Introduction

Management and Organization Studies (MOS) is a multidisciplinary field of research which provides in-depth analyses and knowledge on organizational behavior, culture, psychology, or theory among many others (Augier et al., 2005; Maclean et al., 2016). Today, claims abound within MOS that the field suffers from all kinds of defects, including most of all the production of dissemination of research that is irrelevant, meaningless and parochial, and that members of the field show increasing lack of integrity and focus on the superficial and easily publishable (e.g. March 2005; Bartunek et al. 2006; Pfeffer 2007; Miller et al. 2009; Macdonald & Kam, 2010; Adler & Hansen 2012; Alvesson 2012; Alvesson & Sandberg 2013; Barley 2016; Huzzard et al. 2017; Tourish, 2019). Similar claims have been made before (e.g. Koontz 1961, 1980; Sprague & Sprague 1976; Tinker & Lowe 1982), but seem to become more common.

In addition to all these charges, one development weighs heavily on MOS, namely the development of disconnected academic tribes with their own academic language and an own territory (Becher & Trowler, 2001). Tribalism is defined as the emergence of closed-off communities within an academic discipline with own research styles and values. We suspect these closing-off dynamics are especially prevalent in divergent (regarding the potential to share knowledge) and rural disciplines such as MOS with a low people-to-problem ratio.Footnote 1 Paired with the ongoing differentiation of academic disciplines, this process should result in the emergence of small subcommunities of micro-tribes, which in turn would follow lines of “boxed-in” research “characterized by a strong and narrow focus on some issues within a well-defined, specialized intellectual terrain” (Alvesson & Sandberg, 2014). This, in turn, nudges MOS scholars to the practice of “following of fashion, adapting to a specific sub-tribe and making similar-minded colleagues happy” (Alvesson, 2013), and a loss of “flexibility with regard to problem solving, adaptation, and creative idea generation” (Dane, 2010, p. 579).

Such developments entail three problems besides a boxing-in and a potential growth of irrelevance of MOS. First, tribalism contributes to the fragmentation of MOS, meaning that its ability to integrate different views to tackle problems and research puzzles diminishes. Secondly, tribalism yields the danger of the formation of citation cartels (Tienari, 2012), which would contribute to further fragmentation of MOS, as it prohibits scholars outside the citation circles to join the MOS (sub-)discourses. Finally, this could promote the spread of authoritative citation of dominant actors in the field, which does not provide additions to the knowledge stock (Davis, 2015; Whitley, 2000). Instead, the stark orientation on the most prevalent sources could colonize promising approaches and lead to a crowding-out of solutions of research puzzles as well as practical problems.

Despite the potential dire consequences that would arise if MOS were tribalized, this assumption was not yet subjected to empirical investigation. This is even more puzzling, as there is not only the hypothesis of tribalization present in MOS, but there are also voices that claim that MOS is getting more tribalized (Alvesson & Sandberg, 2014; Macdonald & Kam, 2010). With this in mind, we take both the assumption that MOS is tribalized and the consequences stemming from the associated fragmentation of the discipline into citation circles seriously and ask…

RQ1 …how many tribes make up Management and Organization Studies, and …

RQ2 … how fragmented is its knowledge base?

To trace the potential tribalization of MOS and the fragmentation of its knowledge base due to the formation of citation circles empirically, we employ a diachronic co-citation network analysis on 22,430 papers published in 14 MOS journals between 1980 and 2019. We split the data into timeframes of 5 years each and apply canonical centrality measures (Freeman, 1978), and constraint (Burt, 2004, 2017), and global measures such as modularity and transitivity to trace changes regarding the fragmentation of MOS between these timeframes. We also employ measures that enable us to control for the presence of single- or multiple center–periphery structures in the data (Everett & Borgatti, 1999; Kojaku & Masuda, 2018a, 2018b). The idea here is that tribalism should be present if we identify a multitude of relatively closed-off, densely connected regions of the network with a semi-periphery connecting a number of these regions and a number of peripheries which are adjacent to the tribal centers. We also introduce and use absolute and relative flow between the co-citation communities we have detected, with a time dimension. Thereby, we can identify potential evolving microtribes in MOS, in the shape of closing knowledge circuits. Finally, we adopt node removal strategies (Michele Bellingeri et al., 2020a, 2020b; Boldi et al., 2013; Yang et al., 2018) to check if ritualized citations of the most influential papers holds the otherwise tribalized discipline of MOS together.

To answer the research questions, the remainder of the article is structured as follows. We begin with discussing the theoretical factors driving or prohibiting the formation of tribes in MOS in "Factors driving or prohibiting the emergence of micro-tribes in Management and Organizational Science" section. In the same section, we also present which network structures we expect to observe over time. We then proceed with introducing our research strategy, data and employed methods in more detail in "Research strategy" section, before we present our results in "Results" section. The findings are then discussed in Discussion and Conclusion and linked to the theoretical assumptions outlined in "Factors driving or prohibiting the emergence of micro-tribes in Management and Organizational Science" section. We then close our paper with highlighting the limitations and potential future research avenues regarding tribalism in and beyond MOS.

Factors driving or prohibiting the emergence of micro-tribes in Management and Organizational Science

Theoretical background

In general, the emergence of academic (micro-)tribes is linked to (1) increasing specialization and internal differentiation of scientific disciplines, (2) the intersection of the paradigmatic core with the most relevant journals and subsequent formation of subcommunities and citation circles, and (3) the mono- or interdisciplinary constitution of the discipline under investigation.

(1) Despite recent analyses have argued that specialization is indeed a natural feature of science and therefore an ongoing process (Abbott, 2001; Leahey, 2007), it is linked to fragmentation when driven by a split in research communities regarding basic and applied research (Santos et al., 2022), methodological divides (Schwemmer & Wieczorek, 2020), struggles among schools of thought (e.g. Münch, 2018 on the shism in German sociology), and the absence of chances to collaborate and integrate knowledge, especially in multidisciplinary research fields (Wittek et al., 2023).

Fragmentation, conflict, and specialization are especially witnessed in disciplines characterized by low levels of collaboration, a lack of resources, and without a strong, paradigmatic core (Abbott, 2001; Turner, 2006, 2016; Wieczorek et al., 2021a, b). Every discipline and the scholars therein have a limited attention space, usually restricting the number of in-depth discourses. As academic disciplines branch and differentiate, each branch may develop a limited number of topics suitable for discourse, ultimately severing ties with to other branches, or, to stick with the metaphor, the disciplinary tree from which they grew. Following Collins (2002, p. 24), this may result in “[i]solated communities, where the same lineup of persons are recurrently thrown together”, which “tend to reify their symbols as if they were concrete objects; at the extremes of self-subsistent tribes or deliberately separated cult communities”. Here, the symbols are the ideas, paradigms, outlets, publications to be cited, or the (idolized) forerunner of an academic tribe (ibid. 36). Under these circumstances, increasing differentiation with regards to theories, methods applied, thought leaders, and research questions entail a limited possibility to exchange ideas. In the following, microtribes with boxed-in research (Alvesson & Sandberg, 2014) emerge and limit the flow of ideas even further.

According to Whitley (2000), MOS is no exception. He argued as early as 1984, that MOS is characterized as “fragmented adhocracy” in which “research is rather personal, idiosyncratic, and only weakly co-ordinated across research sites”. Consequently, scholars “do not have to produce specific contributions which fit in to those of others in a clear and relatively unambiguous manner”. In MOS, “there is a wide variety of work techniques, approaches and audiences”, and a “lack of systematic interconnection” between “highly differentiated and specialized subfields” (Whitley, 1984, pp. 798–799). Translated into the tribalistic framework of Becher and Trowler (2001), MOS is a rural, divergent discipline that might yield multiple, relatively small, siloed attention spaces that might contribute to its internal fragmentation and promote the emergence of (micro-)tribes.

However, his descriptions of the state of science and of MOS were made before the recent few decades’ intensified growth in importance of journal publishing, and the stiffening competition among individuals (and, to some extent, groups) for a reputational place in the sun in the shape of articles published in high-impact (“top”) journals (Butler & Spoelstra, 2020; Hallonsten, 2021; Macdonald, 2015; Tourish, 2019). Instead, a citation network analysis by Vogel (2012) indicates that MOS developed into a “polycentric oligarchy” between 1980 and 2009 and is still “characterized by a small number of dominant schools whose intellectual leaders exert strong control over research agendas and resources” (Whitley 1984b, p. 810).

(2) Simultaneously, a number of central outlets relevant for MOS emerged, which nowadays represent its paradigmatic core (see Kuhn, 1962 on the depiction of paradigms, paradigmatic cores and paradigmatic shifts). Due to the increasing publication pressure and the need of MOS scholars to increase their visibility to remain in academia, authors seek to publish in these outlets (Macdonald & Kam, 2007, 2010), and adopt “questionable research practices” (Butler et al., 2017), including inadequate, inaccurate, or authoritative citations. Under the disguise of a grandiose appeal to all authors to make “innovative theoretical contributions”, top MOS journals force researchers into focusing on “narrow issues, frequently of only marginal relevance to the plentiful problems afflicting the world, and, all too often, are inaccessible and unreadable”—the desire of journals to reach the crowned positions of “such elite US outlets as the Academy of Management Review”, makes them co-perpetrators in the “narrowing of academic inquiry”, and urge authors to engage in “alienating displays of metatheorizing, about issues of less and less importance” and in increasingly “higher states of specialization” that could be called “micro-theorizing” (Tourish 2011, pp. 369, 374). Therefore, the co-emergence of a paradigmatic core and central journals should play a significant role in the emergence of microtribes, whose members are then able to lift their likes into the elite US outlets.

A phenomenon directly linked to the need to publish efficiently is the forming of subcommunities. In this regard, Alvesson (2012) claims that the habit of researchers to remain faithful to a subcommunity throughout one’s career in order to avoid serious intellectual challenges is a serious problem not just in MOS but in social sciences generally. Furthermore, Alvesson and Gabriel (2013, p. 251) claim that authors frequently engage in strategic “location of a text in a sympathetic subcommunity” in order to neutralize critique and protecting both paper and author from any serious questioning. This, in turn, gives rise to the coordinated agglomeration of mutual citation among a few high-impact journals and papers (Davis, 2015, p. 182), or, as Whitley (2000, p. 28) noted, “citations are a way of ritualistically affirming group goals and norms, of demonstrating group membership and identity”.

(3) If we now consider the interdisciplinary nature of MOS, we could assume to witness an even larger degree of fragmentation compared other disciplines with a long tradition such as economics. Interdisciplinarity complicates the establishment of a coherent body of knowledge insofar, as the intellectual distance between potential collaborators increases the difficulty to find a common theoretical and methodological ground (Boix Mansilla et al., 2016; Cummings & Kiesler, 2007; Haeussler & Sauermann, 2020). Therefore, MOS had to (1) coordinate multiple knowledge domains at once, while (2) establish an own publication system, and (3) cushioning the negative effects related to interdisciplinary research such as reduced research productivity (Leahey et al., 2017), and potential losses in terms of citation numbers (Unger et al., 2022).

For the reasons discussed above, one would expect to find small pockets of knowledge within MOS subcommunities which roughly follow the boundaries of each discipline from which MOS stems. Yet interdisciplinary fields like MOS can possibly prevent fragmentation, if they manage to do the following. First, a narrow scope of problems and social facts (e.g. organizations as strategic actors) are defined as worthy of investigation. Second, the theories and assumptions previously associated with each discipline are discarded, are used punctual, or in a rather eclectic manner. Third, a common mode of exchange is established with a shared canon of methods. In other words, it is easier to argue about whether a regression coefficient is significant or not than to argue about theoretical vocabulary and epistemological assumptions. Fourth, the combination of the first three assumptions may help a school of thought with the ability to superimpose its paradigm on other schools of thought to emerge. Fifth, the emergence of its own journals and monograph series, which serve as beacons for MOS scholars to rally around. All factors taken together should put the fragmentation of a discipline and the formation of (micro-)tribes to a hold and should help to form a common, citable knowledge base. Following the statements outlined by Collins (2002, p. 24) on the dispersed intellectual communities, this should also be more easy in a multidisciplinary community, as the flux of intellectuals and their stance towards self-reflexivity turns them into critical observers of their own discipline and creates a mixture of organic solidarity (= dependence on the knowledge and expertise of scholars originating in different disciplines and their critical stance) and mechanical solidarity (= based on the belief to be part of MOS as a discipline with shared assumptions, beliefs, and problems to be addressed) among scholars. Table 1 summarizes the factors which drive or prohibit fragmentation of the knowledge base, or the emergence of micro-tribes. However, what can we expect to witness empirically when investigating MOS when we investigate the link between academic micro-tribes and co-citations diachronically?

Table 1 Factors driving and prohibiting the emergence of micro-tribes

Empirically observable patterns of tribalization

If tribalization is present in MOS and micro-tribes emerge, we expect to see a small number of communities in the first observation period t0. These communities consist of a limited number of co-cited articles and are relatively closed. Each community is interpreted as a shared knowledge-domain within MOS, e.g. authors who follow different organizational theories or different methods. Simultaneously, we expect to witness some connections between communities. These connections represent the possibility to link knowledge stemming from different domains in MOS. These domains may even yield different disciplinary origins (e.g. organizational sociology, behavioral economics, management science).

If tribalization exists, we would witness the appearance of subgroups within at least a few clusters at the next time frame t1. From there onwards, these subgroups increase their internal connectivity, and sever connections to other parts of the now larger cluster. This means that despite emergent stronger co-citations in local cores, there are still co-citations between the local cores, albeit to a lesser degree. In other words, we will witness the transition from a single, densely connected core and a periphery, which element’s bears connections to the center but not to others in the periphery, to multiple, disparate cores with peripheral regions each.

Furthermore, articles published in the meantime could either be (1) integrated into the developing subclusters, or (2) may serve as nuclei for novel clusters. In the first case, novel knowledge is integrated a preexisting cluster, which may displace older papers which previously served as glue between emerging subschools in MOS. In the second case, these may introduce either disruptive knowledge (Wu et al., 2019) such as behavioral economics, or indicate a turn, meaning that the same set of problems might be addressed by applying different theories, methods, but ultimately arriving at the same conclusions while using a different vocabulary (Schneider & Osrecki, 2020). If this happens, we expect to see a closure of knowledge circulation in form of citation circles, which in turn might speed up the tribalization process. Nevertheless, there might be seminal works, which might be routinely cited to demonstrate the belonging the MOS (e.g. the paper issued by Meyer & Rowan, 1977). These may serve as a backbone for the discipline and exert the ability to connect different domains of knowledge, but, in fact, do not hinder MOS to progressively fragment into micro-tribes.

At the same time, we expect the following network patterns to emerge from the data which may counteract tribalization. First, articles are added to the network that span bridges between two or more preexisting clusters. Initially, these cite only a few sources from each cluster. However, in a next step, these bridges might be the seedbed for other articles written by different authors to combine even more sources stemming from different clusters and the bridging paper itself. In this case, we should witness the closure of structural holes (Burt, 2017), and the formation of superclusters with abundant co-citations stemming from formerly unconnected knowledge domains. In other words, we expect to see densely connected communities, whereas the local density grows stronger (1) the more paradigmatically unified MOS gets, and (2) the stronger the flow of knowledge gets.

Second, at t+1, papers citing these central papers (= neighbors of the papers mentioned above) would also get a share of citations, meaning that those who read the citing papers but not the core paper tend to cite both at later points in time. Furthermore, we would expect to witness the closure of co-citations between forerunners of a tribe and an inner circle of their followers. This means that the papers issued by the disciples of the forerunners might get so tightly linked to the latter’s contribution at t>1 that they are cited together systematically. In this case, we expect a center–periphery structure to emerge in MOS with one, densely connected core which comprises most of the co-cited publications, and a periphery with little or no connectivity to the core. Yet, if MOS lacks epistemological integrity or if internal conflicts arise, novel subclusters and thus micro-tribes with own territories (= subdomains of knowledge) might emerge. These subclusters, in turn, would still cite the thought leaders as well as their disciples, but would start to criticize them and to gain a cognitive distance at the same time. Over time, this would manifest in citation circles in which ever smaller amounts of literature adjacent to the micro-tribe would be cited, and multiple, smaller local cores and many peripheries would emerge.

If, however, a unification under (1) a set of research questions, (2) a single theory, or (3) a limited set of methods occur, we expect to see the incorporation of smaller clusters into larger ones besides the emergence of a single core–periphery structure or multiple within one partition if MOS is a “polycentric oligopoly” (Vogel, 2012; Whitley, 1984). This process might lead to an overarching network-structure with a single, relatively dense core consisting of a limited number of thought leaders, and a broad (semi-)periphery with more specialized and loosely connected papers and authors. In this case, there might be a number of seminal works, or works citing those, which must then cited by scholars to highlight their belonging to the discipline, in our case MOS. Even if the circle of thought leaders might change over time, an unipolar structure with a strong, integrated core might be stable enough to prohibit micro-tribes to emerge. In other words, the cognitive core would remain intact and the knowledge flow embodied in the citation network would be still in need of the forerunners or their descendants.

Against this backdrop, we would expect four ideal typical trajectories emerge from the data (Fig. 1), two of which could be linked to tribalism. In one case, we expect multiple, sharply distinguishable local cores with little or no connection to emerge (strong Tribalism, upper left panel of Fig. 1). In another case, we expect differentiation, but with limited exchange over time which could be mainly driven by the co-citation of seminal scholars of each tribe and a certain extent of exchange (weak tribalism). This would be the case, if a strong version of the polycentric oligarchy mentioned above with strong boundaries between each center is present. In other cases, we expect a limited amount of branching of (thematically driven) communities. However, in this case they do not systematically exclude publications from scholars of other potential “tribes” from being co-cited. Rather, scholars specialize but within their areas of specialization read and cite other specialists (lower left panel, Fig. 1). This pattern would correspond most clearly to the polycentric oligarchy described by Whitley (1984) and Vogel (2012). Here, different cores emerge and yield the possibility to (a) exclude inconvenient approaches, and (b) to integrate knowledge stemming from other acknowledged centers of MOS. If, however, no tribalism and focus on a limited number of theories, problems, etc. is present in the data, we expect to see a branching from different MOS communities due to specialization at first, followed by a reintegration into a co-citation core (lower right panel, Fig. 1). This would correspond to a single, dominating school of thought in MOS, which is paired with internal, problem-based specialization. This paradigmatic core should consist of a small number of different approaches or specializations in MOS similar to the law of small numbers in the attention space described by Collins (2002, p. 24). For this reason, the core should be able to glue the approaches and potential approaches and specializations intellectually together.

Fig. 1
figure 1

Idealized trajectories of tribalization and (re-)centralization of the discourses in MOS

Research strategy

In the following, we begin with describing our data and the search strategy. We then continue with describing our method, namely the diachronic co-citation analysis. We then introduce the three analytical approaches we combine to identify the emergence or absence of micro-tribes in the diachronic co-citation analysis.

Data

After removing duplicate articles, our dataset comprises of 22,430 papers issued in the 14 MOS core journals listed in Web of Science (WOS) between 1980 and 2019. These include the Academy of Management Review, Journal of Management, Administrative Science Quarterly, Leadership Quarterly, Academy of Management Journal, Strategic Management Journal, Journal of Organizational Behavior, Journal of Management Studies, Organization Studies, Human Relations, Organization, European Management Journal, Organization Science, and Management Science. An overview of the number of publications per timeframe is provided in “Appendix 1”. We used the WOS interface and downloaded all full articles and review articles.

Diachronic co-citation analysis

We apply diachronic co-citation analysis to investigate the development of micro-tribes. Co-citation analysis is a subdomain of social network analysis (Zupic & Čater, 2015), which has previously been used in studies of the knowledge base of MOS, and how it has evolved over time (Acedo & Casillas, 2005; Ozturk, 2021; Vogel, 2012). If an article cites two or more sources (= nodes), this establishes a link (= tie) between those cited sources. These ties, and the absence of ties, together form the overall structure of the co-citation network. Since co-citations are symmetrical two-way connections among cited articles, co-citation networks are undirected, and their edges are also weighted by the number of co-appearances of papers in the reference lists. As we rely on the reference lists to establish co-citations, we might overestimate the paradigmatic coherence of MOS. However, even if tribes exist, the fact that proponents of those cite other tribes indicates that (1) tribes are still in contact and do not ignore each other, and (2) they still know to a certain extent what their competitors in MOS do and, by doing so, still uphold the potential for knowledge evolution.

If co-citations appear systematically in a period of observation, that is 3 times or more, a tie between two papers is established and included in the analysis.Footnote 2 Systematic co-citation of articles hint at the emergence of a specific knowledge domain, and if this domain establishes firm boundaries, this may indicate the emergence of an academic (micro-)tribe.

A diachronic co-citation analysis is a form of longitudinal analysis, in which discrete time windows are defined and the evolution of the co-citation network is depicted between these windows. We decided on 5-year timeframes (1980–1984, 1985–1989 and so forth), as shorter timeframes yield the danger to witness random fluctuations in co-citation patterns, whereas longer timeframes would gloss over the potential development of micro-tribes. Furthermore, articles in the social sciences need approximately 4–5 years to get integrated into the scholarly discourse (Gou et al., 2022). Afterwards, their citation number and thus their potential for being a nucleus for one or more academic tribes starts to decline as their academic age increases (Zhang & Glänzel, 2017). Therefore, focusing on a 5-year timeframe to establish co-citations enables us to witness the formation of (micro-)tribes in MOS, especially if many timeframes are analyzed in succession.

We use articles published in the 14 journals under investigation to establish the diachronic co-citation network, as an assumption regarding the emergence and stability of these micro-tribes relies on the closure of their knowledge base and the necessity to signal the belonging of MOS scholars to these tribes. As these 14 journals represent the core of MOS and define the topics, methods, and theories legit in these fields, the emergence of micro-tribes should be most clearly visible within the lines of research acknowledged in these journals. To provide an overview on the data analyzed, Table 2 lists the number of papers issued per timeframe, the total number of cited items issued in the current and previous observation period, the number of total co-citations, as well as the number of co-citations after removing articles which cited fewer than three times.

Table 2 Overview on the papers and co-citations per time frame included in the analysis

Analytical approach I: analysis of local and global network measures over time

To analyze the changes in the co-citation structure over time and thus the potential emergence of micro-tribes in MOS, we focus first on local and global network measures. The former enable us to identify specificities in regards to the co-cited documents and their neighbors, whereas the latter enables us to identify structures of the overall network, e.g. the degree of fragmentation into different micro-tribes. As a network is a graph object and changes in the local structure impact the global structure and vice versa (Kossinets & Watts, 2006), we must analyze both simultaneously to answer our research question.Footnote 3

Local structures comprise the (more or less) direct neighborhood of a given node (= cited publication) or a set of nodes and indicate the embeddedness of actors within a given network. To capture different aspects of the local structure of our co-citation network, we employ degree centrality, eigenvector centrality, closeness centrality, and betweenness centrality as described by Freeman (1978), and network constraint as described by Burt (2017).Footnote 4

Degree centrality measures the number of ties assigned to each node. In our case, we use the weighed degree centrality measure, which measures the strength of co-citations among a paper and its neighbors (e.g. paper a and b are co-cited ten times together, so this link will have an initial weight of 10) divided by the sum of the edge weights for edges incident to the node under observation. According to Freeman (1978, p. 219f.), degree centrality is associated with the visibility of a node within a network, and thus its potential to be a focus of activity. In our case, MOS papers with high degree centrality yield the potential to become a seed for tribes or are relevant sources of knowledge within different tribes.

Eigenvector centrality measures the relative influence of a node in a network. It transcends degree centrality insofar, as it takes the connectivity (e.g. weighted number of edges) of the neighbors of a given node into account. For example, a night might yield connections to a few other nodes, which, in turn, might be very well connected to multiple other nodes. An example is the collaboration network between eminent scholars described by Abramo et al. (2019), who tend to collaborate with other eminent scholars. However, if a less eminent scholar collaborates with at least one of the more eminent, he or she gains in access to more network resources and thus is potentially able to exert more impact on the network. Translated to our case, papers with high eigenvector centrality are either seminal papers, or adjacent papers that may become the seedbed of novel specialties or knowledge domains in MOS. As stated in "Factors driving or prohibiting the emergence of micro-tribes in Management and Organizational Science" section, these domains might be linked to the emergence of micro tribes. However, if a large number of co-cited papers yield high levels of eigenvector centrality, this might indicate an overall better embeddedness in the MOS discourse and higher levels of co-citedness of the articles.

Betweenness centrality measures the weighted share of shortest paths between two nodes, which passes through a node under observation (ibid. pp. 222–224). In other words, betweenness centrality measures the probability of a node to control the flow of information in a given network. In our case, high betweenness values indicate the possibility of a paper to be co-cited across different tribes or domains. This, in turn, means that a paper might either serve as bridge, or as a seedbed for the emergence of a novel tribe over time.

Next, closeness centrality follows the idea of independence of nodes from intermediaries in a network (Freeman, p. 224f.). It also indicates the possibility of a node to easily spread information to even distant outskirts (= periphery/peripheries) of the network. In our case, the notion of independence may be interpreted literally, as nodes who are close to every region of the co-citation network are not bound to a single paradigm. Therefore, if closeness centrality increases over time, this might indicate dissolving boundaries between tribes in MOS (if they exist in the first place) as knowledge incorporated in the co-cited papers is increasingly easy to co-cite and circulate.

Finally, constraint is a measure of brokerage, or the lack thereof. It is calculated as the share of nonredundant ties of a node and its neighbors (Burt, 2017). As such, it is assumed that a node with low constraint values may be able to span multiple knowledge domains. In our case, high constraint values indicate the local closure of co-citations and are thus an indicator of the emergence of (micro-)tribes.

In turn, the term global structure means that also measures of the overall properties of a network are included. The properties are number of nodes (= cited articles), the share of realized connections on the total number of possible connections, the appearance of fragmentation into isolated subgroups (Newman, 2006), and also relations among clusters, meaning systematically co-cited articles, including their closure, size, and stability over time.

There are several measures of global structure. One is transitivity (Newman, 2006), which means the share of triplets on all possible triplets. A triplet consists of three nodes (triad), which are connected among each other (paper A is co-cited with paper B, B with paper C, and A with C). Transitivity can be seen as a global counterpart to network constraint, as it measures how densely connected the co-citation network is overall. If, for example, regions of the MOS–co-citation network are fully connected and little to no connections are present to other regions, then we would witness high levels of transitivity. This, in turn, would indicate either a tribal structure or a (unipolar or multipolar) center–periphery structure. However, transitivity alone is not sufficient to distinguish between these two cases, which is why we also use the number of detected components as the second measure, and the number of nodes of the largest component as third measure. A detected component is an isolated area of the network, which in our case means a group of (densely or loosely) co-cited papers that are otherwise unconnected with other cited papers in the sample. At the same time, the more nodes the largest component comprises, the more cohesive the network is. Ideal typically, we would expect a multitude of (roughly) equal sized components to emerge if MOS is truly organized in (micro-)tribes, as these tribes would only co-cite papers belonging to the same tribe and ignore papers stemming from different tribes. But if we witness the emergence of a stable, large component which comprises the lion’s share of co-cited papers, then the probability to witness tribalization decreases.

Nonetheless we must check for the presence of densely connected regions or co-cited papers in our network, to make sure we do not under- or overestimate the degree of tribalization or centralization. This is why, at last, modularity measures that a graph is fragmented into different densely connected subgroups. The clearer cut the boundaries between these groups, the higher the modularity.Footnote 5 This is the case, if papers are co-cited systematically, thus establish a community, while links to other communities of co-cited papers are sparse. According to Newman (2006, p. 8578), “modularity is, up to a multiplicative constant, the number of edges falling within groups minus the expected number in an equivalent network with edges placed at random”. In our case, higher values indicate more clearly delineated structures within the MOS co-citation network and thus the potential for tribes to emerge.

The last piece of the puzzle to control whether the MOS co-citation network yields a tribelike structure is to employ center–periphery detection algorithms. As outlined above, if tribes are present, we will find multiple, densely connected cores, a semi-periphery which links a number of cores, and multiple peripheries. This is more likely, if the components are of equal size.

To test for center–periphery (CP) structures, we employ the algorithm developed by Kojaku and Masuda (2018a, 2018b). It is designed to detect a multitude of core–periphery structures. We also controlled if a single CP-structure is present in our data using the algorithm developed by Everett and Borgatti (1999). However, the calculations did not converge, meaning that there is no clean, single CP-structure present in our data for each timeframe. We then employed q–s test for multiple cores developed by Kojaku and Masuda (2018a, 2018b) to check for spurious core–periphery structures. The q-s test assumes that (especially smaller) cores may be artifacts and only weakly distinguished from other parts of the network.

Analytical approach II: community detection and flow of publications between communities at t n and t n+1

The analysis of the local and global structure of the network may hint at the emergence of micro-tribes in MOS. However, these hints alone are not sufficient, as they do not give us details about the potential splitting of knowledge stocks and their embedding into new subgroups, or—in terms of network analysis—communities. A community is defined as a region of the network, in which nodes are densely connected, whereas connections to different parts are low (Girvan & Newman, 2002). In our case, a community is detected if articles are systematically co-cited.Footnote 6

The intention is, that emergent micro-tribes entail a focus on certain papers, which become essential over time for MOS scholars to back their line of reasoning within the micro-tribes. At the same time, scholars belonging to a (micro-)tribe, will progressively refrain from citing sources related to other (micro-)tribes over time. Instead, they will cite focal papers of the forerunners of the tribe and other members, which may be closely linked to the former. This should increase transitivity and decrease betweenness- and closeness values. If this applies, we may witness a schism between (micro-)tribes. In other words, if these sources are assigned to a cluster C1 or CP-structure CP1 at time tn, we will witness a split of this cluster into two or more distinct clusters or CP-structures at tn+1. If we check the intersection between co-cited papers of C1 or CP-structure CP1 at tn and clusters C2 to Cm or CP-structures CP2 to CPn at tn+1, we are able to calculate their flow over time. This pattern indicates the formation of micro-tribes if it reoccurs in the forthcoming timeframes.

As the number of nodes per cluster varies greatly within and between timeframes, we calculate two flow measures: total flow and relative flow. While the former measures the absolute number of nodes which appear at a cluster Cm or CP-structure CPm at tn and Ck or CPk at tn+1, the latter calculates the share of nodes assigned to Cm or CPm at tn which is also assigned to Ck or CPk at tn+1. Total flow is therefore able to depict flows between clusters with large number of nodes, while relative flow is more sensitive to the flow of cited papers assigned to smaller clusters.

Analytical approach III: removal of central nodes

A final building block concerns the argument that the whole diachronic co-citation network is only held together by a few seminal papers which must be cited ritually to demonstrate the affiliation with the MOS research community. In this case, these few seminal papers yield the ability to connect different parts of the network. Consequently, they would gloss over the emergence and establishment of micro-tribes. Therefore, we follow previous studies and employ a node removal strategy to demonstrate the fragility of networks (Michele Bellingeri et al., 2020a, 2020b) and the potential of central nodes to fragment into different communities (Yang et al., 2018).

According to Boldi et al. (2013), and Bellingeri et al., (2020a, 2020b), removing links or nodes with the highest betweenness centrality is most effective in fragmenting the network, if there are few central nodes which yield the ability to bridge different areas of the network. In our case, these nodes equate with seminal papers which must be cited to demonstrate the affiliation with the MOS community. We follow this approach and remove the 5, 10, and 25 most central cited papers regarding their betweenness centrality from the diachronic co-citation analysis. This approach should suffice to increase the number of components, reduce the number of nodes associated with the largest component, and increase the modularity considerably—even if the co-citation network is larger, namely consists of many cited papers and abundant combinations of co-citations, i.e. at later timeframes. In combination with the change in the local and global structure over time, as well as the flow of cited papers between detected communities over time, this strategy should help us to discover potential micro-tribes in MOS.

Results

Local and global measures

Beginning with the local network measures, we witness a decrease in mean and median values of normalized degree centrality, betweenness centrality and eigenvector centrality over time.Footnote 7 To a certain extent, this is unsurprising, as these values are associated with the number of nodes (= cited references) and edges (= co-citations) in the network under scrutiny. Regarding the local structure, these measures indicate (a) that MOS papers are co-cited in increasingly limited circumstances, and (b) that their ability to get embedded prominently in the MOS discourse diminishes over time. At the same time, we witness an increase in mean and median closeness centrality and a decrease of the constraint value between the time-frames. The change over time in the last two measures suggests that the observed co-citation network is becoming more cohesive. On first glance, the findings contradict each other. Nonetheless, these findings might hint at increasing specialization and a broadening knowledge base at the same time; even if MOS gets more local, yet more cohesive at the same time. The change in these measures, excluding outliers, is depicted in the boxplots in Fig. 2, and listed comprehensively in “Appendix 2”. Further sensitivity analyses with different edge thresholds are provided in “Appendix 6”.Footnote 8

Fig. 2
figure 2

Boxplots of local network measures by timeframe

Turning to the global measures depicted in Fig. 3, we witness an increase in the number of nodes and edges over time, which is especially prevalent from the 2005 onwards (upper echelon). At the same time, the number of components rise between 1980 and 1999, reaches a plateau between 2000 and 2009, declines slightly in 2010–2014, before increasing again in the timeframe between 2015 and 2019. Simultaneously, the largest component comprises nearly all nodes present in the network at all points in time. At the same time, modularity declines over the whole period, with the exception of a small increase between 1995 and 1999, which means that during this short period, the initially very closed subgroups in the co-citation network began opening up and enabled idea exchange. We see this also in the transitivity measure, which shows that in the period 1980–1984, almost 40% of all triads were completely connected (e.g. citations A and B, B and C, and A and C are connected). Thereafter, transitivity continuously declined until it plateaued in 1995–1999, with 31.53% triads completely connected, and then declined again, this time steeply to 0.2032 in 2010–2014, before slightly rising again.Footnote 9

Fig. 3
figure 3

Global network measures over time

Next, Fig. 4 depicts the point measures regarding the size of the detected clusters. The orange line depicts the number of nodes assigned to the largest community. It tells us that the largest cluster constantly grows in size and is, with the exception of 2000–2004, at least double the size of the 90% percentile. Furthermore, we witness the steepest rise in the number of cited papers assigned to the largest cluster in the between 2005 and 2014. The 90th percentile grows nearly as strong from 1980–1984 to 2000–2005, but loses ground against the largest component from this timeframe onwards. Especially between 2015 and 2019, less nodes are assigned to the 90th percentile cluster compared to the timeframe 2010–2014. The same, albeit to a much less extend, is seen for the median and mean cluster sizes.

Fig. 4
figure 4

Descriptive statistics regarding the number of nodes assigned to the largest cluster over time

If there were tribalization, we would expect a nearly exponential increase in number of detected components over time, a stagnant number of nodes assigned to the largest component, constantly high levels of modularity, and stable levels of triadic closure. Yet none of these is detected. Furthermore, these findings on the global measure complement the local measures insofar, as the increasing levels of brokerage and closeness indicate the increasing ability of knowledge to flow between different areas of the network and are mirrored in the global structure. Finally, the distribution of nodes regarding the detected communities in the co-citation network does not support the assumption regarding the tribalization of MOS.

Finally, let us turn our attention to the CP-structures extracted by the KM-algorithm. The main findings are depicted in Fig. 5, and a full account is listed in “Appendix 2”, Table 4.Footnote 10 In the upper echelon, we ee the total number of nodes present in our data (upper left) and the share of nodes assigned to core–periphery structures deemed significant by our qs-statistics. This means that, after testing, we can be sure at p < 0.05, that we detected a core–periphery structure within our data. Regarding the latter, we see that in 1980–1984, only a tiny fracture of co-cited papers (≈11.4%) are assigned to CP-structures. Albeit the share rises considerable over the next timeframes, it takes until 2010–2014 until we may speak of a co-citation network in MOS, which is generally ordered into CP-structures (94% of all nodes in 2010–2014 are assigned to significant CP-structures, and 97.3% in 2015–2019).

Fig. 5
figure 5

Descriptive statistics of the core–periphery structure over time

Turning to the left panel in the middle echelon in Fig. 5, we see that approximately half of the nodes present in each timeframe are assigned to a center. The only exception is 1980–1984, where 4 out of 5 nodes assigned to the CP-structure are denoted as central nodes (see “Appendix 2”, Table 5). The number of significant CP-structures (right, middle echelon) detected rises over time; moderately at first (nine in 1985–1989 up to 14 in 1995–1999), then jumping to 24 in 2000–2004. In the following two timeframes, the number stagnates before rising sharply to 68 in 2015–2019. So far, the findings of the CP analysis indicate either an ongoing process of internal differentiation of MOS, or a tribalization over time, which increases its pace over the last two decades and, maybe, lead to the emergence of micro tribes in 2015–2019.

We must now focus on the two panels at the bottom echelon of Fig. 5 to see, which one of the two interpretations is more likely. On the left, we see that in the first timeframes, namely from 1980–1984 to 1995–1999, the CP-structures were only detected in a single network component. Between 2000–2004 and 2010–2014, the number increased to two. This indicates a possible bifurcation of MOS into a dominant core and a subaltern center, which coincides with the establishment of the journals Leadership Quarterly, Organization, and European Journal of Management. A fragmentation into components with a size large enough to house CP-structures on their own occurs only recently in 2015–2019. In this timeframe, 12 components house significant CP-structures.

This impression is covered by the sizes of the CP-structures as depicted in the bottom right panel in Fig. 5. Eyeballing the graph and reflecting the values in Table 5 in “Appendix 2”, this might hint at two simultaneous developments. First, tribalization might occur, which is pronounced in the last timeframe. Secondly, the decreasing size of CP-structures and comparatively small CP-structures extracted from the data might also hint at the emergence of small, maybe short-lived co-citation pockets within the MOS discourse, or the crowding-out of marginalized scholars and their papers. Such closed-off pockets of knowledge yield little changes to develop paradigms on their own and might incorporate either disciplinary pariahs, or scholars who only occasionally publish in MOS journals and get cited by MOS scholars.

Flow of cited papers between clusters over time

Nevertheless, the objection could be raised that we are not mapping the process of tribalization and that the clusters detected at different timeframes are mutually independent of each other. For this reason, let us now turn to the flow of nodes between the detected communities over time depicted in Figs. 6 and 7, and between the CP-structures in Figs. 8 and 9. Tables of the full account of flow between the clusters and CP-structures detected per timeframe is provided in “Appendix 5”. In these figures, communities are arranged per timeframe on the y-axis from the smallest to the largest clusters and are divided by timeframe under observation on the x-axis. In case of absolute flow, we colored the flow from red (low absolute numbers of citation transfer) to blue (high absolute numbers of citation transfer between communities and timeframes). Regarding relative flow, we colored the clusters and the flow between them from green (smallest clusters and CP-structures) over red (medium sized communities and CP-structures) to blue (largest communities and CP-structures) and used the same color to denote the percentage of nodes transferred between these communities from tn to tn+1.

Fig. 6
figure 6

Absolute flow of nodes between the detected network communities over time

Fig. 7
figure 7

Relative flow of nodes between the detected network communities over time

Fig. 8
figure 8

Absolute flow of nodes between the detected core–periphery structures over time

Fig. 9
figure 9

Relative flow of nodes between the detected core–periphery structures over time

Regarding the absolute flow in Figs. 6 and 8, we witness a certain degree of exchange between clusters and CP-structures, which intensifies over time, but is restricted to seven or eight detected clusters and CP-structures per timeframe. Over time, the flow between the largest detected communities at the bottom of Fig. 4 even intensifies. Meanwhile, the flow from larger to smaller communities is almost negligible, and the flow between them even smaller. To this shall be added that if small and novel communities emerge at a time tn, we can see that they usually are incorporated into larger communities at tn+1. Furthermore, there are few or no connections between these communities and communities of articles in earlier timeframes.

In Fig. 6, we see three patterns of relative flow. The first is that nodes that belong to smaller clusters (top of both figures) at tn usually get incorporated in larger clusters and CP-structures at tn+1. Second, our graph indicates that a small share of the nodes present in large clusters and CP-structures at tn flow towards other larger clusters at tn+1. Meanwhile, there is not much flow from large clusters to small and peripheral clusters. Third, we witness a higher fluctuation in the node assignment between CP-structures over time compared to the clusters, which are either caused by the overall larger number of CP-structures present in our data, or from the recombination of knowledge within MOS and research trends therein.

If MOS were dominated by tribalism, we would see a larger flow of cited papers from the large clusters and CP-structures to small clusters and CP-structures among the timeframes, while the flow between the large clusters and CP-structures would become increasingly smaller and eventually fade away. Instead, our findings suggest a polycentric center–periphery structure present in the MOS co-citation network. The center consists of a few large clusters and CP-structures consisting of many co-cited articles published in the ten major MOS journals. Articles belonging the periphery tend to be absorbed into the denser center. Here, a kind of annexation and possible recombination of knowledge takes place, while, at the same time, new co-citation clusters and CP-structures are added. The opposite movement can hardly be observed. This may indicate either that many articles and the knowledge stocks bound to them are not (or no longer) cited, or that certain knowledge stocks are systematically pushed out of the discourse.

Node removal

The next component in testing whether MOS is characterized by a tribalistic structure is in removing the 5, 10, and 25 most centrally cited papers according to betweenness centrality. If MOS is truly tribalistic, we would expect the number of unique components, modularity as well as transitivity to increase, but the number of nodes that make up the largest component in the co-citation network to decrease very sharply, especially if the fragmentation into multiple CP-structures is considered. Figure 10 lists the four above values for the networks in the respective time periods. At first glance, there is mixed evidence in favor and against tribalization.

Fig. 10
figure 10

Changes in the global measures after applying the node-removal strategy

Evidence in favor of tribalization is indicated by the increase in the number of components (top left panel of Fig. 10). This is even true if the number of reduced nodes is small, and becomes more pronounced over time, especially in 2005–2009 and on. This result is in itself a testimony that the central nodes indeed are seminal papers that are cited as a rite of passage into the MOS community. Another indication of tribalization is the increase in modularity (bottom left of the graph). But when removing the 25 most central nodes in the first period observed (1980–1984), the increase in modularity is not seen, since this already meant a removal of a large portion of the nodes that are present in the network. An increase in modularity means that we find increasingly closed network structures when removing the most central nodes.

Evidence against tribalization is conveyed by the values for the number of nodes of the largest component. They indicate the presence of a large, contiguous co-citation area despite the removal of the most central nodes within the network (in the upper right graph). Again, the largest component dominates the co-citation network entirely and includes almost all nodes in each timeframe. Looking at components and modularity, the surge in components is likely driven by small co-citation pockets which participate in the ritual citation of central articles in order to connect to the mainstream of the field. This is also indicated by the transitivity values (bottom right in the graph), which would increase to a larger extent if there were tribal, self-contained structures. This is true just because the central nodes with the highest betweenness centrality are removed, i.e. those that have a large number of connections to all areas of the co-citation network. However, this also that these central nodes are not the nodes with the highest brokerage.

If the findings are combined, the conjecture is strengthened that structure consisting of multiple centers and peripheries is present in the MOS co-citation network, which, with the exception of 2015–2019, clearly concentrates on the largest component. We can once again see this structure in the removed most central nodes listed in “Appendix 4”. Among them are articles published in the Academy of Management Review, Administrative Science Quarterly, and the Academy of Management Journal—the journals in the field of MOS that have the highest impact and therefore can be said to represent the (Anglo-American) center. This means that the results of the edge-removal strategy are complementary to the results of the flow of cited articles between clusters and CP-structures over time. Consequently, the largest component persists in spite of removal of the most central nodes, while transitivity does not increase significantly. Instead, components that are seemingly small, self-contained, and therefore more edge-removed disappear from the largest component when the most central nodes are removed. This means, in the context of our study, that they represent what we can call the semi-periphery of MOS.

Discussion and conclusions

The paper at hand sought to answer the questions of how many tribes make up Management and Organization Studies (RQ1), and how fragmented is its knowledge base is (RQ2). To do so, a diachronic co-citation analysis was conducted. Besides the canonical local and global network measures as well as CP-structures, a measure of flow among of the cited papers between the communities over time was established. Additionally, the most central nodes according to betweenness centrality were removed to investigate whether or not the coherence of MOS is due a limited number of potentially authoritatively cited papers.

If we interrelate all the evidence, a dichotomy between a more interconnected center and a periphery of bodies of knowledge emerges, which becomes more and more pronounced over time. This picture runs counter to the assumption that MOS is breaking up into ever smaller micro-tribes (Alvesson 2012; Alvesson & Sandberg, 2014; Barley 2016; Huzzard et al. 2017; Tourish, 2019). Instead of a weak paradigmatic core, which is seen as one of the main drivers of fragmentation and thus tribalization (Abbott, 2001; Schwemmer & Wieczorek, 2020; Turner, 2006, 2016; Wieczorek et al., 2021a, b), our findings indicate an oligopolist, paradigmatic core which captures nearly all of the co-cited papers in our dataset and might hint at a crowding-out of subalterns or pockets of knowledge in 2015–2019. In this sense, MOS is a convergent discipline in terms of the tribalism approach (Becher & Trowler, 2001), albeit our findings suggest a mixture of the weak tribalism, and core with (Re-)integration scenario depicted in Fig. 1 in "Empirically observable patterns of tribalization" section. If there were no strong paradigmatic core, then, at the very least, our findings indicate a center of power to which peripheral authors orient themselves. This center is structured around a limited number if CP-structures, which are aligned to the limited attention space described by Collins (2002, p. 24), but with a tendency to fractalize into smaller, probably more specialized branches of MOS in line with Abbot’s (2001) argumentation. These interpretations are backed by the absolute and relative flow of citations among detected communities over time. They signal the strong interrelation between a small number of detected communities within the very center of MOS, the negligible outflow of cited papers, and the ability to incorporate peripheral communities of co-cited papers by the detected central co-citation communities.

As indicated by the number of detected communities, CP-structures, and modularity values, we suspect specialization and crowding-out effects in MOS to be present in our data, but it is overshadowed by the integration of knowledge into the largest component in the co-citation network. Also, the increasing number of components detected if the most central nodes are removed might be a hint at this underlying differentiation, but again, potential differentiation is also adherent to the alignment to dominant sources and the need to follow narrow, connective paths of research, as criticized by Alvesson and Sandberg (2014). Furthermore, the oligopolistic center–periphery structure present in the data and the fact that a large number of the most central papers are situated in the reputational place of the sun (Butler & Spoelstra, 2020; Hallonsten, 2021; Macdonald, 2015), namely in the Academy of Management Review, Administrative Science Quarterly, and Academy of Management Journal adds to the impression that MOS is not as prone to form tribes as expected (Alvesson, 2013). Rather, a subaltern (possibly European) core emerges, which is accompanied by small CP-structures in 2014–2019.

These findings also resonate with Alvesson and Gabriel (2013) and Davis (2015) insofar, as the structure of the co-citation network reveals the strategic location of thought products in a sympathetic community, and the practice of ritualized citations in high-impact MOS journals. In turn, the practice of ritualized citation contributes to the detected, stark center–periphery dynamic. In other words: Researchers may become part of the densely connected center embodied in the largest component, otherwise, researchers will have to look for a (possible more fragmented) discipline with a higher tolerance for deviating opinions and research styles.

Taken together, our approach reveals that MOS does not consist of micro-tribes, but an oligopolistic center–periphery structure (RQ1), and yields a low degree of fragmentation of its knowledge base with an only recent increase (RQ2). Due to the limited number of CP-structures in line with Collins’ (2002) line of argumentation, this oligopolistic center of MOS yields the ability to incorporate the influx of novel, yet not fully integrated ideas. In this sense, not only the tribalism approach (Becher & Trowler, 2001) provides us with important mechanisms to study and map the fragmentation of academic disciplines. Also, the multiple CP-algorithm as well as the qs-algorithm developed by Kojaku and Masuda (2018a, 2018b), our employed strategy to remove central nodes (Bellingeri et al., 2020a, 2020b Boldi et al., 2013; Yang et al., 2018), and the introduced approach to investigate the absolute and relative flow of nodes among different detected communities over time proved to be useful tools which for studying the fragmentation and internal differentiation of research communities or academic disciplines.

As usual, our approach yields a number of limitations. The first limitation is linked to the usage of Web of Science as data base. Despite its relevance, Web of Science is biased against publications in the social sciences and the humanities, thus the coverage is not as good as in other databases, e.g. SCOPUS (Aksnes & Sivertsen, 2019). Secondly, network structures are especially sensitive. This is inherent to their relational structure inscribed into the nodes, the connections between those, and the topology of the network itself. If we change one of these, it will immediately have an impact on the others, especially when measures are employed who rely on the topology of the network and the edges between the nodes, in our case the co-citations between published MOS papers (see “Appendix 6”). These two limitations might lead to an overestimation of tribalism, as missing data would translate in both missing nodes and missing edges in our co-citation network (Smith et al., 2017). In this sense, our approach would even overestimate the presence of academic tribes in MOS. Thirdly, some journals were founded in the 1990s. To a certain extent, this explains the growth of the network and the emergence of a second component with CP-structures from the 2000s onwards. Nonetheless, one would expect to find accelerated fragmentation and thus tribalization of the co-citation network from this time onwards, which only occurred in 2015–2019 and could also be accountable to a crowding out effect of peripheral MOS papers. Yet micro-tribes did not only emerge from our data, leading us to the conclusion that MOS is not a fragmented adhocracy anymore (Whitley, 1984), but rather the “polycentric oligarchy” described by Vogel (2012).

Besides the limitations stemming from the choice of the bibliometric database, there are three further limitations linked to the methodology employed in this article. First, we could not adequately capture the displacement of bodies of knowledge despite the hints provided by the CP-analysis. However, the way papers are citated in MOS leaves open the possibility that bodies of knowledge from other disciplines and fields are briefly absorbed, only to become bogged down or transferred back to other disciplines. However, it is also possible that knowledge assets that are published in MOS journals, but are not central, are exported to other disciplines through a crowding-out effect.

Second, we only considered one network layer. However, tribalization could also be reflected in cooperative relationships. Given that these are also unequally distributed according to the prestige of the researchers (collaborations between star scientists) (Abramo et al., 2019; Choi, 2012), the fact that male researchers are generally better networked nationally as well as internationally (Kwiek, 2020), and the U.S. presents itself as a center of management research (Wieczorek et al., 2021a, b), it might possibly be possible to discover more (micro-)tribes. However, their appearance in multiple network-layers would rather reflect (a) the monetary opportunity for international cooperation, or (b) hierarchies within MOS that contribute to the emergence of subcommunities. Nevertheless, it is to be expected that these subcommunities relate to a shared knowledge base, which is why future analyses discussing the MOS from the perspective of cooperation networks should always include co-citation analyses.

Third, we could not include the content of the articles from which we established the co-citation analyses in the present study. This could be realized in the future based on topic modeling approaches such as Structural topic modeling (Roberts et al., 2014) or Word-embedding algorithms (Kozlowski et al., 2019). These could then be used to contrast the present findings, or more precisely: to find out in which topic references the articles were and partly still are cited and which topic mix the clusters we extracted show.

Related to this, we were fourthly unable to discern the type of (co-)citations of the articles. Admittedly, some of the citations could be the result of adverse citations, e.g. which are used by scholars for reasons of distinction against different paradigms or mock competitors in MOS. Unfortunately, we could not take this into account, as we based our co-citation analysis on the articles’ reference lists. For this reason, we might still underestimate the true level of fragmentation between paradigms in MOS. Nonetheless, even if papers assigned to different tribes are cited as strawmen, this could also mean that the different tribes are still in touch and discuss the theoretical approaches, methodological stances, or empirical findings. To account for this shortcoming, future studies should include textual content related to the citation to discern whether is an adverse citation or not. However, this could be achieved if scholars have full-text access to all of the 22,430 articles analyzed in this paper.

Despite these limitations, our study largely debunks the myth, repeated in several recent high-profile publications, that MOS is experiencing increasing (micro)tribalization. The main takeaway of this comprehensive analysis of the claims of microtribalization is therefore that we hereby hopefully have opened up for a host of research into the potential power dynamics of MOS in itself and in comparison with other disciplines. Against this backdrop, a first novel line of inquiry could be linked either to a qualitative content analysis of the central nodes extracted from each cluster or CP-structure, or a topic modeling approach linking extracted topic features to the components, clusters, and cores. The composition of topics assigned to each core etc. might then, firstly, depict the changes in the combination of problems, methods, and theories employed in MOS, and, secondly, what topics dominate the MOS discourse and whether these stem from the disciplinary center or the periphery. Furthermore, we encourage other scholars to replicate our study with different databases (e.g. SCOPUS), and with comparative approaches that include disciplines that display varying degrees of internal struggles and differences in their paradigmatic cores (e.g. Psychology, Sociology as examples of disciplines with high fragmentation and weak paradigmatic core, or physics as example of relatively low levels of fragmentation and a strong paradigmatic core). Finally, we advise scholars to analyze the re-embedding of sources within the sections of the articles to find more evidence of whether citations are ritual or have a real potential of re-integrating knowledge stemming from different branches of MOS. Inquiries of this and other related matters are the only true means by which we can obtain a more comprehensive view of MOS and of mechanisms that can drive tribalization and unification of disciplines, which can then be used in similar studies of other disciplines.