Rigor With or Without Templates? The Pursuit of Methodological Rigor in Qualitative Research

Explicating Rigor in Reasoning

The key question that this reasoning perspective triggers is: On what basis might one assess whether a reasoning process is rigorous? That is, on what basis might one judge whether the inferences that arise from a process of reasoning should be regarded as justified? To assess the rigor of reasoning, we propose and demonstrate the applicability of two broad criteria: the coherence of the argumentation and the process through which researchers have arrived at what they regard as the best explanation for their findings (Lipton, 2004). There may be other criteria, such as transparency (Aguinis & Solarino, 2019; Reinhardt et al., 2018), which others have related to rigor, but we consider these here as secondary to the strength of the core inferential processes that are involved in reasoning (Lipton, 2004).

We recognize that coherence is a contested concept in philosophy and cognitive science but that in general it implies a lack of contradiction within the components that support a claim to truth or a belief (see Thagard, 2000). When we use the term here, we are not pretending to use it in the strict formal sense in which a logician might use it but in the way in which it is commonly used in everyday language and in the applied reasoning practices of management researchers, as referring to consistency or fit between parts of a whole. The extent to which the parts of a process of inferential reasoning can be shown to fit together and not involve any contradictions suggests that the process is internally consistent and thus to that extent rigorous.

We argue that there are two key components of coherence in a process of reasoning. First, there is methodological coherence, which refers to the extent to which ontological, epistemological, and methodological assumptions are consistent with each other and the method employed and the extent to which the resulting theoretical claims also line up with these assumptions. The starting point for methodological coherence is one’s ontological assumptions, which in turn should inform epistemology, methodology, method, and finally, the kinds of theoretical claims that arise from a study (for discussion of the importance of aligning the assumptions of a research study with methods and claims, see Bell et al., 2019, pp. 25-34; Crotty, 1998, pp. 1-12). We do not regard any particular set of ontological assumptions as harboring the potential of rigor more than any other—the key here is that the epistemology, methodology, and so on that follow them should be consistent with them and with one another. Thus, although as noted earlier we are not using coherence in the strict sense that a logician might, nonetheless the same principle applies in the sense that there should not be contradictions between the elements that form the chain from ontology to theoretical claims.

Second, logical consistency as part of coherence refers to the extent to which there are explicit logical links provided in a manuscript that lead from data to theoretical presuppositions, to candidate theoretical frameworks, and to actual theoretical inferences. Again, this implies that each part of an argument leading to theoretical inferences should be consistent with each other part, with no apparent contradictions. Moreover, the links between the steps in the argument should be spelled out clearly in a manuscript.

There is of course overlap between these two components of coherence, but nonetheless, we treat them separately. This is because one is focused primarily on philosophical assumptions and the way that they inform methods and interpretations of data, whereas the other is focused on the explication of the steps that take a researcher from empirical data to firm theoretical conclusions. The same fundamental principle applies to both, as we have suggested, in that there should be a clear explication of linkages between components and a lack of contradiction between them.

Perhaps at face value both aspects of coherence would seem obvious to most qualitative researchers. However, when put to the test in a single piece of research, they are harder to practice. There may be slippages in the inferential steps such that a qualitative researcher ends up, for example, with a variance-based theoretical model that is not consistent with the constructivist or processual underpinnings at the ontological level that the researcher started from (Amis & Silk, 2008). As such, we propose that the general principle of coherence is still an important general criterion to strive for in the conduct and writing up of qualitative research.

Turning to the second criterion, the notion of inference to the best explanation reflects the fact that the theoretical conclusions that a researcher can draw from a given piece of empirical data are, as already mentioned, always underdetermined by the data themselves (Lipton, 2004). Except in genuinely deductive reasoning, which excludes the use of empirical data, researchers always reason inductively and thus must justify the explanatory conclusions that they arrive at in relation to their data (Mantere & Ketokivi, 2013).

This general challenge of inductively warranting theoretical conclusions from data is generally known through the question of “what is this a case of” or as the general conceptual leap from qualitative data (Klag & Langley, 2013). However, besides the general point, there has been very little consideration of the inferential linkages—offering the warrant—between empirical data and theoretical claims as reasonable conclusions. The term inference to the best explanation is often used interchangeably with abductive reasoning (Behfar & Okhuysen, 2018) and implies a process whereby, confronted with a finding or findings from the analysis of empirical data, researchers attempt to develop the “best” explanation that they can offer for the findings from among several possible explanations. This principle of inference to the best explanation has a long history. Thagard (1978) cited Darwin’s evolutionary theory, Lavoisier’s oxygen theory of combustion, and Huygen’s wave theory of light as examples of situations in which based on empirical data, scholars sought to develop and justify theoretical explanations.

We suggest here that this second criterion for rigorous reasoning comprises three subcriteria. Each of these can be understood as principles and practices that can be applied in processes of reasoning in order to seek to develop a robust and compelling—in short, rigorous—explanation for findings. We treat these criteria as distinct from those that comprise the aforementioned criterion of coherence because these latter criteria relate to the specific steps that might be applied in seeking to develop an explanation for findings over and above the coherence of the resulting argument. That is, they specify practices that are more specific than mere logical coherence. Thus, although logical coherence is vitally important, these latter criteria specify practices that may be employed to aid a particular form of logical reasoning that we argue enhances rigor.

The first element of inference to the best explanation is the consideration of several competing explanations for empirical findings rather than simply inferring a single explanation. ¹ As noted previously, the theoretical inferences drawn from empirical data are underdetermined. This implies that it is not credible, in terms of reaching conclusions, to propose an explanation as if there were no alternatives considered as part of the research process. If one is to make the case that one’s explanation of a finding is best (which does not imply that it is the only explanation but rather that a systematic case can be made that it should be preferred over others), it is surely necessary to demonstrate that this has been determined through a systematic consideration of plausible rival explanations. Such explanations may come from a variety of sources, for example, from the overall theoretical framing of the research project, reference to prior work, intuition, or “common sense” (Klag & Langley, 2013). To an extent, we would argue that it does not matter where the explanations emerge from as long as the researcher makes the abductive process explicit and can demonstrate that the proposed candidate explanation is warranted as plausible.

The warrant here is crucial. The warrant in qualitative research projects is an inductive generalization, asking of the researcher to inductively enumerate—that is, list—why the proposed candidate explanation has more correspondences with the data than any reasonable alternatives. It is possible that a warrant may also stem from a specific set of assumptions or an existing theoretical tradition, offering a principled argument in a certain direction (Ketokivi et al., 2017). Yet in qualitative research projects, such a correspondence needs to be compatible with any induced generalizations from the data.

The second component of inference to the best explanation, contrastive reasoning, provides an important means to focus explanation and thereby to narrow the range of plausible theoretical inferences one might draw. Contrastive reasoning is a process whereby, confronted by a finding, the researcher asks, “Why this finding and not another one?” (Folger & Stein, 2017). The logic here is that when researchers make inferences, they are seeking to explain a finding. Typically, researchers wish to explain something specific rather than all aspects of a phenomenon. The more specific researchers can be about what it is that they are seeking to explain, the more likely they can focus on those things that are most relevant in developing an explanation and thus develop an explanation that they can argue is a sound one.

Advocates of contrastive reasoning have argued that such reasoning oftentimes entails a fact (the finding) and a foil (a hypothetical finding) and that researchers should contrast the fact and the foil to gradually focus their theorizing on those explanatory factors that are most useful in explaining why the fact and not the foil was found (Lipton, 2004). For example, if one asks why a manager fired a certain employee, the explanation that one develops is likely to be different depending on whether the foil one poses is that an employee has been fired or that a manager has made a decision (to fire a person). Explanations that are developed through such forms of contrastive counterfactual reasoning are more likely to be “best” because they are more likely to be zooming in on precisely the aspect of the phenomenon that requires explanation and thus to have more explanatory power than if this were not the case (Tsang & Ellsaesser, 2011). One in turn considers the reasoning demonstrated in a qualitative research article more rigorous if rather than offering highly stylized or general explanations, the researchers have engaged in such contrastive reasoning to zoom in on potent explanations. In doing so, the researchers have shown that they have gone through a reflective process of entertaining possible emphases as alternatives (as theoretical “frames”) and in the process, have made a more persuasive case for the explanation that they have formed and put forward.

The third component also provides another criterion for assessing explanations and working towards the best one. An obvious challenge in a situation where there are various plausible rival explanations is how to discriminate between them and decide which one is to be proposed as the preferred explanation. In choosing among competing explanations, it is important not simply to accept the explanation that supports our implicit or explicit theoretical assumptions but rather, to aim for maximum explanatory power (Thagard, 1978). Lipton (2004) conceptualized this in terms of the “likeliest” explanation (the explanation that seems most probable based on one’s assumptions) and the “loveliest” explanation (the explanation that most fully explains the findings) and argued that in making an inference to the best explanation, one must weigh up likeliness and loveliness of possible explanations before reaching a conclusion. Some arguments will be extremely lovely in that they elegantly explain all aspects of a set of findings but are so improbable that they must be rejected. Some arguments may be highly probable but explain so little that they will be rejected. In practice, according to Lipton, often the loveliest and likeliest explanations will converge significantly.

This consideration of the virtues of various explanations within the philosophy of science resembles discussions about the qualities of social science research. Abbott (2004) distinguished between elegant stylized explanations that put form over substance based on their “parsimony” or “generality” versus explanations that are “simple” yet effective in explaining a portion of social reality. In the context of qualitative research, Weick (1999) and Cornelissen (2017) similarly referred to the trade-offs between probable explanations that provide a simple cut on a particular reality (likeliness) versus ones that are accurate enough and describe a setting in rich detail (loveliest). In most instances of qualitative research, we expect the reasoned choice of an explanation to involve a well-articulated balance between theorized abstractions that although simple, still adequately reflect the setting studied (Cornelissen, 2017). In relation to this third subcriterion, we propose that qualitative researchers specify the virtues, or qualities, of the explanations that they put forward, explicating themselves the grounds upon which they selected their theorized explanation (rather than implicitly abiding by the virtues that are tied into a particular template). The more they do so explicitly in a manuscript, the more we consider the resulting reasoning about their inferred explanations rigorous.

Now that the criteria have been mapped out, an obvious question is what this means for qualitative research. Addressing this question will be the task of the next section of the article, in which we will explore the application of these criteria by examining work that in our view demonstrates rigor and by providing some practical guidance for qualitative researchers.

Methodological Coherence

A recent article that provides a useful example of rigor regarding this criterion is Einola and Alvesson’s (2019) process study of teams. Early in their article, the authors consider how others have researched team processes, then make a very clear statement of their ontological position:

We take a radically different approach [from prior work on teams] and work with another ontological view on process. Instead of studying team processes, we study teams as a process…we are more interested in the becoming of teams, or teaming, than teams as static or dynamic entities. (p. 1892)

Having specified their ontological position, they clarify their epistemological position, arguing that from a process perspective, the way to engage with the reality of teams as processes is to understand the sensemaking of participants:

In this study, we consider sensemaking as an organizational process enabling (or disenabling, as we shall see later) the accomplishment of other key organisational processes, in this case the becoming of a team. Sensemaking is thus a core aspect and a way to understand how teams are processed. (Einola & Alvesson, p. 1897)

The methodological approach is also clearly specified and follows logically from the ontological and epistemological assumptions underlying the project. Einola and Alvesson (2019) describe their methodological approach as falling within the “interpretive sensemaking tradition to case studies” (p. 1897). They argue that they adopt a methodology that reflects the assumption that knowledge is generated through understanding the understandings of participants and in which the researcher is recognized as an active participant. Following from this latter point, they also reference “self ethnography,” an approach in which the researcher is familiar with the setting and has direct access to it.

In this study, the setting is provided by a set of 22 teams of graduate business students dispersed geographically and diverse in terms of nationality and working virtually on a series of assignments, which were set by the authors. The authors characterize these teams as “global virtual teams” (p. 1898). Reflecting the ontological, epistemological, and methodological underpinnings, data collection was longitudinal (over a 3-month period) and drew on a range of sources, with a very clear focus on the experiences and understandings of participants.

These sources comprised explanations of the teams’ ways of working written collectively by team members, multiple essays and videos produced individually by team members and reflecting on the functioning of the team, and finally interviews with members of those teams that appeared to be functioning effectively as assessed by the quality of their output of team assignments and lack of reported open conflict. In parallel, one author kept a personal diary in which she recorded informal observational data, communication with participants and course instructors, and discussion of observations with both students and staff. The method is made very explicit in the article, and it appears entirely consistent with the underlying assumptions in the sense that the aim is to explore participants’ understandings of processes without denying the active role of the researchers in co-creating knowledge.

Einola and Alvesson (2019) then explicitly position their analytical strategy with reference to their underlying ontology and epistemology, in the process explaining why they eschew templated coding approaches, which they associate with “inductive positivist research” (p. 1900). They explain their approach as follows:

We applied the principles of live coding, in which the systematic, procedural nature of the inert coding is substituted by a more open-ended and imaginative way of working with research materials, joining validation and discovery that aims at new interpretations and theoretical sightlines, instead of building conceptual walls between validation and discovery…a careful hermeneutic reading and re-reading of our material as it accumulated, gradually conveying a feeling for the parts (team member narratives as they made sense of their team experience over time) and the whole (combining the individual narratives to team narratives, then contrasting the teams)…constant efforts to make sense of specific examples of material that seemed to be representative for significant parts of the whole, while conveying some important clues of something specific and “deeper” that needed to be discovered. (pp. 1900-1901)

Finally, Einola and Alvesson (2019) present their theoretical claims about the processes they studied in a way that is consistent with assumptions, methodology, and method. These take the form of providing accounts of three teams, organized around consistent themes that were developed in the course of analysis and that focus on processes and sensemaking. They cover the dynamics of team processes, the apparent beliefs of members about how the team should function, the apparent meanings of team membership for different members, and so on. In each case, they draw conclusions about how sensemaking appears to contribute to the becoming (or “unbecoming”) of teams. For example, summarizing what appeared to have happened in one of the teams, they write: “Diverging individual sensemaking fuels distance, conflict and the corrosion of the team. Unmaking thus prevails” (p. 1905). On the basis of their comparisons of the processes in the three teams, the authors ultimately reached a series of conclusions about the ways that sensemaking played a role in shaping the dynamics of team processes. The authors explicitly state that they do not seek to generalize their findings but rather to enrich the field of studies of teamwork by showing the insights that can be gained by studying “the becoming of teams” (p. 1915) as an alternative to positivist studies that ignore process and instead focus on generalizable, universal models.

We regard Einola and Alvesson (2019) as a useful illustration of rigor in reasoning when judged against the criterion of methodological coherence. To be clear, we make no judgement of the overall rigor of the article but use it here only to illustrate this criterion. Nor do we make any judgement of the virtue of the ontological assumptions from which the other pieces of the methodological chain follow. Rather, we would suggest that given those assumptions, what follows is characterized by a high level of methodological consistency. Space constraints do not allow such detailed discussion of additional examples, but we list two in Table 1, which start with quite different ontological assumptions. These are Kothiyal et al.’s (2018) study of identity work in Indian business schools and Ozanne and Appau’s (2019) article on the consumption of religion and spirituality in Ghana.

We offer as a practical recommendation that qualitative researchers state their assumptions in their article and work coherently through them so as to offer a clear and logical basis for the theoretical claims that they are putting forward. We recognize here that authors may have various options to demonstrate such methodological consistency and may (compared to the example set out previously) wish to offer fewer references or signposts to their methodological suppositions. The discussed example is based on a self-proclaimed “radical counterfactual or antidote to existing research on teams, which then asks of the authors to provide sufficient detail on their own methodological assumptions and to offer these as a coherent alternative. In other instances, a piece of qualitative research may be part of an established theoretical tradition or stream of research and may thus require less of a detailed and deliberative approach toward seeding novel assumptions (Alvesson & Sandberg, 2011). In such instances, authors may instead weave their ontological and epistemological position into the theoretical framing and review sections of their article as well as in their method section, ensuring that there is a logically coherent thread running through the article from its premise to its logical conclusions in the form of the nature of the theoretical claims that are put forward. The questions we set out in the final column of Table 1 are intended to encourage a focus on attending to these recommendations.

Logical Consistency

Our focal example to illustrate logical consistency in practice is Bucher et al.’s (2016) study of professional associations’ responses to attempts by government to change boundaries between professions in the health care sector. The context for this study was a government initiative in Ontario, Canada, to encourage interprofessional collaboration (IPC) in the health sector. The focus of the research was on how professions struggled over professional boundaries. The specific research questions the authors sought to answer were: “How do professions engage in discursive boundary work in response to initiatives aimed at reshaping professional practice? And How do the field positions of professions influence the types of discursive boundary work that they engage in?” (p. 498).

In our estimation, Bucher et al. (2016) present an extremely logical, clear, and coherent account of the steps they went through in analyzing data, reporting findings, and then moving to a theoretical explanation for the findings. The primary data for the study comprised a report from an advisory council to the Ontario health department proposing regulatory changes and submissions in response from five professional bodies (representing physicians, registered nurses, psychologists, registered practical nurses, and psychological associates). A minor and secondary source was Ontario’s legislation for governing health professions, which was used to cross-check prior published research in the process of classifying professions in terms of field position (defined here in terms of the status of the professional group and its centrality to the health field). The authors specify the documents comprising the primary data set very clearly in the text and then summarize them in a table (p. 503) that specifies author of the text, components of the document, and its length. Crucially, they provide a link to the submissions on the relevant government websites (p. 502), thereby facilitating easy access for readers who wish to assess findings against the data set.

Bucher et al. (2016) then specify clearly the analytical approach: “Data analysis included two stages. First, we examined the professions’ field positions in terms of status and centrality. Second, we inductively analyzed the discursive boundary work strategies used by the professions, and related these to field positions” (p. 502). The analysis of field positions was fairly straightforward and brief, based on published research and government sources. The analysis of the submission to explore discursive strategies, which necessarily draws on the analysis of field position, provides a much richer set of examples of logical consistency, and we focus on it here for this reason. As the following passage shows, the authors set out the analytical steps very clearly and logically.

First, we identified themes that were indicative of boundary work by looking for explicit or implicit aspects related to professional roles and practices, capabilities, relationships, and hierarchy employed by each profession in relation to itself and other groups with whom it interacted. By moving back and forth between data and the literature, we gradually condensed the initially large number of themes into four more abstract ‘foci for framing’, common across the five texts….

In a second step, we focused on each of the professions. We explored their particular use of the four foci for framing in relationship to their field positions, extracting the most salient similarities and differences between professions’ strategies. We iteratively considered extant literature to better ground our distinctions….

In a third step, we looked across the particular uses of framing for the five professions, for patterns in the way field positions manifested themselves in framing strategies. We identified distinctive framing-field position relationships.

In a final step, we returned to the literature and to our data in an effort to better understand the theoretical mechanisms underlying the relationships discovered in the previous steps. (p. 504)

As well as telling what was done, the authors subsequently show how each phase of the analysis was conducted by presenting a series of tables that summarize the findings, surrounded by explanatory text. For example, in the first phase, a table is presented for each of the professional groups in turn, setting out the specific discursive framing strategies of each within each of the “foci for framing” and presenting concrete examples from the submissions to illustrate (Bucher et al., 2016, pp. 504-510). Each of the tables is surrounded by text that explains how the strategies were inferred from the submission, including excerpts from the text. At each stage, there is rich presentation of evidence and explanation that readers can cross-check against the submissions should they wish.

This analysis is then condensed into a single table that maps discursive strategies against field positions (degree of centrality and status) systematically to highlight similarities and differences (Bucher et al., 2016, p. 515). Again, this is buttressed with very detailed explanation of the associations that then leads into the development of theoretical inferences about how associations might be explained.

In our estimation, Bucher et al. (2016) present an extremely logical, clear, and coherent account of the steps they went through in analyzing data, reporting findings, and then moving to theoretical explanations for the findings. Through the process of reasoning that proceeds from empirical data to theoretical inference, they are able to construct an explanation of discursive struggles over professional boundaries in which each step of the process is linked to each other step in a logical and clear way. Moreover, they devote a good deal of their article to demonstrating the inferential process so that readers can follow closely how data analysis led to inferences. They also do this in such a way that readers can easily follow the steps and in ways that logically and tightly connect with one another.

Again, we list two additional example articles in Table 1. The first is Bolander and Sandberg’s (2013) study of selection decisions made by interview panels, and the second is Pouthier’s (2017) study of “griping” and joking at work. We also set out a series of questions that we hope will help scholars focus their thinking on the implementation of this criterion (Table 1). We now turn to the criterion of inference to the best explanation and its subcriteria.

Considering Competing Explanations and Contrastive Reasoning

We distinguished three sub-criteria above, and discussed the first two together in sequence, before turning to the third. The reason that we are considering the first two together is that both aspects often combine as the processes driving a reasoning process (and it would thus be hard to separate them for specific examples); the third subcriterion involves a separate assessment of the warrant for an inferred theoretical explanation. In discussing all three subcriteria, we refer to Gjerde and Alvesson’s (2020) study of middle managers (with additional very brief discussion of Cornelissen et al.’s [2014] article on the so-called Stockwell shooting for one of the subcriteria). The reason for using Gjerde and Alvesson across all three subcriteria is that it is by far the best example we are familiar with, and illustrating the three subcriteria is much easier within a single example. We do, however, list additional examples in Table 1, which we would encourage scholars to consult.

Quite early in their discussion of methodology, Gjerde and Alvesson (2020) write that “it is not possible to work exclusively inductively. Data seldom point in a clear direction” (p. 130). They explicitly reject the use of templated approaches to coding, which they suggest can lead to privileging obvious features of the data rather than digging more deeply for less obvious features that may have more value as a basis for understanding the phenomenon being researched. Our reading of this statement as a starting point for their theorizing is that the authors are explicitly stating that the same empirical data can be understood in different ways depending on how one approaches it and that this necessitates digging deep in considering alternative interpretations and in pursuit of the best explanation for what is observed.

The overall aim of the Gjerde and Alvesson (2020) study is to understand how middle managers (in this case, heads of departments in universities) experience and make sense of their role “sandwiched” between senior management and subordinates. In their interviews with middle managers Gjerde and Alvesson report that a frequently invoked (indeed, the most frequently invoked) metaphor for their identity was that of “the umbrella carrier” (p. 131), whose role was to shelter subordinates from negative effects of directives from more senior managers. The prevalence of this metaphor in the interviews led the authors to focus on it in follow-up interviews and in their analysis, yet they did not simply accept the obvious inference that the best way to explain how middle managers experienced their roles was as protectors of subordinates. In reaching their explanation, they considered alternatives in number of ways.

First, in making sense of respondents accounts, Gjerde and Alvesson (2020) considered the possibility that in characterizing their work in this way, respondents’ responses might be self-serving, rather than simply reflecting how they experienced their roles:

It may also be, as is common in human life, an element of self-serving bias…it may be tempting for heads [of departments] to rationalise their work and importance by portraying others as in need of their…support, facilitating the head as mature, powerful and good. (p. 137)

Finally, in the discussion section of the article, Gjerde and Alvesson (2020) propose the umbrella carrier metaphor as a new and useful way to understand the experience of middle managers and try to develop an explanation for why this metaphor might be invoked by respondents. That is, they move beyond the argument that this metaphor represents a useful theoretical tool for making sense of the experience of middle managers and ask why it seems to be adopted. In doing so, again a range of explanations is considered. Do they project this “protector” role because they identify with subordinates (p. 143)? Is it, as alluded to previously, a self-serving understanding (p. 144)? Are they responding to explicit calls from others to protect the capacity of subordinates to undertake their academic work (p. 144)? Having considered this range of possible explanations, they then make an argument that the umbrella metaphor can best be explained as a way of managing the complexities of the middle-management role and making sense of “middleness,” which positions the respondents between but attentive to both superiors and subordinates. In doing so, they provide a detailed explanation of their reasons for favoring this explanation rather than simply asserting it.

The presentation of reasoning in Gjerde and Alvesson (2020) illustrates one way in which authors have sought to demonstrate this key facet of inference to the best explanation. They explicitly reject the idea that one can simply infer an explanation in a direct linear fashion; then, having developed an explanation, they subject it to challenge by other competing explanations to demonstrate its explanatory power. Even then, they do not propose that it is the only explanation but rather a new one that can be added to the ways of understanding middle management.

A particularly strong feat of Gjerde and Alvesson (2020) is that the authors entertain multiple metaphors as theoretical imagery for making sense of the role of middle managers and their day-to-day experiences. They contrast alternative metaphors, ranging from ones that simply position a middle manager’s role (container, conduit) to ones that describe a particular activity (e.g., “carrying” in the “umbrella carrier” image) and ascribe varying degrees of agency to these subject positions. In doing so, they iterate between the data and what interviewees have told them and broader theoretical imagery, comparing and contrasting alternative images that best typify and explain the middle manager’s situation. As part of their article, the authors not only clearly stipulate the alternatives they have considered but have also contrasted alternatives as a way of shoring up confidence in their candidate explanation of the middle manager as umbrella carrier. Building on this example, the practical recommendation for qualitative researchers is to spell out the inductive or abductive process that they have followed by articulating the alternative explanations they have considered and engaging in a contrastive reasoning process to draw out the warrant for their proposed explanation. The more authors do this as part of their manuscripts and show reflexivity in their reasoning process from data to warranted theoretical inferences, the more robust and thus rigorous we consider their reasoning to be.

Explicating the Virtues of Explanations

The closely related step in the inferential process is to consider the virtues of the explanations that are compared and contrasted and the degree to which a researcher demonstrates awareness and reflexivity of the strengths and limitations of the claimed explanation that they put forward. We distinguished here between explanations that are chosen because they are likely, that is, simple and probable, and ones that are lovely, in that they explain a phenomenon in patently rich detail. In relation to this third subcriterion, we propose that qualitative researchers specify the virtues, or qualities, of the explanations that they put forward, explicating the grounds upon which they selected their theorized explanation. The more they do so explicitly in a manuscript, the more we consider the resulting reasoning about their inferred explanations rigorous.

Returning to the Gjerde and Alvesson (2020) article, although they do not use our terminology, they nicely illustrate this subcriterion. First, they actively question for themselves, as mentioned, whether the umbrella carrier metaphor is the right image to depict the situation and experiences of middle managers. They thus reflect on whether this likely explanation, which has been handed to them by their interviewees, is also at a second-order, theoretical level the preferred explanation of the predicament of middle managers. They conclude in the end that it is, but they do so in ways that reflect a deep consideration of the loveliness of this explanation as well. To illustrate this, what Gjerde and Alvesson do particularly well is to reflect on whether this explanation accurately and richly captures the day-to-day experiences of middle managers. They contrast the image with more stale spatial imagery of middle managers that is prevalent in the academic literature as well as notions of upward voice and representation and downward managerial sensegiving and instruction. They thus conclude that it captures more than previous imagery did. They also consciously go for a metaphor to depict middle managers’ experiences rather than other forms of theorized abstractions (e.g., propositions, process models). Metaphors typify experiences in the form or regular scenarios that as such (and compared to alternative theoretical products) offer rich and elaborate detail on the middle managers’ role and sensemaking in practice. In other words, Gjerde and Alvesson not only reflect on the warrant for their best explanation, but also they do so by considering the virtues of the explanations they are putting forward.

In addition to the Gjerde and Alvesson (2020) article that we have already discussed, another article that nicely illustrates this subcriterion is the study by Cornelissen et al. (2014) of the so-called Stockwell shooting, in which a Brazilian electrician was mistakenly identified as a potential terrorist and killed by armed police. Their study starts with the open question of why an innocent civilian was shot and what had led to this outcome. They state:

In our case, the mystery was tangible: What would explain the tragic outcome of the Stockwell shooting? How could the sensemaking of officers on the job that day, lacking sufficient evidence to support drastic action (i.e., no positive identification), result in such an outcome? (p. 706)

We surmise that most qualitative researchers, working with rich and complex data, face this challenge of settling on an explanation that is still representative of the phenomenon that they have studied as a whole (Cornelissen, 2017). In other words, they struggle to find a justifiable balance between simplicity and richness in their explanations. In other instances, researchers may preference simplicity over richness, for example, by inferring a more abstract and transferable model from a case (Gioia et al., 2013). The key here from our perspective is not to adjudicate and formally suggest that the one virtue is preferred over the other. There is no basis for such a claim (Lipton, 2004). Instead, we offer, as already mentioned, the more practical recommendation that qualitative researchers consider these virtues and actively entertain them as the basis for the candidate explanation that they put forward. The more they convey these considerations in a manuscript, the more rigorous we judge their reasoning to be. The key here is that researchers reflexively show their conscious choices for the “best” explanation and at the expense of alternatives. For example, a choice for a transferable set of propositions or process model (Gioia et al., 2013) at the expense of a thick, patterned description (Cornelissen, 2017) is a defensible choice but one that needs to be clearly articulated and justified. The more researchers can do this in an explicit manner and as part of their own reasoning—as opposed to referring to existing templates or protocols as legitimacy arguments—the more rigorous their research is.

In Table 1, in addition to the two articles we have discussed here, we list three example articles. These are Sasaki et al.’s (2019) study of family-owned businesses in Japan (to illustrate consideration of competing explanations), Rennstam and Karreman’s (2019) study of managerial control in a high-tech firm (contrastive reasoning), and Pratt et al.’s (2019) study of trust among firefighters (explicating the virtues of explanations). Having discussed and illustrated the criteria of rigorous reasoning in qualitative research and enunciated practical lessons, we now turn to discuss the implications of our reframing of rigor. We elaborate the contributions that we have sought to make and why we think they are important.