Methods for the Study of
Structuration in
Information Technology
Marshall Scott Poole
Department of Speech Communication
Texas A&M University
College Station, TX 77843-4234
409-845-5177
409-845-6594 (fax)
Gerardine DeSanctis
Fuqua School of Business
Box 90120
Duke University
Durham, NC 27708
919-660-7848
919-681-6245 (fax)
February 6, 2000
Prepared for the Organization Science Winter Conference, Winter Park, CO, February 6, 2000. Please do not cite without permission.
The impact of the theory of structuration on theorizing in organization science is noteworthy. It is becoming accepted as one of the standard theoretical resources. However, there has been much less attention to how to conduct empirical research on structuration. The purpose of this essay is to address this gap by reviewing the different methods that have been used to study structuration in organizational contexts and considering the options available to researchers.
Structuration is a complex and wide-ranging theory. A review of all applications, even those in organizational studies, would be far more extensive than the current format allows. So we will restrict our consideration to the use of structuration in information systems research. This review specifically focuses on research that attempts to gather evidence on how structuration occurs in information systems and on the impacts of structuration of information technology (IT). Structuration theory has a rather long history of application in this area over the past fifteen years and it can truly be said to be one of the dominant perspective in information systems research. There have been numerous empirical studies, which cover a representative range of modes of inquiry, Hence, a review of studies in this area will cover the field of possible projects pretty well. While IT may seem a specialized and narrow subject, we believe that many of the insights we garner will extend to other applications of structuration theory.
The theory of structuration is fascinating. It is easy to get lost in a contemplative reverie about dualities, the flow of intentionality, the attitude of late modernity, modalities (whatever they are!), and other such issues, and how they apply to information systems research and life as a whole. Because structuration theory attempts to cut across traditional problems connected with the action-structure and stability-change dualisms, it has inspired a number of “calls to arms” that catalog its promise for organization studies in general and information systems in particular (Ransom et al., 1983; Poole, Seibold, & McPhee, 1985; Markus & Robey, 1988; Poole & DeSanctis, 1990; DeSanctis & Poole, 1994). Such exhortations are useful, because they describe the potential of the theory for the field, and point the way for further research. However, without a succession of empirical studies that develop and test the abstract theory in concrete settings, such calls eventually ring hollow. Flesh must be put on the theoretical bones if the theory is to walk and live.
There is another reason that empirical research on structuration theory is important. It is necessary to keep us from letting an attractive theory become a self-sealing. Structuration theory is quite pliable. It is extremely easy for a cognitively complex mind to reinterpret any phenomenon in structurational terms, to see production and reproduction, the mutual entailment of action and institution, etc. in everything. But reading structuration into a phenomenon does not really advance our understanding very much. It is akin to applying one of those other great, seductive theories, exchange theory, to social life. In this case everything becomes an exchange, and the circular exchange theory formula is applied easily and in slapdash fashion:
Why did X do Y?
Because of the exchange: X got something X valued or avoided
something noxious.
How do you know an exchange took place (or was meaningful to X)? Because X did Y.
(or you can start at the bottom and work up; it’s circular!)
In such cases the theory is not really tested, but applied as an article of faith. For our understanding of structuration to advance it is important to have rigorous and critical applications of the framework in empirical research and specification of how structuration happens. This requires concrete research.
Before going further it is necessary to pause to register an important caveat: it is impossible to falsify structuration theory per se. It is one of a handful of “grand” formulations (along with exchange theory) that might be called meta-theories, theories that put forth a particular way of viewing the world which is so general and encompassing that it cannot be refuted or falsified in any definitive way. Meta-theories rest on the theorist’s value choices and are therefore not susceptible to refutation—there is always a way to “save the appearances.” However, what can be empirically explored and (sometimes) tested are structurational theories developed for particular contexts, such as the implementation of information systems. In specific contexts conjectures concerning how structuration occurs, the conditions that influence it, and its consequences or outcomes can be advanced and tested. The context provides the frame—or cage—that confines the theory enough to test it. (However, if it fails the test in any specific context, it can always take flight into the stratus clouds of abstraction and hang there, waiting to pounce upon some other unsuspecting phenomenon.) Perhaps the closest we could come to ”falsifying” structuration theory would be a judgment of its utility on the basis of success or failure across a number of specific phenomena.
Structuration theory comes into contact with the empirical world through research in particular contexts. But how does one conduct research on such a complicated and multilayered process? Where do we start and how do we break the research down into manageable tasks? The following review assumes that the reader is familiar with structuration theory and goes right into methodology. Those unfamiliar with the theory can find a short summary in Appendix A.
The Conduct of Research on Structuration
A General Frame
Giddens (1984) maintains that there are two general strategies for research on structuration: the analysis of strategic conduct and institutional analysis. The analysis of strategic conduct takes institutions as a backdrop and focuses on how actors draw on and reproduce structures in social practices. Institutional analysis assumes that strategic conduct is going on, but focuses on the structural characteristics of institutions and their long-term development. Each approach "brackets" a certain part of the structurational process and uses the rest as an unanalyzed ground for its object of interest. Giddens notes that this bracketing artificially segments structuration, but argues that it is necessary for methodological purposes. While we agree that parts of any structurational process must serve as "ground" against which to discern the "figure", we do not believe that structurational studies have to focus only on action or institutions. An alternative is to shift back and forth between action and institutional levels during an analysis, hoping that the "whole" will convey the nature of structuration. Studies by Smith (1983) and Brewer (1988) do this in a single report.
However, no single study can fill in the whole story behind structuration. Structuration is an encompassing and complex process, and developing a thorough understanding—as opposed to valuable but isolated insights—requires multiple studies that focus on different levels and aspects of a phenomenon. A program of research must explore the processes involved in institutional analysis, strategic action, and the relation of the two realms through multiple studies, each of which captures part of the whole.
An analysis of structuration should address a set of interlocking problems:
(1) It must develop a good analysis of how the system works. This analysis defines the general field in which structuration occurs and its "surface" indicants. It forms a foundation for the analysis of constitution of the group system. By implication, this requires identification of causal links that characterize the system's operation and members' interpretive maps. It also requires determining what features of the context influence structuration.
(2) It must identify the array of relevant structures that are used to constitute the system. This may involve identification of both potential and active structures in both the structuring process and its context. All such identifications are, of course, reifications that "freeze" the modalities of structuration for purposes of analysis.
(3) It must identify structuring moves or processes by which agents appropriate these structures, producing and reproducing them in activity. As Poole and DeSanctis (1992; DeSanctis & Poole, 1994) note, structuration in groups can be studied on at least three levels: the microlevel moves involved in appropriating structures into group interaction, the larger global patterns of appropriation that stretch over several meetings, and at the societal level of general discourse about the relevant structures. Another significant object of inquiry are the impacts of context on structuring moves and processes.
(4) It should also clarify the mediation of one structure by others, as well as the contradictions between structures and their role in the structurational process.
(5) It ultimately should shed light on how social institutions are reproduced or shaped by the process in question. This turns the context issue on its head: how does structuration influence the context itself?
(6) The subjects or actors themselves are produced and reproduced in structuration (Poole et al., 1985). Hence we also have to account for the positioning of the subject in the social system.
(7) It must also undertake critical inquiry into the power dynamics underlying the structuration process and possible relations of dominance among different classes of actors. Power imbalances, covered by ideologies of rationality or equality, can strongly shape structuration, creating relative advantage for certain parties at the expense of other parties and perpetuating preexisting biases in social institutions.
These are not a set of stages for research. Instead they are the “parts” that make up a fully realized structurational theory. Working on any single problem can also produce insights on others. For example, characterizing the structuring moves and structuration in microlevel interaction may lead to identification of additional structures and also to insights as to how the system "works". Insights into how structuration shapes social institutions may shed light on structuring moves through illuminating constraints. Ideally a mature theory of structuration would incorporate these multiple and multilayered analyses into a complex, but coherent whole.
These considerations have several implications for the way in which we go about studying structuration. First, they imply that research on structuration can take two modes. Groups are systems of human interaction, and all action can be described and explained from both the exterior perspective--the viewpoint of the analyst who seeks to understand what causes actions and what makes action effective--and from the interior perspective--in terms of the actions, cognitions, interpretations, feelings, and intentions of the actors. Hence, a first layer of analysis looks at groups from the outside, from the perspective of the observer interested in understanding the factors which cause group behavior and which determine group outcomes. A second layer of analysis attempts to get more of an inside perspective on the group, to study the interpretations that give meaning to events and the actions and interactions that constitute the group, its processes, and member responses to exogenous influences. The goal of the second layer of analysis is to elucidate the processes that constitute the variables, causes, and effects that the first layer of analysis identifies. Each layer of analysis involves a different approach to research, and each reveals different aspects of group processes.
The first layer, functional analysis, focuses on the system itself and depicts it as a network of causal, moderating, and correlational relationships among abstract variables. Exogenous influences, group processes, and outcomes are decomposed into well-defined variables. These variables may be part of the members' lifeworld, but they are defined and operationalized by researchers; they reflect the scientist's perspective and are subjected to analysis of reliability and validity. The validity of a functional analysis depends on its ability to stand up to empirical evaluation; relationships between variables are relevant only insofar as they can be investigated in the laboratory or field. Most current research on groups and GDSSs relies on functional analyses. Hence, most of what we know about groups is stated in causal terms. For example, Poole and Roth (1989b) found that task complexity was inversely related to the complexity of group decision paths. Systems of variables can also be modeled, as this explanation advanced by Homans (1950) and Schacter (1951) illustrates: increased liking among members causes an increase in communication; increased communication causes an increase in perceived similarity among members; increases in perceived similarity causes increased liking (which completes a deviation-amplifying causal loop). However, functional analyses need not be quantitative. Qualitative methods have been used to derive generalizations about behavioral systems since the turn of the century. Interpretive insights into subjects can generate useful and powerful functional theories, such as Lindesmith’s (1947) study of addicts and Glaser and Strauss’s (Strauss, 1987) numerous grounded theories illustrate. And qualitative and quantitative modes of analysis can be combined in functional research, as Weber’s analysis of the Protestant ethic illustrates. Functional analysis is useful, because it permits both explanation and prediction of behavior. It also employs an explanatory scheme that is amenable the application of powerful research and statistical methods to test theories.
But a second layer, constitutive analysis, is necessary to discover how group systems and the variables and relationships that compose them are constructed through group interaction. A functional analysis regards variables and relationships as non-problematic, so long as they satisfy the canons of method. However, any theory grounded in human action is inherently incomplete without an account of how interpretive and interaction processes construct variables and figure in the operation of causal relationships. For example, task is commonly used as a causal variable in systemic studies. However, it is not task per se, but members' interpretation of the task which influences group processes and outcomes. These interpretations may be incorporated in a causal model as "perceived task", but this transforms an interpretive process into a static variable. It cannot capture the role of interpretation in the group's approach to its work. Underlying every variable and relationship in a causal analysis is a process of social construction responsible for making it an active force in the group. A constitutive analysis takes the causal model as a starting point, shows how it operates through member activities, and then adds additional rich detail about these processes.