What a Century of Research Says About Leadership: From Trait Lists and Behavior Grids to Contingency Models and Exchange Theory

Most management training presents leadership theories side by side, as if a manager could simply pick whichever one fits today's problem: a trait checklist for recruiting, a behavioral grid for a workshop, a contingency model for a harder case. That presentation hides something the research itself does not hide, which is that each theory in this lesson was built, in large part, as a correction to the one before it. Trait theory was abandoned because it failed to predict behavior reliably across situations. Behavioral theory replaced it by watching what leaders actually did, but implied, without quite saying so, that one behavioral combination worked everywhere. Contingency and situational theories were built specifically to deny that implication, arguing that the right style depends on the situation, though the theories claiming this disagree sharply on how much evidence actually supports their own specific claims. Leader-member exchange theory then challenged an assumption all of the earlier work shared: that a leader has one style, consistently applied to an entire group.

Knowing this sequence matters for a working manager for a specific reason: it explains why a model that still gets taught in corporate training, Hersey and Blanchard's Situational Leadership in particular, can be enormously popular and empirically weak at the same time, while a less fashionable model, Fiedler's contingency theory, has a far more thoroughly tested record despite being harder to apply day to day. Popularity in a training catalog and survival under empirical testing are not the same credential, and this lesson treats that distinction as a central finding, not a footnote.

The earliest systematic leadership research assumed that leaders were simply a different kind of person, and that the research task was to identify which qualities set them apart. Ralph Stogdill's 1948 review, covering more than one hundred studies conducted between 1904 and 1947, remains the reference point for what that search actually found. Stogdill did not conclude that leaders have no distinguishing traits at all; he found that intelligence, alertness, insight, initiative, persistence, self-confidence and sociability showed up often enough in the average leader to be worth naming. What he could not find was a consistent set of traits that reliably separated leaders from non-leaders across different situations. A trait profile that predicted effective leadership in one organization or one kind of task often failed to predict it in another, and people who scored highly on the traits associated with leadership sometimes never became leaders at all, while people who lacked them sometimes did.

That inconsistency, more than any single disconfirming study, is why trait theory lost standing through the 1950s and 1960s. If a quality cannot be shown to matter reliably once the situation changes, it is hard to build a selection system, a training program, or a theory of leadership around it. The field's attention moved toward behavior and then toward the situation itself, treating the person as only one part of a larger equation rather than the whole of it.

Trait research did not stay abandoned, though what came back looked different from what had been set aside. Shelley Kirkpatrick and Edwin Locke's 1991 article, pointedly titled "Leadership: Do Traits Matter?", argued that Stogdill's own data had been read too pessimistically, and that six traits in particular, drive, leadership motivation, honesty and integrity, self-confidence, cognitive ability, and knowledge of the business, showed up with enough regularity to be taken seriously again, while cautioning that possessing them guarantees nothing by itself. A decade later, Timothy Judge, Joyce Bono, Remus Ilies and Megan Gerhardt's 2002 meta-analysis, organizing dozens of earlier studies around the Big Five personality framework, found that extraversion correlated with leadership emergence and effectiveness more strongly than any other trait, with low neuroticism and high conscientiousness also showing meaningful, if smaller, associations.

The revival does not actually contradict Stogdill, once the two claims are stated precisely. Stogdill's finding was that no trait profile predicted who would lead effectively in a specific situation with the reliability a selection test would need. The later meta-analyses make a narrower, statistical claim: across a large number of studies, certain traits correlate with leadership outcomes often enough that the relationship is unlikely to be chance, even though any individual case can deviate from the average, sometimes substantially. An extraversion correlation of roughly 0.3, the figure Judge and colleagues reported, describes a real but modest tendency, not a law that identifies which specific extravert in a specific role will succeed. Both the 1948 skepticism and the later revival can be correct at once, because they are answering different questions with different kinds of evidence.

If traits could not reliably explain leadership, the next generation of researchers reasoned that observable behavior might. Two research programs, run independently at Ohio State University and the University of Michigan through the late 1940s and 1950s, produced the field's first durable behavioral framework, and both survive in some form in management textbooks today.

The Ohio State studies, built on a questionnaire originally developed by John Hemphill and Alvin Coons and later adapted by Andrew Halpin and B. James Winer into the Leader Behavior Description Questionnaire, identified two dimensions of leader behavior: initiating structure, meaning how far a leader organizes work by defining roles, scheduling tasks and setting performance standards, and consideration, meaning how far a leader attends to subordinates' welfare, shows trust, and respects their input. Edwin Fleishman, who worked on the same research program, published one of the earliest formal taxonomies built around these two dimensions in 1953. The detail that still trips up people encountering this research for the first time is that the two dimensions are independent, not opposite ends of a single scale. A leader can score high on both, organizing work tightly while also attending closely to people, or low on both, or high on one and low on the other; knowing a leader's score on one dimension tells a manager nothing about the other.

The Michigan studies, led by Rensis Likert and summarized in his 1961 book New Patterns of Management, framed a related but distinct contrast: employee-oriented leadership, which emphasizes relationships and the welfare of subordinates, against production-oriented leadership, which emphasizes technical efficiency and output. Likert's research associated employee orientation, combined with general rather than close supervision, with better outcomes on the measures the studies used. That finding has been repeated in management teaching for decades as though it settled the question, but it is worth being precise about what kind of evidence it actually is. Much of it came from comparing already-existing teams and their already-existing supervisors at a single point in time, which makes it difficult to rule out the reverse explanation: a team that was already functioning well for other reasons may have made its supervisor comfortable enough to loosen close supervision, rather than loosened supervision being what made the team function well. The association is real and worth taking seriously; treating it as settled proof of a causal direction is not something the underlying studies can support on their own.

Robert Blake and Jane Mouton took both research traditions and turned them into a single visual tool, the Managerial Grid, published in 1964. The grid plots concern for people against concern for production, each on a nine-point scale, producing eighty-one possible combinations, with five named styles anchoring the corners and center: impoverished management (low on both), country club management (high people, low production), produce-or-perish management (high production, low people), middle-of-the-road management (moderate on both), and team management (high on both), which Blake and Mouton presented as the universally preferable style. The grid is genuinely useful as a diagnostic vocabulary, and it remains common in management training for exactly that reason. It is worth noticing, though, that its central claim, that one position on the grid is best regardless of context, is the same kind of universal prescription that the behavioral research's successor, contingency theory, was built specifically to challenge. Unlike Fiedler's model, discussed next, the grid was developed primarily as a consulting and training instrument rather than tested through the kind of large-sample empirical program the Ohio State and Michigan studies themselves generated; its popularity in management education has outpaced the independent evidence behind its central recommendation.

Fred Fiedler's contingency model, published in 1967, broke from the behavioral tradition on two fronts at once. First, it argued that leadership effectiveness depends on the fit between a leader's style and the situation, not on any single style being best. Second, and less commonly remembered, it treated a leader's style as a relatively stable, trait-like orientation, not a behavior that could simply be dialed up or down the way the Ohio State and Michigan frameworks implied. Fiedler measured that orientation with the Least Preferred Co-Worker scale, asking leaders to rate the single colleague they had worked with least well on a set of descriptive adjectives. A leader who still rated that difficult colleague relatively favorably scored as relationship-motivated; a leader who rated them harshly scored as task-motivated.

Fiedler combined that leader-style score with three features of the situation, leader-member relations, task structure, and the position power the leader formally holds, into a measure of overall situational favorableness, sorted into eight combinations he called octants. His central hypothesis was that task-motivated leaders perform better at the extremes, in very favorable situations where little needs managing and in very unfavorable ones where someone has to impose order, while relationship-motivated leaders perform better in the mixed, moderately favorable middle, where influence has to be built rather than assumed.

Two independent meta-analyses tested this against the accumulated evidence: Strube and Garcia in 1981, and Lawrence Peters, Darrell Hartke and John Pohlmann in 1985. Both found the pattern of results Fiedler predicted held up reasonably well overall, but both also found that the model fit laboratory studies more cleanly than field studies conducted in real organizations, and both flagged specific octants, octant two in particular, where predictions and outcomes did not line up consistently. That gap between lab support and field support matters practically: a model that predicts well in a controlled experiment but less reliably on the ground is a weaker guide for an actual manager than its lab results alone would suggest.

The model's most awkward practical implication followed directly from treating style as stable rather than adjustable. If a task-motivated leader cannot simply become relationship-motivated to suit a new assignment, the model's own logic points toward changing the situation to fit the leader, through reassignment, restructuring the leader's authority, or selecting leaders for situations that already match their style, rather than toward training leaders to flex their style at will. Organizations have mostly not adopted leadership development on those terms, preferring training programs premised on adaptable behavior, which is one reason Fiedler's model is more often taught as history than applied as a staffing method today, whatever its comparatively solid empirical record.

Paul Hersey and Kenneth Blanchard published what they called the Life Cycle Theory of Leadership in 1969, later renamed Situational Leadership as the two authors developed increasingly divergent versions of it separately. The model proposes four leadership styles, telling, selling, participating and delegating, moving from highly directive to highly delegative, matched to a follower's "readiness," a combination of competence and willingness to take responsibility for a specific task. A newly assigned, low-readiness follower is supposed to need a directive, telling style; an experienced, high-readiness follower is supposed to need delegation and comparatively little direction. The model's intuitive appeal and simple visual format made it one of the most widely taught leadership frameworks in corporate training programs for decades.

Its empirical record has not kept pace with its popularity. Claude Graeff's critical reviews, published in 1983 and expanded in 1997, found conceptual ambiguity and internal inconsistency in both Hersey's and Blanchard's separately evolved versions of the model, including unclear operational definitions of "readiness" itself. Robert Vecchio's 1987 field study of high school teachers found that the model's prescriptions held up, at best, for teachers early in their careers, with the predicted relationship between leadership style and follower readiness fading to nothing among more experienced staff. A later study by Geir Thompson and Vecchio, published in 2009 and conducted among Norwegian bank employees, found that a 2007 revision of the model actually predicted worse than the original version, reinforcing their conclusion that the theory still lacks a solid empirical foundation even after four decades of refinement attempts. None of this means the underlying intuition, that newer or less confident employees often benefit from more direction, is wrong; it means the specific four-style, readiness-matched structure the model prescribes has not held up well as a tested scientific claim, which is a different thing from the model being useless as a rough coaching heuristic. A manager using it should treat it the way an experienced clinician treats a rule of thumb rather than a lab result: a starting hypothesis to check against the specific person in front of them, not a validated diagnostic.

Robert House's path-goal theory, first published in 1971 and substantially reformulated in 1996, pursues a related idea on more cautious, more theoretically grounded terms. Drawing explicitly on expectancy theory, House proposed that a leader's job is to clarify the path between a follower's effort and valued outcomes, selecting among directive, supportive, participative and achievement-oriented behaviors, later expanded further in the 1996 revision, based on two sets of moderating conditions: follower characteristics such as ability and need for autonomy, and task or environmental characteristics such as how structured the work already is and how much formal authority already governs it. A directive style adds the most value when a task is ambiguous and a follower lacks experience; it adds little, and can even frustrate a competent follower, when the task is already clear and the follower already knows what to do.

Path-goal theory never achieved anything close to Situational Leadership's footprint in corporate training, and its own empirical record is also mixed, partly because its growing list of moderating variables makes it genuinely difficult to test cleanly in a single study. The useful contrast is not that one theory is proven and the other disproven; both are "it depends" theories whose specific prescriptions have had an uneven time surviving contact with data. The difference is reputational rather than purely evidentiary: one became the basis of a training and certification industry with a commercial incentive to keep teaching it regardless of what independent researchers kept finding, while the other stayed mostly inside academic journals, where critique accumulates quietly and rarely reaches the manager being trained on the other model.

Every theory discussed so far shares an assumption that is easy to miss precisely because it is never stated: that a leader has one style, applied more or less uniformly to an entire group or team. Research beginning with Fred Dansereau, George Graen and William Haga's vertical dyad linkage studies in the mid-1970s noticed that this assumption did not match what was actually happening inside work groups. Leaders, it turned out, routinely developed noticeably different relationships with different subordinates occupying the same formal role, and those differences predicted outcomes, like performance ratings and turnover, better than any single group-level style score did.

George Graen and Mary Uhl-Bien consolidated twenty-five years of this research into a single framework in a 1995 review, describing leader-member exchange as a relationship that develops through roughly three stages: an initial stranger phase, governed by formal roles and contractual obligation; an acquaintance phase, as leader and follower test and expand what they offer each other; and, for some but not all followers, a partner phase, marked by high mutual trust, loyalty and a far more flexible, negotiated exchange of resources, latitude and informal influence than the formal job description specifies. Followers who reach that partner stage, often informally described as the leader's in-group, tend to receive more information, more discretion and more of the leader's attention; followers who remain closer to the stranger or acquaintance stage, the out-group, tend to operate within the stricter boundaries of the formal employment relationship.

LMX theory's real contribution is reframing where the explanatory action sits. Instead of asking what behavior or style a leader generally displays, which the Ohio State, Michigan and contingency traditions all asked in one form or another, LMX asks what relationship exists between a leader and each specific person, and treats that relationship, not the leader's general disposition, as the more useful predictor of how the follower actually experiences being led. It is a genuinely different unit of analysis, and it explains a pattern the earlier theories could not: two employees reporting to the same manager, who would receive an identical score on any group-level behavioral or contingency measure, can have substantially different working relationships with that same manager and substantially different outcomes as a result.

The theory is also honest about a tension it creates rather than resolves. If leaders genuinely do, and perhaps must, build differentiated relationships rather than treating every subordinate identically, the line between a legitimate high-trust relationship, earned through demonstrated reliability, and ordinary favoritism is not something LMX theory settles on its own; it describes the pattern without adjudicating when the pattern is fair. A separate methodological caution applies as well: much of the supporting research relies on self-reported relationship quality measured at a single point in time, which raises the same causal-direction problem noted earlier for the Michigan studies. A subordinate who happens to be a strong, reliable performer may earn a high-quality exchange relationship as a consequence of that performance, rather than the relationship being what produced the performance in the first place; most LMX studies cannot fully separate the two directions from each other.

This relational lens is also where Lesson 27 picks up. Transformational and transactional leadership, and the specific case of servant leadership, can each be read partly as more developed accounts of what a leader actually offers inside a high-quality exchange, vision and individualized attention in the transformational case, a clearer transactional bargain in the other, rather than as a wholly separate line of theory. This lesson leaves that development to the next one.

No single theory in this lesson won the argument outright, and reading the sequence honestly means resisting the temptation to treat the most recent entry as the final answer. Trait theory overclaimed universality and was set aside too completely before returning on narrower, statistical terms. The behavioral studies found real, durable dimensions of leader conduct but implied, through the Managerial Grid especially, that one combination of those dimensions was best everywhere, an implication the contingency theories were built specifically to reject. The contingency and situational theories agreed on the core idea, that fit between style and situation matters, while producing wildly different evidentiary records behind near-identical marketing, Fiedler's model surviving meta-analytic testing considerably better than Hersey and Blanchard's did. Leader-member exchange theory then reframed the whole conversation around the individual relationship, without settling the fairness question that reframing raises.

The practical lesson for a manager is not a formula, because the field's own history argues against treating any one of these models as a complete instruction set. It is closer to a discipline of checking claims against their actual evidentiary standing before leaning on them. A model's popularity in a training catalog, as the Situational Leadership case shows clearly, is not a reliable proxy for how well it survives contact with data. A model's intuitive appeal, as the Managerial Grid shows, is not the same as independent empirical validation of its central recommendation. And a model's elegance, path-goal theory's weakness in practice, can come at the cost of being difficult to apply cleanly to one actual person in one actual situation.

What survives across all five traditions, trait, behavioral, contingency, situational and relational, is a narrower but more defensible claim than any one of them made on its own: that leadership effectiveness depends on some combination of who the leader is, what the leader does, the situation the leader is operating in, and the specific relationship the leader has built with the person in front of them, and that no researcher has yet produced a single model that weighs all four correctly in every case. That is not a comfortable place to end a lesson, but it is where the actual evidence leaves a manager, and claiming otherwise would be exactly the kind of overreach this lesson has spent its length describing in other people's theories.

Identify a real leader-follower situation you have directly observed or can document from a published source (a case study, news report, memoir, or organizational account you can cite). Do not invent a case; if your evidence is limited to your own professional experience, say so explicitly rather than presenting inference as documented fact.

Write a report of 1,500 to 2,000 words that:

1. describes the leader, the follower or followers, and the task or situation factually, distinguishing what you directly observed or what a cited source reported from any inference you are adding;
2. applies at least two distinct theories from this lesson (for example, a trait-based assessment and Fiedler's contingency model, or the Ohio State behavioral dimensions and leader-member exchange) to the same situation, and states explicitly where the two theories agree and where they would recommend different actions;
3. evaluates the strength of the empirical evidence behind each theory you apply, rather than treating all theories as equally well supported;
4. identifies at least one way the situation does not fit neatly into either theory, and explains why forcing the fit would misrepresent the case;
5. assesses whether a leader-member exchange perspective would change your read of the case, particularly whether different followers in the situation appear to be experiencing the same leadership differently; and
6. states, honestly, what remains uncertain in your analysis and what additional information would resolve it.

Cite at least four academic sources, including at least three from this article's reference list. Where your evidence is incomplete, say so explicitly rather than presenting a confident conclusion the evidence does not support.

Assessment rubric (100 marks)

1. Stogdill, R. M. (1948). Personal factors associated with leadership: A survey of the literature. Journal of Psychology, 25, 35-71.
2. Fleishman, E. A. (1953). The description of supervisory behavior. Journal of Applied Psychology, 37(1), 1-6.
3. Likert, R. (1961). New patterns of management. McGraw-Hill.
4. Blake, R. R., & Mouton, J. S. (1964). The managerial grid: Key orientations for achieving production through people. Gulf Publishing.
5. Fiedler, F. E. (1967). A theory of leadership effectiveness. McGraw-Hill.
6. Peters, L. H., Hartke, D. D., & Pohlmann, J. T. (1985). Fiedler's contingency theory of leadership: An application of the meta-analysis procedures of Schmidt and Hunter. Psychological Bulletin, 97(2), 274-285.
7. Hersey, P., & Blanchard, K. H. (1969). Life cycle theory of leadership. Training and Development Journal, 23(5), 26-34.
8. Graeff, C. L. (1997). Evolution of situational leadership theory: A critical review. Leadership Quarterly, 8(2), 153-170.
9. Vecchio, R. P. (1987). Situational leadership theory: An examination of a prescriptive theory. Journal of Applied Psychology, 72(3), 444-451.
10. House, R. J. (1971). A path-goal theory of leader effectiveness. Administrative Science Quarterly, 16(3), 321-339.
11. Graen, G. B., & Uhl-Bien, M. (1995). Relationship-based approach to leadership: Development of leader-member exchange (LMX) theory of leadership over 25 years. Leadership Quarterly, 6(2), 219-247.
12. Kirkpatrick, S. A., & Locke, E. A. (1991). Leadership: Do traits matter? Academy of Management Executive, 5(2), 48-60.
13. 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.

Verification notes: all thirteen sources were checked via live web search against their publisher or secondary citation record (title, journal or book, year, and volume/page details where stated) immediately before drafting; none were read in full. Thompson, G., & Vecchio, R. P. (2009), discussed in Section 5 as a Norwegian bank-employee study finding a 2007 revision of Situational Leadership performed worse than the original, is referenced in the text but not included as a separate numbered entry because its exact publication details could not be independently confirmed to the same standard as the sources listed above; readers wanting that study should verify it directly before citing it further. The Fiedler meta-analysis discussion in Section 4 also references Strube, M. J., & Garcia, J. E. (1981) by name in the text; it is likewise omitted from the numbered list for the same reason. The branch-manager scenario involving Amina, Rafiq and Tanvir in Section 5 is original, fictional teaching material constructed for this lesson and is explicitly labeled as such; it does not describe a real bank, person, or dataset.

Next in The Human Capital MBA series: Lesson 27 examines transformational, transactional and servant leadership in depth.