Knowledge advances with the comparative method across cases, time, and models. for several studies in this special forum, the authors examine not just one process story, but several, allowing theoretical ideas to be tested and deepened in different settings. for example, in their quantitative study, Klarner and Raisch (2013) obtain their data from annual corporate reports of 67 European insurance companies between 1995 and 2014. They code these data into six different temporal patterns of change using an innovative multiply sequence alignment method derived from the biological and then show how regularity in ongoing organizational changes is associated with performance using statistical methods. Although Klarner and Raisch had a large enough sample of cases to use statistical methods of comparison, various forum of analytical replication can also be embedded in qualitative research designs and analyses.
For example, Bresman (2013) uses an interesting embedded multiply case design, focusing on two units in a pharmaceutical company, and examining learning transferred among four successive projects occurring in each unit (for a total of eight units of analysis). His inductively derived four-phase model of vicarious learning is replicated across all his cases. This design reflects Eisenhardt's (1989), Eisenhardtand Graebner's (2007) and Yin's (2009) recommendations for building theory from cases studies. Similarly, Bruns (2013) replicates her model of collaborative research in two different setting involving multiple groups. Maquire and Hardy (2013) also compare two different cases of risk assessment processes, showing how both incorporated similar bundles of normalizing and problematizing practices, but how the differential ordering of these practices led to different consequences for the construction of risk.
Comparing distinct is not however the only way to achieve replication. It is a common misconception that longitudinal case studies represent "samples of one." However, it is important to note that the sample size for a process study is not the number of cases, But the number of temporal observations. Depending on how researchers observation in a longitudinal study can be substantial. For example, van Oorschot et al. (2013) focus on 344 individually coded events in their case history of a failing project to develop an explanatory model that explains their observations. In their archival study, Bingham and Kahl (2013) observed 399 articles and books from 1945-75 showing development of a business computer schema in the insurance industry.
Commonly however, qualitative process researchers rely on more integrative forms of "temporal bracketing" or decomposition (Langley, 1999) to identify comparative units of analysis within a stream of longitudinal data. These temporal brackets (which generally unfold sequentially over time) are constructed as progressions of events and activities separated by identifiable discontinuities in the temporal flow. They enable researchers to examine the recurrence and accumulation of progression. The permits replicating theoretical ideas in successive time periods and also to analyzing how the changing context from previous periods impacts subsequent events in current periods. Thus for example, Jay (2013) considers three successive time periods in the life of the energy alliance he studied that were punctuated by changing definitions of success. Monin et al. (2013) examine the dynamics of sense-making and sense-giving about norms of justice in three periods involving eight different is sues associated with a major merger. Lok and de Rond (2013) examine and compare five successive incidents in which institutional rules were violated and repaired, and Wright and Zammuto (2013) compare two successive incidences of rule change in the game of cricket.
Note that although analyses based on temporal bracketing may lead to propositions about patterns of progression over time in the form of well-defined phases or stages (as in the articles by Monin et al., Bingham and Kahl, and Genman et al.), thiis is not always the case. The power of temporal bracketing actually lies principally in its capacity to enable the identification of spefic theoretical mechanisms recurring over time (Tsoukas, 1989; Van de Ven, 1992). Thus, for example, Howard-Grenville et al. (2013) use temporal bracketing over three periods to show how identity reproduction and resurrection depend on an interactive process of resource mobilization and authentications of experience.
Process Representations: Rethinking Boxes and Arrows
A notable feature of many of the articles in this issue is also how the authors draw on visual maps or diagrams to represent processes and their iterative dynamics. The conventional boxes and arrows of variance studies (representing concept and causal linkages respectively) return in new forms, wherein boxes tend to represent states and arrows relations of precedence or distinctive processual elements or flows. It is also common for researchers to represent processes as "strange loops" (Hofstadter, 2008) that is, processes that depart ever further from their origin, but wind up, curiously, exactly where they stared out, as paradoxical level-crossing feedback loops. The interaction causal loop diagrams of Von Oorschot et al. (2013) are illustrative of this approach, which draws on the well-known process modeling algorithm of systems dynamics (see also Azoulay, Repenning, & Zuckerman, (2010), but diagrams of some kind are omnipresent analytical and communicational tools in the work in this issue.
For example, Gehman et al. (2013) provide a visual map of the emergence and practice of organizational values practices. The map offers a rich picture of events, with events synopsized in boxes, linked by unlabeled directional arrows demonstrating the passage of time and categorized in broad themes. Wright and Zammuto (2013) present us with an elegant depiction of the process of institutional change in county cricket. The model shows states of society, the field of cricket, and organizations (placed in boxes) transformed by various processes represented by labeled arrows. Lok and de Rond (2013) present a model in which strange loops link institutional practices, but they make some progress in resolving the vexed issue of setting complex processes in the context of time. In analyzing complex processes of change in an organization, MacKay and Chia (2013) employ a creative juxtaposition of concresced organizational states (in boxes) transformed by processes of managerial coping and adaptation (labeled arrows). The passage of time is well-represented. Similarly, Howward-Grenville et al. (2013) use diagrams to represent the iterative nature of identity dynamics over time.
The art of representing process diagrammatically still lacks the conventions of variance studies and clearly presents researchers with challenges and trade-offs. The convenience of unlabeled arrows and feedback loops may sometimes obscure the causal complexity that process theorizations are intended to explain. Yet attempts to faithfully capture the complexities of process can result in diagrams that are busy and equally opaque. As the papers in this issue moved through the stages of review to their final versions, we often saw authors struggling to creatively but accurately project the dynamics of living processes onto the static two-dimensional page. These diagrammatic representations are nevertheless often crucial in describing and communicating dynamic process theorizations.
Process Generalization: Abstracting from the Particular
Many of the enable the representation of nuance and ambiguity can be combined with more structured analytical approaches that favor the articulation and replication of more abstract theoretical ideas. At least two forms of inference from the particular to the general can be seen in the research designs represented in this forum. They provide a useful heuristic for extending research on processes of change beyond idiosyncratic stories.
First, the authors identify and make analytical generalizations to the general case of which their study is an instance. This is perhaps the most critical move in theory building to climb the ladder of abstraction by inferring the general throretical phenomenon of which the observed particular is a part (Van de Ven, 2007). This move is nicely illustrated by Van Oorschot et al. (2013), who generalize their specific study of new product terminations to the more general and abstract problem of how decision traps unfold.
Yet making this inference to an insightful general case requires concrete and penetrating understanding of the particular. Tsoukas (2009) noted that the value of small-N studues is profoundly embedded in the ability to capture situated specificity to answer the question "what is going on here?" while building on this to answer the broader question, "what is this a case of?" (Tsoukas, 2009:298). As Merton emphasized, a first step in science is "establishing the phenomenon,"because "oftentimes in science as in everyday life, explanations are provided of matters that are not and never were" (Merton, 1987:21). Maguire and Hardy (2013) exemplify the importance of grounding research problems. They ground the concept of risk in normal everyday use as being "objective" and "calculable." Their act of insight comes with the process description of how the risk of two chemical products comes to be socially constructed and forever changing through normalizing and problematical processes.