Fiedler - The Contingency Model

. ... .’ - ; . ‘:

eadings in social psy-York: Holt, 1952.

, Gurin, G., &:ivity, supervision,d w~~kirs. Ann Ar-.earch Center, 195 1.

& Morse, Nancyion and mode in aninn Arbor, Mich.::r, 1951.leadership in the fit-or, Mich.: Survey

21The Contingency Model: A Theory ofLeadership Effectiveness1Red E. Fiedler

Leadership, as a problem in social psychology, has dealt primarily with twoquestions, namely, how one becomes aleader, and how one can become a goodleader, that is, how one develops effec-tive group performance. Since a num-ber of excellent reviews (e.g., Stogdill,1948; Gibb, 1954; Mann, 1959; Bass,1960), have already dealt with the firstquestion we shall not be concerned withit in the present paper.

The second question, whether agiven leader will be more or less effec-tive than others in similar situations,has been a more difficult problem of re-search and has received corresponding-ly less attention in the psychologicalliterature. The theoretical status ofthe problem is well reflected by Browneand Cohn’s (1958) statement that“ . . . leadership literature is a mass ofcontent without coagulating substancesto bring it together or to produce coor-dination . . .” McGrath (1962), i n

‘The present paper is mainly based on researchconducted under Office of Naval ResearchContracts 170-106, N6-ori-07135 (Fred E.Fiedler. Principal Investigator) and RN 177-472, Noor 1834(36). (Fred E. Fiedler, C. E.Osgood, L. M. Stolurow, and H. C. Triandis,Principal Investigators.) The writer isespecially indebted to his colleagues, A. R.Bass, L. J. Cronbach. M. Fishbein. J. E.McGrath, W. A. T. Meuwese. C. E. Osgood,H. C. Triandis. and L. R. Tucker, who offeredinvaluable suggestions and criticisms ar variousstages of the work.

making a similar point, ascribed this sit-uation to the tendency of investigatorsto select different variables and to workwith idiosyncratic measures and defini-tions of leadership. He also pointed out,however, that most researchers in thisarea have gravitated toward two pre-sumably crucial clusters of leadership at-titudes and behaviors. These are thecritical, directive, autocratic, task-ori-ented versus the democratic, permis-sive, considerate, person-oriented typeof leadership. While this categorizationis admittedly oversimplified, the majorcontroversy in this area has been be-tween the morebrthodox viewpoint, re-flected in traditional supervisory train-ing and military doctrine that the leadershould be decisive and forceful, that heshould do the planning and thinking forthe groups, and that he should coordi-nate, direct and evaluate his men’s ac-tions. The other viewpoint, reflected inthe newer human relations orientedtraining and in the philosophy behindnon-directive and brain-storming tech-nique stresses the need for democratic,permissive, group-oriented leadershiptechniques. Both schools of thoughthave strong adherents and there is evi-dence supporting both points of view(Gibb, 1954; Hare, 1962).

While one can always rationalizethat contradictory findings by other in-vestigators are due to poor research de-

Source: From “The Contingency Model: A Theory of Leadership Ef&ctiveness” by Fred E.Fiedler, in Probh in So&l Psychology, edited by Carl W. Backman and Paul F. Secord, pp. 279-289. New York: McGraw-Hill Book Company, 1970. Reprinted by permission of the author.

2 9 9

. -

sign, or different tests and criteria, suchproblems present difficulties if they ap-pear in one’s own research. We have,during the past thirteen years, con-ducted a large number of studies onleadership and group performance, us-ing the same operational definitions andessentially similar leader attitude mea-sures. The inconsistencies which we ob-tained in our own research programdemanded an integrative theoreticalformulation which would adequately ac-count for the seemingly confusing re-sults.

The studies which we conductedused as the major predictor of group per-formance an interpersonal perception orattitude score which is derived fromthe leader’s description of his most andof his least preferred co-workers. Heis asked to think of all others withwhom he has ever worked, and then todescribe first the person with whomhe worked best (his most preferredcoworker) and then the person withwhom he could work least well (his leastpreferred co-worker, or LPC) . These de-scriptions are obtained, wherever possi-ble, before the leader is assigned to histeam. However, even when we dealwith already existing groups, these de-scriptions tend to be of individualswhom the subject has known in the pastrather than of persons with whom heworks at the time of testing.

The descriptions are typically madeon 20 eight-point bipolar adjectivescales, similar to Osgood’s SemanticDifferential (Osgood, et al., 1957),e.g.,

Pleasant -:-:-:-i,:,:,:,Unpleasant

I .wFriendly -:-:,:-iv:-:-:,Unfriendly

These items are scaled on an eval-

.;1 uative dimension, giving a score of

! 8 to the most favorable pole (i.e.,Friendly, Pleasant) and a score of I to

: I-. .,

Leadership

the least favorable pole. Two mainscores have been derived from these de-scriptions. The first one, which wasused in our earlier studies, is based onthe profile similarity measure D (Cron-bath and Gleser, 1953) between the de-scriptions of the most and of the leastpreferred co-worker. This core, calledthe Assumed Similarity between Gppo-sites, or ASo, indicates the degree towhich the individual perceives the twoopposites on his co-worker continuumas similar or different. The second scoreis simply based on the individual’sdescription of his least preferred co-worker, I-PC, and indicates the degreeto which the subject evaluates his LPCin a relatively favorable or unfavorablemanner. The two measures are highhcorrelated (.80 to .95) and will here betreated as interchangeable.

We have had considerable difficultyin interpreting these scores since theyappear to be uncorrelated with the usualpersonality and attitude measures. Theyare, however, related to the Ohio StateUniversity studies’ “Initiation of struc-ture” and “Consideration” dimensions(Stogdill and Coons, 1957). Extensivecontent analysis (Meuwese and Gonk,1960; Julian and McGrath, 1963; Mor-ris and Fiedler, 1964) and a series ofstudies by Hawkins (1962) as well as re-search by Bass, Fiedler, and Krueger(1964) have given consistent results.These indicate that the person withhigh I-PC or ASo, who perceives hileast preferred co-worker in a relativelyfavorable,’ accepting manner, tends tobe more accepting, permissive, consid-erate, and person-oriented in his rek+tions with group members. The personwho perceives his most and least preferred co-workers as quite different, andwho sees his least preferred coworker ina very unfavorable, rejecting mannertends to be directive, task-oriented and*controlling on task relevant group be- : i,haviors in his interactions. . .*

.;; f.

7%econti7. .j‘\

ASo and LPCwith group perfcety of studies, aabove, not consrection. For examhigh school ba:ciometrically chcorrelated - .6’percent of gam- .51 with the ;civil engineerand the melter frelated - .52 vopen-hearth she1956). These ndicate t h a t lotwere associated

. formance, i.e.performed bettrective leaders tsive, acceptingthe ASo scoreaccepted campalso negativelycome of consboard chairmarsame circumst(Godfrey, FiecThus, groupsseemed to reqitudes. In a mccreativity in HSCOR correlatedin religiously I-formally appoirreligiously hetwhile the conegeneous groupswas -&in1ing emergent (kai;;) . (Fied

The resultsclearly shower

magnitude of ttingent upontask situationto deal. Ourthen into (a)

. Two mainrom these de-I, which wasj, is based onure D (Cron-tween the de-d of the leasts core, calledetween Oppo-the degree to:eives the two.er continuume second scoree individual’spreferred co-

tes the degreeluates his LPCor unfavorablerres are highlyId will here bee.rable difficulty)res since theyi with the usualneasures. Theythe Ohio Stateation of stfuc-,n” dimensions‘57). Extensiveese and Oonk,th, 1963; Mor-and a series of2) as well as re-‘, and Kruegerisistent results.ie person withCI perceives hisr in a relativelyInner, tends tonissive, consid-ted in his rela-ers. The persont and least pre-te different, and-red coworker inzjecting mannerrsk-oriented andevant group be*ms .

The Contingency ‘Model: A Theory of Lade&~ Ejjictiveness 301

ASo and LPC scores correlated highlywith group performance in a wide vari-ety of studies, although, as mentionedabove, not consistently in the same dierection. For example, in two samples ofhigh school basketball teams the so-ciometrically chosen leader’s ASo scorecorrelated -.69 and -.58 with thepercent of games won by teams and- .5 1 with the accuracy of surveying ofcivil engineer teams (Fiedler, 1954),and the melter foreman’s ASo score cor-related - .52 with tonnage output ofopen-hearth shops (Cleven and Fiedler,1956). These negative correlations in-dicate that low ASo or LPC scoreswere associated with good group per-formance, i.e., that these groupsperformed better under managing, di-rective leaders than under more permis-sive, accepting leaders. However, whilethe ASo score of the sociometricallyaccepted company manager correlatedalso negatively ( - .70) with the net in-come of consumer cooperatives, theboard chairman’s ASo score under thesame circumstances correlated + .62(Godfrey, Fiedler, and Hall, 1959).Thus, groups with different tasksseemed to require different leader atti-tudes. In a more recent study of groupcreativity in Holland, the leader’s LPCscore correlated with performance + .75in religiously homogeneous groups withformally appointed leaders, but - .72 inreligiously heterogeneous groups; andwhile the correlation was + .75 in homo-geneous groups with appointed leaders itwas - 64 in homogeneous groups hav-

ing emergent (sociometrically nominatedieadrs). (Fiedler, Meuwese and Oonk,

The results of these investigationsclearly showed that the direction andmagnitude of the correlations were con-tingent upon the nature of the group-task situation with which the leader hadto deal. Our problem resolved itselfthen into (a) developing a meaningful

system for categorizing group-task situ-ations; (b) inducing the underlying the.ore&al model which would integratethe seemingly inconsistent results ob-tained in our studies, and (c) testing thevalidity of the model by adequate re-search.

DEVELOPMENT OFTHE MODEL

Key definitions. We shall here be con-cerned solely with “interacting” ratherthan “co-acting” task groups. By an in-teracting task group we mean a face-to-face team situation (such as a basketballteam) in which the members workirtterde@tly on a common goal. Ingroups of this type, the individual’s con-tributions cannot readily be separatedfrom total group performance. In aco-acting group, however, such as abowling or a rifle team, the group per-formance is generally determined bysumming the members’ individual per-formance scores.

We shall define the leader as thegroup member who is officially appointed or elected to direct and coordi-nate group action. In groups in whichno one has been so designated, we haveidentified the informal leader by meansof sociometric preference questions suchas asking group members to name theperson who was most influential in thegroup, or whom they would most preferto have as a leader in a similar task.

The leader’s effectiveness is here de-fined in terms of the group’s perfor-mance on the assigned primary task.Thus, although a company managermay have, as one of his tasks, the job ofmaintaining good relations with his cus-tomers, his main job, and the one onwhich he is in the final analysis evalu-ated, consists of the long range profita-bility of the company. Good relationswith customers, or high morale and lowlabor turnover may well contribute to

302

success, but they would not be the basiccriteria by this definition.

Tk categorization of group-task sicua-tions. Leadership is essentially a prob-lem of wielding influence and power.When we say that different types ofgroups require different types of leader-ship we imply that they require a differ-ent relationship by which the leaderwields power and influence. Since it iseasier to wield power in some groupsthan in others, an attempt to categorizegroups might well begin by asking whatconditions in the group-task situationwill facilitate or inhibit the leader’s ex-ercise of power. On the basis of our pre-vious work we postulated three impor-tant aspects in the total situation whichinfluence the leader’s role.

1 . Leader-member&lions. Thelead-er who is personally attractive to hisgroup members, and who is respected byhi group, enjoys considerable power(French, 1956). In fact, ifhe has the con-fidence and loyalty of hi men he has lessneed of official rank. This dimension cangenerally be measured by means of so-ciometric indices or by group atmospherescaleS (Cf. Fiedler, 1962) which indicatethe degree to which the leader experi-ences the groups as pleasant and well dis-posed toward him.

2. Tark strucr~re. The task generallyimplies an order “from above” which in-corporates the authority of the superiororganization. The group member whorefuses to comply must be prepared toface disciplinary action by the higherauthority. For example, a squad memberwho fails to perform a lawful commandof his sergeant may have to answer tohis regimental commander. However,compliance with a task order can be en-forced only if the task is relatively wellstructured, i.e., if it is capable of beingprogrammed, or spelled out step by step.One cannot effectively force a group to

LdtTShip

perform well on an unstructured tasksuch as developing a new product orwriting a good play.

Thus, the leader who has a sttucturedtask can depend on the backing of hissuperior organizations, but if he has anunstructured task the leader must relyon his own resources to inspire and mo-tivate his men. The unstructured taskthus provides the leader with much lesseffective power than does the highlystructured task.

We operationalized this dimension byutilizing four of the aspects which Shaw(1962) recently proposed for the classi-fication of group task. These are, (a) de-cision uerijiability, the degree to whichthe correcmess of the solution can bedemonstrated objectively; (b) good ckzr-ity, the degree to which the task require-ments are clearly stated or known to thegroup; (c) goal path multiplicity, the de-gree to which there are many or few pro-cedures available for performing thetask (reverse scoring); and (d) soiutiunspecificity, the degree to which there isone rather than an infinite number ofcorrect solutions (e.g., writing a storyvs. solving an equation). Ratings basedon these four dimensions have yieldedinterrater reliabilities of .80 to .90.

3. Position @UKT. The third dimen-sion is defined by the power inherent inthe position of leadership irrespective ofthe occupant’s personal relations withhis members. This includes the rewardsand punishments which are officially ortraditionally at the leader’s disposal, hisauthority as defined by the group’s rulesand by-laws, and the organizational support given to him in dealing with hismen. This dimension can be operation-ally defined by means of a check list(Fiedler, 1964) containing items such as“Leader can effect promotion or demo-tion, ” “Leader enjoys special rank andstatus in real life which sets him apart

Ti tom”:,,

FlGURE

t+:

(-

from, and aThe mediurfour indepensituations w;

AthreedirGroup-task 5on the basisleader-memtand positiorgroup in a trough categplished by tsions so thacube (Fig. 1whether theattitudes an{each of thesrelatively silrection. If tthe group c

cessfilly ac

L&&l-Ship

m unstructured taskg a new product or,‘.t who has a structuredn the backing of hisens, but if he has anthe leader must relyes to inspire and mo-he unstructured task.eader with much lessIan does the highly

.zed this.dimension bye aspects which Shaw.oposed for the classi-ak. These are, (a) de-the degree to which

’ the solution can bextively; (b) good cIar-rhich the task require-xated or known to theth multiplicity, the de-‘e are many or few pro-

for performing theing); and (d) solution;ree to which there isan infinite number of(e.g., writing a story

lation). Ratings basednensions have yieldedties of .80 to .90.

uer. The third dimen-the power inherent in

Idership irrespective ofersonal relations withis includes the rewardswhich are officially or

re leader’s disposal, hisled by the group’s rulesthe organizational sup-m in dealing with hisrsion can be operation-means of a check listontaining items such asct promotion or demo-njoys special rank andI which sets him apart

TheCmtingencyModel:A TkqofL.eadenhi~ Effectiveness 303

FIGURE 1 . A MODEL FOR THE CLASSIFICATION OF GROUP-TASKSITUATIONS.

(+I Afi ective leader-member relations

from, and above his group members.”The medium interrater agreement offour independent judges rating 35 groupsituations was .95.

A three dimensional group classifiarion.Group-task situations can now be ratedon the basis of the three dimensions ofleader-member relations, task structure,and position power. This locates eachgroup in a three dimensional space. Arough categorization can be accom-plished by halving each of the dimen-sions so that we obtain an eight celledcube (Fig. 1). We can now determinewhether the correlations between leaderattitudes and group performance withineach of these eight cells, or octants, arerelatively similar in magnitude and di-rection. If they are, we can infer thatthe group classification has been sue*cessfully accomplished since it shows

that groups falling within the same oc-tant require similar leader attitudes.

An earlier paper has summarized 52group-task situations which are basedon our previous studies (Fiedler, 1964).These 52 group-task situations havebeen ordered into the eight octants. Ascan be seen from Table 1, groups fallingwithin the same octant show correla-tions between the leader’s ASo or LPCscore and the group performance crite-rion which are relatively similar in mag-nitude and direction. We can thus .inferthat the group classification has beenaccomplished with at least reasonablesuccess.

Consideration of Figure 1 suggests afurther classification of the cells in termof the effective power which the grouptask situation places at the leader’s dis-posal, or more precisely, the favorable-

Such an ordering can be accom-plished without difficulty at the extremepoles of the continuum. A liked andtrusted leader with high rank and astructured task is in a more favorableposition than is a disliked and powerlessleader with an ambiguous task. The in-termediate steps pose certain theoreticaland methodological problems. To col-lapse a three-dimensional system into aunidimensional one implies in Coombs’terms a partial order or a lexicographicsystem for which there is no unique so-lution. Such an ordering must, there-fore, be done either intuitively or inaccordance with some reasonable as-sumptions. In the present instance wehave postulated that the most impor-tant dimension in the system is theleader-member relationship since thehighly liked and respected leader is lessin need of position power or the powerof the higher authority incorporated inthe task structure. The second-most im-portant dimension in most group-tasksituations is the task structure since a

L

?leader with a highly structured task does

f not require a powerful leader position.

4 _ i9

TABLE 1 l MEDIAN CORRELATION BETWEEN LEADER LPC ANDGROUP PERFORMANCE IN VARIOUS OCI’AN’Ts

Number ofRelations

Leader-Member Task Position Median IncludedRelations Strucnlre Power Correlation in Median

Octant I G&d snucnlred Strong - .52 2Octant II Good sttuctured Weak - .58 3O c t a n t I I I G o o d Unstructured Strong -.41 4Octant IV Good Unstructured Weak .47 10Octant V Mod. poor smlctured Strong .42 6Chant VI Mod. poor Structured Weak 0Cktant VII Mod. poor Unstructured Strong .05 10Chant VIII Mod. poor Unsm~ctwed W e a k -.43 12

ness of the situation for the leader’sexercise of his power and influence.

(For example, privates or non-commis-sioned officers in the army are at timescalled upon to lead or instruct officersin certain highly structured tasks such asdemonstrating a new weapon, or, forexample, teaching medical officers closeorder drill-though not in unstructuredtasks such as planning new policies onstrategy.) This leads us here to order thegroup-task situations first on leader-member relations, then on task struc-ture, and finally on position power.While admittedly not a unique solu-tion, the resulting ordering constitutesa reasonable continuum which indi-cates the degree of the leader’s effectivepower in the group.*

As was already apparent from Table1, the relationship between leader atti-tudes and group performance is contin-gent upon the accurate classification ofthe group-task situation. A more mean-

‘Another cell should be added which containsreal-life groups which reject their leader.Exercise of power would be very dificult inthis situation and such a cell should be pkdat the extreme negative end of the continuum.Such cases are treated in the section onvalidation.

:N LEADER LPC ANDS OCTANTS

Number ofRelations

Mediin IncludedCorrelation in Median

-.52 2-.58 3-.41 4

.47 10

.42 60

.05 1 0

-.43 I2

le, privates or non-commis-ers in the army are at timesL to lead or instruct officersighly structured tasks such asIng a new weapon, or, foraching medical officers close-though not in unstructured3s planning new policies onhis leads us here to order thesituations first on leader-

.ations, then on task struc-finally on position power.wittedly not a unique solu-zsulting ordering constitutesle continuum which indi-agree of the leader’s effectiveie gr0up.lalready apparent from Tableionship between leader atti-:roup performance is contin-the accurate classification ofask situation. A more mean-

II should be added which containsups which reject their leader.Jower would be very difficult-inI and such a cell should be placedne negative end of the continuum.re treated in the section on

l 0

l

3 0 6

ingful model of this contingency rela-tionship emerges when we now plot thecorrelation between LPC or ASo andgroup performance on the one hand,against the octants ordered on the effec-t ive power, or favorableness-for-the-leader dimension on the other. This isshown on Figure 2. Note that eachpoint in the plot is a correlation predict-ing leadership performance or group ef-fectiveness. -l-he plot therefore repre-sents 53 sets of groups totalling over 800separate groups.

As Figure 2 shows, managing, con-trolling, directive (low I-PC) leadersperform most effectively either undervery favorable or under very unfavora-ble situations. Hence we obtain negative correlations between LPC andgroup performance scores. Consider-ate, permissive, accepting leaders ob-tain optimal group performance undersituations intermediate in favorable-ness. These are situations in which (a)the task is structured, but the leader isdisliked and must, therefore, be diplo-matic; (b) the liked leader has an am-biguous, unstructured task and must,therefore, draw upon the creativityand cooperation of his members. Herewe obtain positive correlations betweenLPC and group performance scores.Where the task is highly structured andthe leader is well-liked, non-directivebehavior or permissive attitudes (suchas asking how the group ought to pro-ceed with a missile count-down) is nei-ther appropriate nor beneficial. Wherethe situation is quite unfavorable, e.g.,where the disliked chairman of a vol-unteer group faces an ambiguous task,the leader might as well be autocraticand directive since a positive, non-di-rective leadership style under these con-ditions might result in complete inac-tivity on the part of the group. Thismodel, thus, tends to shed some lighton the apparent inconsistencies in our

Leadership

own data as well as in data obtained byother investigators.

EMPIFUCAL TEST-S EXTENSIONOF THE MODELThe basic hypothesis of the model sug-gests that the directive, controlling,task oriented (low LPC) leader will bemost successful in group-task situationswhich are either very favorable or elsevery unfavorable for the leader. Thepermissive, considerate, human rela-tions oriented (high I-PC) leader willperform best under conditions whichare intermediate in favorableness. Thishypothesis was tested by re-analyzingdata from previous studies as well asby a major experiment specifically de-signed to test the model. Both arebriefly described below.

Re-analyses of Previous StudiesAs we indicated before, there is reasonto believe that the relationship betweenthe leader and his members is the mostimportant of the three dimensions forclassifying group-task situations. Theproblem of exercising leadership will bea relatively easy one in group-task situ-ations in which the leader is not onlyliked by his crew and gets along wellwith his group, but in which the task isstructured and the leader has a rela-tively powerful position. The situationwill be somewhat more difficult if theleader under these circumstances has anonly moderately good relationship withhis group members, and it will be quitedifficult if the leader-member relationsare very poor, if the group members re-ject ot actively dislike the leader. Ordi-narily this does not occur in laboratorystudies. It does happen, however, thatreal-life groups strongly reject leaders-sometimes to the point of sabotagingthe task. Since such a situation wouldpresent a very difficult problem in

Tk Conringt

TABLI(AC’s)PAT-T

AC is most 1AC is most iAC is most 1AC is not mAC is not mAC is not m

‘Table adaptr

leadership,formance f

t r o l l i n g l e acorrelationWC scoreThis resultbomber crethe apptopby new ana

Bomberconducted Iler, 1955) Imance conThis is thecuracy of hradar ptocesified on tlbetween tlhis crew. ’whether 01mander wsof the crcommandehis keymer(the radar I

The restsented in 1correlationfonnance :having vergroup relatitive in the

L?adership

data obtained by

S EXTENSION

of the model sug-tive, controlling,‘C) leader will beup-task situationsI favorable or else

the leader. Theite, h u m a n rela-LPC) leader willconditions whichtvorableness. Thisd by re-analyzingitudies as well asnt specifically de-model. Both are

W.

vious StudiesIre, there is reason,ationship betweenembers is the most.ee dimensions fork situations. The: leadership will bein group-task situ-leader is not onlyId gets along welln which the task isleader has a rela-ion. The situation.ore difficult if thercumstances has and relationship withand it will be quiter-member relationsgroup members re-

:e the leader. Ordi-occur in laboratory3en, however, thatgly reject leaders-joint of sabotagingI a situation wouldfficult problem in

The Contingency Model: A Theoq of Lxdership Effectiveness 3 0 7

TABLE 2 c CORRELATIONS BETWEEN AIRCRAFT COMMANDER’S(AC’s) ASo SCORE AND RADAR BOMB SCORES UNDER DIFFERENTPATTERNS OF SOCIOMETRK CHOICES IN B-29 BOMBER CREWS’

RHO N

AC is most preferred crew member and chooses keymen (K) -.81 10AC is most preferred crew member and is neutral to K -.14 6AC is most preferred crew member and does not choose K .43 6AC is not most preferred crew member but chooses K -.03 18AC is not most preferred crew member and is neutral to K -.80 5AC is not most preferred crew member and does not choose K - .67 7

‘Table adapted from Fiedler (1955).

leadership, we would expect better per-formance from the task-oriented, con-trolling leader, and hence a negativecorrelation between the leader’s ASo orLPC score and his group’s performance.rlhis result appeared in one study ofbomber crews for which we already hadthe appropriate data, and it was testedby new analyses in two other studies.

Bomber Crew Study. A study wasconducted on B-29 bomber crews (Fied-ler, 1955) where the criterion of perfor-mance consisted of radar bomb scores.This is the average circular error, or ac-curacy of hitting the target by means ofradar procedures. The crews were clas-sified on the basis of their relationshipbetween the aircraft commander andhis crew. The crew:‘ were ordered onwhether or not (a) the aircraft com-mander was the most chosen memberof the crew, and (b) the aircraftcommander sociometrically endorsedhis keymen on his radar bombing team(the radar observer and navigator).

The results of this analysis are pre-sented in Table 2. As can be seen, thecorrelations between ASo and crew per-formance are highly negative in crewshaving very good and very poor leader-group relations, but they tend to be positive in the intermediate range.

Anti-aircraft Art&y Crews. A sec-ond set of data came from a study ofanti-aircraft artillery crews (Hutchinsand Fiedler, 1960). Here the criterionof crew performance consisted of scoresindicating the “location and aquisi-tion” of unidentified aircraft. Thesecrews were subdivided on the basis ofleader-crew relations by separately cor-relating the leader’s LPC score withgroup performance (a) for the ten crewswhich most highly chose their crewcommander, (b) the ten which were inthe intermediate range, and (c) the tencrews which gave the least favorable so-ciometric choices to their leader. Thesedata are presented in Table 3.

Consumer Cooperative Carpanies. Fi-nally we reanalyzed data from a studyof 31 consumer cooperatives (Godfrey,Fiedier and Hall, 1959) in which thecriterion of performance consisted ofthe per cent of company net incomeover a three-year period. The com-panies were subdivided into those inwhich the general manager was socio-metrically chosen (a) by his board of di-rectors as well as by his staff of assistantmanagers, (b) those in which the gen-eral manager was chosen by his boardbut not his staff, or (c) by his staff butnot his board, and (d) the companies in

Leadmhip

TABLE 3 . CORRELATIONS BETWEEN LEADER LPC SCORES ANDANTI-AIRCm ARTILLERY CREW PERFORMANCE

RHO N

Most highly chosen crew commanders -.34 1 0Middle range in sociometric choices .49 10

Lowest chosen crew commanders -.42 10

TABLE 4 . CORRELATIONS BETWEEN GENERAL MANAGER’S ASoSCORE AND COMPANY NET INCOME

RHO N

Gen. mgr. is mostly chosen by board and staff (ASo perf.) -.67 10

Gen. mgr. is chosen by board, but rejected by staff .20 6Gen. mgr. is rejected by board. but chosen by staff .26 6Gen. mgr. is rejected by board and staff -.75 7

which the general manager was re-jected, or not chosen, by both board ofdirectors and staff. (Table 4.)

As these tables, and Figure 3 show,the task-oriented, managing, low LF’Cleaders performed best under very favor-able and under very unfavorable situa-tions, while the permissive, considerateleaders performed best under conditionsintermediate in favorableness. Thesedata, therefore, clearly support the hy-pothesis derived from the model.

Experimental Test of theContingency Made1hi cooperation with the Belgian NavalForces we recently conducted a majorstudy which served in part as a specifictest of the model. Only aspects imme-diately relevant to the test are here de-scribed. The investigation was con-ducted in Belgium where the Frenchand Dutch speaking (or Flemish) sectors

I of the country have been involved in a

1long standing and frequently acrimon-

!

ious dispute. This conflict centers about

the use of language, but it also involvesa host of other cultural factors whichdifferentiate the 60 per cent Flemishand 40 per cent French speaking popu-lation groups in Wallonie and Brus-sels. This “linguistic problem” which isrooted in the beginning of Belgium’s na-tional history, has in recent years beenthe cause of continuous public contro-versy, frequent protest meetings, andoccasional riots.

The linguistic problem is of particu-lar interest here since a group, consist-ing of members whose mother tongue,culture, and attitudes differ, will clearlypresent a more difficult problem in lead-ership than a group whose membersshare the same language and culture.We were thus able to test the major hypothesis of the model as well as to ex-tend the research by investigating thetype of leadership which linguisticallyand culturally heterogeneous groups re-quire.

Design. The experiment was con-ducted at the naval training center at

Tht?Cona’ngenc

FIGCSCORESOF LEAI:

CREU‘

1.00 r30 -

ou .60 -3i -ati .40ZEzg a-“<$0)$2 a0 -ifi:PI-0’ -.20-6e‘g m -40 -Fs -60 -

-.80--1.ocl L

Ste. Croix-!petty oficeibeen selecton the basiobtained Land languag

‘This invest&collaborationhis &dents YResearch Fell1963-1964. -permission otChief of Stafcarried out a:Navale, Ste.to express hi:appreciationCaptain V. \

Leadership

ZORES AND.NCE

N

1 01010

JAGER’S ASo

U-IO N

-.67 10.20 6.26 6

-.75 7

iut it also involvesUral factors whichper cent Flemishch speaking popu-allonie and Brus-problem” which isng of Belgium’s na-recent years been

ous public contro-zst meetings, and

blem is of particu-e a group, consist-se mother tongue,j differ, will clearly.It problem in lead-) whose members:uage and culture.test the major hy-:I as well as to ex-J investigating therhich linguisticallylgeneous groups re-

,eriment was con-training center at

Tk Crmtingeruy Model: A Theory of -ship Effectiwss 309

FIGURE 3 l CORRELATIONS BETWEEN LEADER LPC OR ASoSCORES AND GROUP PERFORMANCE UNDER THREE CONDITIONSOF LEADER ACCEPTANCE BY THE GROUP IN STUDIES OF BOMBER

CREWS, AN-II-AIRCRAFT ARTILLERY CREWS AND CONSUMERCOOPERATIVES.

1.00 r

-l.oolLeader sociometrically Leader neither chosen Leader sociometrically

highly chosen nor rejected rejected

Ste. Croix-Bruges.’ It utilized 48 career The experiment was specifically de-petty officers and 240 recruits who had signed to incorporate the three majorbeen selected from a pool of 546 men group classification dimensions shownon the basis of a pre-test in which we on Figure 1, namely, leader-memberobtained WC, intelligence, attitude, relations, position power, and taskand language comprehension scores. structure. It also added the additional

‘This investigation was conducted in the personnel and the facilities of the centercollaboration with Dr. J. M. Nuttin (Jr.) and available to us, but whose active participationhis students while the author was Ford Faculty in the planning and the execution of theResearch Fellow at the University of Louvain, project made this study possible. We are also1963-1964. The experiment, undertaken with most grateful to Dr. U. Bouvier, Director ofpermission of Commodore L. Petitjean. then the Center for Social Studies, Ministry ofChief of Staff of the Belgian Naval Forces, was Defense, to Capt. W. Cafferata. USN, thecarried out at the Centre de Formation senior U.S. Naval representative of theNavale. Ste. Croix-Bruges. The writer wishes Military Assistance and Advisory Group,to express his especial gratitude and Brussels, and to Cmdr. J. Robison, U.S. Navalappreciation to the commandant of the center, Attache in Brussels, who provided liaison andCaptain V. Van Laethem, who not only made guidance.

320

dimension of group homogeneity vs.heterogeneity. Specifically, 48 groupshad leaders with high position power(petty officers) while 48 had leaderswith low position power (recruits); 48groups began with the unstructuredtask, while the other 48 groups beganwith two structured tasks; 48 groupswere homogeneous, consisting of threeFrench or three Dutch speaking men,while the other 48 groups were hetero-geneous, consisting of a French speak-ing leader and two Flemish members, ora Dutch speaking, Flemish leader and,two French speaking members. The

_ quality of the leader-member relationswas measured as in our previous studiesby means of a group atmosphere scalewhich the leader completed after eachtask session.

Groug Performance Criterion. Two es-sentially identical structured tasks wereadministrated. Each lasted 25 minutesand required the groups to find theshortest route for a ship which, givencertain fuel capacity and required portsof call, had to make a round trip callingat respectively ten or twelve ports. Thetasks were objectively scored on the ba-sis of sea miles required for the trip. Ap-propriate corrections and penalties wereassigned for errors.

The unstructured task required thegroups to compose a letter to young menof 16 and 17 years, urging them tochoose the Belgian Navy as a career.The letter was to be approximately 200words in length and had to be com-pleted in 35 minutes. Each of the let-ters, depending upon the language inwhich it was written, was then rated byDutch or by French speaking judges onstyle and use of language,. as well as in-terest value, originality, and persuasive-ness. Estimated reliability was .92 and.86 for Dutch and French speaking

! judges, respectively.It should be noted in this connection

i.

that the task of writing a letter is not asunstructured as might have been desir-able for this experiment. The form ofany letter of this type is fairly standard-ized, and its content was, of course,suggested by the instructions. The na-vy officers with whom we consultedthroughout the study considered it un-wise, however, to give a highly unstruc-tured task, such as writing a fable orproposing a new policy, since tasks ofthis nature were likely to threaten themen and to cause resentment and poorcooperation. High and low task-struc-ture is, thereforr, less well differentiatedin this study than it has been in pre-vious investigations.

Results. The contingency model speci&s that the controlling, managing, lowL.EC leaders will be most effective eitherin very favorable or else in relatively un-favorable group-task situations, while thepermissive, considerate, high LPC leaderswill be mote effective in situations inter-mediate in difficulty.

The basic test of the hypothesis re-quires, therefore, that we order thegroup-task situations, represented inthis experiment, in terms of the diffl-culty which they are likely to present forthe leader. Since there are 16 cells inthe de&n, the Size of the sample with-in each cell (namely 6 groups) is,of course, extremely small. However,where the conditions are reasonablyreplicated by other cells, the relation-ship can be estimated from the medianrank-order correlations.

The hypothesis can be tested mostreadily with correlations of leader LPCand group performance in homogeneousgroups on the more reliably storable sec-ond structured task. These conditionsapproximate most closely those repre-sented on Figure 3, on bomber and anti-aircraft crews and consumer coopera-tives. We have here made the fairlyobvious assumption that the p~wirful

Tk confingc

leader or thaccepted fatuation tharthose who Eand tense. Iby two cell:there were thalf the grstructured tunsmcturectask situaticfor the leading results:

High groupatmospherhigh posit!power

High groupaanospherlow positic

I.otv groupaanospherhigh pitpower

low p&tic

These iin size amwhich theleaders areable and utions: the Ihigh LPC 1the intermc

Extendilheterogenemake a mtions for Htask dimen16 cells xordered orhere assigrfavorablesions, i.e.,

l.t?&TShi@

ng a letter is not asIt have been desir-nent. The form ofe is fairly standard-nt was, of course,tructions. The na-lorn we consultedy considered it un-Je a highfy unstruc-writing a fable or

Jicy, since tasks ofely to threaten the:sentment and poorand low task-struc-;s well differentiatedit has been in pre-

:ingency model speciling, managing, lowmost effective eitherelse in relatively un-situations, while thelte, high Lpc leaders‘e in situations inter-

f the hypothesis re-that we order thens, represented in

1 terms of the diffi-3 likely to present forhere are 16 cells inof the sample with-

nely 6 groups) is,4y small. However,Lens are reasonably; cells, the relation-ted from the medianions.can be tested most

ations of leader LPCmce in homogeneousreliably storable sec-k. These conditionsclosely those repre-on bomber and anti-

consumer coopera-Lere made the fairlyn that the powerful

Tk Contingency Model: A Theory of Leadership Ejjectiuenerr 311

leader or the leader who feels liked andaccepted faces an easier group-task sit-uation than low ranking leaders andthose who see the groups as unpleasantand tense. Each situation is representedby two cells of six groups, each. Sincethere were two orders of presentation-half the groups worked first on thestructured task, the other half on theunstructured task, arranging the grouptask situations in order of favor-ablenessfor the leader then gives us the follow-ing results:

Order 1 Order 2

High groupatmosphere andhigh positionpower -.77 -.77

High groupatmosphere andlow position power + .60 +.50

LQW groupatmosphere andhigh positionpower +.16 +.01

L4-m WUPatmosphere andlow position power . - .16 -.43

These are, of Course, the trendsin size and magnitude of correlationswhich the model predicts. Low Z-PCleaders are again most effective in favor-able and unfavorable group-task situa-tions: the more permissive, consideratehigh UC leaders were more effective inthe intermediate situations.

Extending the model to includeheterogeneous groups requires that wemake a number of additional assump-tions for weighting each of the group-task dimensions so that all 48 cells (i.e.,16 cells x 3 tasks) can be reasonablyordered on the same scale. We havehere assigned equal weights of 3 to thefavorable poles of the major dimen-sions, i.e., to homogeneity, high group

atmosphere, and high position power. Aweight of one was assigned to the firststructured task, and a weight of two tothe second structured task on the as-sumption that the structured task makesthe group-task situation somewhat morefavorable than the unstructured task,and that the practice and learning effectinherent in performing a second, prac-tically identical task, will make thegroup-task situation still more favorablefor the leader. Finally, a weight of onewas given to the “second presentation,”that is, the group task which occurredtoward the end of the session, on theassumption that the leader by that timehad gotten to know his group membersand had learned to work with themmore effectively, thus again increasingthe favorableness of his group-task situ-ation to a certain extent.

The resulting weighting system leadsto a scale from 12 to 0 points, with 12as the most favorable pole. If we nowplot the median correlation coefficientsof the 48 group-task situations againstthe scale indicating the favorableness ofthe situation for the leader, we obtainthe curve presented on Figure 4.

As can be seen, we again obtain acurvilinear relationship which resem-bles that shown on Figure 2. Heteroge-neous groups with low position powerand/or poor leader-member relations fallbelow point 6 on the scale, and thustend to perform better with controlling,directive, low LPC leaders. Only underotherwise very favorable conditions doheterogeneous groups perform betterwith permissive, considerate high LPCleaders, that is, in group-task situationscharacterized by high group atmosphereas well as high position power, four ofthe six correlations (66%) are positive,while only five of eighteen (28%) arepositive in the less favorable group-tasksituations.

It is interesting to note that the curve

.

E

FIGURE 4 l MEDIAN CORRELATIONS BETWEEN LEADER LCP AND GROUP PERFORMANCE SCORES PL.O’lTEDAGAINST FAVORABLENESS-FOR-LEADER SCALE IN THE BELGIAN NAVY STUDY.

Code for *digit numbers figure indicating the type of group invofved

COlllpOSitkXl Position Power HighGroup Almos. LowGroup Atmos. Task 1 s1 Pres. 2nd Pres.-

HomogelleOJS HighHomogeneous

k$: i

Slrucfured I 1Structured II

Heterogeneous 3x

Unstructured ii:6

Heterogeneous LOW 4

33

2224

2344 66

72

44

66

” .-1 65-60 63

21

-.W 43

73-1.00 I I I I I

45I 1 I I I I I

12 11 10 9 6 7 6 5 4 3 2 1 0Favorablefor leader

Moderately Veryunfavorable unlavorable

Extremelyunfavorable

for leader

Tk Contingency Model: A Theory of Leadership Effhctiwness 3l3

is rather flat and characterized by rela-tively low negative correlations as we gotoward the very unfavorable end of thescale. This result supports Meuwese’s(1964) recent study which showed thatcorrelations between leader LPC aswell as between leader intelligence andgroup performance tend to become at-tenuated under conditions of relativestress. These findings suggest that theleader’s ability to influence and controlthe group decreases beyond a certainpoint of stress and difficulty in thegroup-task situation.

DISCUSSION

The contingency model seeks to recon-cile results which up to now had to beconsidered inconsistent and difficult tounderstand. We have here attempted todevelop a theoretical framework whichcan provide guidance for further re-search. While the model will undoubt-edly undergo modifications and elabo-ration as data become available, itprovides an additional step toward abetter understanding of leadership pro-cesses required in different situations.We have here tried to specify exactlythe type of leadership which differentgroup task-situations require.

The model has a number of im-portant implications for selection andtraining, as well as for the placement ofleaders and organizational strategy. Ourresearch suggests, first of all, that wecan utilize a very broad spectrum of in-dividuals for positions of leadership.The problem becomes one of placementand training rather than of selectionsince both the permissive, democratic,human-relations oriented, and the man-aging, autocratic, task-oriented leadercan be effectively utilized. Leaders can betrained to recognize their own style ofleadership as well as the conditions whichare most compatible with their style.

The model also points to a variety ofadministrative and supervisory strate-gies which the organization can adoptto fit the group-task situation to theneeds of the leader. Tasks can, after all,be structured to a greater or lesser ex-tent by giving very specific and de-tailed, or vague and general instruc-tions; the position power of the groupleader can be increased or decreased andeven the congeniality of a group, and itsacceptance of the leader can be affectedby appropriate administration action,such as for instance increasing or de-creasing the group’s homogeneity.

The model also throws new light onphenomena which were rather difficultto fit into our usual ideas about mea-surement in social psychology. Why, forexample, should groups differ so mark-edly in their performance on nearly par-allel tasks? The model-and our data-show that the situation becomes easierfor the leader as the group moves fromthe novel to the already known grouptask situations. The leaders who excelunder relatively novel and thereforemore difficult conditions are not neces-sarily .those who excel under thosewhich are more routine, or betterknown and therefore more favorable.Likewise, we find that different types oftask structure require different types ofleader behavior. Thus, in a researchproject’s early phases the project direc-tor tends to be democratic and permis-sive; everyone is urged to contribute tothe plan and to criticize all aspects ofthe design. This situation changes radi-cally in the more structured phase whenthe research design is frozen and theexperiment is underway. Here the re-search director tends to become man-aging, controlling, and highly auto-cratic and woe betide the assistant whoattempts to be creative in giving in-structions to subjects, or in his timingof tests. A similar situation is often

--..* . . -,. .

314

found in business organizations wherethe routine operation tends to be wellstructured and calls for a managing, di-rective leadership. The situation be-comes suddenly unstructured when acrisis occurs. Under these conditionsthe number of discussions, meetings,and conferences increases sharply so asto give everyone an opportunity to ex-press his views.

At best, this model is of course onlya partial theory of leadership. The lead-er’s intellectual and task-relevant abili-ties, and the members’ skills and moti-vation, all play a role in affecting thegroup’s performance. It is to be hopedthat these other important aspects ofgroup interaction can be incorporatedinto the model in the not too distantfuture.

REFERENCES

Bass, A. R., Fiedler, E E., and Krueger, S.Personality correlates of assumed similarity(ASo) and related scores. Urbana, 111.:Group Effectiveness Research Laboratory,University of Illinois, 1964.

Bass, B. M. Leadership psychology and organiza-tional behavior. New York: Harper Brothers,1960.

Browne, C. G., and Cohn, T. S. (Eds.) Thestudy of leadership. Danville. Illinois. TheInterstate Printers and Publishers, 1958.

Cleven, W. A., and Fiedler. E E. Interper-sonal perceptions of open hearth foremenand steel production. J. appl. Psychol. 1956.40. 312-314.

Cronbach, J. J., and Gleser, Goldene C. As-sessing similarity between profiles. Psych&Bull., 1953, 50, 45-73.

Fiedler, E E. Assumed similarity measures aspredictors of team effectiveness. J. &norm.sot. Psychol., 1954, 49, 381-388.

Fiedler, F. E. Leader attitudes, group climate,and group creativity. J. &norm. sot. Psy-chef., 1962,64. 308-318.

Fiedler, F. E. A contingency model of leader-t ship effectiveness. In L. Berkowtir (Ed.)

:Advances in experimental social psychology.New York: Academic Press, 1964. Vol. 1.

Fiedler, E E., and Meuwese, W. A. T. Theleader’s contribution to performance in co-hesive and uncohesive groups. J. &norm.sot. PsychoL, 1963, 67, 83-87.

Fiedier, E E., Meuwese. W. A. T., and Gonk,Sophie. Performance of laboratory tasks re-quiring group creativity. Aa Psychubgica,1961, 18, 100-119.

French, J. R. P., Jr. A formal theory of socialpower. PsychoL Rew, 1956, 63, 181-194.

Gibb, C. A. “Leadership” in G. Lindzey (Ed.)Handbook of socid psychology, Vol. II, Cam-bridge, Mass.: Addison-Wesley, 1954.

Godfrey, Eleanor P., Fiedler, E E., and Hall.D. M. Boards, rmxnagement, and cornponysuccess. Danville, Illinois: Interstate Print-ers and Publishers, 1959.

Hare, A. P. Handbook of small group research.New York: Free Press, 1962.

Hawkins, C. A study of factors mediating a re-lationship between leader rating behaviorand group productivity. Unpublished Ph.D.dissertion, University of Minnesota, 1962.

Hutchins, E. B.. and Fiedler, F. E. Task-ori-ented and quasi-therapeutic role functionsof the leader in small military groups. So-ciomctry. 1960, 23, 293-406.

Julian, J. W., and McGrath, J. E. The influ-ence of leader and member behavior on theadjustment and task effectiveness of negoti-ation groups. Urbana, Ill.: Group Effective-ness Research Laboratory, University of II-linois, 1963.

McGrath, J. E. A summary of small group re-search studies. Arlington, Va.: Human Sci-ences Research inc., 1962 (Litho.).

Mann, R. D. A review of the relationship be-tween personality and performance in smallgroups. Psychol. Bull., 1959. 56. 241-270.

Meuwese, W. A. ‘T. The effect of the leader’sability and interpersonal attitudes on groupcreativity under varying conditions ofstress. Unpublished doctoral dissertation,University of Amsterdam, 1964.

Morris, C. G., and Fiedler, E E. Applicationof a new system of interaction analysis tobe relationships between leader attitudesand behavior in problem solving groups.

Urbana, III.: Group Effectiveness ResearchLaboratory, University of Illinois, 1964.

Osgood, C. A., Suci, G. A., and Tannen-baum, R H. The rneaumt of metin&Urbana, III.: University of Illinois PEW1957.

Tk coluingency Mofle

Shaw, M. E. Annual TeGainesville, Florida:1962 (Mimeo.).

Stogdilf, R. Personal fatleadership: a surveyRychoL, 1948, 25, :

A. T. Thexmance in co-4. J. abnorm.i7.T., and Oonk,ratory tasks re-a Psychologiccr,

.eory of social63, 181-194.Lindtey (Ed. 1

I, Vol. 11, Cam-ley, 1954.

E., and Hall,and companynterstate Print-

soup research.

mediating a re-ating behaviorpublished Ph.D.innesota, 1962.:. E. Task-ori-: role functionsxy.groups. SO-6.E. The influ-

behavior on theveness of negoti-Group Effective-University of II-

small group re-Va.: Human Sci-(Litho.).

relationship be-‘ormance in small9. 56. 241-270.t of the leader’stttitudes on grouponditions ofral dissertation,1964.1 E. Application:tion analysis toleader attitudessolving groups.tiveness ResearchIllinois, 1964.and Tannen-

mm of meaning.>f Illinois Press,