Exploring the “theory is king” thesis in accounting research: the case of actor-network theory

Abstract

Purpose

This paper examines the influence of the normal science tradition, epitomized by the notion that “theory is king”, on contemporary accounting research and the epistemological tensions that may emerge as this idea is applied to particular ways of studying accounting. For illustrative purposes, the authors focus on research informed by actor-network theory (ANT) which can be seen as an “extreme case” in the sense that it is, in principle, difficult to reconcile with the normal science aspirations.

Design/methodology/approach

The paper offers an analysis based on a close reading of how accounting scholars, using ANT, theorize, and if they do engage in explicit theorizing, how they deal with the tensions that might emerge from the need to reconcile its epistemological underpinnings with those of the normal science tradition.

Findings

The findings of this paper show that the tensions between normal science thinking and the epistemological principles of ANT have, in a few cases, been avoided, as researchers stay relatively faithful to ANT and largely refrain from further theory development. However, in most cases, the tensions have ostensibly been ignored as researchers blend the epistemology of ANT and that of normal science without reflecting on the implications of doing so.

Originality/value

The paper contributes to emerging debates on the role of the normal science tradition in contemporary accounting research, and also extends recent discussions on the role of theory in accounting research inspired by ANT. The paper proposes three reasons for the observed blending of epistemologies: unawareness of tensions, epistemological eclecticism and various political considerations.

Citation

Lukka, K., Modell, S. and Vinnari, E. (2022), "Exploring the “theory is king” thesis in accounting research: the case of actor-network theory", Accounting, Auditing & Accountability Journal, Vol. 35 No. 9, pp. 136-167. https://doi.org/10.1108/AAAJ-01-2022-5616

1. Introduction

A widely shared view in contemporary accounting and management research is the idea that researchers need to constantly strive to make a theoretical contribution to prior research in their conceptual as well as empirical work. This position is often seen as testimony to a strongly ingrained normal science tradition, epitomized by the notion that “theory is king” (Straub, 2009), in such research [1]. This normal science tradition is not only the dominant view in the so-called “mainstream accounting research” (Chua, 1986), but also appears increasingly in “alternative”, or inter-disciplinary accounting research (Parker and Northcott, 2016; Richardson, 2018). This tradition has many advantages, such as the systematic building on an existing knowledge base and constant attempts to advance our theoretical knowledge of various empirical phenomena and can, at least in principle, include incremental as well as more radical theoretical advances. However, over the past decade, concerns have increasingly been raised about researchers placing too much emphasis on theory development and that this, at times, leads to overly eclectic, forced or superficial theorizing just for the sake of getting published (Hambrick, 2007; Avison and Malaurent, 2014; Guthrie and Parker, 2017; Tourish, 2020). Such tendencies arguably testify to a fetishization of the role of theory and theorization as an end in itself rather than as a means to an end. An exaggerated emphasis on the notion that “theory is king” may also lead researchers to do violence to particular research traditions by, for instance, “corrupting” the enactment of their underlying epistemological assumptions. However, while the encroachment of the “theory is king” position has been duly criticized in inter alia management (Hambrick, 2007; Tourish, 2020), information systems (Avison and Malaurent, 2014) and accounting research (Guthrie and Parker, 2017), few attempts have been made to examine its manifestations and implications in specific bodies of research where the risk of such “corruption” may be expected to be especially pronounced.

In this paper, we probe these issues through an excursion into accounting research based on actor-network theory (ANT), which can be seen as an “extreme case” when it comes to the problem of reconciling a particular research tradition with the normal science tradition. A particular challenge in this respect relates to ANT's highly empiricist epistemology, which arguably displays a remarkable lack of interest in any conventional strategies of theory development. While ANT constitutes a somewhat heterogeneous body of research (Law, 2009), Latour (1996, p. 131), a leading advocate of this school of thought, neatly captures this seemingly “a-theoretical” stance when he says that the aim of ANT studies is to produce “[o]ne single explanation to a singular, unique case; and then we throw it away” [2]. This view, which is also echoed by other leading contributors to ANT (Callon, 1999; Law and Singleton, 2013), suggests that all empirical phenomena are so unique that theory has no role to play other than as a possible object of analysis that is, itself, implicated in the formation of actor-networks. According to Latour (2005a), an ANT study is nothing more than an ethnographic tracing of associations that is reported as a detailed description of the events of interest rather than as a highly theorized account of reality that can serve as a basis for cumulative theory development or be generalized beyond the specific empirical context where such events take place. If this assertion is taken literally, ANT's epistemological commitments are antithetical to the aim of constantly making a theoretical contribution to prior research that characterizes the normal science tradition.

This lack of compatibility between the epistemology of ANT and that of the normal science tradition creates significant challenges for accounting scholars applying the former as a method theory to their specific domain [3]. As noted by Modell et al. (2017), it is difficult to envisage how accounting scholars with a strong commitment to ANT can escape the pressures for compliance with the normal science tradition. Since the latter tradition seems strongly entrenched in contemporary accounting academia, such scholars may have little choice but to engage in cumulative theory development and, in doing so, run the risk of compromising the epistemological principles of ANT. However, as the ensuing debate between Modell (2020a, b) and Baxter and Chua (2020) makes plain, there are still disagreements as to whether ANT constitutes an “a-theoretical” approach. In rebutting this claim, Baxter and Chua (2020) affirm their commitment to the highly empiricist epistemology of ANT while arguing that this epistemology allows for an element of theory development as researchers narrate the formation of actor-networks in specific contexts. However, they stop short of discussing whether and how this epistemology can be reconciled with the demands to constantly make a theoretical contribution to prior research. Hence, despite Modell et al.'s (2017) observations [4], the questions of whether and how accounting scholars with a strong commitment to ANT actually theorize and whether this mode of theorizing can be reconciled with the normal science tradition remain unanswered.

Extending this line of inquiry, the present paper offers a partially naturalistic [5] examination of whether and how accounting scholars using ANT theorize and, if they do engage in explicit theorizing, how they deal with the tensions that might emerge from the need to reconcile its epistemological underpinnings with those of the normal science tradition [6]. To clarify, our paper is motivated by the epistemological concerns outlined above, to which we return in the concluding discussion, while the literature analysis is naturalistic. While recognizing the somewhat heterogeneous nature of ANT as a larger body of scholarship, prior reviews and reflections on its use in accounting research have tended to treat it as epistemologically homogeneous (Justesen and Mouritsen, 2011; Lukka and Vinnari, 2014, 2017; Vosselman, 2014; Baxter and Chua, 2017; Robson and Bottausci, 2018) and largely distinct from the normal science tradition (Boedker, 2010; Hansen, 2011; Justesen and Mouritsen, 2018). Based on this background understanding, we chose to focus on a smaller but rapidly growing sub-section of this literature, exploring the so-called “performativity thesis” (Vosselman, 2014), that draws on the work of especially Callon (1998a, b) on the performativity of economic theories. The notion of performativity is of central concern to actor-network theorists (Latour, 1986, 2004a) and, as far as we were aware at the outset of our study, there is nothing in Callon's (1998a, b) work suggesting that it should be studied differently from what has been proposed by other actor-network theorists.

However, as our analysis progressed and we received feedback on an initial draft of the paper from a group of colleagues, it was suggested to us that there may be important elements of normal science thinking in Callon's (1998a, b) work and subsequent advances that have sought to extend and refine this work. This led to an “analytical twist” on our part, whereby we deepened our investigation of what this element of normal science thinking might be and whether accounting scholars drawing on Callon's work (1998a, b) have recognized this aspect in his thinking. Even though this “analytical twist” did not change our initial, overriding observations, it led us to reflect further on the extent to which it may be seen as justifiable to combine the epistemology of ANT and that of the normal science tradition. The relevance of doing so was further underlined to us by emerging debates in organization studies, where the growing efforts to advance cumulative theory development on how theories become performative (see Marti and Gond, 2018, 2019) have prompted relatively forceful objections suggesting that this might lead research on performativity to drift too far from the epistemological roots of ANT and that this may reinforce the epistemological tensions we have in mind (D'Adderio et al., 2019; Garud and Gehman, 2019).

Our analysis suggests that, even though accounting researchers are not necessarily unaware of the tensions resulting from combining the epistemology of ANT with that of the normal science tradition, many of them effectively blend these epistemologies and rarely, if ever, offer explicit reflections on the justifiability of doing so. Such blending of epistemologies may be seen as problematic by scholars with a preference for epistemological purity and may, at worst, undermine the scholarly legitimacy of specific bodies of research (Okhuysen and Bonardi, 2011; Modell, 2015; Modell et al., 2017). However, consistent with our partially naturalistic approach, we are reluctant to dismiss such practices out of hand. Although there may be multiple, alternative reasons for the blending of epistemologies, we find that, in practice, this problem is at least partly alleviated by what we call a “strategy of compartmentalization”, which enables researchers to engage in a pronounced element of normal science thinking while remaining reasonably faithful to the epistemological principles that underpin ANT as an empirical method. The fact that this position seems widely accepted in contemporary accounting academia leads us to reflect further on the status of the normal science tradition within this epistemic community. In doing so, we end up problematizing the notion that “theory is king” as the predominant criterion for evaluating scholarly work in contemporary accounting research and extend the discussion about how research exploring the “performativity thesis” may progress.

The paper proceeds in the following manner. In Section 2, we develop the analytical framework of our study by juxtaposing the epistemology of ANT and that of the normal science tradition of continuously advancing theoretical contributions to prior research. In Section 3, we then present our review of accounting research exploring the “performativity thesis”. Finally, in Section 4, we discuss our findings in the light of the epistemological tensions between ANT and the normal science tradition and the broader implications of our study for future accounting research.

2. The epistemologies of ANT and the normal science tradition

ANT is a scholarly approach developed originally through the early works of Bruno Latour (Latour and Woolgar, 1979; Latour, 1987, 1988) and Michel Callon (e.g. Callon and Latour, 1981; Callon, 1986), accompanied particularly by the works of John Law (e.g. Law, 1994, 1999). These first works arguably comprised a coherent body of thought focusing mostly on the construction of scientific knowledge through material-semiotic networks. This “classic” version of ANT is the one that has spread most widely within the social sciences (Law, 2009), including accounting academia [7] (Justesen and Mouritsen, 2011, 2018; Robson and Bottausci, 2018). However, as we came to realize during the course of our study, since the mid-1990s ANT has expanded and become fragmented to such an extent that it can no longer be referred to as a single, unified system of thought. In this vein, Law (2009, p. 141) characterizes it as follows:

Actor network theory is a disparate family of material-semiotic tools, sensibilities and methods of analysis that treat everything in the social and natural worlds as a continuously generated effect of the webs of relations within which they are located. It assumes that nothing has reality or form outside the enactment of those relations. Its studies explore and characterize the webs and the practices that carry them. Like other material-semiotic approaches, the actor network approach thus describes the enactment of materially and discursively heterogeneous relations that produce and reshuffle all kinds of actors including objects, subjects, human beings, machines, animals, “nature,” ideas, organizations, inequalities, scale and sizes and geographical arrangements.

Despite some heterogeneity, the above definition suggests that at the heart of ANT there are, arguably, four key ideas associated with its ontology (see especially Harman, 2009):

1. The world consists of actors (or actants or objects). ANT posits that there is nothing more to the world at a given point in time than its actors with their features. However, the world consisting of actors is always fragile and therefore work is required either to make it hold together or to change it. An actor can be any thing that produces effects.

2. Following the principle of irreduction, no object can be inherently reduced to any other. Herein is embedded the anti-essentialist feature of ANT, which presumes nothing (but actors and their relations) is there in the world just per se. This principle is also connected to the next two key ideas.

3. All objects are linked to each other through the interaction that occurs between them in translations. Through translations, actors associate with each other and form networks and thereby potentially new actors. From this it follows that while nothing is inherently irreducible to anything else, everything is, at least in principle, transformable if the potentially painful translating work will be successfully undertaken.

4. Actors obtain their powers from their alliances, not from their private essences (which are dismissed as not making sense in the ANT context).

What follows, from these key principles is the famous flat ontology and the assumption of absolute concreteness: Every actor, formed through its network of the time, is simply nothing more than what it is at a given moment in time. Actors and their agency are not viewed as embedded in any kinds of pre-existing structures, as would be the case in, for instance, structuration theory (Giddens, 1984), neo-institutional sociology (see Greenwood et al., 2008) and Bourdieusian theorizing (e.g. Bourdieu, 1979). All actors are also locked in trials of strength and can be kept intact, or be transformed, only through translation work related to these trials. In its emphasis on concreteness, ANT might, at first sight, seem to enact naïve realism, but this is not quite the case. Rather than assuming atomistic and isolated objects in an essentialist sense (as per logical positivism), the ontology of ANT should be labeled relational (Harman, 2009).

The concept of performativity is fundamental to the ontology of ANT and implies a certain intertwining of its ontology and epistemology. Latour (1986) contrasted a performative definition of society with the ostensive one that arguably permeates much sociological research. The ostensive definition is based on a realist ontology that views actors as embedded in social realities, which are only partially knowable to them, but which social scientists equipped with appropriate theories and methods can, in principle, piece together and explain. By contrast, a performative definition of society rejects the view of scientists occupying a privileged position as producers of knowledge and rather emphasizes that society is continuously made and remade as scientists and other actors go about constructing social realities (see Latour, 2004a). This position chimes with the constructivist and anti-essentialist ontology of ANT and leads to a conception of knowledge production as intricately intertwined with the formation of inherently fluid relations between scientists and other actors (see also Callon, 1986). Rather than conceiving of the social world as pieced together by social scientists, working in relative isolation from the actors they study, society is seen as being “performed through everyone's efforts to define it” (Latour, 1986, p. 273) and, as such, defying any attempts at definition by scientists alone (Callon et al., 2009). This has led Latour (1986, 2005a) to caution against the adoption of an epistemology based on an etic [8] perspective and the temptation of researchers to impose pre-defined theoretical concepts, such as “class”, “power” and “society”, on other actors in order to make sense of the world.

As we shall see, this pronounced reluctance to adopt an etic perspective in what we may refer to as “classic ANT” differs significantly from the core ideas of the epistemology associated with the normal science tradition. The main differences between these two approaches are summarized in Table 1 and elaborated further below.

The first major difference between the epistemology of the normal science tradition and that of “classic ANT” relates to their over-riding views of the role of theory in research. While theory in itself is an elusive concept – suggesting that it is easier to state what theory is not than what it actually is (Sutton and Staw, 1995) – for the purpose of our analysis, we present an attempt to define it, yet acknowledging how fragile such attempts are. Largely in line with Whetten (1989), for us theories represent attempts to systematically describe, help to understand, explain and, at times, also predict matters of scholarly interest. Theories are interested in patterns and relationships as well as principles that govern the empirical domain of interest. Theories are composed of a structure of inter-related concepts as well as propositions as to what they tell us about the matter of interest. Importantly, theories should by no means be limited to just “grand” theorization of settings and structures, but can also occur at a lower level through the theorization of metaphors, differentiation or concepts (Llewellyn, 2003) [9].

Within the epistemology of the normal science tradition, a rather established view, that is epitomized by the idea that “theory is king” (Straub, 2009), is that any conceptions of theory, such as those described above, form an important input to as well as output of the process of discovery. An underlying assumption of this position, which is widely shared in the social sciences (see, e.g. Corley and Gioia, 2011; Shepherd and Suddaby, 2017; Sandberg and Alvesson, 2021), is that the world is considered sufficiently stable to let us assume and employ in a useful manner existing, yet potentially evolving, structures, such as theories, in our reasoning. Herein lies the general idea of researchers continuously seeking theoretical advances – a position that can be traced back to Ancient Greece Aristotle (1996), where knowledge having broader than just instantaneous bearing was viewed as the aim. The seeking of such knowledge places the researcher in the privileged and somewhat detached position as the ultimate arbiter of what counts as meaningful knowledge.

The assumption of such an etic perspective stands in stark contrast to the epistemology that follows as a logical consequence of subscribing to the anti-essentialist, flat ontology of ANT. Since this ontology assumes that all there is to observe is which actors exist, how they are maintained, and how new actors are created through translations of various kinds and become endlessly embroiled in trials of strength, the only thing that researchers are supposed to do is to follow the actors and describe the events that unfold (Latour, 1987, 1999, 2005a). If taken literally, there is no place for theory, other than the basic conceptual building blocks provided by ANT, either as an input to or an output of the process of discovery in such research (Modell et al., 2017; Modell, 2020a). As emphasized by Latour (1999), ANT is neither interested in offering predictions nor in the development of social laws. In fact, theory is viewed as one of the essentialist structures to be avoided as it is seen as distracting researchers from the production of pure, empirical descriptions of events (Latour, 2005a). Indeed, the analytical adaptability afforded by ANT's lack of theory in general, and the lack of any pre-determined theory of the actor in particular, is considered to be its main strength (Callon, 1999). This position seems to be shared by key advocates of the use of ANT in accounting research, such as Baxter and Chua (2017, 2020) and Justesen and Mouritsen (2011, 2018), who agree that ANT as a method theory is indeed vehemently opposed to researchers using established social categories or appealing to structural forces as the drivers of social action to advance theoretical insights that are, in any way, generalizable beyond specific contexts.

A second dimension where the epistemology of “classic ANT” departs from that of the normal science tradition relates to expected modes of reasoning. Normal science thinking is characterized by the well-established, theory-related modes of scientific reasoning, namely deduction, induction or abduction. By deduction we refer to scientific practice that takes extant theory as a starting point, utilizing it to develop testable hypotheses or, in a more relaxed form, to guide the analysis of empirical data. Induction, in turn, is taken to imply generalization from particular empirical observations without using any preconceived, theoretical frameworks. Finally, abduction is associated with analytical generalizations whereby the analyst moves back and forth between theory and empirical data and constantly seeks to advance extant theory by contrasting emerging, empirical observations with such theory (Lukka and Modell, 2010; Timmermans and Tavory, 2012).

In contrast, from ANT's seemingly “a-theoretical” stance it follows that the mode of reasoning is mainly supposed to be open-ended and descriptive, even if it is analytical. The key point is to follow actors, trace the formation of associations between them, and describe these events with the aim of producing rich process descriptions. ANT's mode of reasoning is hence not meant to be informed or guided by theoretical ex ante assumptions about which actors matter or ambitions to advance theoretically informed explanations (Latour, 1987, 2005a). This follows as a logical consequence of the view of theories, other than ANT as a broad, sensitizing framework, as a potential distraction from the need to nurture an anti-essentialist view of the world. Even though such research may seem reminiscent of an inductive approach, we need to bear in mind that inductive reasoning still aims at drawing conclusions that, to some extent, can be generalized beyond specific empirical contexts. Hence, the mode of reasoning associated with ANT is not easily reconcilable with any of the modes of reasoning associated with the normal science tradition.

A third distinction between “classic ANT” and the normal science tradition relates to the development of research questions. In normal science thinking, the research questions are typically developed through theoretically informed gap-spotting or result from the problematization of either deductively derived tensions and paradoxes or empirically derived constructions of mysteries (Alvesson and Kärreman, 2007; Alvesson and Sandberg, 2011). Gap-spotting relies on rather straightforward identification of theoretical omissions (or gaps) in prior research, often leading to merely incremental contributions aimed at cumulative theory development. Problematization, in turn, implies a fundamental questioning of established knowledge through the identification of theoretical tensions in prior research (e.g. assumptions or results), that can lead to more radical breaks with extant knowledge. Mystery construction, finally, is based on the observation of clashes between extant knowledge and empirical events that generate a sense of surprise or puzzlement. While this strategy for developing research questions is more empirically tuned than gap-spotting and problematization, it still implies the use of extant theory as a reference point for determining what constitutes interesting research questions (Alvesson and Kärreman, 2007). Of these three options, especially gap-spotting tends to dominate the motivation of research questions within the normal science tradition (Alvesson and Sandberg, 2011), although some commentators increasingly see empirical surprises, such as those underpinning mystery construction, informing such questions (Corley and Gioia, 2011).

With regards to ANT research, as theory is not considered an appropriate starting point for such studies, it is not quite clear how scholars select their research topics and, especially, how they develop the research questions for their analyses, if they follow ANT's epistemological principles to the letter. It may be that many ANT studies are motivated by the intuitive appeal of examining “matters of concern” that are somehow encountered in empirical observations (e.g. Latour, 1988, 1996, 2004a). While “matters of concern” evade precise definition, they can be understood in contrast to “matters of fact”, which Latour (2005b, p. 39) describes as being “indisputable, obstinate, simply there” or, in other words, stabilized at the moment of observation. Hence, while “matters of fact” are, at least at the moment, stabilized and closed facts, “matters of concern” are not. Therefore, the latter can be picked by certain actors or their constellations as targets of stabilization efforts. In this context it is useful to note that although focusing on “matters of concern” to a certain extent resembles empirically derived mystery construction (Alvesson and Kärreman, 2007) as a way of developing research questions, the latter is still theoretically informed. Moreover, as “matters of concern” are not even supposed to be examined from a theoretical perspective, ANT studies do not necessarily need explicit research questions but can be motivated instead by a broad empirical purpose.

The final point of epistemological divergence between the normal science tradition and “classic ANT” concerns the nature of research contributions. The ways of contributing to prior research that are characteristic of the normal science tradition include developing a new theory or illustrating, extending/refining and/or testing an existing one (Keating, 1995). In contrast, ANT can be viewed as a full-blown methodology, which includes methodological guidelines derived from its distinctive ontology and epistemology (Latour, 1987, 2005a; Callon, 1999). Hence, especially “classic ANT” is arguably unique in the sense that it is only interested in describing, not theorizing, empirical events unfolding in the world. Since researchers are hence neither supposed to use theoretical inputs nor generate theoretical outputs, ANT research cannot be expected to make a contribution in any of the ways that is characteristic of the normal science tradition. The rejection of the aspiration to generate theoretical contributions that are somehow transferable beyond the specific, empirical context at hand is perhaps most forcefully expressed by Latour (1996, p. 131) when he argues that ANT-inspired scholars are expected to produce one-off empirical descriptions of events and then just “throw them away”. However, similar rejections of attempts to abstract from context-specific events in an effort to advance theoretical contributions of more general relevance can be found elsewhere in the ANT literature, especially in the writings of Latour (e.g. Latour, 1999, 2005a) and Law (e.g. Law and Singleton, 2013).

As should be clear from the above, brief outline of the differences between “classic ANT” and the normal science tradition it is difficult, if not impossible, to combine these approaches in an epistemologically consistent manner. Doing so may be expected to generate significant epistemological tensions and, similar to other combinations of research approaches with diverging epistemologies, this implies a risk of undermining the legitimacy of research inspired by ANT as a worthwhile scholarly endeavor (cf. Okhuysen and Bonardi, 2011; Modell, 2015; Modell et al., 2017). However, following the evolution of ANT into an increasingly fragmented body of research, at least some of its advocates have started to advance research in a direction that shows greater affinity to the normal science tradition. As we became increasingly aware during the course of our study, this tendency is especially notable in Callon's (1998a, b) work on the performativity of economics. This work not only started with an explicit problematization of the theoretical conception of economic relationships as embedded in social structures, that has long dominated economic sociology (Granovetter, 1985), but also borrowed extensively from sociologists such as Erving Goffman (1971) to explain how economic theories become performative and (re-)shape actor-networks in ways that cause such relationships to become embedded in economics. In doing so, Callon (1998b) advanced a theoretical framework that explained how the use of economic theories is implicated in complex framing processes, whereby economists and other actors reconfigure economic relationships, and how such processes do not only cause economic theories to become self-fulfilling but also generate unintended consequences, or overflows, that continuously require such relationships to be re-framed to achieve a temporary degree of stability.

While Callon's (1998a, b) conception of performativity as a process that is continuously (re-) shaping social realities is consistent with Latour's (1986, 2004a) anti-essentialist, performative definition of society, his extensive engagement with other seminal thinkers to advance a distinct theoretical framework as a starting point for empirical studies can be seen as at least implicit testimony to the normal science tradition. Elsewhere, he has continued to extend and refine this framework (Callon, 2007a, b; Callon and Muniesa, 2005) and has even called for more theoretical research on the involvement of emerging, concerned groups in technoscience (Callon and Rabeharisoa, 2008) without ostensibly abandoning his epistemological commitment to ANT. This points to a potential risk of epistemological tensions in his work and that of other scholars who seek to apply and develop it while combining it with tenets drawn from “classic ANT”. Callon's (1998a, b) conception of performativity was further developed by MacKenzie (2006, 2007) and their combined works have spawned a vibrant research program in economic sociology and organization studies that is at least partly underpinned by growing aspirations to advance cumulative theory development (see Marti and Gond, 2018, 2019). However, this rapprochement with the normal science tradition has recently sparked a counter-reaction from scholars who caution that it may cause research on performativity to drift too far from the epistemology of “classic ANT” and reinforce the epistemological tensions between these research traditions (see D'Adderio et al., 2019; Garud and Gehman, 2019). A particular worry in this regard is that excessive concerns with cumulative theory development, aimed at establishing general, theoretical boundary conditions under which performative tendencies are more or less salient, will abstract too much from the rich process descriptions that are typically seen as the hallmark of “good” ANT research and that this will lead performativity to be portrayed as an overly orderly phenomenon, evolving in a linear fashion and generating predictable outcomes. In what follows, we examine whether and how accounting scholars following the works of Callon (1998a, b) and MacKenzie (2006, 2007) have dealt with the epistemological tensions associated with the combination of ANT and the normal science tradition and what implications this has for accounting scholarship.

3. Review of accounting research exploring the “performativity thesis”

3.1 Analytical focus, sample and analytical procedures

The following literature review covers a set of empirical research papers that draw extensively on the works of Callon (1998a, b) and/or MacKenzie (2006, 2007) while using ANT as a broader method theory to shed light on various accounting practices [10]. This research is concerned with what Vosselman (2014) calls the “performativity thesis” and has examined how a range of calculative devices are implicated in the more or less ongoing shaping of social realities. While ours is not the first attempt to evaluate this body of research (see also Boedker, 2010; Vosselman, 2014), our review is distinct from prior studies in that we pay more focused attention to the epistemological tensions that can emerge as ANT is combined with an element of normal science thinking and how accounting scholars have dealt with such tensions.

The sample of studies under review was drawn from a systematic literature search covering eight major accounting research journals [11] that have published, or may be expected to publish, research of interest to us up until the end of November 2021. While initially conducting a very broadly based search for papers containing references to performativity as an analytical concept, we gradually whittled this down to empirical studies that mainly use the works of Callon (1998a, b) and/or MacKenzie (2006, 2007) to investigate this topic. The literature search ultimately generated a sample of 21 papers that were deemed relevant for inclusion in our review. In analyzing these papers, we initially paid attention to how the authors claim to advance research on specific accounting phenomena, or domain theories, by adopting ANT as a method theory and/or to advance ANT itself as a more generally applicable method theory (cf. Lukka and Vinnari, 2014). We developed a coding scheme reflecting the differences between the epistemology of “classic ANT” and that of the normal science tradition discussed in the foregoing (see Appendix). Each of the three authors then coded every paper independently of each other before comparing the individual codes to arrive at an agreed set of codes. Following these coding procedures, we first pilot-coded four of the 21 papers to ensure our coding was based on a shared understanding of how to apply the coding scheme. This step prompted some minor revisions of the coding scheme before we extended the coding to the entire sample.

Following our initial analysis, we sent a draft version of the paper to a relatively large number of colleagues, several of whom are authors of the papers under review and requested feedback on our analysis. The feedback received drew our attention to the possibility that the epistemological commitments of especially Callon (1998a, b) differ somewhat from those of “classic ANT” in that his work on performativity displays an element of normal science thinking while still being rooted in ANT. This prompted us to re-analyze the papers with an eye to whether these epistemological nuances are explicitly recognized and how researchers have dealt with concomitant tensions. While the feedback suggests that at least some of the authors are aware of these nuances, this additional analysis did not change our initial, over-riding observation that the epistemological tensions between ANT and the normal science tradition have received little, if any, explicit attention in the papers under review. We reflect further on this apparent anomaly in our review. Our final analysis of the 21 papers is summarized in Table 2.

3.2 Over-riding view of the role of theory

Starting with the over-riding view of the role of theory, none of the papers under review overtly rejects existing theory as either an input or expected output with reference to the epistemological underpinnings of ANT. The most common positioning is to apply ANT as a method theory to specific accounting topics or domain theories. As such, there is a clear ambition to advance continuous theory development about the role of accounting in organizations and society. In addition to advancing domain theories, five papers also contain an explicit ambition to further develop ANT as a method theory (Millo and MacKenzie, 2009; Vinnari and Skærbæk, 2014; Ferreira, 2017; Warren and Seal, 2018; Boedker et al., 2020). In only two papers is the theory development ambition less explicitly articulated (MacKenzie, 2009; Roberts and Jones, 2009). These papers are rather positioned as contributions to emerging policy debates with far-reaching ramifications for economy and society, such as the emergence of carbon markets (MacKenzie, 2009) and the global financial crisis (Roberts and Jones, 2009), that ANT might inform.

While the papers under review differ somewhat in their over-riding views of the role of theory, it is worth noting that none of them offers any explicit reflections on the justifiability of blending the epistemology of “classic ANT” with that of the normal science tradition and the epistemological tensions that emerge from doing so. Typically, the authors mainly introduce the works of Callon (1998a, b) and/or MacKenzie (2006, 2007) and then discuss the practicalities of data collection and analysis without dwelling on the epistemological underpinnings of their research. Still, all of the papers under review are explicitly rooted in ANT as a method theory and, as will be demonstrated below, the vast majority of them show significant import from “classic ANT” especially insofar as the mode of reasoning is concerned. Hence, despite the lack of explicit reflections on the matter, there is a real risk of epistemological tensions, such as those outlined in the foregoing, permeating this body of research. The lack of such reflections also applies to the normal science thinking that is evident in Callon's (1998a, b) work and the subsequent debate about the risk of research on performativity drifting too far from the epistemological roots of ANT (D'Adderio et al., 2019; Garud and Gehman, 2019). Even though the feedback received on an earlier version of the paper suggests that there is at least some awareness of these aspects, this awareness is not translated into explicit reflections on the implications of combining the epistemology of “classic ANT” with that of the normal science tradition. Hence, with the exception of the papers that effectively avoid epistemological tensions by sticking closely to “classic ANT” (MacKenzie, 2009; Roberts and Jones, 2009), the authors have dealt with such tensions by ostensibly ignoring them and remaining silent about the challenges of combining its epistemology with that of the normal science tradition.

A possible explanation for this may be that the authors together with editors and reviewers have not considered the concomitant epistemological tensions serious enough to merit extensive discussion. If this is the case, it may suggest that accounting scholars, as a broader epistemic community, are rather tolerant of the eclectic blending of epistemologies and perhaps consider it more important “get on” with their research than dwelling too much on abstract philosophical issues. Such a posture has been documented in prior investigations of accounting research blending diverse epistemologies (Modell, 2015). There may also be pragmatic reasons, such as journal space limitations, for why the authors say little about the epistemological challenges of combining ANT with elements of normal science thinking (Modell et al., 2017). Nevertheless, it is striking that the silence on such matters is systematic and it remains unclear to what extent the authors of the papers under review are actually aware of the epistemological tensions that follow from the blending of these research traditions or have found it relevant to consider such tensions as they go about conducting their research. This is potentially a cause for concern if we want to avoid excessive epistemological eclecticism.

3.3 Employed modes of reasoning

Nearly all of the papers under review show at least some evidence of the modes of reasoning typically associated with ANT. Several of them make explicit references to key methodological principles associated with ANT, such as the need to follow multiple human and non-human actors as they go about constructing actor-networks, to advance highly descriptive, open-ended process accounts of how accounting becomes performative (Skærbæk, 2009; Skærbæk and Tryggestad, 2010; Vinnari and Skærbæk, 2014; Georg and Justesen, 2017; Themsen and Skærbæk, 2018; Kastberg and Lagström, 2019; Pucci and Skærbæk, 2020). Other papers are less explicitly grounded in these principles (Christensen and Skærbæk, 2007; Revellino and Mouritsen, 2015; Ferreira, 2017; Jollands et al., 2018; Boedker et al., 2020; Lassila et al., 2019; McLaren and Appleyard, 2020) or only offer limited, if any, methodological reflections (MacKenzie, 2009; Millo and MacKenzie, 2009; Roberts and Jones, 2009; Kastberg, 2014; Cuckston, 2018) but still remain relatively faithful to the methodological tenets of “classic ANT”. The empirical accounts in these papers generally have a highly descriptive flavor and document how the performative effects of accounting emerge and unfold although the depth of their process analyses varies somewhat.

Following the ambition to advance continuous theory development about the role of accounting in organizations and society, the majority of the papers under review combine the mode of reasoning associated with ANT with those of the normal science tradition. Some studies apply a deductive mode of reasoning by mobilizing relatively extensive, a priori deliberations that set out clearly articulated theoretical expectations regarding the performative effects of accounting and under which circumstances such effects are likely to occur (e.g. Millo and MacKenzie, 2009; Skærbæk and Tryggestad, 2010; Warren and Seal, 2018; Boedker et al., 2020). This occasionally entails a blending of ANT with other method theories. For instance, Millo and MacKenzie (2009) mobilized ANT to critique the Granovetterian view of markets as embedded in pre-existing social structures (Granovetter, 1985), that dominates traditional variants of economic sociology, without entirely abandoning this view in elaborating how risk management models become performative. Similarly, Warren and Seal (2018) discuss how the view of performativity inspired by ANT may be elaborated in ways that resonate with the emerging literature on critical performativity (Spicer et al., 2009, 2016) within a broader cultural political economy framework. To justify the adoption of such a perspective, the authors go to some length in discussing the epistemological consistency of the general concept of performativity with their notion of cultural political economy. However, despite drawing significant inspiration from ANT, they do not dwell on whether this broader method theory is consistent with the cultural political economy framework informing their analysis. This represents a piecemeal and somewhat eclectic incorporation of ANT into a deductively derived analytical framework that draws on more than one method theory.

Other papers make explicit references to induction (Revellino and Mouritsen, 2015; Ferreira, 2017) or effectively apply this mode of reasoning to advance generalizable knowledge claims based on emerging empirical observations (Christensen and Skærbæk, 2007; Jollands et al., 2018; McLaren and Appleyard, 2020). For instance, Christensen and Skærbæk (2007) conducted a systematic, comparative study of public sector accountability innovations in Denmark and Australia and drew several general conclusions, based on observed similarities between these contexts, while discussing how their findings might be extended to other contexts. A similar element of inductive, comparative theory development is discernible in Jollands et al. (2018), who contrasted the organizational responses to different, performative effects to advance our theoretical understanding of this phenomenon.

Finally, a large number of papers engage in an abductive mode of reasoning (although none of them use the term explicitly) by comparing their empirical findings with extant, related research in a way that resembles analytical generalization. An illustrative example is Kastberg and Lagström's (2019) study of hybridization in a social care organization. While recognizing that this substantive phenomenon has been extensively examined in prior research drawing on other method theories than ANT, the authors systematically compare their empirical findings with such research to specify the distinct contributions that ANT might make and used this as a basis for refining our theoretical understanding of hybridization. However, the extent to which abductive reasoning is applied varies and, in most of the papers, it does not detract too much from their descriptive process accounts by overloading these accounts with abstract theoretical reflections. Some studies are a little less detailed when it comes to describing the intricate processes through which accounting becomes performative (Cuckston, 2018, 2019; Lassila et al., 2019; McLaren and Appleyeard, 2020). However, in contrast to the concerns that have been raised in the broader literature on performativity (D'Adderio et al., 2019; Garud and Gehman, 2019), we see little evidence of the normal science tradition abstracting too much from researchers' attention to such processes. By and large, theoretical reflections following a deductive, inductive and/or abductive mode of reasoning are mainly found in the early and concluding parts of the papers under review and do not infringe on the detail in which the empirical narratives unfold.

3.4 Strategies for developing research questions

Moving on to the strategies for developing research questions, few of the papers under review rely exclusively on the ANT-inspired strategy of highlighting empirically derived, controversial topics, signifying “matters of concern”, without advancing a more theoretically informed motivation of their research focus. The clearest examples of this strategy are found in the two papers addressing issues of great socio-political importance (MacKenzie, 2009; Roberts and Jones, 2009). To MacKenzie (2009), the main concern is the politics surrounding carbon markets as a vehicle for addressing climate change, while Roberts and Jones (2009) take as their starting point the sense of shock that accompanied the emergence of the global financial crisis in 2008. In the case of Roberts and Jones (2009), there is some evidence of an attempt to use emerging, empirical observations as a basis for constructing a theoretically informed mystery, pivoting on how economics-based conceptions of self-interests become performative (cf. Alvesson and Kärreman, 2007). However, this is a less explicit part of the strategy for advancing their research question and it is not used as a basis for cumulative theory development. Other papers primarily rely on empirically derived research questions while retaining an overall emphasis on theory development resembling that associated with the normal science tradition (Revellino and Mouritsen, 2015; McLaren and Appleyeard, 2020). For instance, Revellino and Mouritsen (2015) primarily motivate their paper based on the empirical observation that innovations may start to “drift” as they are put into practice, but then use their analysis of this phenomenon as a basis for explicitly theorizing the implications this has for accounting later in the paper. A more common strategy, however, is to blend an emphasis on empirically derived motivations, drawing attention to potentially critical or controversial topics, with theoretically informed research questions. Similar to the papers that rely exclusively on “matters of concern” as a basis for advancing research questions (MacKenzie, 2009; Roberts and Jones, 2009), this is especially the case where highly topical and politically charged issues, such as accounting for environmental and social concerns, are at stake (Ferreira, 2017; Cuckston, 2018, 2019; Jollands et al., 2018).

Of the strategies for developing research questions associated with the normal science tradition, gap-spotting is the most commonly used. This strategy is mobilized in 11 of the 21 papers and mainly takes the form of identification of gaps in the domain theories under examination which ANT is then used to fill. This is sometimes combined with an element of problematization of specific aspects of the domain theories concerned (Christensen and Skærbæk, 2007; Georg and Justesen, 2017; Boedker et al., 2020; Pucci and Skærbæk, 2020). For instance, Christensen and Skærbæk (2007) not only identified a need to take a broader range of interests into account in research on public sector accountability innovations, but also problematized the prevailing view that managers have largely unfettered discretion in devising such innovations. This subsequently led to their choice of ANT as a method theory to better account for the complexities associated with the development of such innovations. Six papers exclusively or primarily rely on problematization as a strategy for developing research questions. It is also worth noting that the number of papers relying on problematization has grown considerably in recent years. Even though these papers are still in the minority among those under review, this development contradicts Alvesson and Sandberg's (2011) observation that problematization is relatively rarely used within the normal science tradition and there are some notable examples of ANT being mobilized to problematize firmly established domain theories. For instance, Skærbæk and Tryggestad (2010) motivated their study by questioning the long-standing view that the choice of accounting practices is subordinate to corporate strategy and opened up a discussion of how accounting is actively involved in performing such strategies. Similarly, Pucci and Skærbæk (2020) problematized the use of financial economics as a method theory for the examination of accounting standard-setting, arguing that it often provides uncertain depictions of how accounting operates. This observation was then taken as the starting point for arguing that financial economics can become performative and that such performativity is better understood through the use of ANT as a method theory.

Other papers extend the strategy of problematization to ANT itself (Millo and MacKenzie, 2009; Ferreira, 2017; Boedker et al., 2020). This emerging ambition to problematize ANT as a method theory is especially obvious the case in Ferreira (2017), who draws attention to how the failure to establish markets for biodiversity offsets in the UK suggests that accounting does not always play a powerful, performative role in the shaping of economic activities. This observation is taken as a starting point for problematizing the tendency to perhaps over-emphasize the extent to which accounting becomes performative in research inspired by ANT. Ferreira (2017) then deepens his empirical analysis of how this case of “failed” performativity unfolded and, in doing so, he does not just provide a descriptive account of how this process evolved in a unique empirical context but also seeks to draw more general inferences as to when such failures may be expected to occur. A similar line of argument is discernible in Boedker et al. (2020), who problematized what they saw as a tendency in prior research to over-emphasize the extent to which accounting representations become self-fulfilling and not paying sufficient attention to the conditions under which social realities become less similar to such representations. Even though the tendencies for accounting to generate overflows have been extensively examined in prior studies, Boedker et al. (2020) are the first to explicitly call for more systematic theory development regarding the conditions under which such tendencies emerge as a way of developing ANT as a method theory. In so far as the development of research questions is concerned, this is perhaps the clearest manifestation of the normal science tradition as it points to an interest in systematically mapping the boundary conditions under which accounting representations become more or less self-fulfilling and drawing more generalizable conclusions about this phenomenon (cf. Marti and Gond, 2018, 2019).

3.5 Nature of research contributions

The widespread practice of advancing theoretically informed research questions is mirrored by a pronounced tendency to also articulate theoretical contributions in a way that goes against the epistemology of “classic ANT”. With few exceptions (MacKenzie, 2009; Roberts and Jones, 2009), the papers under review do not merely conclude by summarizing their empirical findings, but also contain at least some reflections on how these findings contribute to cumulative theory development in a manner that is consistent with the normal science tradition. While practically all papers can be said to contain some form of theory illustration, none of them lay claims to having advanced entirely new theories. All of the papers under review accept the received conception of performativity that is associated with ANT and the domain theories under examination have often been extensively researched through the use of other method theories. Also, none of the papers under review can be said to test theory in a conventional sense although some of them contain a pronounced element of deductive, a priori theorizing. However, as explicated below, the vast majority of them make at least some attempts to extend and/or refine extant domain and/or method theories.

Consistent with the tendency for theoretically informed research questions to mainly focus on the domain theories under examination, the theory extension and refinement efforts are primarily geared towards adding new insights into specific accounting topics. Eleven of the 21 papers contain clearly discernible attempts to either extend or refine theory development in relation to specific accounting topics by elaborating how ANT enhances our theoretical understanding of these topics. For instance, Georg and Justesen (2017) mainly positioned their contributions in relation to the social and environmental accounting literature, arguing that ANT provides a distinct alternative to much prior research in this area that can extend our understanding of how accounting representations of natural realities become performative and change people's conceptions of such realities. Other papers display an ambition to both extend and refine domain theories. For instance, Skærbæk (2009) emphasizes how ANT adds new insights into the dynamic processes through which the identity of auditors takes shape, but also proposes some refinements to prior research by stressing how extant notions of the “purification” of auditing (Pentland, 1993) need to be revised in order to better account for the complexities of such processes.

It is rarer for the papers under review to discuss theory extensions and/or refinements in relation to ANT as a method theory. Yet, a growing number of examples of such theory development can be found. Vinnari and Skærbæk (2014), for instance, go to some length in explaining how Callon's (1998b) work might be extended by showing how the performativity of accounting entails the mobilization of multiple frames and how this may enhance our understanding of performativity inspired by ANT. A similar attempt to extend ANT as a method theory is reflected in Ferreira's (2017, p. 1584) conclusion that “markets sometimes fail to become instituted” despite the presence of accounting calculations and his call for more systematic research on why performative effects do not materialize. Similar, albeit more tangential, calls for further research into the conditions under which performativity “fails” can be found in other papers (Revellino and Mouritsen, 2015; Warren and Seal, 2018). Answering such calls for more systematic theory development, at least two recent papers explicitly extend ANT as a method theory by specifying the general conditions under which accounting may be expected to create realities that resemble those depicted by particular accounting representations (McLaren and Appleyard, 2020) and when this is less likely to be the case (Boedker et al., 2020) based on their empirical findings. However, as we have seen above, only Boedker et al. (2020) also mobilize this line of thinking as a way of motivating their research question and the efforts to systematically identify the boundary conditions under which accounting representations become more or less self-fulfilling are still at a nascent stage. The papers under review still owe a significant intellectual debt to ANT and typically rely heavily on their context-specific, empirical observations as a basis for articulating theoretical contributions even though some of them are also aspiring to advance more generalizable conclusions that may contribute to the development of ANT as a theory in its own right.

The growing efforts to not only use ANT to advance various domain theories, but also extend and/or refine it as a theory in its own right can perhaps be interpreted as an indication of the normal science tradition exercising increasing influence on accounting research exploring the “performativity thesis”. In contrast to Lukka and Vinnari (2014), who found very few attempts by accounting scholars to extend or refine ANT when using it as a method theory, our findings suggest that such scholars are increasingly compelled to do so. However, it is worth recalling that, so far, the efforts to this end have been relatively modest and they do not yet amount to a systematic research program aimed at cumulative theory development such as that evolving around the notion of performativity in economic sociology and organization studies (cf. Marti and Gond, 2018, 2019). Hence, even though the normal science tradition is perhaps gaining ground in accounting research exploring the “performativity thesis”, it has not yet resulted in a situation where abstract theorizing is crowding out detailed empirical attention to the processes through which accounting representations become more or less self-fulfilling.

4. Concluding discussion

4.1 Summary of findings

In this paper, we set out to examine the questions of whether and how accounting scholars using ANT theorize and how they deal with the tensions that might emerge from the need to reconcile its epistemological underpinnings with those of the normal science tradition. By way of reminder, it is not our intention to promote or demote either approach; our interest lies in how scholars cope with the notable epistemological differences between the two. Examining accounting research exploring the “performativity thesis” (Vosselman, 2014), we show that, with few exceptions (MacKenzie, 2009; Roberts and Jones, 2009), researchers do engage in explicit theorizing in a manner that is consistent with the normal science tradition. In so far as this is not the case, researchers simply avoid the epistemological tensions between ANT and the normal science tradition by staying faithful, in a full-blown manner, to the highly empiricist epistemology advocated by leading propagators of the “classic” form of ANT (e.g. Latour, 1996, 1999, 2005a). This is manifest in the adoption of a highly descriptive and open-ended mode of reasoning in an attempt to explore empirically derived “matters of concern” without first posing a theoretically motivated research question or making any claims to continuous theory development aimed at generalizing findings beyond the specific empirical context in which the research is situated.

However, the second and much more prevalent strategy for dealing with the epistemological tensions between ANT and the normal science tradition is to blend their epistemologies while ostensibly ignoring such tensions by refraining from offering explicit reflections on the matter. While most of the papers under review remain relatively faithful to the epistemology of “classic ANT” as a basis for empirical research in the sense that they offer fairly descriptive, open-ended and detailed empirical accounts of how accounting becomes performative, they generally combine this mode of reasoning with an element of deductive, inductive and/or abductive reasoning. The vast majority of them also include other normal science elements, such as the articulation of theoretically informed research questions and the advancement of clearly articulated theoretical contributions. Moreover, in addition to deploying ANT to specific domain theories to shed new light on how accounting becomes performative, accounting scholars increasingly aspire to develop ANT further as a theory in its own right. As we have seen, researchers extending their work in this direction are beginning to take an interest in developing more generalizable insights into the conditions under which specific, performative effects may be expected to be more or less salient. Although this development is still at a nascent stage, it is reminiscent of how research on performativity is currently evolving in economic sociology and organization studies (Marti and Gond, 2018, 2019). Hence, it is fair to conclude that the influence of the normal science tradition on accounting research inspired by ANT has been rather prevalent and that it is perhaps growing stronger.

As concerns the potential reasons for such blending of epistemologies, we can think of three possible, alternative explanations. The first of them is unawareness, that is, the authors not being deeply familiar with the epistemological principles of ANT and how they differ from those of the normal science tradition. Such a situation might occur for instance in cases where the authors have made their way into ANT research by reading only or mainly Callon's and MacKenzie's works which, as mentioned above, do not discuss questions of epistemology except in passing. In addition, as noted above, already Callon's (1998a, b) early writings contain indications of a certain drift from the epistemological roots of “classic ANT”. An interesting question that emerges in this context concerns the extent to which scholars wishing to follow the Callon–MacKenzie stream of research on performativity could or should be expected to familiarize themselves with the philosophical works of Latour. Or, more generally put, how far upstream do we need to follow a particular strand of literature?

The second explanation of the observed blending is epistemological eclecticism, where the authors are aware of possible epistemological mismatches but do not consider them important. Eclecticism as a scientific practice has a long pedigree, going back to Ancient Greece and Rome where philosophers sought to find the best possible explanation by combining selected elements from different systems of thought [12]. As drawing a line that separates acceptable from excessive eclecticism is an act of social construction, scholars have considerable discretion in deciding how to combine epistemological elements from different theories and approaches. Indeed, in the case at hand, it may well be that some researchers are using the original writings of “master thinkers”, such as those of Latour, only as sources of inspiration for their own process of thinking [13]. However, such tendencies towards eclecticism need to be weighed against the risk of particular research traditions losing their distinct identity if they go completely unchecked.

The third potential reason for the blending of epistemologies is associated with political aspects. By this we refer to, first of all, authors' potential resistance to methodological rule systems on the grounds that each such system is underpinned by politically tuned assumptions and the authors in question wish to determine their own – especially given that methods, just like theories, are potentially performative; in other words, they bring a certain kind of reality into being (Law, 2004, pp. 149–155). Another political aspect relates to the pressures ANT scholars face during the review process. Many of us have been trained in the normal science tradition to always ask for a study's theoretical motivation and the corresponding contributions. As these are not readily detectable in full-blown ANT studies, and especially if the reviewer is not familiar with ANT's epistemological principles, he/she is likely to push the authors to introduce such theoretical elements at least in the front and back ends of the paper. Resisting reviewers' and editors' demands is perhaps a course of action few are willing to embark upon; many would rather relax their commitment to ANT principles than run the risk of being rejected. In broader terms, it might even be that ANT lends itself better to books or chapters than the rather rigid journal article format [14].

The fact that accounting research blending epistemological elements from “classic ANT” and the normal science tradition in the study of performativity has grown into a relatively substantial and firmly established body of scholarship suggests that accounting scholars, as a broader epistemic community, are rather tolerant of the blending of epistemologies. This might be seen as consistent with the naturalistic perspective on science informing our literature analysis, which implies that we should be more concerned with the scientific practices that become widely accepted within particular epistemic communities than with the establishment and policing of ideal standards for how such practices should evolve. To some extent, it may be also argued that the tensions resulting from the blending of epistemologies are alleviated by what may be described as an epistemological “strategy of compartmentalization”. In most of the papers included in our literature review, this strategy is manifest in the tendency of researchers to confine their theoretically informed reflections to the early and concluding parts while the empirical parts retain much of their descriptive and open-ended flavor and offer rich process descriptions. This helps to preserve the mode of reasoning that is typically seen as the hallmark of “good” ANT research, even though researchers combine this with an element of normal science thinking to develop research questions and advance theoretical contributions. This observation resonates with Lukka and Vinnari's (2017) argument that, inasmuch as accounting scholars merely employ ANT as an ethnographic method to, for instance, “follow the actors”, it can be relatively unproblematically combined with accounting domain theories and, in such a limited form, harnessed to support the typical normal science aim of continuously advancing theory development.

The adoption of a “strategy of compartmentalization” can at least partly alleviate the risk of research on the “performativity thesis” drifting too far from the epistemological roots of ANT and provides a bulwark against excessive compliance with the normal science tradition. While the seeds of such drift were perhaps sown already in Callon's (1998a, b) seminal works on the performativity of economics, the risk of this occurring may have been heightened by the tendencies to increasingly reconcile research on this topic with the ambition to advance cumulative theory development in economic sociology and organization studies (cf. Marti and Gond, 2018, 2019). This risk may be reinforced by the strongly ingrained normal science tradition, epitomized by the notion that “theory is king”, in contemporary accounting academia (Vollmer, 2009; Parker and Northcott, 2016; Richardson, 2018). This may, at worst, lead to the forced, artificial or superficial theorizing that several scholars have warned against (Hambrick, 2007; Avison and Malaurent, 2014; Gurthrie and Parker, 2017; Tourish, 2020). We see a particular risk of this occurring if the lack of explicit reflections on epistemological tensions, which is evident in our literature review, is indicative of an overly tolerant and eclectic posture on the part of accounting academics. We are not the first to raise such concerns in relation to the borrowing of method theories from cognate fields of research in the accounting literature. Indeed, concerns that researchers' nonreflexive borrowing of method theories from other disciplines can cause accounting research to evolve into a philosophically incoherent area of scholarship have been voiced by several commentators before us (e.g. Chua, 1988; Malmi and Granlund, 2009; Modell, 2015; Modell et al., 2017).

4.2 Implications

Before considering the broader implications of our study, it is useful to pause and reflect on the framing of our investigation so far. Although we have juxtaposed ANT with the normal science tradition, we do not wish to convey the impression that the latter is an objectively “better” way of undertaking research and that the former is an anomaly that goes against the grain of everything we have been taught to believe about science. Considering that normal science thinking is a social construction, can we really view it as a neutral, unproblematic take on science, which can be used as a superior yardstick for evaluating research? From a Foucauldian perspective, the answer would be a resounding “no” since normal science, as the word itself implies, is a form of normalization (Foucault, 1977) or a hegemonic project seeking to perpetuate itself by inculcating yet further generations of academics with its doctrines. In doing so, normal science thinking effectively invalidates other ways of knowing the world. While such thinking is no doubt a helpful approach to differentiating science from unsystematic inquiries and conspiracy theories driven by dubious interests, the relentless policing of the boundary between “proper” and “improper” science, that strict adherence to the normal science tradition tends to entail, may end up sidelining creative processes of problem formulation and solving. This is especially so since the current view of normal science is quite narrowly focused on producing theoretical contributions only, thereby relegating other aims of research to the penumbra of scholarly work.

Indeed, of the three intellectual virtues of Aristotle (see Flyvbjerg, 2001), the normal science tradition emphasizes seemingly value neutral theoretical development (episteme) over instrumental applications (tekhné) or research targeted at finding the right or wise course of action in a specific situation based on some form of value rationality (phronesis). While theorization can be perceived as academics' raison d'être, there is a risk that an excessive concern with theorizing might become an end in its own right. This underlines the need to problematize the notion that “theory is king” in contemporary accounting research. We therefore support the views of Lukka and Suomala (2014) regarding the multidimensional relevance of research whereby it is considered equally important for scholars to pursue other goals. Such wider goals include understanding or having an impact on mundane practices, or critically scrutinizing societally relevant matters of concern through evaluations based on both technical and value rationality with a view to complex issues related to the role of power in organizations and society. From this perspective, ANT's (de)constructivist approach (Latour, 2004b), with its seeming preference for inquiries focused on matters of concern and its animosity toward theorization can be viewed as an antidote to the excessive focus on theory and relative neglect of value-rational inquiries that are characteristic of the normal science tradition. One could even go as far as to speculate that ANT has been designed as an anti-program (Latour, 1990) to counter the attempts of normal science advocates to represent the collective views of all scientists.

In broader terms, it seems to us that the entire debate on theory fetishism may have become polarized. At one extreme, we have those who perceive the need for constant theoretical advances, regardless of how forced – if not even outright artificial, many theoretical contribution claims are these days. At the other extreme, some researchers may interpret the recent critiques around theory fetishism as implying that we should even abandon seeking theory contributions or at least notably relax such an emphasis. This view may well, at least partly, motivate those ANT scholars who follow the “classic ANT” take. However, what sometimes gets lost in the debate is that there are other options than compulsive, artificial theorizing on the one hand and ANT's complete dismissal of conceptual frames and categories on the other. These are only the two extreme ends of a continuum; the middle ground covers a variety of balanced approaches which allow scholars to pursue both original theoretical contributions to knowledge and societally relevant outcomes in one and the same inquiry. One such approach is the strategy of compartmentalization discussed further below.

Against the backdrop of these general contemplations on the normal science tradition and ANT, what are the implications of the present analysis for future accounting research? As the preceding sub-section has indicated, there are at least two main strategies for addressing the epistemological tensions between ANT and the normal science tradition. The first of these is to resist pressures to produce theoretical contributions and simply stay faithful to the highly empiricist epistemology of “classic ANT”. Such research may be helpful for developing research in a policy-relevant, phronetic, direction and is well-suited for advancing the mode of critique that is arguably distinctive of ANT (Baxter and Chua, 2017, 2020). Rather than relying on received, theoretically informed notions of what engenders undesirable states of inequality and injustice and how such states of affairs may be remedied, actor-network theorists arguably aspire to grounding social critiques in context-specific experiences and rely on the revelatory capacity of detailed empirical inquiries as a basis for envisaging how alternative realities might be created (Latour, 2004b; Law and Singleton, 2013, 2014). The studies by MacKenzie (2009) and Roberts and Jones (2009) can perhaps be seen as exemplars of such critique given that their primary concerns were with describing highly topical and politically charged issues rather than the advancement of theoretical contributions. This mode of critique differs markedly from the abstract and somewhat esoteric conceptions of critique that have arguably come to be associated with much critical accounting research and can thus help to rejuvenate such research (Dillard and Vinnari, 2017; Baxter and Chua, 2017, 2020).

The second possible trajectory for ANT-inspired accounting research is to continue along the path that is most evident in our literature review and persist in blending the epistemology of ANT with that of the normal science tradition. As indicated above, this blending of epistemologies started already with the works of Callon (1998a, b) and it should perhaps be seen as a development that has been embraced as a “natural” extension of ANT by scholars with a strong commitment to this method theory rather than a violation of some ideal standard for how it should be applied. However, as we have also made plain, we see a risk of the ambition to engage in continuous theory development beginning to encroach on the epistemological principles of ANT and possibly crowding out other intellectual virtues, such as those discussed by Flyvbjerg (2001), if the pressures to comply with the normal science tradition become too pervasive. If taken to the extreme, this may lead ANT-inspired researchers to privilege conceptual sophistication and concerns with the development of generalizable conclusions over the in-depth engagements with context-specific practices and complex processes that arguably constitute the hallmark of “good” ANT research. One way of reducing this risk might be that researchers would use ANT in a limited fashion, by adhering only to its basic principles as an ethnographic method, while combining it with elements of normal science thinking. As mentioned above, this is reminiscent of what we have referred to as a “strategy of compartmentalization” and, as we have seen, it can help to at least partly alleviate the epistemological tensions between ANT and the normal science tradition in research exploring the “performativity thesis”.

For researchers who wish to avoid the epistemological tensions between ANT and the normal science tradition in exploring the “performativity thesis” there is also a third option available, namely to abandon the former as a method theory and, instead, undertake analyses of how accounting becomes performative with the help of other method theories that are more conducive to continuous theory development. Although we have not analyzed this option in detail, we see an emerging tendency among accounting scholars to incorporate Callon's (1998a, b) or MacKenzie's (2006, 2007) conceptions of performativity into other method theories than ANT as a basis for advancing cumulative theory development (Ezzamel et al., 2012; Baker and Modell, 2019; Faulconbridge and Muzio, 2021). However, with the exception of Baker and Modell (2019), this has not been accompanied by explicit reflections on the epistemological implications of abandoning ANT and, unless it is combined with detailed and thoroughly contextualized process analyses, we see a danger of the risks that we have highlighted above being amplified. Although we are not necessarily averse to these emerging efforts to explore alternative ways of studying the “performativity thesis”, we suggest that when doing so, scholars would be wary of fetishizing the “theory is king” view and also pay sufficient attention to the context-specific intricacies involved in the processes through which accounting becomes performative.

Whether one of the three trajectories outlined above will come to dominate over time or whether they will continue to co-exist is currently an open question, which cannot be answered in isolation from deeper considerations of how contemporary accounting academia works as a broader epistemic community. It is quite likely that different approaches to research on the “performativity thesis” will continue to evolve and be deployed to different research tasks depending on the purpose at hand. Such fragmentation of research practice is not necessarily a problem and is, in some ways, compatible with the commitment to heterogeneity and pluralism in the broader inter-disciplinary accounting project (Lukka and Mouritsen, 2002; Baxter and Chua, 2003; Llewellyn, 2003). However, for accounting scholars who wish to remain entirely consistent with the epistemological principles of “classic ANT”, it implies a need for greater awareness of how far their knowledge claims can be stretched than what is evident in our analysis. As we have argued throughout this paper, upholding such consistency is challenging in the face of the pressures for compliance with the normal science tradition, which seem to be growing in contemporary accounting academia. While this challenge is especially pronounced in the case of ANT, it may also be pertinent to other method theories that are widely used in contemporary accounting research. Hence, similar to Modell et al. (2017), we close this paper by calling for greater and perhaps more explicit reflexivity on the part of researchers when it comes to the application of specific method theories to enhance our understanding of accounting as an organizational and social practice. We hope to have laid a foundation for such reflexivity by examining, as an example, the epistemological challenges of advancing theory development in research inspired by ANT and that this example can foster similar reflections on other method theories.

Table 1

A comparison of the epistemology of ANT and that of the normal science tradition

Epistemological dimensions	Epistemology of “classic ANT”a	Normal science epistemology
Over-riding view of the role of theory in research	Theory, other than ANT itself, is not expected to play a role as an input to, or output of, research	Theory plays a notable role in research both as an input and as an output. This may include an intention for cumulative theory development
Expected modes of reasoning	Open-ended, descriptive mode of reasoning that is not guided by any theoretical assumptions about which actors matter or ambitions to advance theoretically informed explanations	Theoretically informed modes of reasoning following the logics of deduction, induction or abduction
Strategies for developing research questions	Empirically derived issues that constitute “matters of concern” in specific contexts	Theoretically informed gap-spotting or problematization based on either deductively derived tensions and paradoxes or empirically derived construction of mysteries
Nature of research contributions	Application of ANT to specific empirical context without any attempts to generalize beyond this context or to make conceptual or theoretical contributions to extant research	Developing new theory or illustrating, testing or extending/refining existing theory

Source(s): ^aBased particularly on Latour (1987, 1990, 2005a) and Callon (1986, 1999)

Table 2

Summary of literature review

Study	Over-riding view of the role of theory		Employed mode of reasoning		Strategies for developing research questions		Nature of research contributions
	“Classic ANT”	Normal science	“Classic ANT”	Normal science	“Classic ANT”	Normal science	“Classic ANT”	Normal science
Christensen and Skærbæk (2007)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology relatively faithful to ANT	Possibly elements of all three normal science modes of reasoning. Particular emphasis on inductive, comparative generalization	No	Gap-spotting and problematization in relation to domain theory	No	Mainly illustration of method theory Perhaps some extension of domain theory
MacKenzie (2009)	No	No clear theory development on either input or output side	Methodology faithful to ANT	None of the normal science modes of reasoning applied	Yes	None of the normal science strategies applied	Yes	None of the normal science modes applied Contributions rather of a practical/policy nature
Millo and MacKenzie (2009)	No	Yes – on the input side in relation to both domain theory and method theory. On the output side more in relation to domain theory	Methodology relatively faithful to ANT (although methodology discussion is limited)	Mainly deductive but with some elements of abduction towards the end	No	Problematization in relation to domain theory Problematization of method theory later in the paper	No	Illustration and extension of method theory Extension of domain theory
Roberts and Jones (2009)	No	No clear theory development on either input or output side	Methodology relatively faithful to ANT but with some conceptual elements borrowed from other method theories	Limited use (if any) of the normal science modes of reasoning	Yes	Possibly an element of mystery construction (although theoretical significance of puzzle less clear)	Yes	None of the normal science modes applied Contributions rather of a practical/policy nature
Skærbæk (2009)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology is faithful to and explicitly anchored in ANT	Somewhat unclear but there seems to be an element of abduction (analytical generalization)	Not explicitly, although the methodology discussion may suggest otherwise	Gap-spotting in relation to domain theory	No	Extension and refinement of domain theory
Skærbæk and Tryggestad (2010)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology is faithful to and explicitly anchored in ANT	Largely deductive but with some indications of abduction	No	Problematization in relation to domain theory	No	Extension and refinement of domain theory
Kastberg (2014)	No	Yes –on input and (less so) on output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology is faithful to ANT (although methodology discussion is limited)	There seems to be a (limited) element of abduction (analytical generalization) although the paper reads like a straightforward illustration of method theory	No	Gap-spotting in relation to domain theory	No	Mainly illustration of method theory, but some attempt to advance more generalizable insights
Vinnari and Skærbæk (2014)	No	Yes –on input side in relation to domain theory (applying new method theory to extant domain theory) and on the output side both in relation to domain theory and method theory	Methodology is faithful to and explicitly anchored in ANT	There seems to be an element of abduction through analytical generalization	No	Gap-spotting in relation to domain theory	No	Extension of domain theory and method theory
Revellino and Mouritsen (2015)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology is faithful to ANT	Both induction (explicitly) and abduction (implicitly through analytical generalizations) applied	Yes	None of the normal science strategies applied (although some implicit problematization in relation to domain theory)	No (but somewhat ambiguous)	Refinement of domain theory and perhaps extension of method theory (including call for more research on failed performativity)
Ferreira (2017)	No	Yes –on input side as well as output side in relation to domain theory as well as method theory	Methodology relatively faithful to but not explicitly anchored in ANT	Both induction (explicitly) and abduction (implicitly through analytical generalizations) applied	Yes	Some problematization in relation to domain theory. Clear problematization of method theory through discussion of failed performativity	No	Extension of method theory through extensive discussion of why performativity fails and calls for further research on this topic
Georg and Justesen (2017)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology is faithful to and explicitly anchored in ANT	None of the normal science modes of reasoning applied	No	Gap-spotting and (limited) problematization in relation to domain theory	No	Illustration of method theory and some extension of domain theory
Cuckston (2018)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology relatively faithful to ANT (but no explicit methodology discussion)	Somewhat unclear, but there seems to be an implicit element of abduction	Yes	Some gap-spotting in relation to domain theory	No	Extension of domain theory
Jollands et al. (2018)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory)	Methodology is faithful to ANT	There seems to be an element of abduction (analytical generalization) and induction (comparative theory building from empirics)	Yes	Gap-spotting in relation to domain theory	No	Extension of domain theory Illustration of method theory (but this is not main contribution)
Themsen and Skærbæk (2018)	No	Yes – both on input and output side especially in relation to domain theory (applying new method theory to extant domain theory) Some attempts to contribute to method theory on output side	Methodology is faithful to and explicitly anchored in ANT	None of the normal science modes of reasoning explicitly applied (but still implicit element of abduction through analytical generalization)	No	Gap-spotting in relation to domain theory	No	Illustration and extension of domain theory Some extension of method theory
Warren and Seal (2018)	No	Yes –on input side both in relation to domain theory and method theory. Less clear on output side	Methodology shows element of ANT reasoning but is not explicitly anchored in ANT	There seems to be an element of abduction as well as deduction	Yes	Problematization of domain theory	Yes, partly, since much of the theorization is given in the early parts of the paper	Extension of method theory to critical performativity Failed performativity discussed as opportunity for further research
Cuckston (2019)	No	Yes – especially on input side in relation to domain theory. Less clear on output side but some critical ambitions in this regard	No	Abduction through analytical generalization	Yes	Problematization of domain theory	No	Extension of domain theory
Kastberg and Lagström (2019)	No	Yes – on input side in relation to domain theory Less clear on output side	Methodology faithful to and explicitly anchored in ANT	Primarily abduction	No	Problematization of domain theory	No	Extension and refinement of domain theory
Lassila et al. (2019)	No	Yes – on input and output side in relation to domain theory	Methodology somewhat faithful to ANT in the sense that the empirics are rich	Less clear but perhaps an element of abduction	No	Less clear but perhaps an element of gap-spotting in relation to domain theory	No	Illustration of method theory Extension of domain theory
Boedker et al. (2020)	No	Yes, on input and output side in relation to domain theory and method theory	Methodology faithful to ANT	Deduction and an element of abduction	No	Problematization and gap-spotting in relation to domain theory and method theory	No	Extension of domain theory and method theory
Pucci and Skaerbaek (2020)	No	Yes, on input and output side in relation to domain theory	Methodology faithful and explicitly anchored in ANT	Abduction through analytical generalization, and an element of induction	No	Problematization and gap-spotting in relation to domain theory and method theory	No	Extension and refinement of domain theory and illustration of method theory
McLaren and Appleyard (2020)	No	Yes, on output side in relation to domain theory. Less clear on input side	Methodology somewhat faithful to ANT in the sense that the empirics are rich	Abduction through analytical generalization, and an element of induction	Yes	None of the normal science strategies applied	No	Extension of method theory

Table A1

Specific indicators of the epistemology of ANT and that of the normal science tradition used in the coding of papers

Epistemological dimensions	Indicators of “classic ANT” reasoning	Indicators of normal science reasoning
Over-riding view of the role of theory in research	Are there any statements in which the authors reject existing domain theory as an input to their research? Are there any statements rejecting theory advancement as the output of the research?	Are there any statements regarding the authors' theory development ambitions based on using theory as an input or expecting theoretical advancement to occur as an output of the research?
Employed modes of reasoning	Do the authors follow an open-ended mode of reasoning (e.g. by “following the actors”) primarily aimed at generating rich empirical descriptions?	Do the authors engage in any attempts to derive hypotheses/predictions from extant domain and/or method theory (deduction), generalize from “grounded” empirical observations to either type of theory (induction), or combine the two types of theorizing (abduction)?
Strategies for developing research questions	Do the authors mainly rely on some empirically derived topic, pivoting on potentially controversial issues (“matters of concern”), to motivate their research questions/purpose?	Do the authors motivate their research questions (or position their purpose) in relation to some claimed theoretical lacuna (gap-spotting) or through theoretical problematization (in relation to extant domain and/or method theories) that is derived either deductively from tensions or paradoxes in the existing literature or starting from an observed empirical surprise/puzzle/mystery?
Nature of research contributions	Do the authors refrain from discussing their findings as (domain or method) theoretical contributions, i.e. are they content to conclude their research by simply summarizing the empirical findings?	Do the authors claim to have advanced any new domain and/or method theory (developing new theory)? Do the authors claim to have tested (theory testing) or illuminated (theory illustration) extant domain and/or method theory? Do the authors propose any additions to (extension) and/or revisions/fine-tuning (refinement) of extant theory or concepts in relation to the domain and/or method theories?

Notes

1. While the notion of “normal science” is often associated with Kuhn's (1962) conception of science as progressing through continuous theoretical puzzle-solving within firmly established research paradigms until challenged by competing paradigms, our use of this term is somewhat broader and denotes all research in which theory plays a role as an input to and/or as an output of the research process. Hence, our notion of normal science thinking refers to all research with a clearly articulated theory development ambition, recognizing that such theory development may evolve in a relatively incremental, cumulative manner, as well as through more radical breaks with extant, theoretical knowledge (see also Vollmer, 2009).

2. By characterizing ANT as “a-theoretical” we by no means intend to say that it is not a theory in its own right. It certainly does provide a set of conceptual building blocks, that are specific to this school of thought, as a basis for making sense of the world. However, there seems to be relatively broad consensus that such conceptualizations mainly aim at capturing particular phenomena as they evolve in specific empirical contexts rather than engaging in any forms of cumulative theory development that are characteristic of the normal science tradition (see Sandberg and Alvesson, 2021).

3. Lukka and Vinnari (2014, p. 1309) define domain theory as “a particular set of knowledge on a substantive topic area situated in a field or domain such as accounting” and method theory as “a meta-level conceptual system for studying the substantive issue(s) of the domain theory at hand”. In addition to economics and psychology, accounting scholars have applied method theories originating from, for instance, organization studies and sociology.

4. Modell et al. (2017) mainly documented how ANT-inspired arguments have been incorporated into accounting research using institutional theory and their analysis was not centrally concerned with the allegedly a-theoretical nature of ANT or how actor-network theorists theorize.

5. By “naturalistic” we refer here to a current stream in the philosophy of science (“naturalism”) characterized by its aim to study science as it is practiced by scholars rather than as it should be according to some normative epistemological rules (Giere, 2000; Kakkuri-Knuuttila et al., 2008).

6. We wish to emphasize that the purpose of our paper is not to promote or demote either normal science or ANT; our focus is on the ways in which ANT-inspired accounting scholars deal with the epistemological differences of the two approaches.

7. Law (2009, p. 142) utilizes the term “actor-network theory 1990” to refer to the version that “tends to find its way into textbooks”. In accounting, according to Justesen and Mouritsen (2011, 2018), the vast majority of ANT-informed research draws inspiration from Latour's works, particularly Science in Action (Latour, 1987).

8. The etic perspective refers to studying human behavior from outside the system, while its counterpart, the emic viewpoint, refers to studying it from the inside (Pike, 1954; Jönsson and Lukka, 2006).

9. Our view of theorizing is hence pluralistic and inclusive, comprising conceptualization, propositional, configurational, process, perspectival and meta-theorizing as well as theoretical provocation and critical meta-theorizing (Cornelissen et al., 2021). However, naturally our definition of theory – even if it is very broad – follows the normal science tradition as defined in this paper, not that of “classic ANT”.

10. Our analysis does not include papers that incorporate Callon's (1998a, b) and/or MacKenzie's (2006, 2007) notions of performativity into other method theories, such as neo-institutional sociology (Ezzamel et al., 2012; Faulconbridge and Muzio, 2021), Gramscian critical theory (Cushen, 2013) and critical realism (Baker and Modell, 2019), since these papers are less likely to reflect the epistemological tensions of interest to us.

11. The journals included in our search are Accounting, Auditing and Accountability Journal, Accounting and Business Research, Accounting, Organizations and Society, Contemporary Accounting Research, Critical Perspectives on Accounting, European Accounting Review, Management Accounting Research and Qualitative Research in Accounting and Management.

12. Encyclopedia Britannica, https://www.britannica.com/topic/eclecticism. Accessed 25 March 2022.

13. We thank David Cooper and one of the anonymous reviewers for this observation.

14. We are grateful to one of the anonymous reviewers for this point.