thatpeople who describe them definitively to any of the theory types or to use them to argue selves as positivist researchers in IS do so primarily when for the legitimacy or value of one type of theory at the theyengage in research that includes testable theoryof these expense of another.
types. Again, this is not to say that only people who call themselves positivists build Types III and IV theory. Should theword theorybe usedfor allfive classes of theory? To reiterate,paradigmatic labels should not be attached toany of the theory types. There is no clear and direct connection Proponents have been identifiedfor each theory type to show between any theory type and any one paradigm. thateach has some support as a formof theorizing. For each class, however, there are also opponents, who would not
Is one typeof theorymore valuable than another? agree that this is an appropriate view of theory. Taking each class of theory in turn, Here again a very strong "no" is the answer to this question.
The discussion of the classes of theoryhas avoided as far as Type I: Analytic theory. Proponents include Fawcett and possible any value-laden words like formal, traditional, Downs (1986) andMcKelvey (1982). Others (Bacharach complete, or "higher" or "lower" theory. Each class of theory 1989; Doty and Glick 1994; Dubin 1978; Kerlinger 1973; can provide importantand valuable contributions. Nagel 1979; Popper 1980) would argue against this class of theorybecause no predictions are made that can be Illustrating the importance of Type I theory,we see examples empirically tested. This class also does not entail rea such as Linnaeus' 18thcentury system for classifying plants soning on thebasis of causality of any type,which would into groups depending on the number of stamens in their furtherdisqualify it from being theory for some (Nagel flowers, which provided a much-needed framework for 1979). identification (Wordsworth 1994). The many examples in IS include Kwon and Zmud's (1987) effort to unify the frag Type IT. Theoryfor explaining. A form of theory recog mented models of informationsystems implementationwith nized by DiMaggio (1995), Guba and Lincoln (1994), a categorization of factors linked to the implementation Klein andMyers (1999), and Schwandt (1994). Again, process. Type II theories have led to importantnew insights the omission of statements thatcan be empirically tested intophenomena related to IS. Type V theoryalso has obvious would disqualify this class of theoryfor some (Doty and value in IS. This typeof knowledge iswhat we impartto our Glick 1994; Dubin 1978; Kerlinger 1973; Nagel 1979; students and what fills our textbooks on applications and Popper 1980). systems development (for example, see the text by Turban and Aronson (2001) on decision support and intelligent Type HI. Theory for prediction. This class matches systems). Dubin's (1978) specification of theory,eschewing state
632 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
ments of causality (explanations) in theory,believing that theories. The goal-related characteristics of theory aremen
theybelong to the field of rhetoric. Friedman's (1953) tioned inmany definitions of theory to varying degrees and
views add support. Others would disagree with thisview thus appear to be a natural place to start in constructing a
of theory,either implicitlyor explicitly endorsing a view taxonomy. The only other structuralcharacteristic identified
that theory should include underlying justifications for as a possible candidate for distinguishing among theories is
causally based explanations, not just covering-law the level of generality. This characteristic is seen as of
explanations on the basis of universal or statistical secondary importance to thegoal, but could be used ifa two
regularities (Kaplan 1964; Merton 1967; Nagel 1979). dimensional classification schema was attempted. Further
subclassifications within the types could possibly be
Type IV: Theoryfor explaining andpredicting. Authors attempted. Structural attributeshave been used as theprimary
who expect both explanation and prediction in a theory classifier for the taxonomy, rather than socio-technical or
includeKaplan (1964), Kerlinger (1973), Merton (1967), epistemological distinctions. A classification schema that
andNagel (1979). Some opposition to this typeof theory relies on attributesof different types at the same level can be
can be detected in thework ofMohr (1982), who held problematic. If classification is attempted on nonstructural
thatprocess-type and variance-type theory should not be grounds, then it should be done separately.
mixed. The examination of a sample of IS articles, how
ever, showed a majority were classified as Type IV This essay did not have a great deal to build on in taxonomic
theory. terms and it does not claim to be the final word on theory
classification. There may be a number of other ways of
Type V: Theoryfor design and action. Simon (1996) is classifying theory,or theory categories that can sensibly be
thewell-recognized proponent of thisformof theoryand added to the current taxonomy. Further, only a limited
others have followed his lead (Gregor and Jones 2004; number of examples could be given for each theory type and
Iivari 1983; Markus et al. 2002; Walls et al. 1992). attempts to classify a largernumber of examples might reveal
Otherwise there is some feeling against recognizing further subcategories of interest.
design principles as theory.March and Smith (1995) and
Hevner at al. (2004) promote design science as a research Should one type of theoryprecede the other?
activity, but reserve the word theory for natural-science
type research (Type III and IV theory). The discussion of the interrelationships among the theory
types suggests that one type of theory could lead to theory
The approach argued for here is one of comprehensiveness, development of another type. Certainly, the components of where, followingWeick (1995, p. 386), theoryType I are necessary before theoryof other types can be expressed clearly. The differenttypes of theory,however, We would like writers to feel free to use theory could each have many origins. The development of theoryor whenever they are theorizing. Modesty is all very conjectures in the first place can occur as a result of obser well, but leaning over too far backward removes a vations ofwhat occurs in therealworld (Godfrey-Smith 2003; good word from currency. Nagel 1979) or from insights or imagination or problems or feelings (Popper 1980). Construction of an artifactand Type An advantage of labeling as varying forms of theory V can from inventiveness and theory theory spring imagination, development is that itforces theorizers to thinkclearly about ahead of good knowledge of supporting theoryof other types. what type of knowledge they are aiming at, and thenature of the statements thatcan and should be made within that theory Are thesefive theory types unique to IS? type. Another outcome of thinkingof theory in thisway is that individual contributions to theory can be considered as This question is difficult. The exploration in this essay has complementary and integrateddevelopment of theory facili been done from an IS viewpoint, based on the underlying tated. Each type of theory lacks something thatanother type philosophy of enquiry in areas thatare seen as relevant to IS: has. Stronger and fuller theory can result ifwe regard our the natural sciences, the social sciences, and the sciences of efforts inbuilding one typeof theoryas part of a largerwhole the artificial. Other fields have a similar base and possibly and remain aware of potential connections between the similar types of theory could be detected; for example, in subparts. economics, accounting, management, or archi engineering, tecture. In none of these fields, however, are thereartifactsof Could there be a betterway of classifying theory in IS? the same type thatare so squarely at the intersectionof human and machine behavior. Thus, it ispossible thatnone of these The taxonomy offered relies on the goals of a theory as the other fields needs to draw so stronglyon each of thephysical primary attribute on which to distinguish among classes of and the behavioral sciences and the sciences of the artificial,
MIS QuarterlyVol. 30 No. 3/September2006 633
Gregor/The Nature of Theory in IS
nor that they are so much in need of an integrativeview of a limited space. Readers are referredformore in-depth treat theory. In addition, it is possible that furtherexploration of ments to recent examinations of thephilosophy of science (for theory for design and action will show that it takes on a example, Godfrey-Smith 2003) and theexcellent anthology of unique form for IS artifacts (Gregor and Jones 2004). original writings on philosophy and technology compiled by Scharff and Dusak (2003). Writers in other disciplines have not considered the range in theory types that has been recognized here. DiMaggio The essay makes contribution at several levels. Novice
(1995), Weick (1995), and others in the social sciences do not researchers should benefit from the depiction of the basic discuss theory for design and action. Simon (1996) con components of theory,helping with theirquestion of "What sidered design science and natural science models but not is theory?" The approach recommended for theorydevelop others. Nagel (1979) discusses natural and social science ment is tobegin with the research problem and research ques theorybut not design science. To sum up, itappears themix tions and thendetermine which type of theory is appropriate of theory types inwhich IS has a strong interestgives us a for the problem, given the current state of knowledge in the distinctive character. The strong theory thatwe can aim at, area and using the classes depicted here as a guide. An which includes all five theory types,will be a unique product epistemological approach and research methods are then of IS. chosen as a further step. Whether one regards oneself as a
positivist or an interpretivistor a scientist should not be the
primaryquestion and indeed thiscategorization may notmake
much sense, given the simplistic and inconsistentmanner in
__ _ __ which these terms are often used. It is suggested that
Concluding Remarks researchers should thinkvery carefully and separately about issues such as and This essay began with an examination of basic problem areas causality, explanation, generalization,
in conceptualizing theory: causality, explanation, prediction, prediction in framing theory. With the realization that and generality. Using thegoals of theoryas a primarymeans stronger theorycan result from combining theoryof different of classification, five interrelated types of theorywere distin types, researchers should keep inmind the potential of their own theory to combine with other types. guished for IS: (1) theory for analyzing, (2) theory for ex plaining, (3) theory for predicting, (4) theory for explaining A furthercontribution of the essay is the support for the and predicting (EP), and (5) theoryfordesign and action. The basic building blocks of theoryhave been shown to include legitimization of each of the five classes of theory.We can see that views on what constitute theory vary considerably the necessary components of means of representation, con and it is possible that our journal editors, reviewers, and structs, relationships between constructs and the specification authors are not aware of this divergence in views, or that their of the scope of the theory. In addition, components thatvary own view of theorymay differ considerably from that of depending on thenature of the theory include causally based others. Recognition that there are diverging views on the explanations (as opposed to covering-law explanations), nature of process-type and variance-type theories and also that verifiable statements (testable propositions), and prescriptive Type IV theory can combine both process and variance statements. Examples of each theory type have been pre aspects isparticularly important.Hopefully the arguments in sented and ithas been shown that there are both proponents this essay can be used as justification by researchers for the and opponents for the attachment of the label theory to each value of theirown theory types. class.
Finally this essay, in addressing the structuralnature of IS
There are some potential limitations to the essay. The disci theory,addresses an aspect of IS research thathas been given plinary areas included for examination were the natural scant attention in comparison with discussion of our disci sciences, the social sciences and the sciences of the artificial. plinary domain, epistemological approaches, and socio There are other disciplines with different traditions that are political issues. If it can provoke furtherattention to this also relevant butwere excluded from this initial exploration. aspect of our research endeavors, then itwill fulfil a useful These disciplines include art, design, architecture, computer purpose. The lingering remnants of logical positivist thought science, law, and mathematics. Mathematics, logic, and com that accord with the omission of causal reasoning from puter science have long been contributors to the foundations theoretical statements should be explicitly confronted and we of Information Systems. Art and design could be relevant if should be very carefulwith our language inproposing theory, we think of the design process for IS as being a creative for example, indifferentiatingbetween generality and predic activity. A furtherlimitation to any essay of this type is the tive power, in defining the scope of a theory, and in the difficulty inpresenting very complex philosophical issues in wording of propositions.
634 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/TheNature of Theory in IS
As this essay had relatively little prior work on which to Ashby, W. R. An Introduction to Cybernetics, Chapman & Hall,
build, therearemany opportunities for further work. Each of London, 1956.
the theory types could be analyzed inmore detail, both for Audi, R. (ed.). The Cambridge Dictionary of Philosophy (2nd ed..), structuralcharacteristics and for instances ofwork performed. Cambridge University Press, Cambridge, UK, 1999.
S., and Wilson, D. IT Uncon
The components of the theory types have been described in Avison, D., Gregor, "Managerial Communications of the ACM fairly general terms. More precision could be attempted by July 2006, sciousness," (49:7), pp. 88-93. specifying the components in terms of symbolic logic or by Avison, D., and Wood-Harper, T. Multiview: An Exploration in structuring each type according to Toulmin's model of Information Systems Development, McGraw-Hill, Maidenhead, argumentation (Toulmin et al. 1979), which contains very 1990. UK, similar constructs and gives a useful model of everyday Bacharach, S. B. "Organizational Theories: Some Criteria for reasoning. Cross-classification of each theory type in terms Review Evaluation," Academy ofManagement (14:4), 1989, pp. of level of generality could be attempted. Empirical investi 496-515. gation of trends inproducing theoryof different types could Baskerville, R, and Wood-Harper, A. T. in Information "Diversity be undertaken. Each of the issues of causality, explanation, Systems Action Research Methods," European Journal of and prediction could be analyzed ingreater depthwith theaim Information Systems (7:2), 1998, pp. 90-107. ofmaking argument about these issuesmore accessible to IS I. "Editorial Research Benbasat, Notes," Information Systems researchers, as Lee and Baskerville (2003) have done with the (12:2), 2001, pp. iii-iv. issue of generalizability. Benbasat, I., and Weber, R. "Research Commentary: Rethinking 'Diversity' in Information Systems Research," Information To conclude, this essay was motivated by theperception that Systems Research (7:4), 1996, pp. 389-399. a fresh approach to the foundations and identity of our Benbasat, I., and Zmud, R. "The Identity Crisis Within the IS
discipline is needed, focusing on the nature of our theory as Discipline: Defining and Communicating theDiscipline's Core a fundamental issue. InformationSystems is a new discipline Properties,"MIS Quarterly (27:2), 2003, pp. 183-194. andmany of us come from very differentbackgrounds?from Bhattacherjee, A., and Premkumar, G. "Understanding Changes in Belief and Attitude Towards Information Technology A physics, chemistry,mathematics, psychology, management, Usage:
sociology, philosophy, and computer science. We need a TheoreticalModel andLongitudinalTest,"MIS Quarterly (28:2), language of our own to talk about theoryand should not adopt 2004, pp. 229-254. or Engi uncritically ideas about what constitutes theoryfrom any one F., and Gregor, Burstein, S. "The Systems Development other disciplinary area. The nature of theory in itself is at neering Approach to Research in Information Systems: An Action Research in Proceedings of the 10th least as important as domain, epistemological and socio Perspective," Australasian on Information Systems, B. Hope and political questions, which to date have attracted a dispropor Conference P. Victoria of Wellington, New tionate share of the discussion of IS research. A common Yoong (eds.), University Zealand, 1999, pp. 122-134. language todiscuss thenature of our theories should facilitate Codd, E. F. "A Relational Model of Data for Large Shared Data the building of sound, cumulative, integrated, and practical Banks," Communications of theACM (13:6), 1970, pp. 377-387. bodies of theory in IS. Cook, T. D., and Campbell, D. T. Quasi-Experimentation Design and Analysis Issuesfor Field Settings, Houghton Mifflin, Boston, MA, 1979.
Acknowledgments Craik, K. The Nature of Explanation, Cambridge University Press, Cambridge, UK, 1943. Thank you toAllen S. Lee and the reviewerswho have helped Cross, N. "Design/Science/Research: Developing a Discipline," a great deal in improving thispaper. Thanks are also due to paper presented at the Fifth Asian Design Conference: themany friends and colleagues who have commented on International Symposium on Design Science, Su Jeong Dang earlier versions of thepaper. Particular thanks are due toM. Printing Company, Seoul, Korea 2001.
Lynne Markus, who has been generous inher support, and to Cushing, B. E. "Frameworks, Paradigms and Scientific Research in Robert Johnston, forhis motivating arguments. Management Information Systems," Journal of Information Systems (4:2), 1990, pp. 38-59. Davis, F. D. "Perceived Usefulness, Perceived Ease-of-Use, and User of Information MIS References Acceptance Technology," Quarterly (13:3), 1989,pp. 319-340. P. Davis, A Technology F. D. Acceptance Model for Empirically Achinstein, The Nature of Explanations, Oxford University Testing New End-user Theory and Results, Press, New York, 1983. Information Systems:
Agre, P. "Computational Research on Interaction and Agency," unpublished Doctoral Dissertation, Sloan School ofManagement,
1-52. Massachusetts Institute of Technology, 1986.
Artificial Intelligence (72:1), 1995, pp.
MIS QuarterlyVol. 30 No. 3/September2006 635
Gregor/The Nature of Theory in IS
Davis, F. D., Bagozzi, R., and Warshaw, P. "User Acceptance of Goode, S. "On the Theoretical Justification of the Organizational
Computer Technology: A Comparison of Two Theoretical Size Construct in Information inProceedings
Systems Research,"
Models," Management Science (35:8), 1989, pp. 982-1003. of 6thPacific Asia Conference on Information Systems, The Japan G. B. "Information Foundations: Society forManagement Davis, Systems Conceptual Information, Tokyo, Japan, 2002. Looking Backward and Forward," inOrganizational and Social Gorry, G. A., and Scott Morton, M. S. "A Framework forManage on Information R. Baskerville, J. Stage, ment Information Sloan Management Review Perspectives Technology, Systems," (13:1), and J. I. DeGross (eds.), Kluwer Academic Publishers, Boston 1971, pp. 55-70. 2000, pp. 61-82. Gregor, S. "Design Theory in Information Systems," Australian Davis, G. B., and Olson, M. H. Management Journal of Information Systems, Special Issue 2002a, pp. 14-22. Information Systems: Conceptual Foundations, Structure and Development Gregor, S. "A Theory of Theories in Information Systems," in (2nd ed.), McGraw-Hill, New York, 1985. Information Systems Foundations: Building the Theoretical
DeLone, W. H., and McLean, E. R. "The DeLone and Mclean Base, S. Gregor and D. Hart (eds.), Australian National Model of Information System Success: A Ten-Year University, Canberra, 2002b, pp. 1-20. Update," Gregor, S., and Benbasat, I. "Explanations from Intelligent JournalofManagement Information Systems (19:4), Spring 2003, pp. 9-30. Systems: Theoretical Foundations and Implications for Practice,"
DeLone, W. H., and McLean, E. R. "Information Systems Success: MIS Quarterly (23:4), 1999,pp. 497-530. The Quest for the Dependent Variable," Information Systems Gregor S., and Jones, D. "The Formulation of Design Theories," in
Research the Infrastructure for the Knowledge (3:1), 1992, pp. 60-95. Constructing Economy: N. K., and Lincoln, Y. S (eds.). Handbook Methods and Tools, Theory and Practice, H. Linger, J. Fisher, W. Denzin, of Qualitative J. Zupancic, K. Vigo, and J.Arold (eds.), Kluwer Research, Sage, Thousand Oaks, CA, 1994. Wojtkowski, P. J. "Comments on 'What Theory is Not'," Academic Publishers, Boston, 2004, pp. 83-93. DiMaggio, Administrative Sciences 391-397. Gribbin, J. Science: A History 1543?2001, Allen Lane, Quarterly (40:3), 1995, pp. D. H., and Glick, W. H. as a Unique Form of Camberwell, Victoria, Australia, 2002. Doty, "Typologies Towards Improved Understanding and Guba, E. G, and Lincoln, Y. "Competing Paradigms inQualitative Theory Building: Review Research," inHandbook of Qualitative Research, N. K. Denzin, Modeling," Academy ofManagement 19(2), 1994, pp. 230-251. and. Y. S. Lincoln (eds.), Sage Publications, Thousand Oaks, CA, R. 1994, pp. 105-117. Dubin, Theory Building (Rev. ed.), Free Press, London, 1978. P. The Aim and Structure Habermas, J. Theory of Communicative Action, Volume 1: Reason Duhem, of Physical Theory, Atheneum, and the Rationalization of Society, Heinemann, London, UK, New York, 1962. 1984. Fawcett, J., and Downs, F. S. The Relationship of Theory and Hempel, C. G., and Oppenheim, P. "Studies in the Logic of Research, Appleton-Century-Crofts, Norwalk, CT, 1986.
Foucault, M. Les Mots et les Choses. Une Archeologie des Sciences Explanation,"Philosophy ofScience (15), 1948,pp. 135-175. translated by A. Sheridan as the The Hevner, A., March, S., Park, J., and Ram, S. "Design Science in Humaines, Gallimard, Paris, Information Systems Research," MS Quarterly (28:1), 2004, pp. Order of Things: An Archaeology of the Human Sciences, 75-105. Tavistock Publications, London, 1971. Hitt, L., and Bryjolfsson, E. "Productivity, Business Profitability Freese, L. "Formal Theorizing," Annual Review of Sociology (6), and Consumer Surplus: Three Different Measures of Technology 1980, pp. 187-212. M. inPositive of Chicago Value," MIS Quarterly (20:2), 1996, pp. 121-142. Friedman, Essays Economics, University Hooker, R. Aristotle: The Four Causes, 1996 (available online at Press,Chicago, 1953. http://www.wsu.edu:8080/~dee/GREECE/4CAUSES.HTM). Friedman, M. "Explanation and Scientific Understanding," Journal Hospers, J. An Introduction to Philosophical Analysis (2nd ed.), ofPhilosophy (71:1), 1974,pp. 5-19. 1967. Routledge and Kegan Paul, London, Gane, C, and Sarson, T. Structured Systems Analysis: Tools and Huber, G., and Van de Ven, A. (eds.).Longitudinal Field Research Techniques, Prentice-Hall, Englewood Cliffs, NJ, 1979. Methods, Sage Publications, Thousand Oaks, CA, 1995. Giddens, A. The Constitution of Society, Polity Press, Cambridge, Hume, D."An Enquiry Concerning Human Understanding," 1748. UK, 1984. in Introduction to Philosophy Classical and Contem Reprinted Glaser, B. G., and Strauss, A. L. The Discovery of Grounded porary Readings (3rd ed.), J. Perry and M. Bratman (eds.), Oxford Theory: Strategies for Qualitative Research, Aldine Publishing University Press, New York, 1999, pp. 190-220. Company, Chicago, 1967. to the Theoretical Iivari, J. Contributions Foundations of Sys Godfrey-Smith, P. Theory and Reality, University of Chicago Press, the Picoco Institute of Data temeering Research and Model, Chicago, 2003. of Oulu, Oulu, Finland, 1983. processing Science, University Goode, S. Size Metrics in IS Research: A Critical "Organizational Iivari, J. "A Paradigmatic Analysis of Contemporary Schools of IS
Survey of the Literature 1989-2000," inProceedings of the 12th Journal
Development," European of Information Systems (1:4),
Australasian on Information G. Finnie, D. Conference Systems, 1991, pp. 249-272. Cecez-Kecmanovic, and B. Lo (eds.), Coffs Harbour, Australia, Iivari, J., Hirschheim, R., and Klein, H. K. "A Dynamic Frame December 4-7, 2001, pp. 257-268. work for Classifying Information Systems Development
636 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
Methodologies and Approaches," Journal of Management temporary Readings(3rd ed.), J. Perry and M. Bratman (eds.),
InformationSystems (17:3), 2000 -2001, pp. 179-218. Oxford University Press, New York, 1999, pp. 139-144.
Iivari, J.,Hirschheim, R., and Klein, H. K. "A Paradigmatic Analy Magee, B. Confessions of a Philosopher, Phoenix, London, 1998. sis Contrasting Information Systems Development Approaches March, S. T., and Smith, G. F. "Design and Natural Science and Methodologies," Information Systems Research (9:2), 1998, Research on Information Technology," Decision Support Systems pp. 1-30. (15), 1995,pp. 251-266. Kambil, A., and Van Heck, E. Making Markets: How Firms Can L. A., and Gasser, L. "A Design Markus, M., Majchrzak, Theory Design and Profit from Online Auctions and Exchanges, Harvard for Systems that Support MIS Emergent Knowledge Processes," Business School Press, Boston, MA, 2002. Quarterly (26:3), 2002, pp. 179-212. Kant, I. "Critique of Pure Reason," 1781, inContinental Philosophy M. and Robey, D. "Information and Markus, L., Technology An Anthology, W. McNeill and K. S. Feldman (eds.), Blackwell, Organizational Change: Causal Structure in Theory and Maiden, MA, 1998, pp. 7-23. Research," Management Science 34(5), 1988, 583-598. Kaplan, A. The Conduct of Enquiry, Harper Row, New York, 1964. G R. "Theories of Explanation," Internet Encyclopaedia Mayes, of Kerlinger, F. N. Foundations of Behavioural Research (2nd ed.), Philosophy, August 11 2004 (available online at http://www. Holt, Rinehart and Winston, New York, 1973. utm.edu/research/iep/e/explanat.htm). Kim, J. "Causation," in The Cambridge Dictionary of Philosophy - McKelvey, B. Organizational Systematics Taxonomy, Evolution, (2nd ed.), R. Audi (ed.), Cambridge University Press, Cambridge, Classification, University of California Press, Berkeley, CA, UK, 1999, pp. 125-127. 1982. Klein, H., and Myers, M. "A Set of Principles for Conducting and Merton, R. K. Social Social Structure, Free Press, New Evaluating InterpretiveField Studies,"MIS Quarterly (23:1), Theory York, 1968. 1999, pp. 67-93. Merton, R. K Theoretical Sociology, Free Press, New York, 1967. Koppius, O. R. Information Architecture and Electronic Market ERIM Ph.D. Series Research inManagement Miles, M., and Huberman, A. An Expanded Sourcebook Qualitative Performance, 13, Erasmus University, 2002. Data Analysis (2nd ed.), Sage Publications, Thousand Oaks, CA, Rotterdam, T. The Structure of 1994. Kuhn, of Scientific Revolution, University 1996. Miles, R. E., and Snow, C. C. Organization, Strategy, Structure and Chicago Press, Chicago, Kwon, T. H., and Zmud, R. W. the Fragmented Models Process, McGraw-Hill, New York, 1978. "Unifying of Information in Critical Issues in Mingers, J. "Combining IS Research Methods: Towards a Pluralist Systems Implementation," Information Systems Research, R. J. Boland and R. A. Methodology," Information Systems Research (12:3), 2001, pp. Hirschheim (eds.), John Wiley & Sons, Chichester,UK, 1987, 240-259.
pp. 227-251. MISQ. MISQ Roadmap. Information for Prospective Authors,.
Latour, B."Technology is Society Made Durable," in_4 Sociology 2004 August (http://wAVW.misq.org/roadmap/standards.html). ofMonsters: Essays on Power, Technology and Domination, J. L. B. Behavior the Limits and Mohr, Explaining Organizational Law (ed.), Routledge, London, 1991, pp. 103-131. Possibilities and San of Theory Research, Jossey-Bass, Lau, F. "A Review on the Use of Action Research in Information Francisco, CA, 1982. Systems Studies," in Information Systems and Qualitative G. E. onto Moore, "Cramming More Components Integrated Research, A. S. Lee, L. A. Liebenau and J. I. DeGross (eds.), Circuits," Electronics (38:8), 1965, pp. 114-117. Chapman & Hall, London, 1997, pp. 31-68. Morrison, J., and George, J. F. "Exploring the Software Engi Lee, A. "Integrating Positivist and Interpretive Approaches to neering Component inMIS Research," Communications of the Organizational Research," Organization Science (2:4), 1991, pp. ACM(38:7), 1995,pp. 80-91. 342-365. Nagel, E. The Structure of Science, Hackett Publishing Co., Lee, A. S. "Editorial," Quarterly MIS (25:1), 2001, pp. iii-vii. Indianapolis, IN, 1979. Lee, A. S., and Baskerville, R. L. "Generalizing Generalizability Neuman, W. L. Social Research Methods (4th ed.), Allyn and in Information Systems Research," Information Systems Research Bacon, Boston, 2000. (14:3), 2003, pp. 221-243. Ngwenyama, O., and Lee, A. "Communication Richness in Lee, B., Barua, A., and Whinston, A. B. "Discovery and Repre Electronic Mail: Critical Social Theory and the Contextuality of sentation of Causal Relationships inMIS: A Methodological
Framework," MS Quarterly (21:1), 1997, pp. 109-134. Meaning," MIS Quarterly (21:2), 1997,pp. 145-167.
K. "The Research Norton, D. F. "Hume," in The Cambridge Dictionary of Philosophy
Lewin, Centre for Group Dynamics at Massa chusetts Institute of Technology," Sociometry (8), 1945, pp. 126 (2nd ed.), R. Audi (ed.), Cambridge University Press, Cambridge, 135. UK, 1999, pp. 398-403.
Little, D. E. of the Social in The Cambridge Nunamaker, J.,Chen, M., and Purdin, T. "Systems Development in "Philosophy Sciences," Dictionary ofPhilosophy (2nded.), R. Audi (ed.), Cambridge Information Systems Research," Journal of Management University Press, Cambridge, UK, 1999, pp. 704-706. Information Systems (7:3), 1990-91, pp. 89-106. Locke, J. "Essay Concerning Human Understanding," 1689. OED (Oxford English Dictionary) Online, August 10 2004 Reprinted in Introduction to Philosophy Classical and Con (http://dictionary.oed.com).
MIS QuarterlyVol. 30 No. 3/September2006 637
Gregor/The Nature of Theory in IS
Orlikowski, W. J. "The Duality of Technology: the Turban, E., and Aronson, J. Decision and Intelli Rethinking Support Systems of Technology inOrganizations," Science Saddle Concept Organization gent Systems, Prentice-Hall, Upper River, NJ, 2001. (3:3), 1992,pp. 398-427. Von Bertalanffy, L. General System Theory (Rev. ed.), George Orlikowski, W. J., and Baroudi, J. J. "Studying Information Tech New 1973. Braziller, York, nology in Organizations: Research Approaches and Assump J.G, G and El O. A. an Walls, Widmeyer, R, Sawy, "Building tions," Information Systems Research 1-28. Information System Design (2:1), 1991, pp. Theory for Vigilant EIS," Orlikowski, W. J., and Iacono, C. S. "Research Commentary: Research Information Systems (3:1), 1992, pp. 36-59. Desperately Seeking the 'IT' in IT Research?A Call to Walsham, G. "Interpretative Case Studies in IS Research: Nature Theorizing the IT Artifact," Information Systems Research and Method," European Journal of Information Systems (4:2), (12:2), 2001, pp. 121-134. 1995, pp. 74-81. Orlowski, W. J., and Robey, D. "Information Technology and the Watson, R. "Research in Information Systems: What We Haven't Structuring of Organizations," Information Systems Research Learned," MISQuarterly (25:4), 2001, pp. v-xv. (2:2), 1991,pp. 143-169. R. "Toward a Theory of Artifacts: A Paradigmatic Base for Weber, Passmore, J. "Logical Positivism," inEncyclopaedia of Philosophy Information Systems Research," Journal of Information Systems (Volume V), P. Edwards (ed.), Macmillan, New York, 1967, pp. (1:2), Spring 1987,pp. 3-19. 52-57. Weber, R. "Editor's Comments," MS Quarterly, (27:2), 2003, pp. Pearl, J .Causality: Models, Reasoning and Inference, Cambridge iii-xi. University Press, Cambridge, UK, 2000. Weber, R. "Editor's Comments," MS Quarterly, (28:1), 2004, pp. Popper, K. Conjectures and Refutations (3rd ed.), Routledge iii-xii. Classics, London, 2002. Weber, R. Ontological Foundations of Information Systems, Popper, K. The Logic of Scientific Discovery, Unwin Hyman, Coopers & Lybrand, Melbourne, Victoria, Australia, 1997. London, 1980. Weick, K. E. "Theory Construction as Disciplined Imagination," Popper, K. Logik der Forschung, Julius Springer, Vienna, 1934. Academy ofManagement Review (14:4), 1989, pp. 516-531. Popper, K. Unended Quest an Intellectual A utobiography, Fontana, Weick, K. E. "What Theory Is Not, Theorizing Is," Administrative Glasgow, 1986. Sciences Quarterly (40:3), 1995, pp. 385-390. Robey, D. "Research Commentary: Diversity in Information Wordsworth. Wordsworth Dictionary of Biography. Wordsworth Systems Research: Threat, Promise, and Responsibility," Infor mation Systems Research UK 1994. Editions Ltd,Hertfordshire, (7:4), 1996, pp. 400-408. E. M. Yin, R. K. Case Study Research Design and Methods (2nd ed.), Sage Rogers, Diffusion of Innovations. (4th ed.), The Free Press, New 1995. Publications, Thousand Oaks, CA 1994. York, Scharff, R., and Dusak, V. (eds.). Philosophy of Technology: The
Technological Condition An Anthology, Blackwell, Maiden, MA, 2003. About theAuthor Schwandt, I. A. "Constructivist, Interpretivist Approaches to Human Inquiry," in Handbook of Qualitative Enquiry, N. K. Shirley Gregor is theANU Endowed Chair in Information Systems Denzin and Y. S. Lincoln (eds.), Sage Publications, Thousand at theAustralian National University, Canberra, where she heads the Oaks,CA, 1994, pp. 118-137. National Centre for Information Systems Research and is Head of Seddon, P. "A Respecification and Extension of the DeLone and the School of Accounting and Business Information Systems. McLean Model of IS Success," Information Systems Research Professor Gregor's current research interests include the adoption (8:3), 1997, pp. 240-251. and strategic use of information and communications technologies, Shannon, C. "A Mathematical Theory of Communication," The Bell intelligent systems and human-computer interface issues, and the System Technical Journal (27), July/October 1948, pp. theoretical foundations of information systems. Professor Gregor 379-423/623-656. spent a number of years in the computing industry inAustralia and Simon, H. The Sciences of the Artificial (3rd ed.), MIT Press, the United before an academic career. She has Cambridge,MA, 1996. Kingdom beginning
Simpson, G. G. Principles of Animal Taxonomy, Columbia Univer published in journals includingMS Quarterly, International Journal of Electronic Commerce, International Journal ofHuman sity Press, New York, 1961.
Sommerville, I. Software Engineering (6th ed.), Addison-Wesley, Computer Studies, European Journal of Information Systems, and
2001. Information Technology & People. Professor Gregor was inaugural
Harlow, England, "What is Not," Adminis President of the Australasian Association of Information Systems Sutton, R. I., and Staw, B. M. Theory 2002-2003 and was made an Officer of the Order of Australia in the trative Sciences Quarterly (40:3), 1995, pp. 371-384. S. The Uses of Argument, Queen's Honours list in June 2005 for services as an Toulmin, Cambridge University Press, Birthday 1958. educator and researcher in the field of Information Systems and for Cambridge, UK, toReasoning, the development of applications for electronic commerce in the Toulmin, S., Rieke, R., and Janik, A. An Introduction
Macmillan, New York, 1979. agribusiness sector.
638 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/The Nature of Theory in IS
Appendix
Classification of Recent Research Articles ^^HHHH^^HI|^^HBHaHIBHg_____l
research articles and research notes are included. Issues and opinions, reviews, and research essays are excluded.
Only
Journal
Issue Title of Research Article Author(s) Theory
Type
MIS Quarterly March 2003 Predicting IntentiontoAdopt Interorganizational H. H. Teo, K. K. Wei, and I.Benbasat IV- EP Linkages: An InstitutionalPerspective Trust and TAM inOnline Shopping: An Integrated D. Gefen, E. Karahanna, and D. W. IV- EP Model Straub
Understanding the Service Component ofApplication A. Susarla, A. Barua, and A. B. IV- EP
Service Provision: An Empirical Analysis of Whinston
Satisfaction withAsp Services
Technology Capability to R. Santhanam and E. Hartono
Issues inLinking Information IV- EP
Firm Performance
CIO Lateral InfluenceBehaviors: Gaining Peers' H. G. Enns, S. L. Huffand C. Higgins IV- EP
Commitment to Strategic InformationSystems
-
June 2003 Reconceptualizing Users as Social Actors in Infor- R. Lamb and R. Kling II mation Systems Research Explanation1 Shaping AgilityThrough Digital Options: Reconcep- V. Sambamurthy, A. Bharadwaj, and V. IV- EP Technology inCon- tualizing the Role of Information Grover temporaryFirms Virtualness and Knowledge inTeams: Managing the T. L. Griffith,J. E. Sawyer, and M. A. IV- EP Love Triangle ofOrganizations, Individuals,and Neale InformationTechnology The Adoption and Use of GSS inProject Teams: A. R. Dennis and M. J.Garfield IV- EP Toward More Participative Processes and Outcomes From theVendor's Perspective: Exploring theValue N. Levina and J.W. Ross II - September 2003 Proposition in Information Technology Outsourcing Explanation Trust and the Unintended Effects of Behavior Control G. Piccoli and B. Ives IV - EP inVirtual Teams The InfluenceofQuery InterfaceDesign on Decision- C. Speier and M. G. Morris IV- EP Making Performance User Acceptance of Information Technology: Toward V. Venkatesh, M. G. Morris, G. B. IV- EP Unified and F. P. a View_Davis, Davis
MIS QuarterlyVol. 30 No. 3/September2006 639
Gregor/The Nature of Theory in IS
Journal
Issue Title of Research Article Author(s) Type
Theory
MIS Quarterly December The Contingent Effects ofManagement Support and R. Sharma and P. Yetton IV- EP 2003 Task Interdependence on Successful Information Systems Implementation The Disruptive Nature of InformationTechnology K. Lyytinenand G. M. Rose IV- EP2 Innovations: The Case of InternetComputing in Systems Development Organizations The Value Relevance ofAnnouncements of B. Defining, V. J. Richardson, and R. IV - EP Transformational Information Technology Investments W. Zmud Sources of Influenceon Beliefs about Information W. Lewis, R. Agarwal, and V. IV - EP Technology Use: An Empirical Study of Knowledge Sambamurthy Workers March 2004 User Heterogeneity and its Impacton Electronic R. Bapna, p. Goes, A. Gupta, and Y. I- Analysis3 Auction Market Design: An Empirical Exploration Jin How Do Suppliers Benefit from Information M. SubramaniIV - EP Technology Use inSupply Chain Relationships? June 2004 GIST: A Model forDesign and Management of Con- T. C. Albert, P. B. Goes, and A. Gupta V - Design tentand Interactivity of Customer-centric Web Sites A Field Study of the Effectof InterpersonalTrust on D. L. Paul and R. R. McDaniel Jr. IV- EP Virtual Collaborative Relationship Performance
Understanding Changes inBelief and AttitudeToward A. Bhattacherjee and G. Premkumar IV- EP
Technology Usage: A Theoretical Model
Information
and Longitudinal Test
The Effect of Relationship Encoding, Task Type, and N. Kumar and I.Benbasat IV- EP
Complexity on Information Representation: An
Empirical Evaluation of 2D and 3D Line Graphs
InformationSystems Research March 2003 Decentralized Mechanism Design forSupply Chain M. Fan, J. Stallaert, and A. B. V - Design Organizations Using an Auction Market Whinston XML-Based Schema Definition forSupport of W. M. P. van der Aalst and A. Kumar V - Design InterorganizationalWorkflow Informational Influence inOrganizations: An S. W. Sussman and W. S. Siegal IV- EP IntegratedApproach to Knowledge Adoption When Subordinates Become itContractors: Persis- V. T. Ho, S. Ang, and D. Straub IV- EP tentManagerial Expectations in ITOutsourcing The Social Construction of Meaning: An Alternative S. M. Miranda and C. S. Saunders IV- EP Perspective on Information Sharing _ Goods Pricing and Copyright Enforce- Information Y. Chen and I.Png V - Design Welfare ment: Analysis|_
640 MIS QuarterlyVol. 30 No. 3/September2006
Gregor/TheNature of Theory in IS
Journal Issue Title of Research Article Author(s) Theory Type InformationSystems Research June 2003 Measuring Information Technology Payoff: A Meta- R. Kohli and S. Devaraj IV- EP analysis of StructuralVariables inFirm-level Empirical Research
Developing and Validating an Observational Learning M. Y. Yi and F. D. Davis IV- EP
Model of Computer Software Training and Skill
Acquisition_
The Impactof Experience and Time on the Use of C. W. Fisher, I.Chengalur-Smith, and IV- EP
Data Quality InformationinDecision Making D. P. Ballou
A Partial Least Squares Latent Variable Modeling W. Chin, B. Marcolin, p. Newsted V -Design4
Approach forMeasuring InteractionEffects: Results
froma Monte Carlo Simulation Study and an
Electronic-Mail Emotion/AdoptionStudy
September Generalizing Generalizability in Information Systems A. S. Lee and R. L. Baskerville I- Analysis 2003 Research _
Replicating Online Yankee Auctions toAnalyze R. Bapna, P. Goes, and A. Gupta III-
Auctioneers' and Bidders' Strategies Prediction5
ImprovingAnalysis Pattern Reuse inConceptual S. Purao, V. C. Storey, and T. Han V - Design
Design: Augmenting Automated Processes with
Supervised Learning
Portfolios of Control inOutsourced Software V. Choudhury and R. Sabherwal II-
Development Products Explanation
December The Influenceof Business Managers' ITCompetence G. Bassellier, I.Benbasat, and B. IV- EP 2003 on Championing IT Horner Reich Synthesis and Decomposition of Processes in A. Basu and R. W. Blanning V - Design Organizations A Research Note Regarding the Development of the C. D. Allport and W. A. Kerler III I -Analysis Consensus on Appropriation Scale March 2004 A Fault Threshold Policy toManage Software I.Robert Chiang and V. S. Mookerjee V - Design Development Projects Economics of an InformationIntermediary with H. K. Bhargava and V. Choudhary V - Design Aggregation Benefits - EP Building EffectiveOnline Marketplaces with P. A. Pavlou and D. Gefen IV Institution-BasedTrust Does Animation AttractOnline Users' Attention? The W. Hong, J.Y. L. Thong, and K. Y. IV- EP Effects of Flash on Information Search Performance Tarn and Perceptions A Practice Perspective on Technology-Mediated U. Schultze and W. J.Orlikowski II- Network Relations: The Use of Internet-basedSelf- Explanation serve Technologies
MIS QuarterlyVol. 30 No. 3/September2006 641
Gregor/The Nature of Theory in IS
Journal
Issue Title of Research Article Author(s) Theory
Type
InformationSystems Research June 2004 ITOutsourcing Strategies: Universalistic, Contin- J. Lee, S. M. Miranda and Y. Kim IV - EP gency and Configurational Explanations of Success Real Options and itPlatformAdoption: R. G. Fichman IV - EP Implications forTheory and Practice An Empirical Analysis of Network Externalities in A. Asvanund, K. Clay, R. Krishnan and IV - EP Peer-to-Peer Music-Sharing Networks M. D. Smith
Impactof Environmental Uncertainty and Task J. Karimi, T. M. Somers and Y. P. IV - EP
Characteristics on User Satisfaction with Data Gupta
Information
Overload and theMessage Dynamics of Q. Jones, G. Ravid and S. Rafaeli IV - EP
Online InteractionSpaces: A Theoretical Model and
Empirical Exploration
Notes: EP = Explanation and Prediction Classified after discussion. Type Iwas also considered. Classified after discussion. Classified after discussion. Classified after discussion. A decision was made forType V as the paper proposes an improvedstatistical method. Classified afterdiscussion. A decision was made forType IIIas the paper focuses on a simulation of auction behavior formodeling purposes (prediction).
642 MIS QuarterlyVol. 30 No. 3/September2006