Article Component Papers (Leadership Measurement)

The Leadership Quarterly 21 (2010) 563–575

Contents lists available at ScienceDirect

The Leadership Quarterly

journal homepage: www.elsevier.com/locate/leaqua

Theoretical and Practitioner Letters

Issues in operationalizing and comparing leadership constructs

George Graen a,⁎,1, Jens Rowold b, Kathrin Heinitz c

a University of Illinois, Urbana-Champaign, United States b University of Münster, Germany c Free University of Berlin, Germany

a r t i c l e i n f o

⁎ Corresponding author. E-mail addresses: [email protected] (G. Graen), je

1 Retired.

1048-9843/$ – see front matter © 2010 Elsevier Inc. doi:10.1016/j.leaqua.2010.03.016

a b s t r a c t

Numerous theories of leadership have yielded a rich array of leadership constructs, yet there has been little comparative work across theories. The authors of this letter exchange compare views around the definition of leadership and four issues related to operationalizing and comparing leadership constructs: the measurement of leadership constructs, overlaps in measures of leadership, follower's ability to differentiate among leadership constructs, and conceptualizing context in studying leadership.

© 2010 Elsevier Inc. All rights reserved.

Keywords: Leadership theories Leadership constructs Measures of leadership

Letter 1 Jens Rowold Professor of Human Resource Development and Change Management University of Dortmund Center for Continuing Education Hohe Strasse 141 44139 Dortmund Germany

Kathrin Heinitz Freie Universität Berlin Fachbereich Erziehungswissenschaft und Psychologie Habelschwerdter Allee 45 14195 Berlin Germany

Dear Jens and Kathrin:

Thank you for taking the time to discuss several issues concerning theory and measurement of competing leadership approaches. I believe we agree that such a scholarly exchange of these issues is long past due. We need to tidy our area by cleaning out the unnecessary ambiguity and complexity in our scientific methods and getting down to the “nitty-gritty” of appropriate measurement. Let me begin with my definition of leadership and then move to the four issues that we agreed to discuss.

For purposes of this letter, leadership is defined as informal emergent influence that can be activated by problem situations demanding leader–member cooperative, extra-role action much beyond the ordinary action feasible from formal supervisor and subordinate or coworker and coworker problem situations, e.g., “when the beehive hits the fan.” In such leadership activating

[email protected] (J. Rowold), [email protected] (K. Heinitz).

All rights reserved.

564 G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

situations, someone steps forward with an action plan and those committed to the emergent leader's efforts willingly follow. Such a phenomenon may involve leader(s) and followers, but not all formal subordinates need be motivated by a leadership relationship or network of relationships and join the cooperative effort. Relational attributes required for successful emergent leadership include critical strong ties (relationships) within the emerging team between leader and followers (e. g., LMX) or partner and partner (e.g., MMX). For example, the LMX ties are characterized by mutual respect for competence, trust in motives, and commitment to the welfare of each other and the mission (Graen, 2007a). Hackman (2002) lists additional helpful attributes of successful leadership teams including clearly defined objectives, effective internal decision processes, adequate resources and authorities, and formal leader as backstop. Finally, the culture of an organization toward the how, when, where, and whys of emergent leadership is critical. Organizational cultures vary from absolute prohibition to trained encouragement of emergent team leadership (Graen, 2007b). In this approach, leadership being an extra organizational phenomenon, formal leadership is an oxymoron. Formal managers also can be leaders, but leaders are selected by those who would follow and not by higher management.

According to this definition, leadership is different than supervision or management of groups of people within an employment situation in that it requires both the leader and followers to go beyond organizational and personal constraints of business as usual and engage in active extraorganizational collaboration. Even outstanding group or team performance by itself does not necessarily indicate the presence of leadership. Recognizable leadership becomes apparent to outsiders after it has emerged as people acting beyond their pay grade toward a common goal. One must understand that leadership should be seen as an extraordinary process. One should not call a person “leader” until he or she has developed the capability to take a committed crew on such an extra-job exceptional voyage. Hackman's (2002) concept of a team leader is one of a “backstop” who lets the team perform without outside intervention or disruption from any lack of resources or red tape. Only when the team seems about to be clearly blocked does the leader emerge to overcome the offending barriers by putting team above self. Using the technical term of “team leader” for any executive performing business as usual duties may be flattering, but this honor should be earned by past performance or clear and present team readiness for leadership emergence.

Finally, the famous statement by Ralph Stogdill that “leadership research” is an American phenomenon may have been descriptive back in the 1950s, but it is far from accurate today. Although leadership may be defined somewhat differently by other nations, scholars from every continent differentiate manager from leader much like legal commander vs. volunteer servant (Gelfand, Erez, & Aycan, 2007).

Issue one: Measurement of leadership constructs

At the Universities of Minnesota and Illinois, we taught students that the measurement of leadership constructs requires that the operations taken to define the construct should map isometrically on the nomological network of the hypothetical construct as closely as feasible. The degree of overlap between the construct and its operational definition was called the epistemic correlation and varied from 0.00 to +1.0, and we should seek the strongest positive correlation. Moreover, scientific hypotheses link the theoretical relationships of at least two different constructs in terms of isometrically strong operational definitions (measures X and Y). Hypotheses are tested in the empirical world using relationships between psychometrically valid measures. Theories with low epistemic relationships are held in limbo as untestable. Although we cannot argue with empirical results, we may disagree with the epistemic correlations and the research methods.

Only with psychometrically sound measures can researchers hope to test a theoretical hypothesis derived from a formal theory. For example, LMX (leader, follower, and team versions) and MMX measures have all been tested empirically and found to possess the appropriate psychometric characteristics, such as reliability and content, concurrent, predictive, and construct validity (Gerstner & Day, 1997; Naidoo, Scherbaum, & Goldstein, 2008; Scherbaum, Naidoo, & Ferreter, 2007). Moreover, these necessary psychometric characteristics used in our functional theory building of leader–member exchange theory have been demonstrated by both predictor and criterion variables. On the dark side, many games may be played with these processes, but they are usually recognized by journal reviewers as Dunnette's “fads, fashions and folderol.” Clearly, many ways exist to get it wrong, but the Platt's strong inference approach helps researchers and theorists to get it right.

Measurement of leader style has a history in the United States from the founding fathers' search to identify characteristics and conditions of emerging leaders. They wondered how such leaders convinced people to become followers to the extent of risking life and limb for a common cause. These questions received rigorous research in the 1950s and 1960s at Ohio State University and the University of Michigan where researchers developed psychometrically sound measures of effective first-line supervisor's behavior in production and service industries in the United States. Ohio State's measure was called Supervisory Behavior Description Questionnaire (SBDQ). Instructions for the SBDQ were to have at least three randomly selected subordinates describe their supervisor's behavior on the job using the questionnaire, and the mean score of the three or more subordinates was the score for each supervisor. The logic of this was that the immediate subordinates of the supervisor possessed the most valid first hand information about the leader's behavior and would give the most accurate descriptions because social desirability pressure was least on them. As insurance against relationship bias, the subordinates were to be picked randomly and the individual reports of at least three were averaged. An example of this proper procedure was recently published (Sosik & Dinger, 2007).

These measurement procedures are still required for all measures of leader style based on follower descriptions. In spite of this, many studies using the Multifactor Leadership Questionnaire (MLQ) have violated this accepted measurement procedure by not averaging. By deleting this necessary step, what is measured is not leader behavior toward the group, but is contaminated by subordinates' perceptions of the leader's relationships with them. If your theory specifies average leader behavior, then averaging is necessary. Remember that the research breakthrough concerning average leadership style (ALS) measurement was that the

565G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

discarded dyadic leadership relationship variance was found to be a major contributor to a leader's influence with his or her team members in addition to leader behavior toward the work group (Graen, 1976).

Issue two: Overlap of measures of leadership

A second issue concerns the overlap among leadership measures that have sufficient data bases to be subjected to meta- analysis. For example, you found in a recent study that the transformational scale of MLQ-5X and the charisma scale by Conger and Kanungo (CKS) correlated .88 (pb.01) and Wang and associates (Wang, Law, Hackett, Wang, & Chen, 2005) found that MLQ-5X and LMX correlated .71 (pb .01). Other studies measuring both LMX and MLQ-5X found similarly high correlations (Graen, 2003). The measured overlap between the two average leadership style (ALS) measures of MLQ-5X and CKS with the dyadic measure of LMX can be expected to decrease with the increasing sample size of unit members included in the average. This is expected because the greater the size of each unit's sample, the less the ALS measure will tap the dyadic relationship of LMX.

In contrast, the overlapping variance between the two ALS measures represents a significant problem conceptually. Is charismatic behavior theory the same as that for transformational behavior? Yammarino, Spangler, and Dubinsky (1998) suggest that, yes, the core of transformational leadership is almost entirely charismatic leadership. If yes, then we should integrate the two conceptually. We should remember that forcing any two measures into separate factors is not a useful test of conceptual overlap, because as long as their correlation is less than unity they can be forced to define different factors via factor analysis. In my opinion, for all practical purposes, MLQ-5X transformational scale (with its five subsets of items as one) and CKS (with its five subsets of items as one) show no differences and should be seen as parallel measures of the same construct of charismatic leadership. If the transformational and charismatic leadership researchers wish to make their MLQ-5X and CKS dyadic and different from each other, they need to have new theories that explain why this should be appropriate. Is it something about the leader or the follower or the relationship? If it is the relationship, okay, we can use more parallel measures.

Issue three: Can followers differentiate between leadership constructs?

Only four published studies have included measures of LMX and of other leadership constructs, and only Wang et al. (2005) showed how transformational and LMX leadership theories are related to each other and work performance. Their studies found that the regression of the descriptions of leaders “transformational” relating to followers and work performance was fully mediated by the LMX relationship. Although their structural equation revealed a coefficient of .80 (pb .01) from transformational to LMX leadership, only LMX was significantly related to both outcomes of organizational citizen behavior (OCB) and task performance. These findings suggest that the behavior of leaders called “transformational” can be related to OCB and task performance. However, this also suggests that some transformational behaviors of leaders are relevant and some are irrelevant to augmenting follower performance and those behaviors that are relevant are related through LMX relations. The question of their study was: Is transformational different from LMX in terms of common variance and in terms of unique predictive variance? Results showed that the transformational style contributed unique valid variance to the prediction of both in-role and extra-role performance. When different leadership constructs each contribute significant predictive power, they are seen as tapping different variance. I recommend this test as useful for answering part of our question above.

Issue four: How should situations be categorized for investigating the role of context in leadership?

As shown in Fig. 1, management problems may vary from safe to dangerous, from routine to complex, and from clear to ambiguous. Placing these three dimensions orthogonally results in eight cells. Each octant has a prescribed leadership structure that contains different distributions from leadership-sharing to leader-following processes. The efficient production situation of clear-routine-safe is the most researched area followed by the fire-rescue, police, or military leadership situation of clear-routine- dangerous. We are beginning to research the clear-complex cells due to the new problems emerging from the knowledge economy. The four ambiguous cells are the challenges of the future. Leadership functions in each of these structures may be different as suggested in the figure.

Considering these eight cells of different problem scenarios, what student of leadership would prepare one team to handle all of these? Clearly, each cell demands different structures to deal with its unique challenges for team leadership. For example, a team trained to solve problems in safe-routine-clear situations may be overwhelmed by an escalation in any one of the three dimensions which is much less than an increase to dangerous-complex-ambiguous. What is needed is thinking outside of the box and seriously considering that problem situations force appropriate problem solving structures to be invented by researchers and kept ahead of the obsolesce curve by proper bursts of new leadership.

I look forward to your advantageous European perspective on our attempts to better understand this critical process within our network-centric societies. Our love–hate relationships with our leaders continue to be significant in our lives. We struggle to love them more and hate them less, but to do this we need a better understanding of how to select candidates, develop them, and keep them putting service above self. I await your scholarly reply.

Respectfully yours, George Graen

Fig. 1. Octants of structures recommended for combinations of dangerousness, complexity and ambiguousness of large management problems.

566 G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

Letter 2 George Graen Professor of Organizational Psychology and Labor and Industrial Relations University of Illinois, Urbana-Champaign (Retired) 10819 Graen B Point Beaver Lake, AR 72745

Dear George:

Thank you for your letter concerning theory and measurement of rival leadership approaches. It even intensifies our impression that an exchange on different views concerning leadership assessment and level of analysis is due. We hope that this issue will not only be a topic in our letter exchange but will encourage further open discussion and empirical research. We were struck by the fundamental differences that we appear to have. Therefore, in turn, we will provide our arguments for each of the four issues that you articulated in your letter.

But before turning to these topics, we would like to give our view on your definition of leadership, as we clearly disagree at this point. In your definition, you focus on emergent leadership. We agree with you that outstanding leadership is more than supervision or management of groups of people. Yet, in most organizations, the leadership role is also defined formally by, for example, the job position. Hence, leadership can be cast either as a formal or an informal role. However, in most work settings and for most cases of daily work routine, subordinates rely on their formal leader. Thus, we think it is also important to focus on day-to- day management and supervision in order to contribute to the important question of “ordinary” leadership. That said, we think that within recent leadership constructs such as transformational and charismatic leadership, several elements focus on the extraordinary side of leadership. For example, in Conger and Kanungo's theory of charismatic leadership, the factors of unconventional behavior and personal risk are examples for this. With regard to your suggestion that informal leadership is important, we would argue that the phenomenon of informal leadership could (and should) be explored by implementing established leadership scales such as the Multifactor Leadership Questionnaire (MLQ-5X) or the Conger and Kanungo Scales of charismatic leadership. For example, in a group setting, each group member would be assessed by these instruments at several points-in-time in order to identify the emerging group leader.

Issue one: Measurement of leadership constructs

We agree that only by using psychometrically sound measures can we test hypotheses and hence test and refine our theories. Likewise, we acknowledge the good psychometric properties of the LMX scales that you mentioned. However, this issue remains problematic for at least two reasons. First, if not for LMX, then for many other leadership constructs, the factorial validity is a major concern. In essence, this raises doubts about the construct validity of these theories of leadership. Second, both theory as well as

567G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

assessment tools for LMX are limited as they only focus on one single perspective on the leader-led relationship. Clearly, the last decades of leadership research demonstrated the importance of many other perspectives as well as leadership factors.

In your letter, you mentioned the necessity to control for relationship bias when dyadic leadership style is assessed. We think that two problems are posed by your arguments, namely, the problem of the analysis level and the problem of positive affect.

First, as for dyadic versus group level leadership, we could not follow your definition of leadership. In your definition of leadership style and in your argument regarding level of analysis, it is assumed that one leader shows the same behavior towards all subordinates. However, both transformational leadership theory and empirical research demonstrate that transformational leadership is mainly a dyadic phenomenon. Most notably, Yammarino and colleagues (Yammarino, Dansereau, & Kennedy, 2001; Yammarino, Spangler, & Dubinsky, 1998) demonstrated that transformational leadership is mostly a dyadic level phenomenon. Apparently, leaders treat followers individually, as every follower has different needs, motives, abilities, and so on. It should be noted that—as other empirical work has demonstrated—transformational leadership is also related to outcome criteria at the group level (e.g., profit).

Also, it is important that transformational leadership is measured at the dyadic level as this pays respect to one important context (or situational) factor: Leadership style varies across subordinates. For example, in one workgroup there may be subordinates with different gender, age, race, job position, etc. Effective leaders will pay respect to these contextual factors and thus exhibit different leadership behavior towards these different subordinates. Also, there is now a large amount of empirical research that demonstrates that the effects of transformational leadership are moderated by followers' characteristics such as romance of leadership, personality, etc. Finally, if one takes leadership as an attributional phenomenon, it becomes evident, that, on the one hand, the followers are the best source for leadership assessment, and that, on the other hand, leadership needs to be assessed individually because everyone can attain different attributions.

It should also be noted that it is the merit of LMX to have emphasized the dyad perspective. Since the 1980s, increasing empirical evidence emerged that demonstrated that the dyad between leader and led is slightly more important than other perspectives (e.g. group level) for any leadership construct. Thus, the question emerges: Is LMX synonymous with the dyadic leadership perspective? Do all leadership constructs (e.g. transformational leadership, initiating structure and consideration) have some effects on the dyadic and also on the group level? In sum, we think it might be necessary that all leadership theories should explicate their dyadic and their group level effects. For example, in our recent Leadership Quarterly publication (Rowold & Heinitz, 2007), we analyzed the effect of transformational, transactional, and charismatic leadership on both the individual and the group level. On both levels, transformational leadership showed stronger relationships to outcome criteria than the other leadership styles.

Second, as for the affect towards the leader, we postulate that the emotions between leader and led are important for leadership research. For example, Brown and Keeping (2005) demonstrated that questionnaires for the assessment of leadership behavior are heavily influenced by followers' positive affect towards the leader. While these first empirical results are noteworthy, we only have begun to explore the effects of emotions on leadership style assessment. Nevertheless, these empirical results have serious implications for leadership research in general. One might ask critically “To what degree are leadership styles independent from positive affect?” Especially in one or two-dimensional approaches to leadership, such as LMX or consideration/initiating structure, it has to be empirically proven that LMX is in fact different from positive affect (i.e. liking). Otherwise, we see the construct validity of LMX in danger, that is, LMX might be redundant to an emotional-colored, dyadic leadership perspective. On the group level, it could be critically asked if LMX is different from positive team climate.

In sum, the two methodological advancements mentioned above (i.e. levels of analysis and controlling for positive affect) allow for a more critical “compare and contrast” approach to leadership theories. As in every field of science, these varying leadership theories should be compared with methodological rigor. Furthermore, a network perspective and time-lagged research can help to understand leadership processes.

Issue two: Overlap of measures of leadership

In your letter, you addressed the issue of the overlap between LMX and transformational leadership. You said that “[…] the measured overlap between the […] average leadership style (ALS) measures of MLQ-5X […] with the dyadic measure of LMX can be expected to decrease with the increasing sample size of unit members included in the average. This was expected because the greater the size of the sample of each individual unit, the less the ALS measure will tap the dyadic relationship of LMX.” From our perspective, two problems exist with this proposition. First, this proposition still has to be supported by empirical data. Second, and more importantly, it could be used as support for our argument provided above that leadership is more of a dyadic phenomenon. From a methodological perspective, effects of team size should be explored in future research as a contextual dimension of leadership.

Also, you mentioned that the conceptual overlap between similar approaches to leadership (e.g. transformational and charismatic leadership) would be problematic. Overall, we would say that this is an understatement. Clearly, we agree with you that transformational (as assessed by the MLQ) and charismatic (as assessed by the CKS) leadership should be integrated at the individual level. Several subscales of these constructs are conceptually close and reveal similar item content. The Full Range Leadership Theory is an excellent base for conducting this kind of theoretical—and consequently, empirical—research. However, we go one step further and also call for an integration of other leadership styles such as consideration and initiating structure, LMX, and so forth. From a methodological perspective, we can use the tool of factor analysis to pursue this goal. For example, one advantage of the Ohio State and University of Michigan studies was the effective approach that was based on factor analyses. This

568 G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

class of statistical procedures allows for a closer examination on the factor structure of any given construct. Within leadership research, this approach is urgently needed as several conceptually overlapping theories have evolved separately in the last decades. This often led to an uncritical encompassing of new leadership theories.

However, it should be noted that this does not mean that one discards one paradigm for the other. You are absolutely right in pointing out the high correlations between existing leadership paradigms. However, previous research shows that aspects of other leadership styles add to the prediction of outcome criteria in addition to for example transformational leadership. Hence, certain subscales of charismatic leadership (i.e., CKS) are valuable for transformational leadership (i.e., MLQ), as are new additions to the Full Range Leadership Theory that have been proposed recently by other scholars. Hence, and we hereby also follow the argument of Wang et al. (2005), MLQ and CKS are not only parallel measures of the same construct. The first step in the process towards a new leadership model, however, has to be a theoretical integration of those leadership constructs.

Our colleagues John Antonakis and Robert House (2002) conducted empirical research to further extend the Full Range Leadership Theory. They argue that additional leadership behaviors such as instrumental leadership have to be included into this at the moment most encompassing leadership paradigm. From our perspective, this is a very valuable avenue of research as this “compare and integrate” approach will combine and extend effective leadership paradigms.

The main point of recent research is that we do not yet know whether (a) leadership constructs overlap or (b) subscales of leadership scales are redundant to each other. A lot more research is necessary before we can arrive at a more advanced leadership theory. We can by no means agree with your statement that it might be suitable to view the MLQ and the CKS as parallel measures “for practical purposes.” First comes theory, then research, and then practice!

In sum, it is time for a new Ohio-State-type study in order to arrive at a more condensed set of leadership factors. These studies could yield condensed sets of leadership factors. Nevertheless, in addition to these analyses which would focus on factorial (and construct) validity, the criterion-oriented validity of these (competing) leadership factors has to be established. That is, rivaling leadership styles should be retained even when they are highly correlated to other styles if they contribute to leadership effectiveness over and above other leadership styles. As we are currently conducting such an Ohio-State-type study, we will return to this issue later.

Issue three: Can followers differentiate between leadership constructs?

We agree with you that the effects of transformational leadership might be mediated by LMX. However, from our perspective, the empirical research that you cite cannot support this assumption. In your argumentation, you put considerable weight on the Wang et al. (2005) study. However, several major methodological weaknesses of this study should be kept in mind. First, within their study, these authors assessed both leadership constructs at the same time. From our perspective, this methodological procedure does not allow for a strong test for the mediating effect of LMX. The key question then is whether the order of constructs could be reversed: Does transformational leadership mediate the relationship between LMX and performance? Clearly, longitudinal or experimental studies are needed before we can assure any causal order of constructs.

Also, these authors never tested whether followers could differentiate between the leadership styles that were assessed in their empirical study. We would suspect that characteristics that are common to both LMX and transformational leadership (e.g. positive affect, sympathy, or positive communication quality) are the reason why we cannot distinguish between these two leadership constructs nor establish a mediating effect for any of these constructs at the individual level. With regard to the arguments provided for issue #2, we would argue that LMX is a theory at the group level, but not for the individual level. However, as transformational leadership and LMX are highly intercorrelated at the group level, the question emerges whether it would be better to merge transformational leadership and LMX at the group level.

As we already discussed in issue two: We do not know very much about the potential overlap of rivaling leadership styles. History of leadership research has demonstrated that from the followers' perspective, only very few leadership factors can be distinguished. In addition to the famous Ohio State studies, the Full Range Leadership Theory arrived at transformational, transactional, and non-leadership, hence three factors (with nine subscales). The key question is: Can followers distinguish between more than two or three leadership styles? If yes, how many leadership styles are there? If no, is it due to followers' cognitive limitations or to the nature of leadership? Determining the number, content, and criterion-related validity of leadership styles would not only advance leadership theory, but would also have important implications for leadership practice (e.g. leadership assessment, feedback, and development).

Overall, we had the impression that you perceive and discuss transformational leadership through the LMX lens. In contrast, we try to see both theories as separate theories that are worth integrating. We assume that LMX is an affect-based outcome perspective (i.e., quality of the relationship “in effect”) of the leader-led relationship, whereas transformational leadership is a behavior process (i.e., what forms the relationship) perspective.

Issue four: How should situations be categorized for investigating the role of context in leadership?

We applaud your contextual model of leadership as the issue of context is one of the key leadership issues and has been neglected considerably in the past. Thus, models like the one that you suggested are needed in order to fully understand the contingencies of leadership effectiveness.

However, we need a lot more basic context factors. For example Antonakis, Avolio, and Sivasubramaniam (2003) were able to provide evidence that (leader's and follower's) gender is one important contextual variable. Other “basic” context factors are age,

569G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

heterogeneity of the work group, and so on. While one could take virtually all possible demographic as well as other factors and combine them into a “leadership contexts” list, we believe that basic factors such as gender and age are most important for future leadership studies.

We encourage researchers to collect as many contextual factors as possible, most notably demographic characteristics of both follower and leaders, and information about the work situation. As we are currently conducting a large-scale research project that addresses the issue of context (as well as the other issues we are addressing in this letter exchange), we briefly introduce you to this project.

The project Context and Leadership Styles (CoLeS) was initiated in order to close several of the above mentioned research gaps. First, in all empirical data assessments within this project, indicators of major leadership theories are assessed. Thus, scales for the assessment of transformational, transactional, laissez-faire, and ethical leadership as well as initiating structure, consideration, and LMX are implemented. The goal is to contribute to the factorial and construct validity of these leadership styles. More importantly, the empirical overlap and thus the nomological network of leadership theories is explored. We hope that through the assessment of several major leadership theories, a “fuller” Full Range Leadership Theory, to say it in Antonakis and House words, will be feasible.

As recent empirical work demonstrated that the assessment of leadership styles is heavily biased by positive affect, we control for this construct in all our data assessment projects. We think it would be very important to see which constructs are related—or redundant—to positive affect and which are independent hereof (i.e., behavioral), as this would allow for a more thorough understanding of the role of positive affect in the leader-led relationship. We assume that the high correlations between leadership styles that have been frequently observed in past empirical research will decrease if positive affect is controlled for.

Along with the above mentioned leadership styles, we also assess outcome criteria such as commitment and performance. In essence, our goal is to establish the criterion-oriented validity of the various leadership styles. As we follow a compare and contrast strategy, we aim to test the relative influence of various leadership styles on outcome criteria of the leadership process.

Finally, as the issue of context is of predominant importance, we assess as many context factors in our empirical research as possible. For example, we always include leader's and led's demographic characteristics (e.g. gender, age, and education) as well as characteristics of the work (e.g. profit vs. non-profit, heterogeneity of the work team, and team size) into our lists of variables. Our strategy is to test the contextual boundaries of effective leadership: Which leadership style will be associated most closely with various indicators of effective leadership under which contextual conditions? Which leadership styles are best within heterogeneous work groups? Are there more effective leadership styles in high uncertainty environments? With regard to this last question, we see your model of leadership contexts as a valuable theoretical framework.

In sum, we hope that this research will shed more light on the issues that—as this letter exchange reveals—are not solved as yet. We are therefore looking forward to your reply.

Yours sincerely, Jens Rowold and Kathrin Heinitz

Letter 3 Jens Rowold Professor of Human Resource Development and Change Management University of Dortmund Center for Continuing Education Hohe Strasse 141 44139 Dortmund Germany

Kathrin Heinitz Freie Universität Berlin Fachbereich Erziehungswissenschaft und Psychologie Habelschwerdter Allee 45 14195 Berlin Germany

Dear Jens and Kathrin:

Thank you for your response to my first letter in our exchange concerning our agreed on four issues. I am delighted that we agree to disagree in such a cordial yet clear and direct manner. I also hope that our exchange will stimulate some critical thinking about asking the right questions and following strong inference methodology so that the meta-contractual force that we call “emergent group leadership” (EGL) does not end up on the clipping room floor or in the waste bin.

I would like some closure on your position on designating a formal position the “team leader.” My question is, if an organization designates you as a formal leader, but no one follows you, what are you? I don't think that you can command your subordinates to become followers. I appreciate your right to view my definition of “emergent group leadership” as too exclusive, but please understand that Ralph Stogdill's infamous 62 different definitions of leadership in the United States included everything and thus excluded nothing. Kerr and Jermier (1978) were prophets when they foresaw that leadership defined as “all influence on people in

570 G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

organizations” was all encompassing and they protested. In an attempt to save something of value from this metaphysical wastebasket, our LMX leadership group decided to emerge from the pack and attempt to save the soul of the live baby that we had personally witnessed in performing our research, namely, “emergent group leadership.” To make our theory clear, as Jim March suggested, we purposely defined “managership” as people executing their employment contract responsibilities which often includes organizational change and other responsibilities but excludes emergent group leadership.

According to the Ohio State model, effective supervisory behavior toward their work group was a combination of two clusters of behavior: “initiation of structure” on subordinates and “giving them due consideration.” When transformational leadership and charismatic leadership were first introduced they used the same concept of consistent behavior of a leader toward all subordinates in the work unit (House, 1977). The consistent charismatic behavior specified by House were (a) sets strong role model, (b) shows competence, (c) articulates goals, (d) communicates high expectations, (e) expresses confidence, and (f) arouses motives. This was an “average leadership” model (ALS) clearly as pointed out by me and my students at the University of Illinois (Graen, 1976). In addition to this ALS, our team found a new leadership construct which is now called “Leader–Member Excellence Working Relationship.” LMX was measured at the vertical dyadic level from both up and down perspectives. What our team's research showed was that both the charisma of the leader and the LMX between leader and follower helped predict over time many leadership outcomes from work and job attitudes to organizational citizen behavior to behavior to performance and turnover (Graen, 2003). The measure of average leadership style (ALS) I prefer is described by Kark and Van-Dijk (2008) under which leaders are oriented toward either promotion of team mission through developing talent or protection of team quality and quantity standards. Other ALS constructs has been researched including solidarity and paternalism.

Since our series of longitudinal studies published in the late seventies, both transformational and charismatic leadership researchers have stopped averaging and began use dyadic level data. However, it's not that easy to take a theory about leader's animal magnetism and make it about leader–member relationships. Clearly, supervisor can perform the behaviors that they think will give them high evaluations on the five subscales defining transformational leadership and on those defining charismatic leadership, but if their subordinates are not paying attention, they will not follow. A supervisor may earn subordinate's attention after they have begun developing a more engaging future through a promotion-oriented approach. We have a mountain of studies that are consistent with our relationship definition of leadership (Uhl-Bien, 2006). We welcome the researchers of charisma, transformational, path-goal, solidarity, structuring and consideration, grid, paternalism, LPC, team excellence, pygmalionism, and others to go dyadic with their measures after they retrofit their theories to the relationship definition.

Issue one: Measurement of leadership constructs

I enthusiastically agree with the necessity of using psychometrically sound measures to test hypotheses and refine our theories. I also agree that it would be tidier if theoretical components of our measures were independent of one another such that a clear factor structure would appear. Unfortunately, nature appears more complex, for example, the big three underlying the LMX construct are highly correlated in the work contexts that have been studied. We should not throw out the baby with the bath water because our components of a construct are highly correlated. The questions is can they be separated by experimental design without changing their definitions. This was done by Sweeney (2007) for LMX. Sweeney interviewed U.S. Army combat teams in Iraq about what components of their relationships with their leader they have and related this was to combat effectiveness. Three factors emerge from the analysis of components: respect for competence was first, trust in welfare of troops was second, and commitment to the welfare of people in the combat team was third.

Thank you also for agreeing that we can put trust into the LMX scales. I agree that many other leadership constructs need to be subject to similar psychometric analyses as were the LMX measures. I disagree that LMX is limited to a single perspective (Naidoo et al., 2008). LMX has been measured repeatedly over time from multiple perspectives of top–down, bottom–up, and outside peers.

Your mention of the Yammarino, Spangler, and Dubinsky (1998) article reminded me of their taking the MLQ and factoring it down to nine items: five for transformational leadership and four for contingent reward managership. Their five core items of transformational leadership were: (1) I have faith in person, (2) person's vision spurs me on, (3) person has a special gift, (4) person is symbol of success, and (5) I trust person to overcome obstacles. The first two were from the inspirational scale and the last three were from the charisma scale of MLQ. This description looks like animal magnetism or charismatic leadership.

Thank you very much for noting the confusion surrounding the two issues of level of analysis and positive affect. Your position that average leadership style is typical behavior within a unit, but interacts with member's behavior to produce something different for each dyadic relationship is the definition of the LMX working relationship that depends on members' independently determined plans for their careers. If a supervisory style is to share or not share with network members, success in each dyad depends on the compatibility of the leader's style (promotion or protection) and the member's career plan. When members choose not to grow out of their jobs, managers cannot command and control them to do so. Research shows that the big three are antecedent conditions for members to agree to grow (Gerstner & Day, 1997).

Recent research on engineering project teams shows that in three separate longitudinal investigations, group means (including all up, down, and horizontal relationships rated by everyone on the team) strongly predicted team performance and team development on professional design projects for client organizations. These findings suggest that relationship development can facilitate team performance and demonstrate a time delayed contribution over leader's ability (Naidoo et al., 2008).

571G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

As for the affect toward the supervisor question, LMX research has experienced the empirical problem that many items tested as candidates to add to our LMX measures needed to be rejected because they correlated highly with satisfaction with supervisor but were not predictive of performance. It seems that many aspects of one's working relationship with one's immediate supervisor have positive affect but few validly predict performance be it hard, soft, organizational citizen behavior, or termination. Therefore, the items we use to assess the quality of the relationship are predictive of both performance and satisfaction with one's supervisor.

Issue two: Overlap of measures of leadership

I agree that a strong inference approach to compare and contrast measures in the same study is to be recommended for future research. We also need go back according to the strong inference approach to tidy-up our theories to separate charisma and transformational and all of those mentioned above. We may well find that at the average style or work group level that structure and consideration accounts for most of the variance. We could test this.

At the dyadic level, we also need to compare and contrast in the same study different psychometrically sound measures of working relationships whether LMX partnerships, charismatic magnetism, transformational teaching, or consideration and structuring reaction evaluations by both leaders and followers. Such studies should use multiple methods for measuring each construct over time. These research designs should also allow the dependent variables of individual and team development and performance so that the relative contributions of each measure could be assessed using an importance analysis over time (Naidoo et al., 2008).

Issue three: Can subordinates differentiate between leadership constructs?

As I understand the Wang et al. (2005) study, the researchers ran their mediational analysis both ways: With LMX first and MLQ-5X second and the reverse order and in both analyses LMX mediated transformational. From LMX theory perspective, supervisor's transformational behavior toward subordinates cannot be related to performance or OCB unless the subordinates are paying serious attention to the missionary's pitch (strong LMX).

Finally, I disagree with your assumptions in your last paragraph of “Issue Three” in which you assume that LMX theory is fixated on outcomes and is not focused on the process that forms the relationship. For example, Naidoo and his associates (2008) found that over the life of the project team, ability was most important early in team processing and LMX-type relationships became most important near the end of the project in terms of project team performance and team development. This is clearly a focus on both process and outcomes.

Issue four: How should situations be categorized for investigating the role of context in leadership?

Thank you for the kind words about my contingency model of leadership contexts, but this is only a crude beginning. I invite researchers and theorists to expand, test, and refine this taxonomy. I think that we need to start to think about how organizations may change structures to better cope with changing contexts. How do top management teams with or without emergent change groups move from a safe, routine and clear context to a more complex one? To a more dangerous one? To a more ambiguous one? This is the new direction for LMX network theory. I applaud your project Context and Leadership Styles (CoLeS), because it appears to be a strong inference approach. I like your plan to control for positive affect. I suggest that you include OCB as a performance measure. Finally, may I offer a friendly caveat that the number of variables contained in a study at the upper values may reduce the quality of data collected, and let me wish you good fortune with your interesting and promising research project.

It has been my pleasure to correspond with you across the Atlantic. I hope that we have shed more light than heat and I agree that much more needs to be done to save the phenomenon that we call leadership from the wastebasket of untestable theories.

Respectfully yours, George Graen

Letter 4 George Graen Professor of Organizational Psychology and Labor and Industrial Relations University of Illinois, Urbana-Champaign (Retired) 10819 Graen B Point Beaver Lake, AR 72745

Dear George:

Thank you for your most interesting letter. We appreciate that you emphasized your point of view regarding the definition and importance of emergent group leadership. Taking into account that not only leadership research but also social psychology research contributed to this field clearly underlines its importance. However, we would like to emphasize that while ad hoc groups

572 G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

with emergent group leadership do exist, we should not forget the strong organizational factors that surround, form, and influence groups in vocational contexts. Answering your question of what you are if an organization designates you as a leader but no one follows, we could say this situation might be a good example for poor personal assessment and placement and that this leader might not stay in that position for long. While the phenomenon of emergent group leadership is, no doubt, important in itself, we need a more contextually rich theory of the processes and effects of leadership to explain the complex leadership phenomena in today's modern organizations. Effective leadership in today's organizations is shaped by contextual factors such as hierarchies, formal job positions, rules, traditions, and many more, that also provide a certain basis of power. Within this set of contextual boundary conditions, leadership—including emergent group leadership—exists. Hence, we would like to emphasize that only by taking these boundary conditions into account, we can possibly arrive at some thorough understanding of the leadership phenomenon. A paradox of leadership in organizations is: Someone is designated as a leader, and due to formal job positions, there are others who have to follow the new leader more or less. In short, the organization is a context that shapes the leadership phenomenon to a large degree.

Having said this, we would like to suggest a more pluralistic approach to leadership. As you said, numerous definitions of leadership exist. However, while you say that they end up in a “metaphysical wastebasket,” we would say that these definitions are valuable because each hints to important elements of the leadership process. You also mentioned that LMX was found in addition to the average leadership model. Over the last decades, various leadership theories such as LMX, transformational and charismatic leadership, as well as situational approaches, learned from each other and, as a consequence, evolved. Today, each of these theories can explain more of the leadership phenomenon than they were able to when first formulated (Hunt, 1999). For example: as LMX highlighted the dyadic perspective on leadership, other leadership theories learned from this perspective. Over the last 25 years or so, transformational leadership theory was researched at the individual, the dyadic, the group, and the organizational level (Bass, 1999). This breadth is valuable both for theory and practice. We agree with you in that we also would call for extensions of leadership conceptions in general so that they can explain leadership processes at the individual, dyad, team, and organizational level. Also, appropriate instruments for the assessment of the respective constructs are necessary.

Let us sum up: It is a natural process of science to first start with seemingly divergent approaches and mutually excluding theories. As a consequence, we have a set of partial diverging definitions of leadership. Subsequently, research can identify unifying elements underlying these various theories. Consequently, we believe that more integrative leadership research is necessary. Let's compare and contrast these seemingly diverging theories and definitions of leadership and see what's valuable and useful from each of them! As the field of leadership developed considerably over the last decades, we might arrive at an integrated, unified leadership theory some day.

Issue one: Measurement of leadership constructs

We agree with you that most measures for the assessment of leadership constructs have subscales that are highly intercorrelated. This is partly due to a theoretical overlap between different leadership constructs, but also displays a methodological phenomenon. Due to single-source, single method, and single point-in-time biases that afflict the “typical leadership study” (Hunter, Bedell-Avers, & Mumford, 2007), the correlations between subscales of leadership are mostly high. One solution to this problem is to do methodologically sound studies. For example, implementing different perspectives of leadership (e.g., self- and other-rating of leadership behavior) or multitrait-multimethod approaches are viable options.

While measures for LMX (i.e., we refer to the LMX-7 measure of LMX throughout this letter) are valid for many perspectives (e.g., top–down, bottom–up, etc.), they still focus on one facet of the leadership phenomenon: The relationship between leader and led. While this facet is of course a very important one, one can't neglect that other facets are important as well. For instance, from our perspective, transformational and transactional leadership might explain how leaders should behave in order to build up and maintain a positive relationship with their subordinates (i.e., high LMX). This positive relationship can motivate followers to perform beyond expectation. As one can see from this example, LMX is but one, though important, aspect in the overall leadership process. Nevertheless, we acknowledge the amount of research that has been conducted, demonstrating that the instruments for the assessment of LMX are valid from various perspectives.

Thank you for your argument regarding the issue of subordinates' positive affect towards the supervisor. We agree with you that many leadership constructs, not only LMX, are biased by positive affect. For example, we would speculate that at least several items from the recommended LMX-7 scale are implicitly associated with positive affect. In their meta-analytic study, Gerstner and Day (1997) found a correlation of r=.71 between LMX and satisfaction with supervisor. Also, in their experimental study, Turban, Jones, and Rozelle (1990) reported that supervisor liking positively influenced the supervisors' expectation of subordinates perception of LMX. Is there a theoretical rationale between differences between LMX and liking? This point has been made by other researchers as well (House & Aditya, 1997). With regard to parsimony, the question arises, does LMX-7 explain more variance in objective and subjective outcome criteria, over and above positive affect.

The underlying problem is that many questionnaires that were designed to assess leadership behavior explicitly or implicitly refer to positive affect towards the supervisor. Thus, researchers cannot distinguish between the behavior that was demonstrated by the respective leader and the positive affect this behavior elicits within the led. We agree with Brown and Keeping (2005) who argued for the inclusion of positive affect scales in every leadership research project. In addition to research which implements longitudinal or experimental designs, research is necessary that controls for positive affect.

573G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

Issue two: Overlap of measures of leadership

In your last letter you said that when one would compare various leadership styles, “we may well find that at the average style or work group level that structure and consideration accounts for most of the variance.” Clearly, we disagree here. As we are currently conducting meta-analytic research with Ron Piccolo, Joyce Bono, and Tim Judge in this area, let us present some of our findings. Most notably, we found that leadership styles are only partially redundant. Hence, when compared directly, several leadership styles (e.g., transformational leadership and consideration) show importance for predicting outcome criteria. That is, one leadership paradigm—either transformational and transactional or initiating structure and consideration—is not enough for explaining all variance in outcome criteria. Overall, we argue that for a more thorough understanding of the leadership process, we need parts of several leadership theories. This brings us back to our first two paragraphs of this letter: Each definition of leadership might appear divergent to other definitions, but this is both a problem and a solution. It is a problem in so far as we don't have the knowledge to understand all basic underlying dimensions and processes of leadership—not yet. It is a solution because it helps us to explain the effects of the leadership process such as performance and satisfaction.

In sum, and as we proposed in the introduction of this letter, each leadership theory has a value of its own. The “compare and contrast approach” to leadership provides one possibility to demonstrate the overlaps between leadership theories. However, more theoretical work is necessary before hitherto diverging leadership theories can be integrated. We agree to you that more empirical work is necessary. At each level (i.e., individual, dyadic, and group), measures of various leadership styles should be compared. Moreover, multiple perspectives or multiple methods and multiple points-in-time should be essential to these future studies as only this kind of studies will enable a real progress within the field of leadership research. We welcome your idea to utilize statistical procedures such as importance analysis. In the research we are conducting with our colleagues from the U.S., besides meta-analytical procedures, we use techniques such as dominance analysis (Budesko, 1993), which is to our understanding similar to importance analysis.

Thank you for informing us about the new study conducted by Naidoo et al. (2008). It was interesting to see that LMX predicts long-term team development. If one keeps in mind that leadership styles might be partially redundant, these results call for additional work. We would like to see research like the one provided by Naidoo et al., but with measures of additional leadership styles other than LMX. Only then we are able to see whether one respective leadership style is redundant.

Please allow us to shortly comment on your term “animal magnetism.” While you use this term to depict elements of transformational or charismatic leadership, we think the inclusion of emotional elements into transformational leadership in the 1980s was one important step in the evolvement of leadership research. While prior to the 1980s, the predominating approach within leadership research was to focus on leader's behaviors, the advent of charismatic and transformational leadership theories led to a focus on behaviors and emotional processes such as excitement. It was the merit of researchers such as House, Bass, Conger, and Kanungo to define charismatic and transformational leadership in scientific precise terms, and thus, rescue charisma from descriptions such as “animal magnetism.”

Issue three: Can subordinates differentiate between leadership constructs?

As you emphasize the importance of followers' attention in the LMX-theory, we would like to shortly discuss the matter of attention. On the part of subordinates, attention is certainly an important cognitive process necessary before subordinates can listen and follow. However, we would say that each of the predominant leadership theories such as transformational, charismatic, and initiating structure and consideration includes leaders' behaviors that elicit followers' attention. This process is certainly not exclusive to LMX. Furthermore, if you have attentive followers but what you do, say, or express has nothing to do with leadership, you will not reach a goal for all the attention you have.

Earlier in this letter, we discussed the overlap between the constructs of liking and LMX. Other leadership researchers summarized their concerns regarding the focus of LMX in this way: “…LMX theory may be more accurately viewed as a theory of dyadic relationships and their subjective consequences, rather than a theory that focus primarily on leadership” (House & Aditya, 1997; p. 435). In line with this statement, we would categorize the LMX perspective on leadership as having a focus on the affective outcome of the relationship between leader and led. It is difficult for us to see observable leadership behaviors being part of LMX. Instead, we would suggest that LMX is partially a result of certain leadership styles (e.g., transformational leadership, consideration). At the same time, based on the empirical research discussed above and on theoretical considerations, we see LMX partially redundant to positive affect. However, more process research including the construct of positive affect is necessary. Especially, we agree to your idea that longitudinal (or experimental) research is necessary to settle down this issue. As an example, the longitudinal study conducted by Keller (2006) where several leadership constructs were compared, is of great value. Only these kinds of research designs can determine within the longitudinal process of leadership, which leadership style influences which facets and outcomes of the leadership process.

Overall, we would subscribe to the statement by Hunter, Bedell-Avers, and Mumford (2007) who said: “Finally, with regard to leadership measures and their use, it is clear that we must be careful about the widespread implementation of pre-developed measures employed in the typical leadership study. It should be noted that it is not the widespread implementation of these measures, per se, that is problematic. Rather, it is the singular implementation of these measures that is potentially dangerous. By failing to pair current leadership tools with additional approaches and considerations of leadership, it is implied that all relevant behaviors are assessed in commonly used measures. By pairing such tools or using alternative approaches together, however, a more complete view of leadership may be gleaned” (p. 438).

574 G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

In addition to our research efforts, other leadership scholars around the world follow this idea. For example, Antonakis and House (2002) analyzed the Full Range Leadership Theory, and suggested important extensions to this theory. For example, the construct of instrumental leadership was suggested to supplement transformational, transactional, and laissez-faire leadership.

Finally, our project Context and Leadership Styles (CoLeS) revealed first insights about the potential overlap of various leadership constructs. For example, it was found that in the context of recreational sports, the only leadership style of sport trainers that was predictive of trainees' sport-related self-efficacy was laissez-faire (negative correlation; Rowold & Borgmann, unpublished manuscript). To our surprise, no other leadership measure (e.g., transformational, transactional, LMX, etc.) was significantly associated with self-efficacy. As we are still analyzing additional data sets both from the profit and the non-profit sector, these results are somewhat preliminary. They demonstrate, however, that in certain contexts, several leadership constructs might be redundant.

Let us return to the question central to this issue: Can subordinates differentiate between various leadership constructs? Based on the results discussed above, and in combination with our own empirical results and those from colleagues such as Keller (2006), we would say that subordinates can distinguish between leadership constructs indeed. However, and this is an important caveat, the number of leadership constructs that can be distinguished empirically is smaller than the total number of leadership constructs that have been discussed in the leadership literature so far.

In sum, it seems that the various leadership constructs overlap to a certain degree. We would speculate that issues which are central to LMX, such as a positive relationship between leader and led, are also implicated in other leadership constructs, such as transformational leadership and consideration. It remains open whether this overlap might be due to the identical underlying psychological mechanisms (e.g., evoking followers' motivation) or due to methodological issues (e.g., leadership styles are partially redundant to positive affect).

Issue four: How should situations be categorized for investigating the role of context in leadership?

We would like to thank you for your kind words about our research project (CoLeS). Indeed, and as you suggested, we do include a measure for OCB in our studies. With regard to the importance of the organizational factors that you mentioned, we extended our list of factors for project CoLeS. In its current form, project CoLeS now not only includes “classic” organizational factors such as level of hierarchy, but also a number of factors that tap the organizational structure. Those were integrated from various perspectives that can be found within the organizational and economic literature. We were surprised to find many factors within these literatures that were never explored with respect to their relevance to leadership. There was, however, one “old friend”: structural facets that were nearly identical to the taxonomy that was articulated within Kerr and Jermier's (1978) substitutes for leadership theory. As a conclusion to this issue, we agree to you that researchers should have a look at the plethora of potential contextual conditions of leadership. Currently, we focus our efforts on the generation of a “maximum” list of organizational context factors describing boundary conditions for effective leadership.

From our perspective, it has been a pleasure to correspond with you. Also, we learned a lot from exchanging our points of view about the complex phenomenon we call leadership.

Respectfully yours, Jens and Kathrin

References

Antonakis, J., Avolio, B. J., & Sivasubramaniam, N. (2003). Context and leadership: An examination of the nine-factor full-range leadership theory using the Multifactor Leadership Questionnaire. Leadership Quarterly, 14, 261−295.

Antonakis, J., & House, R. J. (2002). The full-range leadership theory: The way forward. In B. J. Avolio, & F. J. Yammarino (Eds.), Transformational and charismatic leadership: The road ahead (pp. 3−34). Amsterdam: JAI.

Bass, B. M. (1999). Two decades of research and development in transformational leadership. European Journal of Work and Organizational Psychology, 8, 9−32. Brown, D. J., & Keeping, L. M. (2005). Elaborating the construct of transformational leadership: The role of affect. Leadership Quarterly, 16, 245−272. Budesko, D. V. (1993). Dominance analysis: A new approach to the problem of relative importance of predictors in multiple regression. Psychological Bulletin, 114,

542−551. Gelfand, M. J., Erez, M., & Aycan, Z. (2007). Cross-cultural organizational behavior. Annual Review of Psychology, 58, 479−514. Gerstner, C. R., & Day, D. V. (1997). Meta-analytic review of leader–member exchange theory: Correlates and construct issues. Journal of Applied Psychology, 82,

827−844. Graen, G. B. (1976). Role making processes in complex organizations. In M. D. Dunnette (Ed.), The handbook of industrial and organizational psychology

(pp. 1201−1245). Chicago: Rand McNally. Graen, G. B. (2003). Interpersonal workplace theory at the crossroads: LMX and transformational theory as special cases of role making in work organizations. In G.

Graen (Ed.), LMX leadership: The seriesDealing with diversity, 1. (pp. 142−182) Greenwich, CT: Information Age Publishing. Graen, G. B. (2007). New genotype for enhancing shared network leadership. In G. B. Graen, & J. A. Graen (Eds.), New multinational network sharingLMX leadership:

The series, Vol. 5. (pp. 1−22) Charlotte, NC: Information Age Publishing. Graen, G. B. (2007). Shared network leadership for coping with dangerous 21st century contexts. Invited address at the Global Leadership Conference, West Point, NY. Hackman, J. R. (2002). Leading teams: Setting the stage for great performances. Boston: HBS Press. House, R. J. (1977). A 1976 theory of charismatic leadership. In J. G. Hunt, & L. L. Larson (Eds.), Leadership: The cutting edge (pp. 189−207). Carbondale, IL: Southern

Illinois University Press. House, R. J., & Aditya, R. M. (1997). The social scientific study of leadership: Quo vadis? Journal of Management, 23, 409−473. Hunt, J. G. (1999). Transformational/charismatic leadership's transformation of the field: An historical essay. Leadership Quarterly, 10, 129−144. Hunter, S. T., Bedell-Avers, K. E., & Mumford, M. D. (2007). The typical leadership study: Assumptions, implications, and potential remedies. Leadership Quarterly,

18, 435−446. Kark, R., & Van-Dijk, D. (2008). Birds of a feather flock together: The relationship between leader–follower self regulation congruency, LMX, and outcome. In G. B.

Graen, & J. A. Graen (Eds.), Knowledge driven corporationLMX leadership: The series, Vol. 6. (pp. 183−209) Charlotte, NC: Information Age Publishing.

575G. Graen et al. / The Leadership Quarterly 21 (2010) 563–575

Keller, R. T. (2006). Transformational leadership, initiating structure, and substitutes for leadership: A longitudinal study of research and development project team performance. Journal of Applied Psychology, 91, 202−210.

Kerr, S., & Jermier, J. M. (1978). Substitutes for leadership: Their meaning and measurement. Organizational Behavior and Human Performance, 22, 375−403. Naidoo, L. J., Scherbaum, C. A., & Goldstein, H. W. (2008). Examining the relative importance of leader–member exchange on group performance over time, Knowledge

driven corporation: A discontinuous model. In G. B. Graen, & J. A. Graen (Eds.), Knowledge driven corporationLMX leadership: The series, Vol. 6. (pp. 211−230) Charlotte, NC: Information Age Publishing.

Rowold, J., & Borgmann, L. (unpublished manuscript). Relationships between leadership styles and trainees' self-efficacy. Journal of Applied Sport Psychology. Rowold, J., & Heinitz, K. (2007). Transformational and charismatic leadership: Assessing the convergent, divergent, and criterion validity of the MLQ and the CKS.

Leadership Quarterly, 18, 121−133. Scherbaum, C. A., Naidoo, L. J., & Ferreter, J. M. (2007). Examining component measures of team leader–member exchange: Using item response theory. In G. B.

Graen, & J. A. Graen (Eds.), New multinational network sharingLMX leadership: The series, Vol. 5. (pp. 129−156) Charlotte, NC: Information Age Publishing. Sosik, J. J., & Dinger, S. L. (2007). Relationships between leadership style and vision content: The moderating role of need for social approval self-monitoring, and

need for social power. Leadership Quarterly, 18, 134−153. Sweeney, P. J. (2007). Leader character, competence and relational trust. Paper presented at Global Leadership Conference, West Point, NY. Turban, D. B., Jones, A. P., & Rozelle, R. M. (1990). Influences of supervisor liking of a subordinate and the reward context on the treatment and the evaluation of that

subordinate. Motivation and Emotion, 14, 215−233. Uhl-Bien, M. (2006). Relational leadership theory: Exploring the social processes of leadership and organizing. Leadership Quarterly, 17, 654−676. Wang, H., Law, K. S., Hackett, R. D., Wang, D., & Chen, Z. X. (2005). Leader–member exchange as a mediator of the relationship between transformational leadership

and followers' performance and organizational citizenship behavior. Academy of Management Journal, 48, 420−432. Yammarino, F. J., Dansereau, F., & Kennedy, C. J. (2001). A multiple-level multidimensional approach to leadership: Viewing leadership through an elephant's eye.

Organizational Dynamics, 29, 149−163. Yammarino, F. J., Spangler, W. D., & Dubinsky, A. J. (1998). Transformational and contingent reward leadership: Individual, dyad, and group levels of analysis.

Leadership Quarterly, 9, 27−54.