Making Great Theories
回应Cornelissen关于管理研究中定性研究受压制的问题，指出理论进步的关键障碍是不同方法论部落间缺乏对话，并提出促进跨部落交流的五条路径。
When I was asked to develop a counterpoint to Joep Cornelissen's commentary ‘Preserving Theoretical Divergence in Management Research: Why the Explanatory Potential of Qualitative Research Should Be Harnessed Rather than Suppressed’ I thought it might provide a great opportunity to enter into a debate. However, after reading through the article, I found myself agreeing with much of what he had to say. Thus, rather than debate the ‘relative’ value of qualitative versus quantitative research, my goal is to extend this discussion into what might be a productive path forward. Let me begin by noting that one would accurately characterize my background as being highly quantitative both in my training and my academic publishing. Over the past 10 years I have developed a greater appreciation for and involvement in more qualitative research projects, but by no means consider myself expert. With that in mind, let me discuss Prof. Cornelissen's article and my observations. Prof. Cornelissen rightly observes that qualitative research seemingly faces greater obstacles to publication in top tier journals. As editor of Journal of Management I know that we do not have a policy against qualitative studies, but such studies rarely succeed in the review process. He also argues that even within the qualitative domain, a trend toward a more quantitative orientation has emerged, and rightly warns of some of the associated dangers. As one who takes a ‘big tent’ approach to research, I believe that Prof. Cornelissen's observations and concerns have validity. However, I believe a bigger issue faces our field that continues to serve as an obstacle to theoretical progress. To address these I will first discuss theory and its importance. Second, I will propose measurement as central to theoretical progress and suggest that the criteria of reliability serves as both an obstacle to qualitative research appearing in top research journals and the reason for qualitative researchers promoting more quantification of qualitative research. Finally, I will point to directions that might better promote theoretical progress in our field. ‘A theory is a set of interrelated constructs (concepts), definitions, and propositions that present a systematic view of phenomena by specifying relations among variables, with the purpose of explaining and predicting the phenomenon’ (p. 7). ‘… the answer to queries of why. Theory is about the connections among phenomena, a story about why acts, events, structure, and thoughts occur. Theory emphasizes the nature of causal relationships, identifying what comes first as well as the timing of such events. Strong theory, in our view, delves into underlying processes so as to understand the systematic reasons for a particular occurrence or non-occurrence’ (p. 378). In other words theory provides an understanding of the causal relations among a set of focal phenomenon. Knowingly or unknowingly, practitioners make decisions based on theories developed from their education and experience, but seldom formally tested. Scientists, on the other hand, take a more systematic and rigorous approach to how they both develop and test theory. It is through this systematic development and testing of theory that knowledge within a scientific field can advance. Van de Ven wrote, ‘Good theory is practical precisely because it advances knowledge in a scientific discipline, guides research toward crucial questions, and enlightens the profession of management’ (1989, p. 486). Kerlinger (1986) noted that theory helps us to explain, understand, predict and control. The value of these goals is illustrated by a story articulated by Kanter's (1983) parable of how mankind discovered cooking. In the story, a son accidentally burns down the family house with the pig inside. When going through the rubble, the father accidentally pokes into the cooked pig's still hot flesh and the burn causes him to put his finger in his mouth. Having discovered how delicious this is, every time they wanted cooked meat they burned down their house. The story illustrates that predicting and controlling may work in pragmatic sense, but may not lead to deep understanding of a phenomenon. Thus, good science consists of a continual process to refine theory in a way that provides better explanation, understanding, prediction, and control. Regardless of the style of theorizing, I like to think of theoretical progress in three stages. Mantere and Ketokivi (2013) propose three forms of reasoning. First, induction entails drawing conclusions from observations. Second, deduction consists of predicting what observations are expected. Third, abduction entails inferring alternative rules given an observation. In essence induction provides the foundation for theory development, or the process through which individuals form their mental models of the relevant concepts and the relations among them usually based on some form of observation. Theory testing corresponds to deduction through examining the extent to which their model fits with a different set of observations. Abduction provides the foundation for the third phase necessary to theoretical progress: Theoretical refinement. Theoretical refinement consists of modifying the theory based on either the failure of the new observations to match the theory, or the desire to develop an even deeper understanding of the phenomenon through the addition of more concepts. This is illustrated by Whetton, (1989, p. 493) who stated ‘theoretical insights come from demonstrating how the addition of a new variable significantly alters our understanding of the phenomena by reorganizing our causal maps’. I suggest that science advances when theories are developed, tested, and refined in an iterative process moving toward deeper explanation and understanding. I would note that I am agnostic as to methodologies for doing so, and appreciate that while certain methodologies may be better suited to certain phases of the process, they all can be used in any phase. However, all methodologies require some form of measurement. Measurement serves as a central component to any of the previously discussed phases of theoretical progress. In the theory development stage, researchers observe phenomena in a setting, and from those observations, attempt to explain what they observed by identifying the main constructs of interest and proposing how those constructs relate to one another. In the theory testing stage researchers attempt to reliably and validly measure the constructs and use the theory to predict how those measures should be empirically related to one another. Based on the observed results, researchers may attempt to refine the theory to more specifically account for those results. The importance of measurement transcends different methodologies. All measures must be reliable (consistent… usually across time, items, and/or raters) and valid (accurate) (Nunnally, 1978). Reliability across raters and time is particularly relevant to this discussion. Reliability across raters requires that two or more raters, observing the same phenomena, come to the same rating or conclusion. Reliability across time entails that the same rating or conclusion would be arrived at regardless of the time at which the observation was measured. Reliability acts as a necessary, but not sufficient condition of validity. In other words, if measures are not consistent, they cannot possibly be accurate. The importance of reliability can help explain the difficulty of publishing qualitative research in the top journals as well as why a number of qualitative researchers, as Prof. Cornelissen bemoans, increasingly call for the ‘quantification’ of qualitative research. Reliability must be assessed in order to evaluate whether or not a measurement can be considered valid. Without some demonstration that the measures (or observations) have some consistency, reviewers and editors must accept ‘on faith’ that they are valid, and scientific journals seldom rely on faith. For instance, reliability across raters would require showing that two or more individuals provide the same (or very similar) assessments of a particular phenomenon, and to do so requires putting a quantitative number to the assessments. It does not require that the individuals use numbers, but that their agreement be assessed with a number. For instance, two researchers could observe 10 phenomena and assess each one categorically (e.g., weak/strong, energetic/lazy, engaged/disengaged, etc.) For each assessment we would say they were in 100 per cent agreement or 0 per cent agreement, and across all the assessments we might find that they agreed on 80 per cent of them. However, if only one researcher rates the 10 phenomena, there is no way to tell if s/he was accurate because there is no way to assess the reliability. Note this does not mean the researcher is NOT accurate… s/he may be 100 per cent accurate… but there is no way to know that because reliability has not been assessed. Consequently, top tier journals may often hesitate to publish qualitative studies that cannot provide some empirical evidence of the reliability (and consequently, the validity/accuracy) of the observations reported. However, as qualitative researchers increasingly quantify those observations in ways that allow editors and reviewers to have faith in their reliability, that may result in increasing the prevalence of qualitative research in top research journals. This seems to be the trend that Prof. Cornelissen finds troubling, yet given the importance of reliability, it should not be surprising. However, this does not take away from the seriousness of the problem he identifies as, while reliability of measurement is always important, its relative importance may vary depending upon the purpose or goal of the research. For theory testing and refinement, the importance of reliability and validity of measures take primary priority… without them one does not know if the theory truly was supported or not. However, at the theory development stage, the contribution of a study may stem more from the innovative perspectives and ideas presented than from the reliability of the observations. Their accuracy (based on the accuracy of the theory) can be later tested when the theory is tested in a more quantitative manner. For instance, at the beginning of strategic HRM research I remember the editor of one of the top journals saying to me that the research was not that good in terms of methodological rigor, but that the journal continued to publish it because of the innovative ideas. Prof. Cornellisen links different explanatory programmes to different styles of theorizing and provides examples of typical methodologies for each. His larger point seems to be that research characterized as thick and pattern descriptions seems to be on the wane both in the journals and in doctoral training. I do not have sufficient knowledge to determine the truth of his assertion, but if correct, I would agree that this would be a disturbing trend. As stated in the previous section, such research may result in innovative ideas and theories forming the foundation of later research testing those theories. However, I believe the greater obstacle to theoretical progress stems less from the decreasing popularity of any one methodology/mode of theorizing and more from the lack of discourse across the various methodologies. Not unlike organizations, as professions grow, differentiation and specialization begins to emerge. In our case, researchers increasingly specialize in qualitative, quantitative, or extremely quantitative approaches to science, and this has resulted in different tribes within the profession. Luckily these tribes seldom go to war, but sadly, the also seldom work together. They seem to happily co-exist, each with their own networks, their own conferences (or divisions within conferences), and their own journals. Their attitudes toward other tribes ranges from moral superiority to apathetic indifference. Seldom, if ever, does it move into respect or even fascination. The result is a chasm between those skilled in developing theory and those skilled in testing theory. This may explain why Kacmar and Whitfield (2000) found that only 9 per cent of theoretical presentations in AMR articles were ever tested. In other words, the transfer of knowledge from theory developers to theory testers seems something close to non-existent. This raises the important question as to the extent to which the knowledge gained from theory testing similarly transfers back to theory developers for further refinement. While having no data, I suspect such transfer happens much less frequently than it should. I would note that this chasm is not unique to the theory development/testing/refining paradigm. Similar lack of communication between macro and micro researchers has been revealed to be, in large part, due not to the focus on different levels of analysis, but rather to different disciplinary backgrounds of researchers (Molloy et al., 2011). Similarly, even within the same field of study different disciplines with different methodological paradigms fail to communicate well. For instance, within the strategic human capital literature, researchers trained in econometric methods and those trained in psychometric methods fail to communicate (Wright et al., 2014). I would suggest that these obstacles, while significant, are far from insurmountable. I suggest five avenues that might facilitate the cross-tribe dialogue necessary to increase theoretical progress: Structural, Procedural, Intellectual, Interpersonal, and Continual. The current structure of departments in academe reinforces the tribal mentality decried in this commentary. Often those specializing in qualitative methodologies work in different departments, and sometimes different schools within the university. As any productive faculty member knows, the time commitments to our own teaching, research, and service obligations provide little time to seek out colleagues outside our building, or sometimes even beyond our hallways. To the extent that we can build departments containing diverse approaches to research and co-locate them, greater opportunity for discourse will occur. In addition to rearranging organizational and physical structure, redesigning processes to encourage cross-fertilization can promote theoretical progress. Bringing together researchers from varied disciplines and methodologies to focus on a particular phenomenon can promote collaboration, communication, and appreciation. This can be done via conferences and special issues as is being done in the strategic human capital literature. For instance, Journal of Management's special issue on ‘Strategic Human Capital’ specifically aimed at providing a platform and process through with those from strategy and HRM paradigms could share their research perspectives (Wright et al., 2014). Or more recently, a workshop held at the Darla Moore School of Business at the University of South Carolina invited over 20 strategy and HRM researchers to spend a day discussing their perspectives in a face-to-face format followed by a special issue of Academy of Management Perspectives (Nyberg and Wright, 2015). Prof. Cornellisen rightly suggests that our doctoral programmes need to train students in the variety of qualitative methodologies. They also need to train students in the variety of quantitative methodologies. I do not suggest pure balance (i.e., half qualitative and half quantitative). Programmes specializing in qualitative approaches certainly should emphasize deep expertise in those, but at the same time, promote an adequate level of knowledge of quantitative methods. Similarly, those with a more quantitative bent should continue to develop strength in those methods, but provide more than shallow exposure to qualitative approaches. Students should develop a knowledge of and appreciation for the variety of methodological approaches that exist within our field. More difficult than any of the above, theoretical progress will require an attitudinal change among the different tribes. Human nature leads us to all value our own strengths more than those of others. Unsurprisingly, members of each tribe tend to think of their own tribe and its customs as superior. This arrogance acts as a primary obstacle to valuable discourse. I once heard a senior executive emphasizing the importance of humility, and he noted that humility is not thinking less of yourself, but rather thinking more highly of others. Only to the extent that we can replace the arrogance regarding the superiority of our own methods with the humility to recognize the value of other methods will the foundation for productive discourse be built. Finally, the structural, procedural, intellectual, and interpersonal avenues discussed above cannot be considered one-time events, but must be characterized by continual effort. Many of them are not new and have been proposed and tried before, yet without observable long-term success. Systems theory's entropic principle suggests that organizations tend toward disintegration over time, and thus require energy to work against this tendency. Similarly, the integration of disciplines and perspectives is not a natural process; differentiation and tribalism are. Thus, if we hope to see accelerated theoretical progress through the integration of qualitative and quantitative discourse, continual effort will be required. Prof. Cornellisen does the field a service by warning of the potential negative outcomes of de-emphasizing qualitative methods in general, and more specifically for increasingly emphasizing quantitative logic on qualitative methodologies. I have built upon this warning, by broadening the issue to the lack of discourse across and appreciation of alternative methodological approaches. My hope would be that we, as a profession, not only provide adequate opportunity for each approach to maintain its uniqueness, but that each approach's uniqueness might contribute to the other in a way that promotes overall theoretical progress. Then, perhaps we will burn down fewer houses to cook our pigs.