Clarifying the Dominant Logic Construct by Disentangling and Reassembling its Dimensions

Since its introduction, Prahalad and Bettis's concept of dominant logic has informed a variety of scholarly conversations in management and strategy research. However, scholars have interpreted dominant logic in different ways, emphasizing different aspects, such as managerial mindsets, administrative tools and management functions, as defining elements. Similarly, empirical studies have captured various aspects, such as meanings of entrepreneurs, observable strategic decisions and business model similarity, as indicators of dominant logic. Consequently, the concept lacks analytical clarity, and it is difficult to compare or generalize findings from this diverse set of studies. The aim of this review is to improve conceptual clarity by analysing, comparing and evaluating the existing interpretations and assessments of dominant logic in 94 studies. In the first part of the review, by disentangling the interpretations of the concept, we show that dominant logic consists of four defining dimensions: (i) shared mental models; (ii) values and premises; (iii) organizational practices; and (iv) organizing structures. In the second part, we reassemble dominant logic into an integrative model and theorize about how these dimensions operate in concert to produce a firm's dominant logic. Thus, our main contribution is a clarification and synthesis of the literature, which comes with implications on how future research can conceptualize and operationalize dominant logic more consistently.


Introduction
More than three decades ago, Prahalad and Bettis (1986, p. 491) introduced the concept of dominant logic, which they defined as 'a mindset or a worldview or conceptualization of the business and the administrative tools to accomplish goals and make decisions in that business'. Their aim was to complement the extant economic perspectives on the relationship between diversification and firm performance by adding a cognitive explanation. Since then, the concept has been further refined (e.g. Bettis and Prahalad 1995;Bettis et al. 2011;Côté et al. 1999;von Krogh and Roos 1996) and associated with a diverse set of variables, such as technological change (Zyglidopoulos 1999), entrepreneurial identity construction (Downing 2005), joint venture success (Guidice and Mero 2007) and dynamic managerial capabilities (Kor and Mesko 2013). Empirically, dominant logic has been linked with a variety of outcomes, such as strategic acquisitions (Côté et al. 1999), joint venture decisions (Lampel and Shamsie 2000), entrepreneurial firm performance (Obloj et al. 2010) and attention to stakeholders (Crilly and Sloan 2012). Moreover, methodological reflections have addressed opportunities to empirically operationalize dominant logic (e.g. Laukkanen 1994;Phillips et al. 2008;Schraven et al. 2015). All these studies 324 A. Engelmann et al. underscore the richness of the concept and its impact on the scholarly conversations in management and strategy research.
What is striking when examining the literature, however, is the plethora of interpretations and empirical approaches to investigating dominant logic -although most researchers have used Prahalad and Bettis's (1986) original definition. For example, some researchers interpret a firm's dominant logic as 'habitual modes of functioning based on prior successes and failures ' (Côté et al. 1999, p. 921); others describe it as 'a system of expectations, beliefs, and priorities that are embedded in the firm's routines, procedures, and resource commitments' (Kor and Mesko 2013, pp. 235-236); still others suggest that it consists of sensemaking, choices, learning and routines (Obloj et al. 2010). Similarly, empirical studies include different phenomena, such as the decisionmaking patterns of globally distributed business units (Lampel and Shamsie 2000), shared orientations toward childcare in firms (Kossek et al. 1994) and business model similarity between subsidiaries and headquarters (Monteiro 2015) -all of which are interpreted as indicators of a firm's dominant logic.
The success of the concept is not surprising, given its intuitive nature, providing scholars and practitioners with an appealing term for describing an organization's unique ways of operating. However, the concept also comes across as vague and ambiguous. It seems that researchers have seized upon the room for interpretation that Prahalad and Bettis's (1986) original definition left, by choosing at their own discretion what actually constitutes dominant logic (cognitive schemas, management tools, attitudes, etc.) and where to look for it empirically (verbal statements of managers, annual reports, investment decisions, etc.). The obvious drawback is a lack of shared understanding of dominant logic, which impairs both its analytical clarity as a theoretical concept and its predictive power in empirical studies. Because dominant logic continues to inform important conversations, for example in entrepreneurship (Su and Wang 2018), family businesses (Lumpkin and Brigham 2011) and strategic resource management (Combs et al. 2011;Matysiak et al. 2018), it is particularly timely to elucidate the extant perspectives on the concept and its inner workings.
In this paper, we carry out a systematic literature review to take stock of the vast number of theoretical and empirical studies building on Prahalad and Bettis's (1986) original ideas. Our main purpose is to disentangle how scholars have used dominant logic by identifying, classifying and interrelating existing interpretations and assessments of the concept. Our systematic analysis of the interpretations reveals more than 70 concepts referring to dominant logic. These can be categorized into four constituent dimensions of dominant logic that inform us about where it can be found empirically: (i) shared mental models; (ii) values and decision premises; (iii) organizational practices; and (iv) organizing structures. A closer look shows that all of them are covered by Prahalad and Bettis's (1986) original definition. Hence, even if the existing terminological confusion would suggest the opposite, the interpretations and corresponding empirical approaches seem to revolve around a coherent core. To elaborate on this core, we depart from our analysis of the literature and reassemble the four dimensions into an integrative model of dominant logic.
Our main contribution is a clarification and synthesis of the literature, as well as a revised conceptualization of dominant logic that is in line with its original definitions and which would facilitate further theoretical development. By delving into its cognitive and manifested aspects, we go beyond previous perspectives on dominant logic by theorizing about how exactly the identified dimensions can be integrated. We argue that, because the dimensions operate in concert, alignment across the four dimensions is an integral feature of dominant logic -an aspect that has been neglected in previous research. Building on this argument, we call for a more rigorous application of the concept in empirical studies and provide concrete suggestions on how to capture the four dimensions of dominant logic, both separately and in combination. Taken together, our synthesis of the literature facilitates a more theoretically and empirically sound application of dominant logic in the future and contributes to elevating the concept from a vague metaphor to a more tangible construct.
Before we describe the processes of literature review and analysis, we begin with a brief introduction to dominant logic. The main body of our paper is split into two parts. First, we disentangle dominant logic into the four defining dimensions and describe the ways in which the concept has been captured empirically. Second, we reassemble the dimensions and provide an integrative model of dominant logic. We then discuss implications for theory and empirical assessment, as well as avenues for future research.

The concept of dominant logic
The origin of the concept of dominant logic lies in strategic cognition, a field that focuses on the linkage of organization members' cognitive structures with strategic choices and actions (for reviews, see Kaplan 2011;Narayanan et al. 2011). Along with other researchers in the 1980s and 1990s (e.g. Porac et al. 1989;Reger and Huff 1993;Walsh 1995), Prahalad and Bettis (1986) explored ways in which economic approaches to strategic issues (such as firm diversification) can be enriched with an additional cognitive perspective. Therefore, they introduced the dominant logic as 'the way in which managers conceptualize the business and make critical resource allocation decisions' (Prahalad and Bettis 1986, p. 490). More precisely, they argued that the difficulties of managing diversified firms not only arise from industry structures or the number of distinct businesses, but also from organizations' cognitive structures, which are determined by the knowledge and experience of the corporate management team. These cognitive structures are not purely 'invisible' in the sense of latent managerial thinking; they are also embodied in a firm's 'visible' infrastructure and administrative tools, such as the choice of key individuals and processes of budgeting, control or compensation (Bettis and Wong 2003;Grant 1988). Hence, the notion of dominant logic covers both invisible (cognitive) and visible aspects.
In their 1995 article, Bettis and Prahalad published a 'retrospective and extension', where they highlighted the role of dominant logic as an information filter and provided a new theoretical grounding within the theory of complex adaptive systems. This broader view of dominant logic detached the concept from the context of diversification and suggested that it is an emergent property of human systems, predisposing a firm to certain kinds of strategic problems and interacting 'with organizational systems and structures in a complex way' (Bettis and Prahalad 1995, pp. 8-9). Although cognition remained an important carrier of dominant logic, this quotation also highlights its embeddedness in practices and structures.
Von Krogh and Roos (1996) critically reflected on Bettis and Prahalad's development of the concept and added two important attributes to bolster its system-theoretical underpinning. The first attribute, self-reference, means that social systems tend to interpret upcoming issues by relying on prior knowledge and experience. This tendency makes the dominant logic enduring and history-dependent. The second at-tribute, scale, implies that the dominant features of a particular logic can be found at different levels within an organization (e.g. individual, group, unit, organization). One important aspect related to scale is self-similarity, meaning that similar patterns become visible across different scales.
The idea that human systems operate based on a shared, underlying logic is not unique to Prahalad and Bettis's concept, but is also at the core of the institutional logic view (Friedland and Alford 1991;Thornton et al. 2012). According to this perspective, societal institutions (e.g. family, religion, government, market) shape how social actors interpret the world due to shared 'underlying assumptions, deeply held, often unexamined, which form a framework within which reasoning takes place' (Horn 1983, p. 1). The institutional logic and the dominant logic views approach a similar phenomenon from different perspectives. The institutional logic view assumes that firmlevel logics originate from larger systems of meaning, such as the state, markets or professions (Thornton et al. 2012). This perspective is primarily concerned with field-level logics, such as the fields of finance (Lounsbury 2002), healthcare (Scott et al. 2000) and higher education (Washington and Ventresca 2004). In contrast, the dominant logic perspective, in Prahalad and Bettis's sense, starts from an individual firm's shared cognitions and its history, chosen paths and experiences with success and failure. Indeed, a dominant logic is partly derived from institutional features, but it is mostly firm-specific, meaning that certain elements of a concrete firm will be similar across firms in the same industry (Spender 1989), but other elements will be individual to the firm (Côté et al. 1999). Hence, dominant logic is a phenomenon that originates from managers' cognition and emerges within organizations.
Many scholars in the field have contributed to improving the precision and rigour of the notion of dominant logic, be it through clearer theorizing (Bettis and Wong 2003;Boisot and Li 2005;von Krogh and Roos 1996), sound empirical studies (e.g. Bouwen and Steyaert 1990;Côté et al. 1999;Obloj et al. 2010) or methodical considerations (e.g. Laukkanen 1994;Schraven et al. 2015;van Rekom et al. 2006). Due to a lack of shared understanding, however, the literature on dominant logic is still ambiguous. The aim of this review is to discern the perspectives that exist and provide an integrative perspective on dominant logic. Hence, we ask: How is dominant logic interpreted and operationalized in the literature?

Process of the literature review
To review how scholars have interpreted and operationalized an organization's dominant logic, we performed a systematic literature review (Tranfield et al. 2003) in the ABI/INFORM Global, Web of Science (WOS), SCOPUS and EBSCO databases. In these databases, we input the query 'dominant logic*' and searched across titles, abstracts and keywords. Because the query yielded many papers from the servicedominant logic perspective in the field of marketing (Vargo and Lusch 2004), we adapted the query to exclude these papers. After merging duplicates, we obtained 503 texts (see Figure 1).
As shown in Figure 1, before elaborating on the content, we applied formal criteria for exclusion (language other than English, proceedings/working papers/announcements), leading us to remove 38 articles. The sample of the remaining 465 articles was manually refined in two steps. In the first step, by reviewing the titles and abstracts, we observed that 295 studies were from a different subject area (e.g. biology, clinical psychology, politics, programming), leading to a straightforward exclusion.
In a second step, we analysed the full texts of the remaining 170 articles. Along the way, we defined explicit content-related exclusion criteria. First, many studies used the term 'dominant logic' in a metaphorical sense (e.g. as a substitute for paradigm), without referring to the original concept. Examples of this include a dominant logic of research (Woodside and Baxter 2013), of executive education (Lockhart 2013) or of the international disability field (Meyers 2014). Another set of excluded studies came from different fields in organization, management or strategy research, for instance, research on organizational boundaries (Santos and Eisenhardt 2005), category divergence (Alexy and George 2013) and conceptualizations of innovation (D'Auria et al. 2017). We excluded these articles because dominant logic was not at the core of the studies and was used only to support particular arguments. Finally, we excluded studies that employed an institutional view on logic. We discarded these texts because they did not focus on managerial agency, but instead on more abstract, industry-specific belief systems. After employing these three content-based exclusion criteria, 94 articles remained that explicitly addressed the dominant logic of individual firms ( Table 1).

Review of theoretical conceptualizations of dominant logic
To answer the question of how dominant logic is interpreted in the literature, we employed in-vivo coding (Saldana 2009) of the definitions and operationalizations in all 94 (i.e. conceptual and empirical) papers on dominant logic. This coding step revealed more than 70 different concepts. To impose order among them, we applied a card sorting technique (Rugg and McGeorge 2005): we wrote each conceptualization on a card and sorted these cards into higher-order dimensions that should cover all conceptualizations. Whenever a 'miscellaneous' category remained, we skipped the categories and started again. After three trials, all of the cards were sorted into four meaningful higher-order dimensions (see Table 2): (i) shared mental models; (ii) values and premises; (iii) organizational practices; and (iv) organizing structures.  (2000) 1 Articles in which dominant logic is embedded in the entire narrative of the paper. 2 Articles in which dominant logic is part of a theoretical argument, and/or used to deduce hypotheses, and/or used to discuss findings, but not used in the entire narrative of the paper. * Theoretical foundations, refinements and discussions of dominant logic. † Articles that suggest and/or discuss methods particularly suitable for capturing dominant logic.

Review of empirical approaches to capture dominant logic
To answer the question of how dominant logic has been treated empirically, we started from the 58 empirical studies in our sample and analysed the interpretations and operationalizations of dominant logic therein. This analysis revealed that only 28 of the 58 empirical studies explicitly operationalized dominant logic (see Table 3). In the other 30 studies, the concept was used as a theoretical backing for particular arguments (e.g. for hypothesis development). For example, Leiponen and Helfat (2010) used dominant logic to argue that firms tend to search narrowly for new innovations. However, they assessed innovation objectives (e.g. improve product quality) and knowledge sources (e.g. competitors) to explain variance in innovation success, but did not operationalize dominant logic in their empirics. We found a similar usage of dominant logic in 30 of the empirical studies (superscripts in Table 1 et al. 2000) and those that explore the dominant logic of a certain firm (e.g. Apple Inc., Boivin and Roch 2006) or specific context (e.g. entrepreneurial firms in China; Obloj et al. 2013). Moreover, across all four dimensions of dominant logic, we find two broad empirical strategies: inductive and deductive (see column (c) in Table 3). Inductive studies favour interpretative methodologies, particularly case studies, and collect mainly verbal data, such as formal and

340
A. Engelmann et al. informal interviews or managerial statements in public settings (e.g. speeches, press), in annual reports or on firm websites. The authors in this tradition usually portray dominant logic as a holistic attribute of a human system comprising patterned cognitive and manifested components. Deductive studies usually start from an operational definition of dominant logic, which is then quantitatively measured, either through surveys or through more objective data. These studies seek to find explicit indicators of dominant logic to relate them to other variables, such as performance measures.
In all 28 empirical studies, human actors produce, maintain and/or change dominant logic. However, studies have focused on different agents (i.e. who carries a dominant logic), ranging from individuallevel (e.g. entrepreneurs, HR managers, CEOs) to collective-level actors (e.g. management teams, business units, firms). Researchers who focus on cognitive structures or values and premises argue that dominant logic can be found in invisible, subjective rationalities. They employ either inductive strategies based on verbal data (e.g. interviews, verbal statements of managers in press articles or on websites) or deductive, survey-based instruments. In contrast, scholars who investigate the more visible dimensions of dominant logic, that is, practices and organizing structures, argue that dominant logic can be found in objective (visible or measurable) manifestations, such as similarities in cost structures or business models. Accordingly, they mostly employ deductive, quantitative strategies.

Disentangling interpretations and empirical assessments of dominant logic
We now present the four dimensions of dominant logic that we found in the literature and delineate the theoretical rationale for each of the perspectives. Regarding the conceptualizations, we build on the coding and categorization of the entire sample of 94 papers. For each dimension, we provide a brief definition and illustrative examples of the terms and concepts the authors used. Then, we discuss the empirical approaches that were employed to capture dominant logic. As can be seen from column (e) in Table 3: 21 of the 28 studies employed methods to capture only one of the four dimensions of dominant logic; 4 out of 28 combined two dimensions; and 3 out of 28 combined three dimensions of dominant logic in their empirics.
Hence, a few of the quoted studies will recur in the empirical (i.e. 'Capturing . . . ') sections of different dimensions.

Dominant logic as a shared mental model
Conceptualization. This dimension is the closest to the main definitional component of dominant logic as 'the way in which managers conceptualize the business' (Prahalad and Bettis 1986, p. 490). In cognitive psychology, mental models are defined as 'organized packets of information about the world, events, or people, stored in long term memory' (Eysenck and Keane 2005, p. 564).
Examples of interpretations of dominant logic as shared mental models include terms such as mindset or worldview (Prahalad and Bettis 1986), meanings (Bouwen and Steyaert 1990), cognitive styles to frame problems (Bouwen and Fry 1991), dominant patterns of thinking (Laukkanen 1994), set of shared schemata (Zyglidopoulos 1999), information filter (von Krogh et al. 2000), frames of reference (Downing 2005) and belief systems and mental models (Kor and Mesko 2013). As these examples show, this category includes definitions of dominant logic as mental representations of 'the world' and of 'how things are'.
Operationalization. Although Grant (1988) argued that it might be difficult to empirically investigate dominant logic because cognition is hard to capture, researchers have developed rigorous approaches to investigate a firm's dominant logic as a shared mental model. More explicitly, Laukkanen (1994) operationalized a firm's dominant logic (as 'dominant patterns of thinking') through a multistage procedure of causal mapping that allowed for identifying and visualizing the managers' cognitive maps of a domain. Similarly, Schraven et al. (2015) developed a cognitive mapping technique including statistical tests and graphical approaches to account for the two main characteristics of dominant logic that were highlighted in von Krogh and Roos's (1996) conceptual article: self-reference and scale. Specifically, Schraven et al. (2015) proposed that the only means-end links that are based on dominant logic are those that: (i) are reflected in multiple responses by one and the same respondent; and (ii) re-occur across multiple respondents at different positions within the same firm. However, Laukkanen's (1994) and Schraven et al.'s (2015) methods were rarely used in empirical research on dominant logic. Crilly and Sloan's (2012) analysis of annual reports from eight enterprises is one of the rare exceptions that applied causal mapping.
Whereas scholars in the cognitive mapping camp recommend employing specific structured interview techniques, Phillips et al. (2008) suggested identifying a firm's dominant logic by means of critical discourse analyses. They argued that 'organizational logics can be studied as cases of intersubjective meaning-making that use the discursive resources from outside the organization to achieve consensus around its strategic ends and the means adopted to achieve those ends' (Phillips et al. 2008, p. 776). However, we are only aware of one empirical study that employed discourse analysis as a tool for capturing dominant logic as a shared mental model. Hadida and Paris (2014) coded mission and vision statements of 21 new ventures in the digital music industry to examine the discourses of entrepreneurs.
Instead, most studies in this category employed interpretive approaches that follow one of two strategies. First, some have used dominant logic as a 'container' that can be filled with empirical content; this content is then regarded as 'the dominant logic of . . . '. Examples of this include: 'entrepreneurial firms in China' (Obloj et al. 2013); 'strategic practices of further education colleges' (Smith et al. 2002); 'innovation projects' (Bouwen and Fry 1991); 'a large broadcasting company' (Maijanen 2015b); and 'the Belgian fashion industry' (Jacobs et al. 2016). These authors mainly used case study designs based on interviews.
The second interpretive strategy to capture dominant logic as a shared mental model is to first dimensionalize it and then associate its dimensions with specific outcomes. For instance, von Krogh et al. (2000) derived internal (people, culture, product and brand) and external (competitor, customers, consumer and technology) categories from annual reports, which -in their view -fully capture dominant logic as a shared mental model; subsequently, they created a numerical measure of the bandwidth of these categories and related it to performance data from Nokia and Erikson. Another example is the study by Ellonen et al. (2015). To capture the dominant logic of four divisions of a large media corporation, Ellonen and colleagues interviewed managers, focusing on their perceptions of the role of print, of the business that they are in, and of the industry and its effects on the corporation. Then, they associated these different perceptions in each area with the divisions' sensing, seizing and transforming capacities. Côté et al. (1999), in their single-case study, elicited cognitive concepts from annual reports (i.e. strength in management of large projects, multiculturalism and Canadian identity) in a large engineering company and linked them with that company's acquisition choices and management approaches.
In summary, in this category, the agents of dominant logic are typically top managers and entrepreneurs, or top management teams. This is in line with the idea that cognition, even if it is shared among multiple actors, is bounded to individuals. When researchers interpret dominant logic as shared mental models, they must find ways to capture latent cognitive structures. Whereas explicit methods have been proposed to do so (e.g. Laukkanen 1994;Schraven et al. 2015), scholars often employ inductive strategies and collect verbal material (including interviews, as well as existing speeches, annual reports, text on websites, etc.) to infer a firm's dominant logic. The interpretive studies differ with regard to the degree of 'induction'. Some choose a priori categories to pin down dominant logic (e.g. Ellonen et al. 2015), while others study dominant logic more inductively (e.g. Côté et al. 1999). Our analysis reveals that in interpretive procedures, scholars do not always specify what exactly they regard as dominant logic (for an exception, see Côté et al. 1999). Instead, they often seem to 'freely choose' those contents from their empirical data that are then referred to as shared mental models. Moreover, studies in this category usually argue that dominant logic primarily surfaces in cognitive components, irrespective of whether they label them as concepts, schemes or mental models, and what they truly focus on (e.g. patterned action). These studies in particular would benefit from a more rigorous differentiation of mental models and other constituents of dominant logic (i.e. values and premises, practices and structures) to prevent an arbitrary blending of their data.

Dominant logic as values and premises
Conceptualization. This category comprises interpretations that focus on dominant logic mainly as a vehicle for conveying ideas and beliefs of 'how the world should be'. Values can be defined as 'central and enduring tenets of the organization' (Collins and Porras 2005, p. 73) or as 'concepts or beliefs that pertain to desirable end states or behaviours, that transcend specific situations, and guide selection or evaluation of behaviour and events' (Schwartz 1992, p. 4).
Examples include views of dominant logic as a set of unseen assumptions (Bettis and Prahalad 1995), local meaning framework based on historically rooted values (Bouwen and Fry 1991), set of premises (Lampel and Shamsie 2000), core values (van Rekom et al. 2006), institutionalized beliefs (de Holan 2011), meta-heuristic (Lumpkin and Brigham 2011), template that defines values and beliefs (Vardaman et al. 2012) or fundamental strategic beliefs, assumptions and intentions of the CEO and senior management (Kor and Mesko 2013).
Values can be general (e.g. motivate people, achieve a good result; van Rekom et al. 2006) or specific to a topic. For instance, specific values may concern beliefs about the necessity of childcare initiatives within an organization (Kossek et al. 1994) or the need for flexibility and individual autonomy versus standardization and formalization in patient care (Vardaman et al. 2012). As such, values build the underlying determinants of decision-making preferences and thus serve as -mostly unconscious -decision-making premises. For example, they can mediate the interdependent choices of a firm's strategic orientation and its managerial resource development and deployment practices (Kor and Mesko 2013). This relationship with decision-making is also reflected in Prahalad and Bettis's original definition, describing dominant logic as 'the way in which managers conceptualize the business and make critical resource allocation decisions' (Prahalad and Bettis 1986, p. 490; emphasis added). An example that explicitly emphasizes the role of dominant logic for decision-making is Lumpkin and Bringham's (2011) description of long-term orientation as a higher-order heuristic, in that it assists in filtering information during strategic decision-making in family firms.
Operationalization. Similar to the cognitive mapping approaches, van Rekom et al. (2006) presented a sophisticated method for capturing dominant logic as core values and premises. Specifically, they suggested a means-end analysis, an interview technique that aims to identify important ends (i.e. goals) within an organization and enquires into the means necessary to achieve these ends. The result is a collection of means-end statements. Starting from the idea that core values define what is considered a desired end, van Rekom et al. (2006) further argued that ends which are important to an organization will have multiple means-end paths leading both toward them and away from them. Hence, their method starts with an explorative phase to identify means-end re-lations. In a second step, based on these means-end relations, a standardized questionnaire is presented to the firm's members. The responses enable a calculation of the relative dominance of each of the values in the firm's value system. However, despite its rigour, none of the studies addressing dominant logic empirically have used van Rekom et al.'s (2006) method.
Instead, most empirical methods in this category have built on the assumption that values and premises surface as core orientations, that is, as organization members' 'usual, general, or lasting direction of thought, inclination, or interest' (Merriam-Webster Online Dictionary, cited in Covin and Lumpkin 2011, p. 857). To capture such orientations, scholars have collected cross-sectional survey data. Kossek et al. (1994), for example, assumed that particular configurations of the dominant logic of human resource (HR) managers lead to the adoption of childcare initiatives. They captured that overall dominant logic with an idiosyncratic item scale that operationalized the organization members' overall orientation toward childcare. Likewise, Garg et al. (2003) described and operationalized dominant logic as orientations toward either the internal (i.e. innovation vs. efficiency) or the external environment (i.e. general vs. task) in which manufacturing firms operate. Similarly, Su and Wang (2018) viewed dominant logic as being the result of an entrepreneurial orientation: a specific 'entrepreneurial mindset' of new ventures that affects how they make decisions.
In summary, in this dimension of values and premises, dominant logic is assumed to be located within human agents. To capture a firm's (shared) dominant logic as values and premises, scholars have mostly employed straightforward deductive approaches, such as surveys, that operationalize dominant logic as concrete orientations 'toward something'. Interestingly, empirical studies in this category often shift the agent of dominant logic from an individual or managerial actor to a more collective level, by surveying constructs relevant for the whole firm, such as firm-level proactiveness (e.g. 'Our firm tries to influence directions of changes'; Obloj et al. 2010, p. 165). The threat of capturing dominant logic as an orientation is that the two concepts are quasi-equated. For example, Kossek et al. (1994), who argued that dominant logic manifests as values and premises, defined a firm's orientation toward childcare as its dominant logic. The obvious advantage is that dominant logic then lends itself to be the theoretical backing for that specific orientation (toward childcare, toward the internal or external environment, etc.). However, if the dominant logic is considered to be the same as an organization's orientations, then the value added by the dominant logic as a theoretical concept is lost.

Dominant logic as organizational practices
Conceptualization. This category accounts for Prahalad and Bettis's (1986, p. 490) idea that dominant logic is 'a set of elicited management processes', or a form of process knowledge, and is also in line with the view that cognition should be considered as embedded in organizational practice (Bettis et al. 2011;Burgelman et al. 2018;Nicolini 2011). Organizational practice has been defined as 'routinized types of behaviour including bodily and mental activities, and the materials and tools used to carry out these activities' (Reckwitz 2002, p. 249). Accordingly, this category comprises all definitions that link dominant logic with patterned, recurring actions of organizations and its members, mainly in the shape of organizational processes, standard operating procedures, routines or recurring 'ways of operating'. Examples of interpreting dominant logic as organizational practice include Bouwen and Fry's (1991) interpretation of dominant logic as managerial practice, Côté et al.'s (1999, p. 40) description of it as 'entrenched habitual modes of functioning based on prior successes and failures' and Verbeke's (2010, p. 40) view of dominant logic as 'stable patterns of decisions and actions that coordinate the productive use of resources'.
Operationalization. Although many of the empirical studies in the literature view a firm's dominant logic as recurring organizational practices (e.g. Boivin and Roch 2006;Obloj et al. 2010Obloj et al. , 2013, none of these studies empirically assessed actions or routines in a process study (as described by Langley et al. 2013). Instead, organizational practices are mostly inferred from retrospective interviews or are queried through surveys. One typical example of the former is the single-case study conducted by Bouwen and Steyaert (1990). The authors interviewed the founder and key employees of Newcom, a new venture, to understand how meanings (e.g. doing exciting and independent work, being a salesman) and corresponding actions (e.g. opportunity seeking, planning) affected the development of the young firm. They coded verbal transcripts and documents and found tensions between an opportunistic (i.e. sales and client orientation, technical competence) and an emerging bounding logic (i.e. controlling the inner workings of the firm) that were resolved through dialogue.
In a different approach, Lampel and Shamsie (2000) investigated how a firm's dominant logic manifested as recurring organizational practice. They argued that a mission statement communicated by a former CEO of General Electrics (i.e. being the number one or two player in the focal market) represents the firm's dominant logic, since it shapes the corporation's joint venture decisions and designs. They then identified 70 joint ventures undertaken by the business units of General Electrics between 1984 and 1993 and assessed whether the business units restricted joint venture efforts with large partners, instead preferring joint ventures with small, national partners to adhere to their dominant logic of remaining the number-one player. Similarly, Boivin and Roch (2006, p. 412) employed a qualitative historical analysis -'a process of assembling, critically examining and summarizing the records of the past' -to extract dominant logic. More precisely, they assessed the prevailing logics of Apple Inc. with regard to its experiences with collaboration. From their qualitative analysis, Boivin and Roch concluded that Apple's inspirational logic impeded the firm from successfully establishing internal and external strategic alliances.
Finally, survey-based approaches have been employed in this category. Obloj et al. (2010, p. 165), for example, used survey items such as 'We develop efficient procedures in the early stage of our firm's operation' and 'Important pieces of information mainly pass through formal channels in our firm' to capture dominant logic as enacted cognition.
In summary, in line with Nicolini's (2011, p. 603) idea that shared cognition 'is located within the relationships between the people participating in specific practices', the location of dominant logic is an organizational practice. Accordingly, the agent of dominant logic is typically a collective actor (e.g. a business unit, an entrepreneurial firm, Apple Inc.) who is likely to act in a certain way (e.g. by preferring specific joint-venture constellations) over time, in line with an underlying mental model and set of values or premises. Whereas an appropriate method to capture logic as practices would be observational studies or ethnographical approaches, and earlier research has called for such approaches (Bettis et al. 2011), the empirical studies in this category tend to infer patterned action from (retrospective) interviews, historical analyses or surveys rather than actual behavioural data.

Dominant logic as organizing structures
Conceptualization. The idea of dominant logic as organizing structures is inherent in the original definition, which includes 'the administrative tools to accomplish goals and make decisions' (Prahalad and Bettis 1986, p. 491; emphasis added) and von Krogh and Roos's (1996) view of dominant logic as selfsimilar, meaning that organizational structures at different scales resemble each other. Moreover, theorists have claimed that structural or configurational arrangements, such as specialization, formalization, decentralization and hierarchy, serve as both carriers and manifestations of cognition (Zyglidopoulos 1999; see also Giddens 1984).
Examples of definitions that interpret dominant logic mainly as organizing principles or structures include Grant's (1988) management functions (i.e. resource allocation, strategy formulation and control of performance targets), the texture of organizing (Bowen and Steyaert 1990), the embodiment of parts of dominant logic in organizational structures and systems (Bettis and Prahalad 1995), similarities between lines of businesses in advertising or capital intensity (D'Aveni et al. 2004) and the view of dominant logic as dominant themes and configurations (Obloj et al. 2010).
Operationalization. Methods to capture dominant logic as organizing structures exist in both the interpretive and the deductive camps. As an example of the former, Côté et al. (1999) employed an interpretive strategy based on extensive documentary evidence and in-depth retrospective interviews to elaborate on the management and organizing principles of a large Canadian engineering firm. They performed a historical, longitudinal case study where they analysed the core activities and history of the firm as described in annual reports to identify both patterns of organizing and the underlying organizing principles (individual autonomy and development, ad-hoc collaboration and fluid structures). Similarly, Robertson and Swan (2004) focused on organizing templates, which they defined as modes of organizing and governance. Through a longitudinal case study, including interviews and observations of day-to-day work in a knowledge-intensive firm, they tracked how these organizing templates, as reflectors of managerial logics, changed from an adhocracy to a soft bureaucracy over time. Another example of an interpretive approach is the study by Obloj et al. (2013), who coded verbal interview data to investigate the members' perceptions of the environment, choices and actions, routines and codification of learning. By analysing similarities and differences among six firms from different industries, Obloj et al. (2013, p. 300) found specific characteristics of management systems, such as 'centralization, paternalistic leadership, simple planning or little formalization' to be core organizing principles of the studied firms.
Regarding deductive approaches to capturing dominant logic as an organizing structure, implicitly building on ideas from von Krogh and Roos (1996), a set of scholars strived to capture dominant logic as structural similarity. For example, D' Aveni et al. (2004) studied intra-firm similarities and proposed that a firm's dominant logic is reflected in structural resource congruence (i.e. the degree to which the resource allocation patterns of a line of business resemble those of a sister company's line of business). The authors captured numerical data (e.g. cost structures) and calculated an intensity congruence score for advertising, research and development (R&D), selling and capital intensity. As another example, Monteiro (2015) viewed dominant logic as similarities between business models of subsidiaries and headquarters. In his single-case study of a large multinational enterprise, he collected subjective (i.e. surveys) and objective (i.e. firm-internal database) data to show that dominant logic exerts its impact by being a source of unconsciously preferred technologies. The author operationalized dominant logic as the firms' business models and predicted that strategic options in the subsidiaries that diverged from the headquarters' business model would be discarded. The level of disagreement between business models was assessed by asking managers to rate the level of fit along five dimensions (e.g. market segment addressed by a new technology vs. corporate business model). Similarly, Gentry et al. (2016) hypothesized that the dominant logic of family-influenced firms is characterized by long-term orientation. To test whether family influence predicts long-term orientation, the authors measured slack resources, strategic risk-taking and bankruptcy risk as indicators of the dominant logic of long-term orientation.
Lane and Lubatkin (1998) examined knowledge and research similarities of firms in R&D alliances. Specifically, they assessed inter-firm similarity by counting the number of research communities in which both partners had published, with more shared communities indicating a greater overlap of their problem sets (i.e. their dominant logic). Also building on the idea of similarity, Combs et al. (2011) argued that dominant logic is a strategic resource, and that managers who have experience in managing similar strategic and operational demands develop dominant logics that are beneficial for franchise strategy.
In summary, dominant logic -in this categoryis mostly assumed to surface in organizational structures (e.g. cost structures, R&D structures). Accordingly, the agent of dominant logic is a collective that enacts these structures on a day-to-day basis, in line with their underlying mental models, values and premises. Empirical scholars investigating organizing structures often focus solely on structural similarity, without taking into account other aspects. In many cases, they operationalize specific, rather narrow variables, such as the number of similar research communities or cost structures, and view these numbers as reflectors of dominant logic. Here, a dominant logic is located in manifested dimensions enabling operational measurement. In the empirics of this category, the underlying cognition of actors exists as a background assumption that helps explain why variance in the phenomenon occurs (e.g. why cost structures between several lines of businesses differ). However, cognition is not captured explicitly. While this approach enables large-scale studies and more complex quantitative modelling, the drawback is that much information, especially with respect to more invisible aspects of dominant logic, is lost with these rather narrow operationalizations.

Summary
Our systematic literature review of the extant 94 studies of dominant logic that explicitly build on Prahalad and Bettis's original ideas revealed a variety of interpretations and empirical approaches for capturing dominant logic. In their conceptualizations, researchers have used a multitude of terms and concepts (e.g. information filter, mental model, managerial practice, organizational configuration) in a somewhat arbitrary manner to reveal their interpretation of dominant logic. Depending on where scholars expected to find dominant logic (e.g. in mental models vs. organizing structures), they used different empirical strategies (i.e. inductive vs. deductive) and various methods. This poses threats to the validity of findings and prevents comparisons across studies. Through our content analysis of the literature, we have disentangled the extant interpretations and operationalizations and identified four different dimensions of dominant logic. For each of the dimensions, we have described how researchers have interpreted the concept, where it is assumed to be located and how exactly it was captured, thereby facilitating conceptual clarity.
At the same time, a closer look at the original definition of dominant logic shows that the four dimensions are already covered in Prahalad and Bettis's (1986) original theorizing. Hence, they do not contradict each other, but merely take different angles and are thus consistent. However, researchers seem to struggle with the multidimensionality of the construct: they often emphasize one particular dimension (e.g. shared mental model) at the expense of others, thereby neglecting how the dimensions interrelate. Hence, in the next section, we will reassemble the four dimensions to develop a more coherent understanding of what a dominant logic actually is and how it can be assessed more rigorously.

Reassembling dominant logic: an integrative model
Revisiting the definition of dominant logic by Prahalad and Bettis (1986, p. 491) as 'a mindset or a worldview or conceptualization of the business and the administrative tools to accomplish goals and make decisions in that business', it becomes clear that the authors introduced it as a multifaceted construct. It is neither a mere 'mindset' nor exclusively the 'administrative tools'; all these aspects are interrelated. However, many scholars have neglected to acknowledge this interrelation. Our analysis of the interpretations and assessments of dominant logic enables us to increase the granularity of theorizing by zooming in on the concept's structure and inner workings.
As shown in Figure 2, we regard the four dimensions that we have identified as constituents of dominant logic which in concert determine the concept's structure. Thereby, we consider the dominant logic as a multifaceted entity that is rooted in (invisible) cognition, but simultaneously surfaces in manifested (visible) structures; in Prahalad and Bettis's (1986) terms, it is both a mindset and the according administrative tools to bring that mindset to life. As indicated by the dashed lines in Figure 2, shared mental models, values and premises represent the invisible, cognitive dimensions (scholars have referred to them as schemas, orientations, meanings, beliefs, etc.); organizational practices and organizing structures are the visible, manifested dimensions (also referred to as management tools, administrative structures, etc.). Over time, and with accumulated experience, the cognitive (invisible) shared mental models and values become  (Bettis et al. 2011;Zyglidopoulos 1999), meaning that the invisible side has a generative function because it underlies and affects the visible side. Conversely, (visible) structures and practices provide feedback into the underlying (invisible) mental models and values, thereby both reinforcing and modifying the underlying cognitive dimensions. Accordingly, all four dimensions are tightly coupled through consistent relationships. In what follows, we will consider how the constituents of dominant logic operate in concert.

Link 1: Shared mental models, values and premises
Whereas shared mental models permit actors to categorize an event, assess its consequences and consider appropriate actions (Prahalad and Bettis 1986), values and decision premises make it possible for organization members to choose amongst existing alternatives and accordingly act upon them. Values and premises encapsulate the information filter function of dominant logic because they guide managerial attention by determining what is important. They also affect how events (e.g. technological innovations) are interpreted and evaluated.
For example, Bingham and Kahl (2013) showed that actors in the insurance industry interpreted the emergence of the computer (from 1947 to 1975) through the mental model of either a machine or a brain. These interpretations were contingent on different premises about what a computer is good for (i.e. calculating vs. decision-making). This interlinkage between mental models and values and premises is also reflected in Bettis and Wong's (2003) notion that dominant logic governs both the search space associated with problems (i.e. a mental model determining 'how the world is') and the key features of acceptable solutions (i.e. decision premises determining 'how the world should be').
Even if mental models and values are related, it is important to differentiate between them more carefully. Shared mental models surface in frames, categories and vocabularies (Joseph and Gaba 2020). They are general perceptual mechanisms. In contrast, values and premises pertain to desirable end states, thus accommodating an evaluative function (e.g. by determining what is right/wrong). Values are the carriers of what scholars have described as deeply rooted, taken-for-granted or culturally shaped beliefs, that define the important goals within an organization, as well as the appropriate means to achieve those goals (van Rekom et al. 2006). In summary, while they are closely interrelated, mental models and values play different roles within an organization and are thus separate dimensions of dominant logic.

Link 2: Cognitive dimensions to organizational practices
Over time, mental models and values of individual actors become 'embedded in a firm's routines, procedures and resource commitments' (Kor and Mesko 2013, p. 236; emphasis removed; see also Grant 1988). Whereas at the birth of an organization, the dominant logic may be largely grounded in an entrepreneur's cognition (e.g. her vision about the firm), it eventually becomes enacted over time by the organization's members (Bettis et al. 2011). For example, values become desired ends and as a result, particular organizational practices emerge as appropriate means to achieve those ends (van Rekom et al. 2006). Thereby, 'background knowledge in the form of understanding, know-how, states of emotion and motivational knowledge' shapes practices within an organization (Reckwitz 2002, p. 249; see also Feldman and Pentland 2003;Tsoukas 1996). In other words, cognition generates and shapes organizational practices, which leads to a pattern in how things are typically done (see also Rerup and Feldman 2011;Tsoukas 1996).
Critical strategic management tasks, such as resource allocation, are then performed by relying on negotiated and agreed-upon premises and programmes (Luhmann 2018), surfacing in a 'nexus of interconnected human practices' that span 'knowledge, meaning, human activity and sociality' (Nicolini 2011, p. 602). In its most formalized version, this may lead to standard operating procedures or (automated) processes. In summary, invisible cognitive dimensions of dominant logic become visible in organizational practices.

Link 3: Cognitive dimensions to organizing structures
The cognitive dimensions of dominant logic may solidify into organizational structures. Prahalad and Bettis (1986) termed them 'administrative tools' and thought of these tools as history-dependent process knowledge. In this spirit, Grant (1988) argued that cognition becomes effective through the management functions of resource allocation, strategy formulation and controlling of performance targets. These functions manifest in hierarchical roles, vertical communication and rules (e.g. means and ends) for their enactment. The argument that organizational structures are rooted in cognition has been recognized by scholars who interpret and assess dominant logic in structural terms, such as similarity in cost structures and business models (e.g. D' Aveni et al. 2004;Lane and Lubatkin 1998;Monteiro 2015).
Even if they have not directly assessed cognition, some authors have acknowledged that structural dimensions of dominant logic are deeply rooted in cognitive processes (e.g. attention, informationprocessing or decision-making). Monteiro (2015), for example, concluded that the headquarters of large corporations favour confirmatory and proven external knowledge. They do so because selective atten-tion mechanisms arising from dominant logic are at play, leading to overseeing the potential of new technologies. Similarly, resource relatedness (as measured through a resource congruence score) between lines of businesses, as D' Aveni et al. (2004, p. 367) have stated, reflects 'a set of cognitive simplifications, analogies, conventional wisdom, and intuition about successful strategies'. Thus, the cognitive dimensions of dominant logic underlie and shape organizing structures.

Link 4: Organizational practices and organizing structures
Theorists of dominant logic have also mentioned both practices and structures in their analysis of dominant logic (e.g. Bettis and Wong 2003). Even if certain scholars, like von Krogh and Ross (1996, p. 734), have suggested that it is only a matter of perspective as to whether 'a social phenomenon may appear as processual; [ . . . ] on another, arbitrarily chosen scale, the same social phenomenon may appear sufficiently stable to be called structural', we argue that practice and structure should be seen as two highly interrelated but distinct dimensions.
As we have outlined, practices are routinized types of behaviour (e.g. procedures for hiring new members), and organizing structures refer to the ways in which an organization divides its labour and puts its parts together to achieve certain goals, resulting in structural and configurational arrangements (e.g. hierarchies, specialization and decentralization) -hence, practice and structure are different entities, at least for the sake of operationalization. A recent comprehensive review also confirmed the assumption that what organizational actors do is strongly dependent on the structural configurations in which they operate (Joseph and Gaba 2020). This idea is also in line with Giddens' (1984, p. 25) view that 'the structural properties of social systems (i.e. rules and resources) are both medium and [the] outcome of the practices they recursively organize'. Bowen and Steyaert's (1990) study illustrates this argument by showing how dominant logic (i.e. the meanings and actions of an individual entrepreneur) become embedded over time in the development of the 'texture' of organizing. In terms of the dimensions of dominant logic, this means that human actors within an organization produce and reproduce regular practices and organizational and social structures, thereby establishing aligned systems of how things are typically organized.

Link 5: Feedback from manifested dimensions to cognitive dimensions
We propose two feedback mechanisms from the manifested to the cognitive dimensions (i.e. reinforcement and modification). For the former, we assume that dominant logic has a self-reinforcing nature that unfolds through reproducing existing cognitive and manifested structures (Bettis et al. 2011): through recurring organizational practices and established organizing structures, existing mental models and values are maintained and hence cognitive variance is suppressed (Bettis 2000). Successful business models can become so deeply ingrained in a firm's organizing structures that the underlying cognitive structures become hard to change (Bettis et al. 2011). An example of the reproduction function of dominant logic is what happens when new organizational members, who share similar mental models and values, are recruited and socialized, thereby further condensing the extant dominant logic and potentially leading to cognitive inertia (Bettis and Wong 2003;Kor and Mesko 2013).
This self-reinforcing function of dominant logic may lead to both high performance -because routines and structures are means for ensuring efficiency within an organization -and a state of inertia, when environmental conditions change (Prahalad 2004); because mental models and values continue to endure even when contradictory evidence is at hand (Balogun and Johnson 2004). For example, the successful razor-blade business model of Polaroid (i.e. its organizing structure) was so prevalent that it prevented its top management from changing their worldview and bringing in digital imaging as a new business model (Tripsas and Gavetti 2000). In other words, in the case of Polaroid, the extant structures and practices seem to have reinforced the prevalent mental models and premises.
Nevertheless, as a second feedback mechanism, existing practices and structures may also contribute to modifications of the underlying cognitive dimensions. Although it is not very prominent in the literature on dominant logic, this mechanism is well documented in literatures on schema change (e.g. Bartunek and Moch 1987;Labianca et al. 2000), proposing that any implementation of change in an organization requires a corresponding adjustment of its underlying shared mental models in order to be effective.
Moreover, evidence on the relationship of routines and mental models bolsters the argument that practice may contribute to changes in mental models and values. First, performing routines is not mindless, but contains self-reflective and other-reflective behaviour (Feldman and Pentland 2003), and '[r]outine participants gain a sense for what they ought to do through acting' (Dittrich and Seidl 2018, p. 134). The reflection about 'what one is doing' potentially affects the mental models of the actors involved in the routine. Second, routines do not just give shape to organization members' actions, but also function as normative constraints (Feldman and Pentland 2003;Tsoukas 1996). Because actors do not perform routines exactly the same way, but can 'alter the potential repertoire of activities that creates and recreates [ . . . ] the routine' (Feldman and Pentland 2003, p. 108), over time, new practices may be seen as legitimate. Changes in legitimacy may impact decision premises, values or mental models. In this way, the dominant logic as a whole can become adaptive, as it may incrementally change over time (Bettis and Prahalad 1995).

A revised conceptualization of dominant logic
In summary, based on our review and analyses, we regard dominant logic as a multidimensional construct that describes how (invisible) managerial cognition is related to (visible) practices and structures. More precisely, we think of dominant logic as a system of shared mental models, values and decision premises that manifest in corresponding organizational practices and organizing structures; all these dimensions are aligned.
Our revised conceptualization of dominant logic is still in line with Prahalad and Bettis's (1986) original theorizing of dominant logic, but further refines it. Whereas Bettis et al. (2011) have discerned 'invisible' and 'visible' elements in their theorizing, we zoom in once again and are now able to precisely name the four different dimensions that underlie the 'mindset' and 'administrative tools' covering a dominant logic. By introducing the notion of alignment across dimensions, we highlight that the dimensions need to be consistent with each other, in order to be part of a firm's dominant logic. For example, 'long term orientation' can only be a feature of a firm's dominant logic if it is reflected in the shared mental model of the organization's members, its decision premises (e.g. little risk-taking), practices (e.g. conserving slack) or HR structures (e.g. long-term employment). We will return to this aspect in the discussion section to highlight implications for capturing dominant logic.
Our integrative model still fits the complex adaptive systems theory -an important grounding of dominant logic (Bettis and Prahalad 1995). Complex adaptive systems theory assumes that organizational-level behaviour is produced by agents within an organization: (i) who act based on their mental models; (ii) who are dependent on each other and interact; and (iii) whose mental models and behaviour co-evolve over time such that observable patterns emerge (Anderson 1999). Accordingly, our integrative model suggests that these mutual dependencies lead to agreedupon decision premises that condense into organizing structures. Then, through their repeated interaction, organization members form routinized practices.

Discussion
In this review, we have taken stock of the existing research on dominant logic and provided a comprehensive overview of the field. We have delved into the concept and discerned four dimensions that reflect how scholars have interpreted dominant logic. Moreover, we have elicited and evaluated the empirical strategies employed to capture these four dimensions. Because scholars seemed to struggle with that multidimensionality, we have reassembled the four dimensions into an integrative model that reveals how the dimensions interrelate to produce a firm's dominant logic. Our more detailed perspective makes explicit what has been mostly implicit in earlier research -namely, that dominant logic links shared mental models, values and premises ('invisible' cognition) with recurring organizational practices and organizing structures ('visible' manifestation). This conceptualization highlights the need for future studies to be more careful in the application of the appropriate terminology, speaking of dimensions of dominant logic rather than of dominant logic in general. As a consequence, this review paves the way for a more precise application of the concept. In what follows, we discuss theoretical and methodical implications and provide directions for future research.

Theoretical implications
Our revised conceptualization of dominant logic describes it as a system of four interrelated dimensions; it should be understood as a unique, recurring and discernible pattern of strategic thinking and decisionmaking, and the corresponding organizing structures and practices. This idea that dominant logic should be seen as a pattern extends previous considerations of self-similarity: whereas self-similarity means that '[t]he way managers conceptualize and manage technologies, product development, distribution, advertising or people can be similar [ . . . ] throughout the company' (von Krogh and Roos 1996, p. 735; see also Schraven et al. 2015), we argue that, in addition to seeking similar patterns across scales, scholars should also consider alignment across dimensions. In more simple terms, a certain aspect should only be considered part of dominant logic if it recurs in shared mental models, values and premises, routinized practices and organizing structures.
Future researchers may further examine the microfoundations underlying each of the four dimensions. For example, regarding mental models, sharedness is considered an important characteristic (e.g. von Krogh and Roos 1996); however, it is unclear who shares what, and to what extent that sharing occurs. Similarly, it is unclear how -in the case of competing values within a firm -certain values become more 'dominant' than others (e.g. von Rekom et al. 2006). For example, which roles do conflict, discourse and power play in the evolution of a dominant logic (Bowen and Steyaert 1990;Phillips et al. 2008)?
Moreover, although we have taken a first step toward integrating the four dimensions into a model, future researchers may further illuminate these interrelations. For example, regarding the feedback mechanism of reinforcement between an organization's structure and the cognitive dimensions of dominant logic, Joseph and Gaba (2020) recently established that structure impacts an organization's decision premises -this link may be specified in more detail. Furthermore, although the mechanism of reinforcement has previously been acknowledged (Bettis et al. 2011), the way in which structures and practices may contribute to modifications in cognitive dimensions represents another interesting opportunity for future research.
In this context, our revised conceptualization also implies a sensitive reflection of the inherent temporality of dominant logic. Based on earlier work, we have suggested that cognition transcends into the organization over time, as it is enacted in recurring organizational practices, and solidifies in organizational structures (Bettis and Wong 2003;Kor and Mesko 2013;van Rekom et al. 2006; see also Giddens 1984). Hence, as Figure 2 suggests, as the dominant logic of an organization develops, it becomes more and more visible. Once established, organizational practices and structures feed back to the cognitive dimensions, thereby reinforcing or modifying existing mental models and values. Future work may study those temporal dynamics more explicitly, and longitudinal analyses may reveal mechanisms of how firms cultivate their unique 'mentalities', practices and structures.
On a more general note, scholars could investigate micro-foundations not just of its constituting dimensions, but of dominant logic as a whole. The dominant logic concept implies that actors inside a firm develop similar understandings that help them to collectively 'make sense of, attribute meaning to, and interpret internal and external events' (Rentsch et al. 2008, p. 144); these understandings then shape the organization's recurring practices and its entire structure. Future research could examine general cognitive mechanisms, such as sensemaking (Weick et al. 2005), attribution (Festinger 1957) and collective interpretation (Gavetti and Warglien 2015), which are at play when a dominant logic is formed. These general micro-foundations of dominant logic have so far been neglected in the scholarly community. Similarly, with only a few exceptions (Schraven et al. 2015), scholars of dominant logic have neither theorized about what dominance actually means, nor reflected upon why their empirical content represents a dominant logic. Hence, an interesting avenue for future research would be to theorize in more depth on the notion of 'dominance'.

Implications for empirical operationalization
It has been suggested in the literature that dominant logic can only be thoroughly assessed empirically when the characteristics of self-reference and scale are taken into account (Schraven et al. 2015;von Krogh and Roos 1996). Empirically investigating self-reference implies that a firm's history needs to be considered, like in the examples of Bouwen and Steyaert (1990) and Cote et al. (1999). However, most studies in our sample did not tackle the role of history in a systematic way. In order to investigate scale, von Krogh and Roos (1996) have argued that researchers have to look for patterns across different levels and scales within an organization. In our review, only a few studies have employed methods to account for scale (D'Aveni et al. 2004;Lane and Lubatkin 1998;Schraven et al. 2015). Hence, future empirical research should focus more strongly on the firms' histories, as well as on recurring patterns across scales.
Our revised conceptualization of dominant logic comes with some further implications for empiri-cal assessment. One implication is that a reductionist analysis (i.e. focusing on one dimension) is not sufficient; dominant logic needs be studied in a holistic way. This does not necessarily mean that all dimensions must be taken into account simultaneously, but scholars may combine at least some of them. So far, only a handful of scholars have combined, for instance, recurring organizational practices with values and premises (e.g. Kossek et al. 1994;Lampel and Shamsie 2000), or shared mental models and organizational practices (e.g. Bouwen and Fry 1991). Others have even identified three aspects of dominant logic in one and the same study (e.g. Bouwen and Steyaert 1990;Côté et al. 1999;Obloj et al. 2013; Table 3). However, in general, empirical researchers have not explicitly and purposefully discerned and combined the constituent dimensions of dominant logic. Our revised conceptualization and literature review may serve as a starting point for scholars who aim to combine dimensions of dominant logic, specify how they operate in concert and choose or design appropriate operationalization methods.
Importantly, as we have argued, empirical scholars should aim to discover themes that recur across dimensions. For example, for a certain value (e.g. long-term orientation) to be considered as part of a dominant logic, there should be observable practices (e.g. risk-averse investment practices) that reflect that value. Accordingly, empirical methods -both deductive and inductive -should be suitable to identify such recurring themes across mental models, values and premises, organizational routines and organizing structures. Mixed-method approaches (e.g. combinations of surveys, interviews, observations and objective data) may be fruitful for assessing patterns across invisible (e.g. values) and visible dimensions (e.g. practices) in one and the same study (e.g. see Côté et al. 1999). In this context, despite explicit claims for a more process-oriented perspective (Bettis et al. 2011), existing studies have also largely ignored how dominant logic functions as a process (e.g. how strategic decisions are actually made in practice). Hence, a process-based perspective and the corresponding empirical approaches (e.g. observational studies, process analyses) remain opportunities for future research. As another possibility, empirical scholars could use van Rekom et al.'s (2006) means-end analysis, or Schraven et al.'s (2015) cognitive mapping approach to operationalize core values and mental models, respectively, and combine them with methods to capture patterns in practices (e.g. Pentland et al. 2010). Using more standardized methodological approaches such as these to capture dominant logic could also contribute to comparability of findings in the field.
A more general point of discussion is how empirical scholars have used dominant logic so far. One group of (mostly deductive) studies used dominant logic as an explanation for variance in other (mostly numerical) variables. The concept of dominant logic provided the theoretical basis for operationalizations, for instance, as similarity in the cost structures of different lines of business (D'Aveni et al. 2004) or similarity in research communities (Lane and Lubatkin 1998). In their empirics, this group has mostly designed operationalizations based on the visible aspects of dominant logic (i.e. structures and practices). The invisible, cognitive side of dominant logic, as well as its evolution and history, were used as background assumptions. This group may benefit from our integrative model, because it provides a clear, plausible reasoning for relationships between invisible and visible aspects, which may help explain why, for example, certain numerical measures reflect a firm's dominant logic.
The second group of (mostly interpretive) studies aimed to study the dominant logic of a certain empirical context, for example, of entrepreneurial firms in China (Obloj et al. 2013), or of new ventures in the digital music industry (Hadida and Paris 2014). In these studies, dominant logic was often the target of explanation, which seems particularly suitable for exploring new empirical terrains. So far, some scholars in this group have used dominant logic as an 'open-ended container' for their various observations. Instead of freely choosing the contents of their empirics, interpretive scholars may also look more systematically for recurring themes across dimensions. Our integrative model also provides guidance here, as it outlines the different dimensions that need to be covered in an empirical study.