Introduction

Leadership plays a pivotal role in the success or failure of an organization (Einarsen et al. 2007; Larsson and Vinberg 2010). When it is successful, leadership inspires, empowers, supports, and stimulates the members of an organization, secures external resources, visualizes a desired future, and enhances personal and organizational performance (Bass 2008; Behrendt et al. 2017; Collinson 2012; Karakitapoğlu-Aygün and Gumusluoglu 2013). By contrast, there is a potential dark side of leadership that risks leading to organizational failure caused by abusive, bullying, destructive, and tyrannical leadership (Pelletier 2010; Tourish 2014). To succeed in the former endeavor and avoid the latter outcome, a large portion of research on leadership has been devoted to empirical studies of how leaders behave.

Observations of the behaviors of leaders and, consequently, theories of leadership behavior based on observations e.g., by Yukl (2002, 2012), Burke et al. (2006), and DeRue et al. (2011) have advanced our understanding of leadership behavior that contributes to either success or failure. Still, researchers like Avolio (2007), Van Knippenberg (2013), and Dinh et al. (2014) are concerned about the parts of research based on observations of leadership behavior reported by lay persons, such as the members of the organization. They argue that such observations, including surveys and interviews with peers, subordinates, and superiors, are flawed because they confuse leadership behavior with the followers’ perceptions of leadership behavior. The argument is that if studies are based on observations by members of the organization, they risk the bias of defining leadership behavior “in terms of its effects, and such a definition would disqualify the concept from studying its effects” (Van Knippenberg and Sitkin 2013: 11). The cornerstone of their argument is that it is not the actual behavior that is observed but rather the effects of the behavior, and such observations are likely to be affected by an organizational member’s general impression of the person being observed - a phenomenon also known as the halo effect (Rosenzweig 2007). With this critique, Avolio (2007), Van Knippenberg (2013), Dinh (2014), and Derue et al. (2011) encouraged the research community to develop more sophisticated theories of leadership based on methodology and research beyond those used in the core literature on leadership. Behrendt et al. (2017) answered this call and developed what they called the Integrative Model of Leadership Behavior (IMoLB).

In addition to integrating Yukls’ taxonomy of leadership behavior, which Behrendt et al. (2017) refer to as the gold standard because of its comprehensive and integrative overview of prevalent research on leadership behavior, the IMoLB integrates several theories proposed over more than 60 years of research. According to Behrendt et al. (2017) the IMoLB leads to more objective observations of leadership behavior. The argument is that it relies on observations by professionally trained observers “who are equipped with rating manuals and specific descriptions of the intended behaviors and are intentionally focused on observing these behaviors” (Behrendt et al. 2017: 231). Using highly trained observers who know the “art” of observing behavior, Behrendt et al. (2017) claim to avoid laypeople observation errors such as the halo effect, confirmation bias caused by implicit leadership theories, and lay observers’ intention to answer questionnaires or interviews consistently. This criticism of research based on observation studies and the integrative model which Behrendt et al. (2017) develop from is the starting point of our work. We use it to make a case for a more varied consideration of philosophical and multidisciplinary differences in the conceptualization and perception of leadership behavior, instead of buying too strongly into what can be called neo-positivist psychological science. We furthermore explore the methodological implications of such a move.

We do, of course, acknowledge the conceptual difference between the perceptions of leadership behavior and the “behavior itself,” as Behrendt et al. (2017: 231) call it. Nevertheless, we will question the extent to which researchers should be criticized for not distinguishing between leadership behavior and perceptions of it in empirical studies. We argue that the perceptions of leadership behavior not only matter but are the most important component of leadership behavior theory because they are what followers react to and act upon. Hence, by dismissing perceptions and going for a more “objective” notion of leadership behavior, as Behrendt et al. (2017) have proposed, one risks ignoring crucial aspects of leadership behavior in practice.

The aim of this paper is, first to show that the distinction between leadership behavior and the perceptions of it is not only interesting from philosophical and epistemological points of view, but also has methodological implications. Second, we suggest an approach to studying leadership behavior that acknowledges perceptions and show that leadership behavior and perceptions of it are formed by and depend on the interpersonal relationships in and around the organization.

We shall argue that leaders, followers as well as the relation between them are co-constitutive of leadership behavior. This perspective has methodological consequences that we discuss after having presented our approach.

The remainder of this article is structured as follows: First, we briefly review previous research on leadership behavior. Second, we suggest a middle ground between an objective and a subjective approach to research on leadership behavior. Third, we discuss the methodological consequences and limitations of our proposed approach, before finally concluding where this should lead us and suggesting directions for future research on leadership behavior.

Previous Leadership Behavior Research

Personal Traits and Factor Models

Over the past century, research on leadership behavior has gradually shifted in focus from a single factor (e.g., a leader’s charisma or rationality) to a more differentiated set of behaviors. Although the factors considered have differed from time to time and researcher to researcher (Langford et al. 2017), a large part of early studies on leadership behavior were devoted to two meta-categories (factors) of behaviors. In some like the Ohio state studies, the focus was on initiating structure and consideration (Bass, 2008). Researchers subsequently focused on leadership behavior in the contexts of task and relations until Yukl, Gordon, and Taber in 2002 (Yukl 2002) added a third category: change-orientation. Since then, these three meta-categories have played a dominant role in the behavioral approach (Borgmann 2016).

A common feature of the studies that have focused on any two or all three categories is that they have examined how individual leaders behave. Thus, the focus has been on the leader and how the leader acts against or in front of the people being led (Uhl-Bien et al. 2007). But because a leader is defined as someone who shows the way to followers (OED 2016) and someone who depends on followers, the research needs to acknowledge that leadership is fundamentally a relational phenomenon between a leader and their followers. Thus, leadership behavior cannot be separated from the people who experience or observe the behavior, as has been highlighted in the Hawthorne studies, the Human Relations School, and the early Michigan State studies (Bass, 2008; Langford et al. 2017; Yukl 2013). The effects of leadership have been discussed and investigated by many researchers from this perspective to understand how, in some organizations, leadership impacts individual and organizational performance. For instance, transformational leadership and visionary leadership (the concepts are often used synonymously) focus on how the behavior of a leader might transform, inspire, and engage members of the organization (Bass, 2008). But of course, leadership is bidirectional: In fact, it describes the complexity of the interaction between the “leader” and the “follower.” This is why transactional leadership (Bass, 2008) focuses on how a leader uses transactions (e.g., reward or punishment), whereas emotional intelligence (Goleman 1996; Goleman et al. 2013) uses the leaders’ own emotions and those of their followers to examine leadership. Along the same line of thought, leader–member exchange (LMX) theory discusses both how the leader’s and the follower’s personal characteristics affect the relationship that arises between the two parties (Gooty and Yammarino 2016). Still more importantly, it is the perception of the follower that impacts their behavior toward the leader, who in turn reacts to the perceived behavior of the follower (Carsten et al. 2018). This is essential, as it acknowledges that different members of the organization cannot be expected to perceive and react to perceptions in the same way. Another central aspect of LMX theory is that people’s perceptions and judgements of leadership behavior may not be based on observations of leadership behavior as such, but on observations of other, more or less loosely correlated, phenomena like successful outcomes, as Behrendt et al. (2017) have noted in their recent work. Before returning to this point, however, we need to investigate the relational process between a leader and their followers in more detail.

The Relational Model

According to LMX theory, if a follower has a certain perception of the behavior of a leader, they will react according to this perception. The interpersonal relationship (consisting of affect/liking, ingratiation, self-promotion, assertiveness, and trust) between the parties affects how each of them perceives the other. According to LMX theory, leadership behavior is influenced by and practiced in the interpersonal relationship that arises between the leader and the members. Furthermore, LMX theory contains contextual variables, such as the work setting, the participant’s location, and cultural dimensions (Dulebohn et al. 2012). These variables are secondary to the leader–member relation but their inclusion is still important as they interfere and indirectly play an important role in the consequences of the leader–member exchange relationship. However, LMX-theory fails to consider the contextual variables in the initial process, where the leader and the follower perceive each other’s behaviors. If viewed as a temporal process—with the leader, the member, and an interpersonal relationship between them—the contextual variables are added in step two as an intervening variable. We return to this later. At this stage, the question is what happens when the leaders’ and the followers’ perceptions, or their perceptions of each other’s perceptions, differ? This appears often to be the case, as known from the literature on self–other agreement (SOA) (Atwater and Yammarino 1997; Fleenor et al. 2010). It might not be a problem until we start measuring the perceptions and building theory upon them. In its extreme, the measurement might not measure anything other than judgements of (each other’s) perceptions. It might not say anything at all about leadership behavior, but it still says something about how a leader’s behavior is perceived—not objectively, but subjectively by the members of the organization who are affected by and perceive the interpersonal relationship. This is important because from our point of view, the recognition of (potential) differences in perceptions should inform research on leadership behavior and should motivate a cautious and critical attitude to the data obtained when measuring leadership behavior. Before elaborating our point of view, one research approach to leadership behavior remains to review. This is the integrative model by Behrendt et al. (2017).

The Integrative Model

While acknowledging their intellectual debt to Yukl (2012) and his integration of more than 60 years of research, Behrendt et al. (2017) proposed reducing Yukl’s four meta-categories of leadership behavior (task oriented, relations oriented, change oriented, and external) to two: task-oriented and relations-oriented leadership behavior. Yukl’s two remaining meta-categories—change-oriented and external leadership behavior—are integrated into the IMoLB by two continua describing different degrees of (1) internal vs. external relations-orientation, and (2) degrees of routine vs. the orientation of change regarding tasks. Furthermore Behrendt et al. (2017) suggested using professionally trained observers to overcome the biases of previous perception-based observational studies. They envisaged a research methodology that presumably goes more directly to the subject matter without proposed a model that avoids relying on perceptions of any kind. Theories of leadership behavior must be empirical and thus necessarily maintain a link to perceptions. In addition to suggesting trained observers, Behrendt et al. favor an approach that is more top-down in that it imports its taxonomical categories from another, more general, and presumably more fundamental, branch of science (viz. psychology). They recommend that these categories (the taxonomical categories) be applied and tested empirically by the trained observers. Lay observations and descriptions play no role in this approach.

As such, the integrative model of leadership behavior represents an etic (studying behavior from outside a particular system) rather than emic (studying behavior from inside a system) (Headland et al. 1990; Pike 1967). It obtains and systematizes its empirical findings according to a general theoretical framework couched in terms not usually employed by the persons being studied. It pretends to be able to identify a kind of behavioral mechanism that underlies the more superficial manifestations of leadership behavior. The project as such seems to be very much in line with some of the canons of good science: It aims at a high degree of objectivity and generalizability, and it promises to provide deeper explanations in terms of real mechanisms rather than superficial correlations. A further epistemic virtue to which it aspires is, as the name of the model suggests, integration: By linking the theory of leadership behavior to the well-established fields of motivation, action, and group psychology, it promises a significant theoretical unification.

However, the success of the integrative model depends on the extent to which the subject matter allows for a bypassing of lay perceptions and descriptions. An etic approach to the study of human affairs risks ignoring domain-specific but significant or even essential details. One reason to ignore such details is the excessive generality of the theoretical framework, that is, a mismatch of scale. For example, certain attempts to apply evolutionary psychology to the humanities, especially the study of literature, may yield correct but uninformative and almost trivial results because the evolutionary framework is unable to distinguish between the finer nuances of expression or literary content. Another way in which an etic approach can lead to an ignorance of relevant details has to do with the ontology of the subject matter. Objectivity is the goal as far as possible, but when it comes to human affairs, the way in which things are experienced or understood can be partially constitutive of the object under study. Bypassing perceptions may thus lead not to a clearer focus on leadership behavior as such, but to a preoccupation with an overly reduced and abstract notion of such behavior.

The integrative model of leadership behavior thus seems not just to be in line with the canons of good science when it comes to its aims and overall structure (e.g., unification and parsimony). As it draws on several influential contemporary psychological theories as well as a reflection on the very essence of leadership (Behrendt et al. 2017), it appears both theoretically and empirically to be well founded. The six categories of leadership behavior that it posits (enhancing understanding, strengthening motivation, facilitating implementation, fostering coordination, promoting cooperation, and activating resources) (Behrendt et al. 2017: 235) are prima facie relevant, and seem consistent and clearly delineated.

Still, the superiority of the integrative model can be doubted, where this doubt can be stated both generally and more directly as questioning its central claims and concepts. In general, the model grants the researcher considerable authority and power as they are assumed to know best how to identify and conceptualize that which is important to, and manifests, leadership behavior. This authority derives from the well-established psychological theories underlying the model. Nevertheless, how well established and empirically corroborated these theories are in their own fields it is open to discussion. For example, the “Rubicon model” of action phases has been criticized for being too rigidly sequential and unidirectional as well as for considering only cognitive factors and neglecting emotional motivating factors (Bruch 2003; Goschke 1997). Perhaps more importantly, how much of their authority is preserved in the process of translating them into categories of leadership behavior is an open question.

As mentioned above, Behrendt et al. presented their integrative model to minimize bias, but we suspect that the model itself is subject to, and perhaps even a product of, biases. It may exhibit a bias in the way in which psychological concepts are synthesized and translated into empirically applicable categories. Regardless of whether it does, it becomes a potential source of bias through its very empirical application by demanding that the researcher employ this specific taxonomy, with its implicit assumptions regarding what is central to leadership behavior. There is reason to fear that a strongly theory-driven (or “top-down”) approach might easily become subject to confirmation bias.

Apart from the risk of bias, the categories may prove difficult to empirically apply, at least in the way in which they are intended by Behrendt et al. The two crucial meta-categories in the integrative model adopted from Yukl are those of a task and a relation orientation (Behrendt et al. 2017: 235). While they may seem clearly delineated in theory, the overlap that they are intended to avoid is likely to reappear in practice. Organizational tasks are identified and defined in many different and complementary ways, and at many different levels. If one focuses not merely on “official” general tasks, which can be defined by simple goals like minimizing costs, raising productivity, and improving performance, tasks are very much about relation-building, and vice versa. Fostering coordination and promoting cooperation are important tasks of leadership and enhancing understanding and strengthening motivation involve relation-building actions in practice, at least in complex organizations.

This does not mean that the categories of the integrative model are misleading or inferior to those of other models. They may very well be of good use in empirical studies. However, this provides reason to doubt that the integrative model has moved significantly beyond more traditional, lay perception-based taxonomies, or even that it is free of overlaps between categories. Furthermore, we do not disagree with the positive methodological suggestions made by Behrendt et al., for example, that video analysis, diary studies, and longitudinal designs should be used more extensively (Behrendt et al. 2017: 241). We do think, however, that they should be used as complementary to questionnaires and interviews rather than as competing alternatives. While it may be appropriate, in one sense, to speak of such methods as more “objective” because they rely less on the judgments of lay observers, one should not infer from this that they are necessarily better at measuring what leadership researchers are (or ought to be) interested in, which, in this case, is leadership behavior.

This is, admittedly, still very hypothetical. So far, we have noted only that the case developed by Behrendt et al. depends on the assumption that “leadership behavior” is best understood, and consequently best studied, in abstraction from how it is perceived and reacted to by organization members, and that this is far from obvious. We have not shown that the opposite view, viz., that leadership behavior is dependent on its perceivable effects, is true or even likely to be true, but have merely stated some reasons for why it should still be considered a serious alternative. We have neither made any substantial case for the reliability of the traditional interview and survey-based methods, nor provided any specific reason for doubting the adequacy of the “integrative model.” In the remainder of the article, we will take some steps toward a more substantial, albeit qualified and still somewhat tentative, defense of the traditional approach.

Moving Research Forward from a Middle Ground Between Objective and Subjective Approaches to Leadership Behavior

Rather than abandoning survey- and interview-based research because of its perception-based foundation, and taking a top-down, allegedly objective, approach by using highly trained observers, as suggested by Behrendt et al. (2017), we advocate for an approach that incorporates them. We see our suggestion as a middle ground between an allegedly objective (Behrendt et al. 2017) and a subjective approach (laymen-based interviews and surveys).

As noted earlier, according to LMX theory, leadership behavior is influenced by and practiced in the interpersonal relationship that arises between the leader and the members. Furthermore, LMX theory takes contextual variables, such as the work setting, the participant’s location, and cultural dimensions into account (Dulebohn et al. 2012). However, these variables are secondary to the initial leader–member relation instead of being constituent variables in the leader-member relation. Viewing the contextual variables as constituent elements of the leader-member relation affect the understanding of how both leaders and members perceive each other’s behaviors. Acknowledging this plethora of leaders and members and their correlating interactions have methodological consequences that we discuss in the following paragraphs. From this perspective, multilevel assessments and methodological triangulation might give us much more information on perceptions of leadership and leadership behavior than the reductive view of leadership behavior as the actions of a single individual (the beginning of the behavioral approach), as an interpersonal relationship (LMX theory) or as something that can only be studied by trained professionals (the integrative model, IMoLB).

It should be noted that our approach fits well with a phenomenological approach, and that phenomenology can provide further justification for preferring our approach to the neo-positivist integrative model of Behrendt et al. (2017). Phenomenology generally considers the complex ways in which different entities (e.g., objects, phenomena, relations, or variables) appear and are acted upon and take subjects’ perceptions and actions to be determinants of how the entities “really” are. From a phenomenological perspective, entities or phenomena “are” what they are perceived to be by individual subjects (though these perceptions may well be influenced by, and contain implicit reference to, the perceptions and actions of others). Thus, phenomenology also strikes a middle ground between exaggerated objectivism and subjectivism. It does not privilege the perspective of trained professionals, nor does it conceive of behavior in abstraction from interactions with others (see especially Waldenfels 1980). But this also does not imply that lay perceptions should be taken at face value. It maintains a clear, albeit relative, distinction between things as they are in themselves and things as they appear to a particular observer, or in a particular perspective, as the latter need not adequately reflect the intersubjectively invariant pattern of appearance and interaction, which it takes to be constitutive of how a thing is “in itself.” This tallies well with a multimethod or mixed methods approach, as described in the next section. It further corrects the neo-positivist model in that leadership as a phenomenon is arguably not just a set of individual behaviors or dispositions toward behavior. In one of the few attempts at a phenomenological study of leadership, Souba (2014) captured this aspect well, and suggested that a phenomenological approach to leadership should not focus primarily on the “attributes of a leader.” However, Souba’s positive analysis may be criticized for turning attention too far away from features that can still be recognized as distinctive of leadership, as he focuses instead on ontological structures that make leadership possible. The analysis also relies mostly on an allegedly intuitive insight into leadership. While this is far from non-typical, we think, by contrast, that a phenomenological understanding of leadership calls for an approach no less scientific or methodologically rigorous than that favored by the proponents of the integrative model. Its primary implication is rather the need for a mixed methods’ approach.

Methodological Consequences of a Middle Ground Approach

Having presented an approach to perceptions of leadership behavior involving the leader-member relation, the contextual variables, and its theoretical corollaries, we now discuss the methodological consequences of the approach. The basic question pertains to how to explore and measure leadership behavior as it is perceived in and from multiple relations and perspectives.

The use of multilevel assessment, or mixed methods is by no means a ground-breaking suggestion. We propose them anyway because we acknowledge the need for complementary and less lay observer-dependent methods while refusing to discard the importance of personal perceptions altogether. As argued above, what followers act upon are their perceptions of a leader. Thus, in our view, the perceptions of a leader’s behavior do play a pivotal role when we study leadership behavior. At the same time, we must acknowledge the insight from the literature on self–other agreement (SOA) (McKee et al. 2018), which states that the perceptions of a leader’s behavior vary across sources of rating, such as supervisors, peers, and subordinates (Braddy et al. 2014; Fleenor et al. 2010; McKee et al. 2018). This is because several variables affect both the self-ratings and the ratings of others—e.g., biographical characteristics like gender, age, position, race, and education (Fleenor et al. 2010). Personality and individual characteristics, like the big five personality factors, dominance, empathy, self-esteem, and narcissism, also affect the self-ratings. On the contrary, the cognitive processes of other raters, their ability to assess, motivation, job experience and performance, mood, attitude, and commitment are known to affect their ratings (Fleenor et al. 2010). Thus, while there is good reason to include personal perceptions, of both the person being studied and others, they are not enough. This is where or mixed methods might offer a way forward.

The general idea of mixed methods is to increase the validity of data by crosschecking, comparing, or supplementing them with other types of data and/or methods (see Bryman 2016 for a general introduction). Such methods contain and maybe compare different types of data: qualitative–qualitative, qualitative–quantitative, or quantitative–quantitative data. Using this approach, make it possible to include data on perceptions with more objective data, e.g., as suggested by Behrendt et al. (2017). However, combining different types of data (e.g., subjective and objective) also means that one must be careful with cross-organizational comparison or longitudinal studies, as types and behaviors of leadership, and key features of organizations change over time.

We are aware that mixing methods is not a straightforward matter. One must decide how to compare different types of data. E.g., can they be merged, integrated, or embedded to interpret and discuss convergence, divergence, and the relevant relations? And how does this affect the reliability and validity of research? Although it is not without problems, we still suggest this as a way forward that may solve the problems in research on leadership behavior outlined above. In this case, it may be possible to combine subjective data (e.g., perceptions of self and those of others from interviews, observations, and/or surveys regardless of whether they stem from highly trained professionals or non-professionals) with other types of data, like video recordings as proposed by Behrendt et al. (2017), or performance measures. This allows the researcher to include both perceptions and objective variables related to leadership behavior and contextual variables.

Future Research on the Perceptions of Leadership

Conspicuous by its absence in the general (as well as our) debate on the perceptions of behavior and the way it influences behavior is a consideration of the origin of a perception. It may even be the case that this origin is more important than the perception itself. An important challenge in this regard pertains to aggregation. Researchers in the social sciences measure individual perceptions, as a first step, to calculate aggregated results, in a second step, that go beyond individual perception. However, in the context of leadership, the fact that the leader is a leader not in relation to one follower but in relation to many is very often abstracted from in empirical research. The perceptions of followers might differ, but the leader’s behavior toward their followers might also differ. To better understand this, social scientists look for statistically significant differences between groups of followers and the perception of behavior (Braddy et al. 2014; Fleenor et al. 2010; McKee et al. 2018). This categorization nuances a blunt aggregation but significantly reduces our understanding of leadership behavior: What are we observing? In addition, how do the perceptions relate to one another? Is the perception of the middle management toward higher management of equal importance as the perception of newcomers of the higher management? What then determines the behavior of the management? More importantly, what does it say about the aggregation that researchers try to arrive at? This shows that perceptions are complex, and behavior is much more subtle than is usually assumed. Understanding how and why certain perceptions are created is therefore important as it would not only allow us to define which perceptions of behavior exist, but also why certain perceptions exist and where they are created. Our hypothesis is that behavior is manifested in a playing field of influential perceptions of behavior between followers, between leaders, and between followers and leaders. As researchers, it would be interesting to understand what happens in that field, how perceptions and behavior change, and how robust the relevant findings are over time. Indeed, does behavior change because stereotypes based on perceptions are not sustainable, or is leadership behavior stable because perceptions are passed on from old staff to new staff? Apart from this, more research is needed on meta-perceptions: How do leaders perceive the perception of their leadership behavior by followers, and how do they react to this? In other words, researchers need to exit the “laboratory context,” wherein simple perceptions are measured, and step into the contextualized field to understand the origins of perceptions, forces influencing them, and shifts in them over time to come to more accurately measure and better understand leadership behavior. As theory on leadership acknowledges the importance of contextual variables as constituents of how leadership behavior is perceived, methodologists will need to study this context to properly understand leadership behavior.

Perhaps the problem in prevalent research is that we focus excessively on behaviors as an objective entity and neglect the context in which certain behaviors occur and are subjectively perceived. Leadership is a normative concept and phenomenon. The way leaders behave has to do with both how they would like to be seen and how they think they can behave according to (some of) their followers. Compare this with a country’s president. We can all criticize a president, but they behave in such a way that some of their followers perceive it to be positive, and an example of strong leadership. Effective leadership in that case is different from that in another setting. Perhaps our error lies in measuring the perceptions of behavior of only followers, or in too one-sided a way. What if we analyzed the context of a given behavior to explain why it occurs? In this case, we should (a) observe the behavior and (b) analyze the context (political, social, economic, and a combination depending on the scenario) to (c) better understand why behaviors are reported in this or that way by the followers. Leadership theory (e.g. Hunt and Dodge 2000; Langford et al. 2017) recognizes that it is context dependent, but we assume that we do not measure the context in which the leader–follower relationship is found because it is too complex. Instead, we reduce organizational complexity and content ourselves with measuring the perceptions of followers.

Acknowledging this complexity as well as the mutual dependence between leadership behavior and perceptions of it does not undermine the terminological distinction between behavior and perceptions of it, which is rightly emphasized by Behrendt et al. In a way, it rather reinforces their point, for it makes it highly likely that the behavior, constituted in as complex a manner as it is on our view, differs from any specific perception of it (this is also vindicated by phenomenology). Yet, contrary to what Behrendt et al. imply, this is not because the behavior is more austere and simpler than perceptions of it but, to the contrary, because leadership is a normative concept and leadership behavior both shapes and is shaped by perceptions distributed among several agents and affected by contextual variables. Insofar as this is the case, it makes less sense to talk about the actual behavior, or the behavior “itself,” as Behrendt et al. (2017) do as something “behind” or “underneath” the different perceptions (again, this would also be supported by a phenomenological understanding of behavior). Hence, while the distinction between leadership behavior and perceptions of it should be upheld, it turns out to be less important than previously thought and gives us no reason to discard the perspectives of lay persons.

Our proposal seems to have two theoretical advantages over the integrative model of Behrendt et al., and this also makes it more suited to capture crucial aspects of leadership behavior in contemporary settings. First, it represents a less one-sided, individualist perspective as it considers interaction, including the propensity of leadership behavior to cause and be affected by different perceptions in different followers, to be constitutive of leadership as such. Contrary to this, the integrative model appears to be committed to a more traditional individualism (even when Behrendt et al. briefly draw on group psychology, they characteristically restrict their focus to the properties of the leader that may enable them to—unilaterally—influence others (Behrendt et al. 2017: 36). It thereby not only matches the increased complexity of organizations, but also tallies with recent theoretical developments aimed at capturing the social and distributed nature of intentionality, agency, and knowledge (Huebner 2013; Hutchins 1995; Klausen 2015; List and Pettit 2011; Tollefsen 2002). Moving away from individualism in this way also calls for a more multidisciplinary approach because it entails that psychology alone (including social psychology) is far from sufficient to understand leadership.

Second, by doing so, it stakes a better claim to be a theory of genuine leadership behavior (what leaders actually do, intentionally or unintentionally), whereas Behrendt et al. seem to be concerned with the much narrower concept of action or agency. This is especially pertinent in their adherence to the “Rubicon model,” which is formulated in terms of intentional action (for an instructive account of the distinction between action/agency and behavior, and a convincing argument for the practical importance of the latter, broader notion, see Dretske (1988)).

Once acknowledged, the myriad of perceptions and contextual variables raises further methodological questions that cannot be solved simply by mixing methods. How exactly to mix and balance the different kinds of data is an open question, one that reflects the deeper ontological question of how the diverse perspectives of managerial and non-managerial employees, and diverse views of researchers as well as other people inside and outside the organization contribute to co-determining leadership behavior. Without providing any specific answer, understanding leadership behavior in this way may also indicate how to proceed with framing and responding to the questions. This is because the weights assigned to different factors ought to reflect their contributions, and knowledge of this can be gained only by studying perceptions and meta-perceptions in their specific contexts.

Conclusion

Organizations are evolving to take more complex forms and many organizations, like public sector organizations, are even characterized as hybrid forms of organization (Manning 2013; Stacey 2012; Uhl-Bien and Marion 2009; Uhl-Bien et al. 2007) that have an increasing complexity because of the involvement of many different stakeholders, and the pursuit of multiple and conflicting goals. Important about this is the fact that to study leadership in this kind of organizations, a contextualized view, considering multitudes of stakeholders and associated expectations is key. In line with this perspective, we have argued against previous studies on leadership behavior that have prioritized the actions of individual leaders (Uhl-Bien et al. 2007). We also argue against research that treats the leader–follower relationship as if it were in a vacuum and neglects the context (Hunt and Dodge 2000) (for a phenomenologically inspired analysis of the relevant sense of “context,” see Dohn et al. (2018)). The literature mirrors the fact that a transformation, as a paradigm shift, has occurred from leader–follower perspectives to more bi-lateral perspectives of leadership with mutual influence. From our point of view, research on leadership behavior has not followed this fundamental shift because we are still measuring perceptions in a leader–follower relation. Leader–member exchange theory (LMX) comes close due to the importance of interpersonal relationships (and the perceptions of them). Nevertheless, LMX is still static in its approach as the contextual variables are not constituent variables of leadership behavior. We have aimed here to establish the need for research on leadership behavior that considers the increasing complexity of organizations by putting leadership back into the context of organizations. We have argued that contextual variables, together with the dynamic relationship and the mutual influence that leadership behavior and perceptions of it at multiple levels have on each other are essential to understand leadership behavior. We have taken some inspiration from phenomenology, without committing ourselves particularly to this approach or its ontological and epistemological implications. Our approach is proposed to eliminate some of the shortcomings in the model offered by Behrendt et al. (2017) as well as LMX theory. We acknowledge their contributions to solving some of the problems in previous research by criticizing researchers for not distinguishing between leadership behavior and perceptions thereof. We also acknowledge the strengths of Behrendt et al.’s integrated model, in which they build on and elaborate the lifelong work of Yukl (2002, 2012, 2013; 2005; 1992). Still, we have argued that they tend to exaggerate and overuse the conceptual distinction between leadership behavior and perceptions of it, and the research model that they develop based on this distinction misses important aspects of leadership in practice. Organization members are affected by their relations to one another and by the context in which their relations are established and maintained. This plethora of factors affects not only how leadership behavior is perceived, but also how it is formed. Perceptions cannot be ignored or bypassed because they are an indispensable source of empirical information and are, to some extent, co-constitutive of leadership behavior itself. Insofar as we have succeeded in arguing for our approach, it has methodological consequences that may make researching leadership behavior more challenging. Researchers need to mix methods, measure at different levels, and, perhaps most importantly, accept that what we measure may not be considered completely replicable by other researchers. Organizational life and, thus, leadership behavior are a flux—constantly changing and characterized by multiple actors and variables. Acknowledging this flux does not make the task of leadership behavior, or of studying it, easier, but it is a challenge that must be dealt with, nevertheless.