In 1944, the United States Army faced a practical crisis. The war demanded officers faster than traditional processes could produce them. Psychologists were asked to build a selection system that would identify who had the natural qualities of a leader -- the traits that distinguished generals from privates, the characteristics that could be measured in young men before they had led anyone.
The Army assembled assessment centers, personality batteries, and situational exercises. Raters observed candidates across multiple scenarios and evaluated their leadership potential. The system was expensive, elaborate, and scientifically serious.
It also predicted almost nothing.
The Office of Strategic Services, forerunner to the CIA, evaluated its own assessment center in 1948 and found that ratings of leadership potential showed near-zero correlation with actual performance in the field. The same year, Ralph Stogdill published a review of 124 studies attempting to identify universal leadership traits, concluding that no consistent set of traits distinguished leaders from non-leaders across situations.
This was the beginning of a scientific reckoning with leadership that is still unfolding. The question "what makes a great leader?" turns out to be among the most researched and most confused topics in all of organizational psychology. What seven decades of research has produced is not a simple answer but a progressively more sophisticated understanding of why the question itself was badly posed.
Key Definitions
Trait theory of leadership: The approach holding that effective leaders possess stable personality characteristics that distinguish them from non-leaders and from ineffective leaders. Now understood to be an incomplete account.
Initiating structure: The leadership behavior dimension describing the degree to which a leader defines roles, organizes tasks, and establishes clear communication -- the task-oriented dimension identified in the Ohio State leadership studies.
Consideration: The leadership behavior dimension describing the degree to which a leader shows concern for followers' wellbeing, builds trust, and maintains warm working relationships -- the relationship-oriented dimension.
Transformational leadership: Bernard Bass's 1985 model describing leadership through inspiration, idealized influence, intellectual stimulation, and individualized consideration of followers. Distinguished from transactional leadership, which operates through contingent reward and management by exception.
Leader-Member Exchange (LMX): George Graen's theory that leaders do not treat all followers the same but form individualized dyadic relationships, creating in-groups with high-quality exchange and out-groups with lower-quality exchange.
Psychological safety: Amy Edmondson's concept describing the shared belief among team members that the team is safe for interpersonal risk-taking -- that expressing disagreement, admitting errors, or asking questions will not result in punishment or humiliation.
Romance of leadership: James Meindl's 1985 finding that followers systematically over-attribute organizational outcomes to the leader, a cognitive bias that inflates leadership's perceived importance.
Collective intelligence: Anita Woolley's 2010 finding that teams have stable general ability factors that predict performance across diverse tasks, analogous to individual general intelligence but not reducible to the intelligence of individual members.
The Collapse of Trait Theory
Ralph Stogdill's 1948 review reached a conclusion that seemed to devastate the trait approach: "A person does not become a leader by virtue of the possession of some combination of traits." What he found was that specific combinations of traits were relevant in specific situations, but no trait or combination predicted leadership effectiveness across contexts.
This was not a finding that leadership researchers were willing to accept. The trait approach was intuitive, practically useful (if traits exist, you can select for them), and aligned with popular beliefs about the nature of great men. So research continued, methodology improved, and the picture became more nuanced.
By the 1990s and 2000s, meta-analyses using better statistical methods had rehabilitated a partial version of the trait approach. Timothy Judge and colleagues published a 2002 meta-analysis in Journal of Applied Psychology examining the Big Five personality dimensions and leadership. Extraversion was the most consistent predictor (r = 0.31 with leadership emergence), followed by conscientiousness (r = 0.28), openness to experience (r = 0.24), and neuroticism (r = -0.24, negatively). Agreeableness showed weak relationships.
These correlations are meaningful but modest. More important, the same study found that the relationship between personality and leadership effectiveness (actually leading well) was substantially weaker than the relationship between personality and leadership emergence (being perceived as a leader or being selected for leadership roles). The traits that get people into leadership positions are not identical to the traits that make people good leaders.
This distinction -- between emergence and effectiveness -- turns out to be among the most important in all of leadership science.
Dark Triad Leaders: Why Dangerous Personalities Rise
The emergence-effectiveness gap is most pronounced and most consequential for the dark triad personality traits: narcissism, Machiavellianism, and psychopathy.
Timothy Judge, Jeffery LePine, and Bruce Rich published a meta-analysis in 2006 in Journal of Applied Psychology that quantified the narcissism paradox precisely. Narcissism predicted leadership emergence with r = 0.34 -- a substantial correlation. The same trait predicted leadership effectiveness with r = -0.11. Narcissists are significantly more likely than non-narcissists to be chosen as leaders and significantly less effective once they occupy leadership roles.
The mechanism is not difficult to understand. Narcissists are socially skilled in the specific dimensions that drive first impressions. They speak with confidence, project vision, dominate the room, and perform leadership behaviors that observers associate with competence. In selection contexts -- job interviews, initial meetings, early performance reviews -- these qualities advantage them. The costs of narcissistic leadership emerge over time: poor listening, credit-claiming, blame-shifting, deterioration of team psychological safety, and the departure of talented subordinates who recognize they will not receive credit or development.
The full dark triad creates what researchers call the "leadership selection paradox." Selection processes favor the appearances of leadership rather than its substance. Psychopathic traits (emotional callousness, fearlessness, impulsive behavior) predict willingness to take visible risks and make bold decisions, which appears decisive and strong in selection contexts. Machiavellian traits (strategic manipulation, long-term deception) predict effective self-promotion and coalition-building during selection. Both sets of traits predict poor outcomes for followers and organizations.
"The same qualities that help certain individuals attain leadership positions can undermine their effectiveness once in those roles." -- Timothy Judge, Journal of Applied Psychology, 2006
The Behavioral Revolution: Ohio State and Michigan Studies
The failure of trait theory to fully account for leadership effectiveness drove researchers in the late 1940s and 1950s toward behavior. Rather than asking who leaders were, they began asking what leaders did.
Two independent research programs reached compatible conclusions.
The Ohio State University leadership studies, beginning in the late 1940s under the direction of Ralph Stogdill and Carroll Shartle, reduced observed leader behavior to two independent dimensions: initiating structure (task-focused behavior: clarifying roles, setting goals, planning work) and consideration (relationship-focused behavior: showing concern for followers, building trust, respecting feelings). These dimensions were identified by factor analysis of the Leader Behavior Description Questionnaire administered to hundreds of followers in military and industrial settings.
The University of Michigan Institute for Social Research, working under Rensis Likert at roughly the same time, identified similar dimensions: production-centered behavior versus employee-centered behavior. Multiple studies found that the most effective leaders showed high levels of both -- a finding that challenged the intuitive assumption that task focus and relationship focus are trade-offs.
The behavioral framework's practical appeal was its implicit promise: if leadership is behavior, it can be taught. You do not need to select for mysterious personality traits. You need to train specific behaviors. This belief powered leadership training programs for decades and remains influential in management development.
The limitation of the behavioral approach was its context-insensitivity. Whether task-focused or relationship-focused behavior is more effective depends heavily on the situation, the maturity of followers, the nature of the task, and organizational culture. This led to the third major wave of leadership research: contingency theories.
Situational Leadership and Contingency Models
Paul Hersey and Kenneth Blanchard's Situational Leadership model, developed from the late 1960s onward, proposed that the appropriate leadership style depends on the "developmental level" of the follower on the specific task. Their model maps four styles (directing, coaching, supporting, delegating) to four follower development levels defined by the combination of competence and commitment.
The model's core insight -- that effective leadership requires flexibility, and that the same style applied to a novice learner and an experienced expert will produce different results -- is well-supported by common sense and consistent with research on feedback, autonomy, and skill development. Meta-analytic support for the specific prescriptions of the model is moderate rather than strong, in part because the developmental level construct is difficult to operationalize and the matching prescriptions are hard to test with precision.
Fred Fiedler's Contingency Model proposed a different kind of matching: between a leader's motivational orientation (task-motivated or relationship-motivated) and situational favorability (how much control the situation gives the leader). Fiedler was unusual among leadership researchers in arguing that leaders cannot substantially change their fundamental orientation -- the intervention should be to place the right leader in the right situation, not to develop leaders to be more flexible.
The enduring contribution of contingency theories is the principle they share: leadership effectiveness is not a property of the leader alone but of the leader-situation interaction. The same person may be highly effective in one organizational context and ineffective in another. This principle is now well-established and should fundamentally shape how organizations select and develop leaders.
Transformational Leadership: The Dominant Paradigm
James MacGregor Burns introduced the transforming/transactional distinction in political science in 1978. Bernard Bass adapted it for organizational psychology and developed the Multifactor Leadership Questionnaire (MLQ) in 1985, creating the measurement infrastructure that has made transformational leadership the most-studied model in leadership research.
Transformational leaders motivate followers through four mechanisms, now known as the "four I's":
| Component | Description |
|---|---|
| Idealized Influence | Leader as role model; followers identify with and want to emulate leader |
| Inspirational Motivation | Articulating compelling vision; creating shared sense of purpose |
| Intellectual Stimulation | Challenging assumptions; encouraging innovation; reframing problems |
| Individualized Consideration | Attending to individual follower's development needs; coaching; mentoring |
This stands in contrast to transactional leadership, which operates through contingent reward (rewarding performance, correcting errors) and management by exception (intervening only when standards are not met).
The accumulated meta-analytic evidence supports transformational leadership as a meaningful predictor of outcomes. A 2003 meta-analysis by Timothy Judge and Ronald Piccolo in Journal of Applied Psychology covering 626 correlations found that transformational leadership correlated r = 0.44 with follower satisfaction with the leader and r = 0.36 with follower motivation. The relationship with objective performance was weaker: transformational leadership accounts for approximately 15% of variance in team performance in studies that use objective rather than self-reported measures.
The weaker relationship with objective performance is important context. Transformational leadership is substantially more effective at generating followers' subjective experience of good leadership than at generating measurable outcomes. This is useful information: leaders can be perceived as transformational by their followers while delivering mediocre results, and can deliver excellent results while being perceived as transactional.
Psychological Safety: The Team-Level Variable That Matters Most
In 1999, Amy Edmondson published what would become one of the most cited papers in organizational psychology. Her study, conducted at a Harvard teaching hospital, examined the relationship between team functioning and medical error rates.
The initial hypothesis was straightforward: better-functioning teams would make fewer errors. The data produced a puzzle. When she correlated team functioning scores with error rates, she found the opposite of her prediction: the best teams appeared to make more errors, not fewer.
Edmondson reconsidered her measurement. She had been measuring reported errors, not actual errors. She then found that the best teams were not making more errors. They were reporting more errors -- because team members in high-functioning units felt psychologically safe enough to admit mistakes, near-misses, and concerns without fear of blame. The worst-performing teams were suppressing error reports, creating an illusion of error-free operation while underlying problems festered.
The variable that distinguished high-reporting from low-reporting teams was psychological safety: the shared belief that it is safe to speak up, take interpersonal risks, and be honest about problems without fear of punishment or humiliation. Psychological safety was subsequently linked not only to error reporting but to learning behavior, performance improvement, and innovation.
Google's Project Aristotle, an internal research initiative that studied 180 teams over two years using data from extensive surveys and performance measures, sought to identify the team dynamics that predicted team effectiveness. The findings, published in 2016, were consistent with Edmondson's hospital research. Of five team norms studied (psychological safety, dependability, structure and clarity, meaning, and impact), psychological safety was the most robust predictor of team performance, emerging as important across teams in different functions and at different levels of the organization.
"The most important key to improving team performance... was psychological safety." -- Google re:Work, Project Aristotle report, 2016
Leader-Member Exchange: The Inconvenient Reality of In-Groups
George Graen's Leader-Member Exchange theory, developed in the 1970s, challenged a central assumption of most leadership models: that leaders relate to their followers as a group. LMX proposed instead that leaders form individualized dyadic relationships, and that the quality of these relationships varies substantially within any given leader's team.
High-quality LMX relationships (the "in-group") are characterized by mutual trust, respect, and obligation beyond the formal employment contract. In-group members receive more interesting assignments, greater development opportunities, more access to the leader's time, and higher performance ratings. Out-group members receive more formal, contractual treatment.
The research literature consistently finds that LMX quality predicts employee outcomes (satisfaction, commitment, performance ratings, turnover intention) more strongly than many other leadership variables. This is concerning for two reasons. First, differential treatment within teams creates perceptions of unfairness that damage overall morale. Second, in-group membership is not randomly assigned. Research shows that in-group members tend to be more demographically similar to the leader -- same gender, ethnicity, educational background -- raising serious equity concerns.
LMX theory does not tell leaders that they should form differential relationships. It tells them that they do, whether they intend to or not, and that awareness of this dynamic is the first step toward managing it more equitably.
The Romance of Leadership and the Attribution Problem
James Meindl, Sanford Ehrlich, and Janet Dukerich published a paper in 1985 in Administrative Science Quarterly that attacked a foundational assumption of leadership research: that leaders substantially cause organizational outcomes.
Their research examined the strength of the relationship between leadership quality and organizational performance in published studies, and compared this with the causal importance followers attributed to leaders when explaining organizational successes and failures. They found a systematic mismatch: followers attributed dramatically more causal importance to leaders than the correlational evidence could support. They called this "the romance of leadership."
The phenomenon is a cognitive attribution bias. When organizations succeed, we credit the leader. When they fail, we blame the leader. This attribution occurs regardless of whether the leader's actions were the actual cause of the outcome, which is often determined more by market conditions, regulatory changes, technological shifts, or economic cycles than by any specific leadership behavior.
The romance of leadership has practical consequences. It drives organizations to pay excessive premiums for leadership talent, particularly for charismatic or visionary figures, whose actual impact on organizational performance is far smaller than the premium implies. It drives outsized executive compensation justified by attribution of organizational success to the CEO. And it systematically obscures the organizational and environmental factors that actually drive outcomes.
What Actually Predicts Team Performance
Anita Woolley, Christopher Chabris, Alexander Pentland, Nada Hashmi, and Thomas Malone published a 2010 paper in Science with a striking title: "Evidence for a Collective Intelligence Factor in the Performance of Human Groups."
Their research administered diverse cognitive tasks to 699 people in 192 groups and then tested whether performance on one task predicted performance on others -- the group-level equivalent of asking whether general intelligence exists at the individual level. They found it does. A stable group-level general factor accounted for substantial variance in performance across tasks, analogous to the g factor in individual intelligence.
The question was then what predicted this collective intelligence factor. Three variables emerged:
| Predictor | Relationship to Collective Intelligence |
|---|---|
| Average social sensitivity (Reading the Mind in the Eyes test) | Strong positive |
| Equal distribution of conversational turn-taking | Strong positive |
| Proportion of women on the team | Strong positive |
| Average individual IQ of team members | Weak positive |
| Presence of highest-IQ individual | Not significant |
The finding about individual intelligence is striking. Teams with higher average IQ did not substantially outperform teams with lower average IQ on the collective intelligence measure. What predicted collective intelligence was social sensitivity, conversational equity, and gender composition -- not the presence of high-intelligence or high-charisma individuals.
The gender effect was explained in part through social sensitivity: women on average score higher on the Reading the Mind in the Eyes test, and teams with more women therefore had higher average social sensitivity. But gender composition predicted collective intelligence beyond the social sensitivity effect, suggesting that gender-diverse teams benefit from perspective diversity as well.
The implication for leadership selection is substantial. Hiring the most individually impressive leader may matter less than building the team conditions -- psychological safety, equal voice, social sensitivity -- that enable collective intelligence to emerge.
Conclusion
The science of leadership has moved through several revolutions since Stogdill's 1948 reckoning. Each revolution expanded the frame of what needed to be explained.
Trait theory revealed that some individual characteristics matter but that no universal trait profile predicts effectiveness across contexts. The emergence-effectiveness gap -- which dark triad research has documented precisely -- means that the people who look like leaders are often not the leaders who produce the best outcomes.
The behavioral revolution established that what leaders do matters more than what they are, and that the two core behavioral dimensions (task focus and relationship focus) are not trade-offs but additive in effective leadership.
Situational and contingency models established that the right leadership approach depends on the context -- the follower's development level, the task structure, the organizational culture, the degree of situational control.
Transformational leadership research established that inspiring followers through vision, intellectual stimulation, and individual consideration produces meaningful but moderate effects on team performance -- accounting for approximately 15% of outcome variance in objective studies.
LMX theory revealed the inconvenient reality that leaders form differential relationships with followers, with equity implications that most leadership development programs ignore.
The romance of leadership research established that followers systematically over-attribute organizational outcomes to leaders, creating selection premiums and cultural expectations that are not justified by the actual magnitude of leadership's causal impact.
Psychological safety research -- Edmondson's hospital study, Google's Project Aristotle -- established that team-level climate predicts performance more robustly than individual leader characteristics, and that the leader's primary function may be creating conditions for team effectiveness rather than driving outcomes through personal influence.
Woolley's collective intelligence research established that the properties that make teams smart are social -- distributed voice, emotional attunement, gender diversity -- not the individual brilliance of any member including the leader.
What makes a great leader? Mostly: the ability to create environments where team members feel safe to speak, take risks, and contribute fully. The traits and behaviors that accomplish this are substantially less dramatic, and substantially less visible in selection contexts, than the traits and behaviors that produce leadership emergence.
The great leaders may not look like great leaders. That is not a paradox. It is the most important finding in the field.
References
Stogdill, R. M. (1948). Personal factors associated with leadership: A survey of the literature. Journal of Psychology, 25(1), 35-71. https://doi.org/10.1080/00223980.1948.9917362
Judge, T. A., Bono, J. E., Ilies, R., & Gerhardt, M. W. (2002). Personality and leadership: A qualitative and quantitative review. Journal of Applied Psychology, 87(4), 765-780. https://doi.org/10.1037/0021-9010.87.4.765
Judge, T. A., LePine, J. A., & Rich, B. L. (2006). Loving yourself abundantly: Relationship of the narcissistic personality to self and other perceptions of workplace deviance, leadership, and task and contextual performance. Journal of Applied Psychology, 91(4), 762-776. https://doi.org/10.1037/0021-9010.91.4.762
Bass, B. M. (1985). Leadership and performance beyond expectations. Free Press.
Judge, T. A., & Piccolo, R. F. (2004). Transformational and transactional leadership: A meta-analytic test of their relative validity. Journal of Applied Psychology, 89(5), 755-768. https://doi.org/10.1037/0021-9010.89.5.755
Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350-383. https://doi.org/10.2307/2666999
Meindl, J. R., Ehrlich, S. B., & Dukerich, J. M. (1985). The romance of leadership. Administrative Science Quarterly, 30(1), 78-102. https://doi.org/10.2307/2392813
Woolley, A. W., Chabris, C. F., Pentland, A., Hashmi, N., & Malone, T. W. (2010). Evidence for a collective intelligence factor in the performance of human groups. Science, 330(6004), 686-688. https://doi.org/10.1126/science.1193147
Graen, G. B., & Uhl-Bien, M. (1995). Relationship-based approach to leadership: Development of leader-member exchange (LMX) theory of leadership over 25 years. The Leadership Quarterly, 6(2), 219-247. https://doi.org/10.1016/1048-9843(95)90036-5
Hersey, P., & Blanchard, K. H. (1969). Life cycle theory of leadership. Training and Development Journal, 23(5), 26-34.
Frequently Asked Questions
Is there a universal set of leadership traits?
Ralph Stogdill's 1948 meta-analysis of 124 leadership trait studies concluded there is no consistent universal set of traits that distinguishes leaders across all situations. Subsequent research has refined this: some traits show moderate relationships with leadership effectiveness, but situational factors dominate.
What is transformational leadership and does it work?
Transformational leadership, developed by Bernard Bass in 1985, involves inspiring followers through vision, intellectual stimulation, and individual consideration. Meta-analyses show it predicts roughly 15% of variance in team performance, making it one of the most validated leadership models.
Do narcissistic leaders perform better?
No. Timothy Judge's 2006 meta-analysis found narcissism predicts leadership emergence (r=0.34) but negatively predicts effectiveness (r=-0.11). Narcissists appear confident and visionary but their actual team outcomes are worse than non-narcissistic leaders.
What is psychological safety and why does it matter?
Amy Edmondson's 1999 study at a Harvard hospital found that the best-performing teams reported making more errors, not fewer. Closer analysis revealed high performers were in psychologically safe environments where errors could be discussed openly. Google's Project Aristotle replicated this as the top predictor of team performance.
What actually predicts team performance better than the leader?
Anita Woolley's 2010 Science study on collective intelligence found that team performance is predicted by social sensitivity (reading others' emotions), equal turn-taking in conversation, and proportion of women on the team -- not individual IQ or the quality of the leader.
What is the romance of leadership?
James Meindl's 1985 research found that followers systematically over-attribute organizational outcomes to the leader, whether good or bad. This cognitive bias -- the romance of leadership -- inflates both credit and blame assigned to individual leaders beyond what their actual influence justifies.
Does leadership style need to change by situation?
Hersey and Blanchard's situational leadership model holds that effective leaders adapt their style -- directing, coaching, supporting, or delegating -- to match the developmental level of the follower. Meta-analytic support is moderate, but the core principle that one style does not fit all contexts is well-supported.
![Pensar rápido, pensar despacio [Think Fast, Think Slowly]](https://images-na.ssl-images-amazon.com/images/P/B085NZ4HVD.01._SL200_.jpg)